hckrnws

What also happened in the same time frame according to this website:

  - [1] 1366 x 768 was the strongest growing monitor resolution
  - [2] Dual- and Quad core CPUs went up, 6-16 Cores are down
  - [3] 4 GB and 8 GB of RAM went up, 16/32 GB fell

[1] https://www.pcbenchmarks.net/displays.html

[2] https://www.pcbenchmarks.net/number-of-cpu-cores.html

[3] https://www.memorybenchmark.net/amount-of-ram-installed.html

[from 3dcenter.org : https://www.3dcenter.org/news/news-des-12-februar-2025-0 [German]]

I feel like it could simply be that some big PC recycler, perhaps in a poorer nation reselling Windows Vista machines for $15 a pop, has decided to run this benchmark as part of their recycling process and it has skewed all the results.

For me it sounds more like cheap ChromeBooks because that is enough for most people's need.

Very little in education needs more computing power than that.

We've just told ourselves that computing power will always grow and that we will always need that growth.

That might not be be true and computers might be like cars: we don't need (want) faster cars, we want cheaper and safer cars. Top speed no longer sells.

Are they currently selling new Chromebooks with 1366 x 768 screens?

Almost half the Chromebooks on newegg have that resolution: https://www.newegg.com/p/pl?d=chromebook&Order=8&Submit=ENE&...

It helps that the web scales well

Screen resolution on windows/Mac is important because scaling support is so bad that a low resolution is barely usable when all the icons end up huge.

On the web, which scales well, it really is just a matter of blurryness. Plenty of people can live with a slightly blurry screen to save $50.

As of maybe a year ago, I think essentially all new models are at least 1080p, except maybe that lenovo is still technically making/updating one older model with a 1366 x 768 specifically for the educational market.

However, 1366 x 768 screens were very common in cheaper chromebooks until 1-2 years ago, and most retailers are still selling new old stock of older models with 1366 x 768 screens and n4020 cpus.

(Now a lot of the really cheap models have 1080p screens and n100 or similar cpus which is a massive improvement.)

They're pretty common at low price points. Good displays are expensive, it's an easy thing to cheap out on to lower cost.

Yes

Explain the desktop and server numbers.

Ehhh

Having tried to install a modern Linux on a 2012 laptop, it simply is not as cozy as your memory records it.

For reference even Linux designed for old hardware takes about 6 minutes to boot and even just using the console lags

I have a couple of ThinkPad t420s's from 2011 in active use. One has windows 10 the other kubuntu. Key part was installing a $40 sata 2.5" ssd, and both operating systems boot in well under a minute, completely unoptimized.

Big revolution of last 20 years was spinning drives to ssd. Otherwise, my daily drivers are all 8th generation Intel CPUs - I just got a ThinkPad x1 carbon for $300 cad from Facebook marketplace. 7 years old and runs brilliantly. My wife's daily driver is a t580.

Which is all to say I have no idea how Linux is taking 6 minutes to boot or even worse, lagging on terminal - something is horribly wrong there :-(

Does not reflect my experience at all. T420 from 2011 is still completely usable. Haven't clocked the boot-up time but definitely not minutes, more like a couple dozen seconds. Everything worked out of box with latest stable Debian. Can do 2560x1600 over the DP connector, too. This is not even a top-spec model but the slowest i5 available back then.

To be fair, they didn't say /what/ computer from 2011 ran terribly. The Intel Core CPU's were the best back then. There were also Atom netbooks, terrible AMD chips, etc. AMD solidly has the performance crown now, but in 2011 they did not.

Pretty much anything from 2011 that was not a Core i5 or better will probably not run well today unless you use a purpose-built ultralight distro.

How is battery life?

I used a 2011 i3 dual core laptop with only HDD storage until ~2021 running linux. Boot time was nowhere close to 6 minutes and that was with luks encryption. It was maybe 90 seconds at most.

It also worked perfectly fine for basic web browsing, including youtube.

Biggest issue honestly was battery life.

You're doing something very wrong. Debian Bullseye boots in under 30 sec on my Dell XPS L321X, which was released in 2012 (timed with a stopwatch). I use the machine regularly.

SSD installs were, for the longest time, the single best ROI on upgrading a laptop. I think a SATA SSD in that 2012 bad boy could considerably boost its performance.

I still do this for friends and family: put in an SSD and maybe add a bit of RAM.

Either, clone the OS or else do a fresh install. The old drive goes into a USB enclosure if it's a laptop.

It feels like a brand new computer, going from minutes to seconds to do anything.

(Going from a Raspberry Pi 4 on a MicroSD to a pi5 with an nvme drive is a similar night and day difference. The pi5 feels like a snappy modern computer. Previous pis always felt slow.)

My 2008 MacBook with T8100 and 2GB ram boots Devuan 4 (debian11) in <30 secs and there's no lag. What hardware are you on?

He's probably on some old hardware with a dead fan that is throttling the shit out of the whole system to keep it from letting all the smoke out.

On what laptop and what distro? Does that PC lack an SSD? I have Ubuntu on a Thinkpad W520 and it does not take anywhere close to 6 minutes to boot, nor does the console lag but it also has an SSD installed.

I was just using a 2012 X230 running Arch and it's plenty snappy on the command line or even running an electron based editor (or as good as such a thing can be)

There was no mention of what distribution of linux, just "modern" whatever that means. If the person installs the lastest testversion of some distribution that tries to jam everything in there but does not have actual grapics drivers for the right card, it will be very sluggish. In the meantime my raspberry pi has no problems whatsoever with the latest linux.

I suspect those so-called, "zero-cost abstractions".

I am still using a 2012 X230 and it taking 40 seconds to boot already feels slow. Six minutes sounds like an HDD boot, but console lagging? I don't know how that could even happen - I occasionally run OpenBSD on an even older Atom netbook, and even (un-accelerated) X doesn't really lag.

That’s wild and disheartening. Those machines ran great at the time. I had a 2011 MacBook Air which I adored - it was fast, small and for the day, it had great battery life.

There is no reason for computers to get slower over time. Especially on linux.

Bodhi Linux is the one I used on a Celeron based system. The distro is purpose built for retro hardware.

I wish I could explain what is going on but it is physically painful to use, like you are remote desktoping over an iridium satellite

Technically the chipset was released in 2008 but I believe this is a 2012 laptop (*)

The computers don't get slower, and light Linux distros should still work fine.

But what matters is the software you're using, and that will be modern software (mostly). Modern browsers visiting JS-heavy websites, or a heavier office suite than 15 years ago.

And if this is the context, then computers must get slower over time, right?

> There is no reason for computers to get slower over time. Especially on linux.

https://en.wikipedia.org/wiki/Wirth's_law

Make sure that:

  - your framebuffer is hardware accelerated
  - your systemd doesn't have any timeouts or service hangups
  - run a lightweight environment such as fluxbox

i host mail with spamassassin, a static webpage, and roundcube.. on a single core 32 bit atom with 2gb ram, 500gb seagate sshd, and debian... via a wireguard tunnel to a gcp instance.

in fact i only admin it via ssh. first to the gcp instance, then to the atom inside the wg tunnel.

its not that horrible.

Apparently Windows XP is the dominant desktop OS in Armenia. They were savvy enough not to go to Vista I guess. (Not that the country is anywhere near large enough to skew global benchmarks.)

It has to be something like this. I could believe slower machines growing in market share during high inflation, but nobody's making new laptops or monitors with 1366 x 768 panels.

As other commenter said, prob has more to do with growth of computing in the developing world. Very little labor now that can't be done by someone living in a country where breakfast costs less than 25 cents. I mean, theoretically at least. Most of those people probably aren't learning from professors who spend their whole day teaching and doing research, and that kind of activity can only be supported by a relatively wealthy society. But, the internet has also democratized a lot of that activity. Well, its complicated either way.

It seems like a similar challenge as looking at the steam hardware survey stats, it's probably the best public source of information but it's too much of an overview to make anything but the most general conclusions from it. Within that data there's going to be millions of different stories if you could divide it up a bit - what are common combinations of hardware, what are common combinations in different age ranges, what is used for playing games of different ages, what's common in different locations in the world. Valve could possibly tie hardware to software purchasing activity as well what is played.

> Most of those people probably aren't learning from professors who spend their whole day teaching and doing research, and that kind of activity can only be supported by a relatively wealthy society.

Nope, as your next sentence says, they're learning off forums, Stack Overflow, YouTube and ChatGPT like the rest of us. :P

Possibly a shift from bulkier older laptops to tablet hybrids?

1366 x 768 definitely sounds like it's just an increase in older machines though.

1366x768 laptops I can find in 10 minutes of searching, all being sold as new, running from $150 to $900

HP 15-fd0023dx ($280, 12th-gen i3, Win 11 S, 8 GB RAM, 15.6" touchscreen)

HP 15-fc0025dx ($580, Ryzen 5 7520U, Win 11 S, 8 GB RAM, 15.6" touchscreen)

HP 15-fd0067nr, 15-fd0077nr ($510–$900, 13th-gen i5 or i7, Win 11 Home or Pro, 8 GB RAM, 15.6" non-touch)

HP 14-DQ0052DX Stream 14 ($230–$280, N4120, Win 11 Home or Pro, 4 or 8 GB RAM, 14" non-touch)

Lenovo IdeaPad 3 ($350, 11th-gen i3, Win 11 Home, 8 GB RAM, 15.6" touchscreen)

ASUS Chromebook CM14 ($155–$280, MediaTek Kompanio 520, 4 GB RAM, ChromeOS, 14" non-touch)

"12th-gen i3" is longer then "i3-1215U" or "1215U", but removes all important Information[1]. Why do you (and people in general) do this? I dont get it.

[1] a search engine will give you the intel ark page if you search for "1215U" which tells you thats a 2P+4E 15W Alder Lake CPU, however "12th-gen i3" could be a 4P+0E, but also a 2P+8E or a 4P+4E, and anything between 9W and 60W Basepower.

As a developer I have not felt the need to buy a performance machine anymore, everything happens on remote machines and build farms.

  As the chart is updated bi-weekly, but the data point that may change is for the current year. The first few days or weeks of a new year are less accurate compared to the end of a year.

January 2024: https://web.archive.org/web/20240123174954/https://www.cpube...

January 2023: https://web.archive.org/web/20230130185431/https://www.cpube...

Still, both previous years were notably up after the first month already. This is different and perhaps notable regardless. Either the release schedule is skewed forward or the hardware is genuinely stagnating. Or perhaps the benchmark is hitting some other CPU unrelated bottleneck.

http://web.archive.org/web/20240221001449/https://www.cpuben...

They have more than twice as many samples on their Feb 10 update this year (47810) as they did last year on Feb 14 (22761). They have shown some growth year on year in sample size, but nowhere near doubling.

That suggests this month has been an outlier in having a strangely large number of samples already, which could all be related—maybe their software started being bundled with a specific OEM or was featured on a popular content creator or whatever.

As a sibling notes, less accurate just means less accurate, not necessarily skewed upward. There simply is less data to draw conclusions on than there will be later, so any medium-sized effect that adds 20k extra samples will have a larger effect now than later.

I had the same thought. The common reasons for such a decline would be something like change of CPU architecture prioritizing energy efficiency over raw performance, limitations in manufacturing processes, or changes in benchmarking methodologies wouldn't had such steep decline. So I would guess either change of methodology or as you said weaker CPUs in general are measured by that.

So if I understand right, more old hardware makes it into the sample now than before, increasing the unreliability of early data? That makes sense I guess.

Not increasing the unreliability—it's unreliable no matter what this early in the year—but it's possible that more old hardware made it in this past month than in previous early years which would explain why the number went down this time.

This is just still a sample size of three for a data period that is stated to be less accurate" not "has lower values"

Just because the error isn't presented the same here doesn't mean it's error free...

Really makes you wish the chart showed some measure of variance

Or just have the last data point include everything from the full 12 month period before (as the title "year on year" would suggest) and maybe even put it in the correct place on the x-axis (e.g. for today, Feb 12, 2025, about 11.8% of the full year gap width from the (Dec 31) 2024 point).

And then write an article every month about how the year on year difference has(n't) gone up/down yet.

variance. Have you noticed that the mainstream public-level discussion of almost any topic never progresses farther than a point estimate? Variance implies nuance, and nuance is annoying to those who'd just rather paint a story. Variance isn't even that much nuance, because it is also just a point estimate for the variability of a distribution, not even its shape. Public discourse is stuck at the mean point estimate, and as such, is constantly misled. This is all an analogy, but feels very true.

> Have you noticed that the mainstream public-level discussion of almost any topic never progresses farther than a point estimate? [...] Variance isn't even that much nuance, because it is also just a point estimate for the variability of a distribution, not even its shape.

Next time you're doing something that isn't a professional STEM job, see how far you can get through your day without adding or multiplying.

Unless you're totting up your score in a board game or something of that ilk, you'll be amazed at how accommodating our society is to people who can't add or multiply.

Sure, when you're in the supermarket you can add up your purchases as you shop, if you want to. But if you don't want to, you can just buy about the same things every week for about the same price. Or keep a rough total of the big items. Or you can put things back once you see the total at the till. Or you can 'click and collect' to know exactly what the bill will be.

You don't see mainstream discussion of variance because 90% of the population don't know WTF a variance is.

Sometimes even an average is too much to ask. One of my pet peeves is articles about "number changed". Averages and significance never enter the discussion.

Worst are the ones where he number is the number of times some random event happened. In that case there's a decent chance the difference is less than twice the square root, so assuming a Poisson distribution you know the difference is insignificant.

Seems like there's only bandwidth of 4bit that can be reserved for this, and variance couldn't make the cut. It's actually already generous since sometimes only 1 bit can barely make it through, whether something is "good" or "bad".

This reminds me of the presidential election where most credible institutions talked about 50/50 odds and where subsequently criticized after Trump's clear victory in terms of electoral votes.

Few people bothered to look at the probability distributions the forecasters published which showed decent probabilities for landslide wins in either direction.

We still know the current data point has about three times the standard error of the previous point, but I agree it's hard to say anything useful without knowledge of the within-point variance.

So they have a dataset, and they’ve cherry picket a time period to time period, and not done any location based sorting etc, to suit a narrative.

Whatever pays the bill I suppose.

I wonder why? Some possibilities:

1. Extra silicon area being used by NPUs and TPUs instead of extra performance?

2. Passmark runs under Windows, which is probably using increasing overhead with newer versions?

3. Fixes for speculative execution vulnerabilities?

4. Most computers are now "fast enough" for the average user, no need to buy top end.

5. Intel's new CPUs are slower than the old ones

> 1. Extra silicon area being used by NPUs and TPUs instead of extra performance?

I'm not an expert in silicon design, but I recall reading that there's a limit to how much power can be concentrated in a single package due to power density constraints, and that's why they are adding fluff instead of more cores.

>2. Passmark runs under Windows, which is probably using increasing overhead with newer versions?

This is a huge problem as well. I have a 2017 15" MacBook Pro and it's a decent computer, except it is horribly sluggish doing anything on it, even opening Finder. That is with a fresh install of macOS 13.7.1. If I install 10.13, it is snappy. If I bootcamp with Windows 11, it's actually quite snappy as well and I get 2 more hours of battery life somehow despite Windows in bootcamp not being able to control (power down) the dGPU like macOS can. Unfortunately I hate Windows, and Linux is very dodgy on this Mac.

That is WILD that Windows 11 runs faster than the machine's native OS...

Could this suggest that modern Windows is, dare I say it, MORE resource efficient than modern macOS?! That feels like a ludicrous statement to type, but the results seem to suggest as much.

My first thought is that Apple is throttling performance of older machines on purpose (again). As they did with the phones.

Would explain why Windows runs faster.

> My first thought is that Apple is throttling performance of older machines on purpose (again).

The battery life argument was relevant, but where one sits on the issue is going to depend on the poison one picked.

https://en.m.wikipedia.org/wiki/Batterygate

Modern macOS might be optimized for Apple Silicon CPUs. But even when it was Intel only, there were probably times when Windows was lighter, albeit bad in other ways.

It’s a microkernel OS, it’s going to be slower by its very nature. And on ARM it wastes memory due to its 16KiB pages.

Neither of them use microkernels. They are monolithic kernels with loadable module support ("hybrid" kernels). The cores of both Windows NT and XNU were originally microkernels, but then they put all of the servers into the kernel address space to make them into the monolithic kernels that they are today.

XNU apparently stands for "X is not Unix." Well that's confusing.

It actually was a recursive acronym that stood for XNU is Not UNIX. Then at some point Apple got it UNIX certified as part of MacOS.

> It’s a microkernel OS, it’s going to be slower by its very nature.

Yawn. Ever heard of https://en.wikipedia.org/wiki/L4_microkernel_family ?

Or https://en.wikipedia.org/wiki/Genode ?

(Wenn man keine Ahnung hat...)

You're comparing a modern microkernel to something originating from the 80s?

Keep your rude attitude to yourself.

It's from 1993. Also be more precise about what exactly you are meaning by writing "It’s a microkernel OS, it’s going to be slower by its very nature."

Otherwise don't be offended when people are offended by regurgitated and meanwhile disproven non-sense.

kthxbye

Why is this convo about kernels so angry? And macOS isn't even based on a microkernel.

i forgot ;)

The pessimistic viewpoint is the hardware vendor would not mind if you felt your machine was slow and were motivated to upgrade to the latest model every year. The fact that they also control the OS means they have means, motive and opportunity to slow older models if their shareholders demanded.

Or they simply keep supporting old models with new versions of the OS, even though newer software versions are optimized, and contain new features, enabled by newer hardware improvements.

If you machines have more RAM, you can use a more RAM intensive solution to speed many things up, or deliver a more computationally intensive but higher quality solution, or simply add a new previous challenging feature.

What would be interesting, is how fast to old but high spec'd models slow down. What slow downs are from optimizing for newer architectures, vs. optimizing with expectation of higher resources.

The problem seems to be CPU bound, it's got 16GB of memory and memory pressure is low. The CPU is a i7-7820HQ, though I thought that it was interesting that my iPhone XR (Apple A12) scores higher on a synthetic benchmark than my top of the line MacBook Pro from the same time.

I'm just as surprised. Also, I was using Windows 11 LTSC 2024, not the standard version, which could impact the validity of my comparison.

This doesn't feel terribly surprising to me. MacOS has always had impressively performant parts, but their upgrades always generally lower responsiveness. On modern hardware it's less perceptible obviously, and they want to sell machines, not software. But the last iteration that felt like it prioritized performance and snappiness was Snow Leopard, now twenty years ago.

I will say the problem was a lot worse before the core explosion. It was easy for a single process to bring the entire system to a drag. These days the computer is a lot better at adapting to variable loads.

I love my macs, and I'm about 10x as productive as on a windows machine after decades of daily usage (and probably a good 2x compared to linux). Performance, however, is not a good reason to use macos—it's the fact that the keybindings make sense and are coherent across an entire OS. You can use readline (emacs) bindings in any text field across the OS. And generally speaking, problems have a single solution that's relatively easy to google for. Bad-behaved apps aside (looking at you zoom and adobe) administering a mac is straightforward to reason about, and for the most part it's easy to ignore the app store, download apps, and run them by double clicking.

I love linux, but I will absolutely pay to not have to deal with X11 or Wayland for my day-to-day work. I also expect my employer to pay for this or manage my machine for me. I tried linux full-time for a year on a thinkpad and never want to go back. The only time that worked for me was working at google when someone else broadly managed my OS. macs the only unix I've ever used that felt designed around making my life easier and allowing me to focus on the work I want to do. Linux has made great strides in the last two decades but the two major changes, systemd and wayland, both indicate that the community is solving different problems than will get me to use it as a desktop. Which is fine; I prefer the mac-style to the ibm pc-style they're successfully replacing. Like KDE is very nice and usable, but it models the computer and documents and apps in a completely different way than I am used to or want to use.

I love Linux and run it on my servers, but on the desktop, it requires too much tinkering—I often spend more time troubleshooting than working. Laptops are especially problematic, with issues like Wi-Fi, sleep, and battery life. Windows is a mess—I hate the ads, bloat, and lack of a Unix-like environment. WSL2 works but feels like a hack. macOS, on the other hand, gives me full compatibility with Unix tools while also running Office and Adobe apps. Also Command+C for copy, Command+V for paste is much nicer than Control+Shift+C and Control+Shift+V. macOS does have absolutely terrible screen snapping though in comparison to Windows 11. The hardware is solid—great screen, trackpad, speakers, and battery life. I considered getting a high-end PC laptop (based on Rtings’ recommendations), but every option had compromises, either a terrible screen, terrible processor or terrible battery life. By the time I configured it to a non crappy config, it would’ve cost more than a MacBook Pro 16.

> Laptops are especially problematic, with issues like Wi-Fi, sleep, and battery life.

100% true, but I love Linux as a daily driver for development. It is the same os+architecture as the servers I am deploying to! I have had to carefully select hardware to ensure things like WiFi work and that the screen resolution does not require fractional scaling. Mac is definitely superior hardware but I enjoy being able to perf test the app on its native OS and skip things like running VMs for docker.

Same! That's One less OS that I have to remember how it works!

I'm not sure I understand the whole tinkering thing. Whenever I tinker with my Linux, it's because I decided to try something tinkery, and usually mostly because I wanted to tinker.

Like trying out that fancy new tiling Wayland WM I heard about last week...

I feel like the main times Linux requires tinkering are:

1. You're trying to run it on hardware that isn't well-supported. This is a bummer, but you can't just expect any random machine (especially a laptop) to run Linux well. If you're buying a new computer and expect to run Linux on it, do your research beforehand and make sure all the hardware in it is supported.

2. You've picked a distro that isn't boring. I run Debian testing, and then Debian stable for the first six months or so after testing is promoted to stable. (Testing is pretty stable on its own, but then once the current testing release turns into the current stable release, testing gets flooded with lots of package updates that might not be so stable, so I wait.) For people who want something even more boring, they can stick with Debian stable. If you really need a brand-new version of something that stable doesn't have, it might be in backports, or available as a Snap or Flatpak (I'm not a fan of either of these, but they're options).

3. You use a desktop environment that isn't boring. I'm super super biased here[0], but Xfce is boring and doesn't change all that much. I've been using it for more than 20 years and it still looks and behaves very similarly today as it did when I first started using it.

If you use well-supported hardware, and run a distro and desktop environment that's boring, you will generally have very few issues with Linux and rarely (if ever) have to tinker with it.

[0] Full disclosure: I maintain a couple Xfce core components.

First, thanks for Xfce. I'm a (tiny) donor.

Two kinds of linux tinkering often get aliased and cause confusion in conversations.

The first kind is the enthusiast changing their init system bootloader and package manager and "neofetch/fastfetch" and WM and... every few weeks.

The second kind is the guy who uses xfce with a hidpi display who has to google and try various combinations of xrandr(augmented with xfwm zoom feature), GDK_SCALE, QT_SCALE_FACTOR, theme that supports hidpi in the titlebar, few icons in the status tray not scaling up(wpa_gui), do all that and find out that apps that draw directly with OpenGL don't respect these settings, dealing with multiple monitors, plugging in hdmi and then unplugging messing up the audio profile and randomly muting chromium browsers, deciding whether to use xf86-* Or modesetting or whatever the fix is to get rid of screen tearing. Bluetooth/wifi. On my laptop for example I had to disable usb autosuspend lest the right hand side USB-A port stop working.

If our threshold for qualifying well-supported hardware is "not even a little tinkering required" then we are left with vanishingly few laptops. For the vast vast majority of laptops, atleast the things I mentioned above are required. All in all, it amounted to couple of kernel parameters, a pipewire config file to stop random static noise in the bluetooth sink and then a few xfce setting menu tweaks (WM window scaling and display scaling). So not that dramatic, but it is still annoying to deal with.

The 2nd kind of tinkering is annoying, and is required regardless of distro/de/wm choice since it's a function of the layers below the de/wm, mostly the kernel itself.

I think that for your example of having problems with HiDPI you might have had in mind another desktop environment than XFCE.

I have been using XFCE for more than 10 years almost exclusively with multiple HiDPI monitors.

After Linux installation I never had to do anything else except of going to XFCE/Settings/Appearance and set a suitably high value for "Custom DPI Setting".

Besides that essential setting (which scales the fonts of all applications, except for a few Java programs written by morons), it may be desirable to set a more appropriate value in "Desktop/Icons/Icon size". Also in "Panel preferences" you may want to change the size of the taskbar and the size of its icons.

You may have to also choose a suitable desktop theme, but that is a thing that you may want to do after any installation, regardless of having HiDPI monitors.

I have a 2016 MBP-15 (sticky keyboard). I suspect Apple changed something so the fans no longer go into turbo vortex mode. Normally it isn't sluggish at all, but when it overheats, everything grinds to a halt.[1] (Presumably this is to keep the defective keyboard from melting again.[0]) Perhaps old OS/bootcamp still has the original fan profiles.

[0]Apple had a unpublicized extended warrantee on these, and rebuilt the entire thing twice.

[1] kernel_task suddenly goes to 400%, Hot.app reports 24%. Very little leeway between low energy and almost dead.

I think it's that the measure of modern CPU performance- multithreaded performance is worthless and has been worthless forever.

Most software engineers don't care to write multithreaded programs, as evidenced by the 2 most popular languages - Js and Python having very little support for it.

And it's no wonder, even when engineers do know how to write such code, most IRL problems outside of benchmarks don't really yield themselves to multithreading, and due to the parallizable part being limited, IRL gains are limited.

The only performance that actually matters is single thread performance. I think users realized this and with manufacturing technology getting more expensive, companies are no longer keen on selling 16 core machines (of which the end user will likely never use more than 2-3 cores) just so they can win benchmark bragging rights.

> The only performance that actually matters is single thread performance. I think users realized this and with manufacturing technology getting more expensive, companies are no longer keen on selling 16 core machines (of which the end user will likely never use more than 2-3 cores) just so they can win benchmark bragging rights.

How can you state something like this in all seriousness? One of the most used software application has to be the browser, and right now firefox runs 107 threads on my machine with 3 tabs open. gnome-shell runs 22 threads, and all I'm doing is reading HN. It's 2025 and multicore matters.

Those threads don't necessarily exist for performance reasons - there can be many reasons one starts a thread from processing UI events, to IO completion etc. I very much doubt Firefox has an easy time saturating your CPU with work outside of benchmarks.

If Firefox smears its CPU usage over multiple threads, that leaves more single-threaded performance on the table for other apps that may need it. So there could still be an effect on overall system performance.

Well yeah - what CPU bound task do you need to be performant? Beyond many tabs - which is embarrassingly parallel - it's all either GPU, network or memory bound.

Firefox failing to saturate your CPU is a win-state.

>multithreaded performance is worthless and has been worthless forever.

I have a very opposite opinion; single threaded performance only matters to a point where any given task it's doing isn't unusable. Multithreaded performance is crucial from keeping the system from grinding to a halt because users always have multiple applications open at the same time. Five browser windows with 4-12 tabs on each, on three different browsers, 2-4 Word instances, some electron(equivalent) comms app, is much less unusual than I'd like it to be. I have used laptops with only two cores and it gave a new meaning to slow when you tried doing absolutely anything other than waiting for your one application to do something. Only having one application open at a time was somewhat useable.

> The only performance that actually matters is single thread performance.

Strong disagree, particularly on laptops.

Having some firefox thread compile/interpret/run some javascript 500 microseconds faster is not going to change my life very much.

Having four extra cores definitely will: it means i can keep more stuff open at the same time.

The pain is real, particularly on laptops: i've been searching for laptops with a high-end many-core cpu but without a dedicated gpu for years, and still haven't found anything decent.

I do run many virtual machines, containers, databases and stuff. The most "graphic-intensive" thing i run is the browser. Otherwise i spend most of my time in terminals and emacs.

The fact that multithreaded performance is worthless for you does not prove that this is true for most computer users.

During the last 20 years, i.e. during the time interval when my computers have been multi-core, their multithreaded performance has been much more important for professional uses than the single-threaded performance.

The performance with few active threads is mainly important for gamers and for the professional users who are forced by incompetent managers to use expensive proprietary applications that are licensed to be run only on a small number of cores (because the incompetent managers simultaneously avoid cheaper alternatives while not being willing to pay the license for using more CPU cores).

A decent single-threaded performance is necessary, because otherwise opening Web pages bloated by JS can feel too slow.

However, if the single-threaded performance varies by +/- 50% I do not care much. For most things where single-threaded performance matters, any reasonably recent CPU is able to do instantaneously what I am interested in.

On the other hand, where the execution time is determined strictly by the multithreaded performance, i.e. at compiling software projects or at running various engineering EDA/CAD applications, every percent of extra multithreaded performance may shorten the time until the results are ready, saving from minutes to hours or even days.

Lots of problems can be nicely parallelized, but the cost of doing so usually isn't worth it, simply because the entity writing the software isn't the entity running it, so the software vendor can just say "get a better PC, I don't care". There was a period when having high requirements was a badge of honor for video games. When a company needs to pay for the computational power, suddenly all the code becomes multithreaded.

Yes but look at the chart in the article. Both multi-threaded and single-threaded performance is getting slower on laptops. With desktops, multi-threaded is getting slower and single-threaded is staying the same.

Superscalarity is largely pointless yes, given that memory access between threads is almost always a pain, so two of them rarely process the same data and can't take advantage of using a single core's cache at the same time. It doesn't even make much sense in concept.

But multicore performance does matter significantly unless you're on a microcontroller running only one process on your entire machine. Just Chrome launches a quarter million processes by itself.

Please open your task manager and see how much stuff is running.

The times where you saw one core pegged and the rest idle are a decade ago.

Programmers failing to manufacture sufficient inefficiency to force upgrade.

Can't force upgrade with money you don't have. For a non-enthusiast, even the lowest specs are more than good enough for light media consumption and the economy would affect what they invest in.

As an experiment, I've tried working with a Dell Wyse 5070 with the memory maxed out. Even for development work, outside of some egregious compile times for large projects, it actually worked ok with Debian and XFCE for everything including some video conferencing. Even if you had money for an upgrade, it's not clear it's really necessary outside of a few niche domains. I still use my maxed out Dell Precision T1700 daily and haven't really found a reason to upgrade.

I love my wyse 5070 boxes. I keep telling people to buy them instead of pis

As an enthusiast, I’ve found mid-tier hardware from over a decade ago can run the majority of games on medium/high without much problem. And that the majority of people in my life only need a laptop that can consistently stream 1080p and run a modern browser, maybe with some extra bits like Microsoft Office (although most younger users use G Suite, presumably because schooling preferred it growing up).

I very much doubt that is the cause.

I think it's reason #4. I bought a PC 12 years ago, the only upgrade I did was buying an SSD, and now it's being used by my father, who's perfectly happy with it, because it's fast enough to check mail and run MS Office. My current gaming PC was bought 4 years ago, and it's still going strong, I can play most games on high settings in 4k (thank you DLSS). I've noticed the same pattern with smartphones - my first smartphone turned into obsolete garbage within a year, my current smartphone is el cheapo bought 4 years ago, and it shows no signs of needing to be replaced.

ok I lied, my phone is only two years old, but what happened is that two years ago my phone experienced a sudden drop test on concrete followed by pressure test of a car tyre, and it was easier to buy a new one with nearly identical specs than to repair the old one.

there's 4 which has a couple different component

* battery life is a lot higher than it used to be and a lot of devices have prioritized efficiency

* we haven't seen a ton of changes to the high-end for CPU. I struggle to think of what the killer application requiring more is right now; every consumer advancement is more GPU focused

COVID. People start working from home or otherwise spending more time on computers instead of going outside, so they buy higher-end computers. Lockdowns end, people start picking the lower-end models again. On multi-threaded tasks, the difference between more and fewer cores is larger than the incremental improvements in per-thread performance over a couple years, so replacing older high core count CPUs with newer lower core count CPUs slightly lowers the average.

I don't buy this really. COVID to now is less than 1 laptop replacement cycle for non-techie users who will usually use a laptop until it stops functioning, and I don't think the techie users would upgrade at all if their only option is worse.

A lot of laptops stop functioning because somebody spills coffee in it or runs it over with their car. There are also many organizations that just replace computers on a 3-5 year schedule even if they still work.

And worse is multi-dimensional. If you shelled out for a high-end desktop in 2020 it could have 12 or 16 cores, but a new PC would have DDR5 instead of DDR4 and possibly more of it, a faster SSD, a CPU with faster single thread performance, and then you want that but can't justify 12+ cores this time so you get 6 or 8 and everything is better except the multi-thread performance, which is worse but only slightly.

The reticence to replace them also works against you. The person who spilled coffee in their machine can't justify replacing it with something that nice, so they get a downgrade. Everyone else still has a decent machine and keeps what they have. Then the person who spilled their drink is the only one changing the average.

> (increased overhead with newer versions of Windows)

> Fixes for speculative execution vulnerabilities?

I don't know if they'll keep doing it, but hardware unboxed had been doing various tests with Windows 10 vs 11, and mitigations on vs off, as well as things like SMT on vs off for things like AMD vcache issues, or P cores vs E cores on newer intels... it's interesting to see how hardware performs 6-12 months after release, because it can really go all over the place for seemingly no reason.

It's either the Chromebook effect: people are just buying slower computers, possibly in poorer-ish countries.

Or: The average speed of new computers just isn't going up. People are just buying lower end computers because they're good enough. I think it's mostly that: the average is just trending a bit lower because lower is fine. The thirst for the high end is disappearing for most segments, because people don't need it.

4 has been true for over a decade.

Literally the only reason you need a computer less than 10 years old for day to day tasks is to keep up with the layers of javascript that are larded onto lots of websites. It's sad.

Edit: oh and electron apps. Sweet lord, shouldnt we have learned in the 90s with java applets that you shouldnt use a garbage collected language for user interfaces?

I use a GC'd language for UI every day -- Emacs -- and it runs well even on potato-tier hardware. The first GUI with all the features we'd recognize was written in a GC'd language, Smalltalk.

GCs introduce some overhead but they alone are not responsible for the bloat in Electron. JavaScript bloat is a development skill issue arising from a combination of factors, including retaining low-skill front end devs, little regard for dependency bloat, ads, and management prioritizing features over addressing performance issues.

There's also a lot of fast software written in Golang. GC has an undeserved bad reputation.

It's all nice and fast until the GC eats up all the CPU time under memory pressure (e.g. container memory limit). GC deserves its bad rep, even if GC implementations have gotten a lot better over time.

GC also disqualifies a language from ever being used for anything but userland programs.

Containers for Go software have no sense. You can cross compile stuff from anywhere to anywhere. And the binaries are self contained.

Which isn't really going to happen or be noticeable in this case. Yeah GC is slightly slower, but it's like blaming a 10-minute mile on your shoes. SDN or DBMS written in a GC'd language is probably not a good idea.

It's not that they lack the skill. It's that they don't care.

If they do care then they probably depend on someone who doesn't.

I hate the framing that developers are the people not caring. It's a ridiculous misrepresentation of the dynamics I've seen everywhere. Developers usually love optimizing stuff. However, they are under pressure to deliver functional changes. One could blame "the business" for not caring and not making time for shipping a quality product. However, these business decisions are made because that's what wins customers. Customers might complaint that that wasn't more lightweight software. But clearly their stated preference doesn't match their revealed preference.

My laptop is 11 years old and handles Slack quite well, and Jira. I eventually maxed it out at 32 GB and that probably helps. It's only an i7-4xxx though.

Java in the 90s was really slow. It got much faster with the JIT compiler. JavaScript and browsers got many optimizations too. My laptop feels faster now than it was in 2014 (it has always run some version of Linux.)

The other problem of Java was the non native widgets, which were subjectively worse than most of today's HTML/JS widgets, but that's a matter of taste.

I think even a blank Electron app is rather heavy because it's basically a whole web browser, which btw is written in C++. BUT my 2015 baseline MBP still feels fine.

I hate bloat and don't use Electron apps, but garbage collection is such a silly thing to pin it on. Tossing stuff like Lua and D in the same bin as the massive enterprise framework du jour is throwing the baby out with the bathwater.

For me the big improvement from upgrading was better responsiveness while on a video call (no small improvement in the modern world)

What is wrong with garbage collection and UI?

Latency spikes, probably. But most UIs shouldn't be churning through enough allocations to lead to noticeable GC pauses, and not all of today's GCs have major problems with pause times.

I don't think Electron's problems are that simple.

From a non expert point of view, it’s about bringing in so much dead code. Like support for every os audio subsystem for a graphic app, or the whole sandboxing thing while you’re already accessing files and running scripts. Or even the table and float layout while everything is using flex. All those things need to be initialized and hooked up. At least Cordova (mobile) runs only on the system web engine which was already slimmed down.

I work on an Electron app, and we have a ticket open to investigate why the heck it's asking the OS for Bluetooth permissions even though we'd never use it. There are, of course, higher priority things to get to first, bugs that have larger effects. I'd love to be able to get to that one…

The renderer isn't going to pause on JS execution or GC.

[dead]

I have a 2005 Core 2 Duo desktop running Windows 7. The only things that make it unusable are today's Internet and Electron apps. Everything else (including Visual Studio 2019!) runs at a performance comparable to my 2020 ThinkPad with Core i7. I know it's anecdotal, I'm just saying, if we fixed the bloat on the web, most people could just keep running their 10-15 year old PCs (using Linux if you want).

i use a 10 years old macbook air and honestly i can do everything i need, statistics, light programming, browsing and photo editing. perfect.

i wait it to break before moving on, changing for the sake to changing feels like waste.

I use a 10 years old macbook pro and the only thing I dislike is the battery life and the heat generated.

I simply can't afford the replacement 16" MBP right now. Hopefully it lasts another couple of years.

Surely, the new MacBook Airs offer better performance than a 10 year old MacBook Pro. The M3 is very reasonable, 16GB/512GB for $1,100 or $1,300 for 15 inch I think. Could be going lower too with M4 MacBook Air purportedly releasing in Mar or Apr.

I have cheapest air 8gig and its just fine for anything i do if i dont open 80tabs. But for sure wait for M4 as 16gig ram will be the base model. Coworker just got MacMini m4 and i dont see why that machine wouldnt be enough for 90% of people.

My ten year old computer has an SD card slot and an HDMI port.

I would like to replace it with something that has an SD card slot and an HDMI port, as I use those frequently and don't want to deal with adapter solutions.

I just wish they upped the storage a bit. 512GB on an $1100 machine in 2025 just feels like a bad deal, especially if you already have a machine from 10 years ago that has the same amount.

But has it been true for the type of person who runs CPU benchmarks?

4 and 1.b: extra silion being used by GPUs. CPU performance isn't that important anymore.

I know, I know: all the software you use is slow and awful. But it's generally bad thanks to a failure to work around slow network and disk accesses. If you use spinning rust, you're a power user.

It's also a minority of video games that rely more on CPU performance than GPU and memory (usually for shared reasons in niche genres)

You’re forgetting the most likely possibility: it’s an artifact of data collection or a bug in the benchmarking software.

> 2. Passmark runs under Windows, which is probably using increasing overhead with newer versions?

Shouldn't be an issue. Foreground applications do get a priority boost; I don't know if Passmark increases priority of itself, tho. Provided there isn't any background indexing occurring, i.e. letting the OS idle after install.

As someone obsessing over small performance gains for his processor recently, there is definitely overhead to consider. Note by default you get things now like Teams, Copilot, the revised search indexing system, Skype, etc. Passmark in many tests will max out all cores at once; this tends to make it sensitive to 'background' processes (even small ones).

People are buying more HP Streams and fewer ASUS ROGs than in previous years?

ARM?

Maybe it's all the new Intel CPUs that failed?

It’s everyone jumping ship and switching from 14th gen to AMD

The number of people who swap CPUs or even pay attention to the "Intel inside" sticker on a PC must be very small.

I mean the CPUs were tested early on, but later they failed, and no longer raise the average. It should be visible in more detailed statistics if it is so.

People spending less on new computers.

My hunch is that more people in countries like India use cpubenchmark / buy computers and try out cpubenchmark; and that in countries like those lower performing laptops outsell more powerful laptops by an enough large number that it shows.

It doesn’t have to be literally India, it’s an example for illustration.

Or the ‘slow CPUs’ are actually truly Fast Enough. Intel N100 and N150 systems are low key amazing value.

I bought into the N100 hype train because of Reddit. Sure its an amazing value, but I hope anyone reading this isn't convinced it's remotely fast. I ended up going with a minisforum ryzen for about 2x the price and 4x the performance.

I was a bit bummed since I wanted to use it as a kind of streaming box, and while it can do it, it is definitely slow.

I got a 16GB RAM 512GB SSD N100 minipc for less than $150 last month. Yes it could be faster (always true for any computer), but I feel I got way more than I paid for. Certainly a much better deal than an RPi 5.

PS. If you need more power than that I have a hard time coming up with a better deal than M4 Mac Mini, perhaps with some extra usb storage. It's possibly the only non-workstation computer that is worth buying in the not-inexpensive category (in the base version, obviously).

> it could be faster (always true for any computer)

The question isn't whether it could be faster, but whether it's slow enough to be annoying for the tasks you use it for.

I bet most people would find the N100 annoyingly slow when opening documents or browsing the web, or doing crazy things like opening a few programs at the same time.

> I bet most people would find the N100 annoyingly slow when opening documents or browsing the web, or doing crazy things like opening a few programs at the same time.

I'm actually not convinced. If you're used to a fast box, yes; I'd argue most people actually aren't.

I’m annoyed opening a doc pretty much daily on a year old MacBook Pro.

However the blame lies squarely with Microsoft. What the hell is their office suite up to?

Sounds possible, office laptops in India are generally shit-tier. But the largest non-bulk retail buyers of laptops in India seem to be (from my armchair market observation) school/college kids, and school/college kids increasingly prefer gaming laptops (India is the largest consumer of gaming laptops after the US) that will likely not give below-average benchmark results.

Could it be related to world-wide inflation and decreasing real wages? Salaries have not kept up, hence people can't afford as powerful hardware as they used to. It also seems that top-of-the-line hardware don't depreciate as fast as it used to. Many several year's old GPUs are still expensive, for example.

Someone already gave the answer:

https://news.ycombinator.com/item?id=43030465

It's caused by people failing to read the fine print on the graph.

I had the same instinct as you, but this comment changed my mind:

https://news.ycombinator.com/item?id=43030614

Nah, I read that but I'm not convinced. The reply to them from palijer is correct: inaccurate does not mean lower, it just means inaccurate. Fewer samples means less data means more room for weird sampling errors skewing the data. There's nothing inherent in that that suggests we ought to assume an upward skew.

The best we can say is that the sampling skew this month is different than the sampling skew in past Januaries. That could be due to some interesting cause, it could be totally random, and there's zero point in speculating about it until we have more data.

> Many several year's old GPUs are still expensive,

They are propped up by demand and the fact that most of the new GPUs are marginally better than previous ones.

New GPUs are quite a bit better than previous ones but perf oer dollar has been flat for a while now.

Also, if you're talking gaming GPUs old ones work fine given there has not been a new PlayStation or Xbox in many years. Min spec for many games is 5 years old tech

For gaming GPUs the 5090 is something like 20-50% better than the 4090 in raster performance.

Apparently also 50% more likely to make things catch on fire! ;-)

Yep. The whole fake frames thing seems like a witch hunt to me. If you can actually get a 5090 and were previously fully utilising a 3090 or 4090, it is a significant upgrade, as long as it doesn't burn your house down.

In situations where the budget matters, you need to take that 20-50% better performance and drop it by the 25% higher launch price.

And the generated frames are misleading because they don't work very well if your source framerate is low.

The generated frames thing is interesting as it's an obvious example of the whole discussion around halo products (the x080 or x090 cards), "the rich get richer", it takes up a lot of air in the room but is only appropriate for a small vocal proportion of the market.

As you say you need a good framerate to start with before generated frames make sense so either you're already running well with a high tier card and are further able to show off the combination of framerate/resolution/detail level, or running a lower tier card at lower settings turning the graphics down further which can be a very obvious trade off. Less demanding games which are generally online competitive usually wouldn't do as well with any extra latency introduced, and the situation where I've heard it would be a good fit is emulators or any kind of game where the CPU is the limiting factor

I just don't think the 5090 is for gamers; it's for hobbyist AI people who want to buy multiple to run high parameter LLMs. Fake frames are irrelevant to this crowd, even if they take one of their 5090s aside to game once in a while.

thats the only GPU that is worth the upgrade. All the other ones are meh.

Sure they are better... but once you factor for die size, core count, wattage and so on... the improvements being made are less impressive than they first seem.

That would not change the expectation of having faster hardware with time at the same/lower price.

Discussion from a day or so ago. 59 comments:

https://news.ycombinator.com/item?id=43017612

Given the strong prior pattern, I forecast this is due to

(a) changes in methodology/non-random sampling, with 60 % probability,

(b) random sampling effects that go away on longer timelines, with 20 % probability,

(c) any actual change in CPUs, with 20 % probability.

In my experience, when a metric shows weird behavious these are usually the reasons with roughly those proportions.

Friend of mine bought Zephyrus G16 with Intel 185H and 32GB RAM. He thought this will help him studying and he chose it because some software he has to use is only available for Windows. He called me to have a look as the laptop has been sluggish for him.

This thing barely can handle Office, Teams and the browser. Hot, noisy as hell and performance wise I see no difference over laptops from decade ago. Tragic.

To be fair I don't think I could do anything. Task manager showed CPU as under utilised, yet fans were blasting, editing Word document looked like a slideshow.

ASUS tools showed no problems.

I don't know, feels like people are getting scammed with these new laptops.

I still have M1 Max and never experienced anything like this.

Had something similar with my work laptop today. Lenovo X1 Carbon, couple of years old. Got reinstalled very recently.

Been fine, but suddenly it was slow af. Near zero CPU, GPU and disk usage in Windows Task Manager, but I could feel it was burning hot, which it wasn't just 15 minutes prior.

Did a reboot and all was fine again.

Surely some firmware that messed up, though no idea why.

Anyway, I'd start by removing crap, that goes for 3rd party anti-virus and vendor tools especially. Use something like Bulk Crap Uninstaller[1] or Revo[2], and then reinstall drivers.

Totally agreed on the sad state of laptops these days.

[1]: https://www.bcuninstaller.com/ (open source)

[2]: https://www.revouninstaller.com/products/revo-uninstaller-fr...

What were the disk specs? HDD or SSD? GPU?

Windows is a heavy operating system compared to the others and can cause that problem but that likely isn't Windows alone causing the problem (though bloatware / other AV solutions could also have something to do with it)

1TB NVMe and 4060 GPU

Sounds a lot like a dedicated GPU is running and heat-throttling the whole thing, since you don't see CPU usage.

I wanted to say gaming laptops are a scam, but the older Intel MBPs with dedicated GPUs suffered too. More trouble than they're worth.

One thing I notice is that Google Drive was using 15% of ARC GPU (laptop also has nVidia 4060). When I shut down that process, it cooled down a bit after a while, but system is still sluggish and fans work all the time, even if I switched it to "Silent" mode.

Does it have an option to shut off the dGPU entirely and rely on integrated? Just as a test.

Uninstall all security except for Windows defender and see how it feels.

It's an Intel powered gaming laptop with discrete GPU running Windows. Of course it's gonna heat like crazy. But that box can also do things an M1 can't do. Like play games... And heat up a small room.

The discrete GPU is barely used, so it shouldn't be getting hot.

Yeah, that's the Intel part. They do that. AMD would have been much better although still not M1 cool.

Interestingly, so did server cpu thread performance, I guess we're now starting to see more direct prioritization of core count versus core performance, which I kind of understand, it's superficially simpler to think about horizontal scaling, and for very many workloads, it's ideal, and for many it's at least doable, though I fear the invisible overhead in both compute and complexity

I have a theory that cpu performance in laptops plateauing is something we'll see more of in the future, as manufacturers further optimize for battery life and being very thin. 15 years ago I wanted a very powerful MacBook pro. Now I'm fine with a MacBook air and if I need to do something that requires heavy lifting, I have remote access (ssh, vnc over ssh, etc) into a loud, beefy hypervisor located somewhere else with 512GB of RAM in it.

On the consumer level and for non technical end users, as more functions are offloaded to "the cloud" with peoples' subscriptions to office365, Google workspace, iCloud, whatever, having a ton of cpu power in the laptop also isn't as necessary anymore. Along with things like h.265 and av1 video encode and decode being done with GPU assist rather than purely in software.

Some 15 years ago I was playing DOTA on what you'd call today a gaming laptop (HP HDX9200) which would occasionally overheat causing BSOD and everytime it happened it was hail mary, because the boot time took around 5 minutes and it was the exact time for servers to flag me as leaver and ban me from future games.

These days I have two main machines which boot to Windows in about a minute.

Somehow I remember my old 486 machine was able to boot to DOS within the snap of a power button.

My experience with recent hardware is that it is very poorly designed, not as "computational hardware", but as a "consumer product".

Buying anything new for me is out of question since a few years ago, because it just cannot be guaranteed to survive half a year, hence now I always buy second-hand, because this way at least it was tested by the previous owner.

But even so, modern hardware is hellishly unreliable.

I bought a Dell 7 series laptop (17 inches), and:

  1. Already had to replace the battery back panel three times, just because it is being held by tiny plastic hooks which are easily torn away. 
  2. It's assembly-reassembly process is a nightmare (which used to not be the case for Dell in the past).
  3. Already had to replace the fans (the laptop is not even 5 years old).

  4. When running with a smaller battery (official), it CANNOT RUN THE GPU AT FULL SPEED EVEN WHEN PLUGGED INTO THE MAINS. Really? This is just ridiculous. A laptop using the battery when plugged in is just insane.
  5. You MUST use a smaller battery to install a 2.5" SATA drive. So you much choose: either a SATA drive, or a GPU.
  6. The power on/off button does not work if your BIOS battery has low charge. This is just maddening! What is the connection?

The amount of cacti you have to swallow to get used to modern hardware is just staggering.

Here's my thoughts,

- The "Top CPU Performance to Date" has flatlined, so even if the hypothesis that developing countries are running PassMark is true, it doesn't explain everything.

- The recent failures of high end Intel chips might be a factor.

- This data is after the initial round of developers being replaced by AI and outsourcing.

- PassMark doesn't include AI or ML benchmarks that the next generation hardware prioritizes.

- CPU top clock speeds have been going down for more than a decade now, instead prioritizing the number of cores.

- HPC labs are less likely to use PassMark because of licensing issues whereas free and open source alternatives exist.

- The average consumer doesn't know or care about the difference. Instagram is still limited to around 1000x1000 pixel images and at most 1080p video with 30 fps.

Personally, this signals that PassMark is not a good benchmark because I've seen a significant increases in my applications from recent CPU improvements like with AMD's EPYC offerings.

The chart also shows that single-thread performance hasn’t improved that much in the past twelve or so years, compared to Moore’s law. And this is compounded by Wirth’s law [0].

[0] https://en.wikipedia.org/wiki/Wirth%27s_law

It's always interesting when I install Linux on a machine, and it can fit all running processes on a single page in top.

Arguably a majority of improvements in 2025 haven’t happened yet, seems premature to conclude anything in February.

Looking at the top end chart, which flat lined, it seems that the biggest contributor would be a slower release cycle of the most performant chips. It looks like the score of the top end chips were dragging up the average. With the top spot not changing, the average is falling.

I'm curious what median and std dev look like.

Hyperthreading and speculative execution are going out of fashion?

Now if we could just get rid of fake frames on the GPU side.

It just got introduced and NVDA needs to inflate its stock more.

But yes, I agree on principle: https://gist.github.com/bazhenovc/c0aa56cdf50df495fda84de58e...

I'd suspect some sort of economic factor, or sampling … e.g. just people with lower specs submitting and bringing down the average

Well, Intel abandoned hyperthreading with Lunar Lake, so it's definitely going out of fashion.

For Intel, yes.

I'm using a thinkpad t480 and it runs everything fine

My favorite computer is my t530. It's showing its age a bit, but the newer thinkpads I've used all feel like a step down quality wise

I'll probably use the t530 as my primary machine for years to come.

It’s remarkable how carefully the page is worded to not incorporate Apple’s desktop and mobile CPU performance changes over time - but they screwed up because they said Linux, which doesn’t exclude Apple Silicon like it used to thanks to Asahi.

Ultimately Apple has hit the same dead end too. Their designs and TSMC's processes aren't putting Moore's law back on the map, you can see it in the year-over-year performance. If tariffs end up hitting any harder, Apple can't just go to TSMC's Arizona plant and get faster silicon than what Taiwan has today.

> they screwed up because they said Linux, which doesn’t exclude Apple Silicon

In the context of the second graph, that means Apple Silicon is also failing to push the envelope forwards.

Apple has one advantage that PCs do not: each silicon release cycle is accompanied by a coordinated operating system and compiler release for CPU, GPU, and NPU workloads; so that they can hardware-accelerate functions inside the operating system and programs compiled to target the hardware far more rapidly than the PC industry. I remember a tweet from a couple years ago where the hardware-accelerated IPC calls for a system-wide performance gain (and thus a latency reduction slash benchmark improvement) of several percent within one of those year-long intervals — but such gains would be invisible to Linux, even Asahi, because it would require libc and kernel adaptations for different hardware generations, as well as reverse engineering those adaptations from the various macOS kernels and frameworks each year. I don’t think all of those gains can be measured relative to PCs at all, as there may not be any comparable equivalent of what’s being accelerated for PCs to measure — but it still stands out that they simply chose not to mention Apple at all, when they could have theoretically shown either the same slowdown or lack thereof in that domain now that multiple generations are available for easy testing.

That's neither a new concept, nor unique to Apple. GCC and LLVM are both regularly updated by AMD and Intel who also optimize and patch their own microcode as well. When you compile the Linux kernel, and moreover Linux software, you will be doing it with the optimizations of your chosen architecture. For x86 that means OEMs provide optimizations for stuff like AVX and SSE, whereas Apple works in their equivalent proprietary frameworks like Accelerate. You can credit Apple (or more accurately, Lattner) with starting the trend, but Linux can be compiled using the same LLVM backend MacOS uses with whatever optimizations your OEM supports.

> I remember a tweet from a couple years ago where the hardware-accelerated IPC calls for a system-wide performance gain

Most kernels aren't bottlenecked by IPC throughput like XNU is. A modern Linux system might need IPC for dbus, a few journaling systems and systemd. Apple's entire kernel relies on the throughput of a message queue, so feasibly speeding that up matters quite a lot.

Price increases due to multiple factors from higher interest rates to trade wars, etc. are tempering the financial enthusiasm of customers to buy at the higher end, maybe?

Even in data centres, I'm starting to see performance regressions: I just benchmarked the recently released Azure Ddsv6/Edsv6 series of virtual machines, and in several metrics they're slower than the Dadsv5/Eadsv5 series that are more than three years older!

IT hardware performance has definitely been levelling off. The exponential curves have become logarithmic.

If I'm reading the model correctly, that's comparing Intel machines to AMD machines. I'd imagine it makes sense that there are workloads that AMD does better than Intel (and hopefully vice versa). How do Dadsv6 compare?

Azure doesn't have Dadsv6 in Australia yet. Any year now... any year.

Fundamentally, they're both x86 CPUs, both "high end server chips", both in nearly identical (maxed-out) configurations, etc...

I just got used to computers getting so much faster over a three-year time period that it was almost pointless to benchmark them. The vendor hardly mattered, not within such a narrow market segment.

Sigmoid. That's the word you want.

I am somewhat surprised this has not happened earlier! In my circles there was a period around when the Eee PC and MacBook Air was new and manufacurers started cranking out ultra low power laptops that I felt we regressed in performance, but I suppose the majority of laptops sold were corporate things that didn't follow trends as closely.

People got smarter and buy only what they need?

My main desktop at home is a 2011 hp envy. I sometimes think of upgrading, but why? It works. my main laptop is a roughly 10yrs old thinkpad. Outside of macbooks, PCs haven't really given any reason to upgrade unless one is a hardcore gamer.

About time! Every bit of compute performance increase I've had personally, has been eaten up by software bloat. Admittedly as software goes to use 10x as much compute, it does get better, by maybe 10%, but the explosive increase in compute performance up to, say, the mid 2010s mostly just served to encourage bloat and obsolete older devices.

My main machine is from about 2012 - a garage sale hand-me-down ultra high end game machine of the time (24GB in 2012!) It does suck a fair bit of power, running 24/7 but it's comfortably adequate for everything, and that's with two simultaneous users. Such hand-me-down laptops as have come my way (garage sales or free) are also adequate.

So if the continuing evolution is, instead, in compute per per joule, rather than absolute compute, I'm all for it. Graphics card power connectors melting at 570W... not for me.

Is the steep climb in 2021 due to Apple Silicon entering the scene? Or maybe not, servers seem to follow the same pattern. It’s a remarkable increase in performance on a short time.

Could not get to the page.

But, I dunno. I just bought an Asus Zenbook with an intel 9 ultra with Arc graphics 32Gb ram, 1tb SSD, and it seems pretty nifty to me, especially at the price.

so 9 is fast? The rest of the specs are not CPU.

My i7-13700H is shitting the bed due to poor software. Permanently on power-saving while charging because of "rapid charging". It'd be ok if it didn't render the whole computer useless at 400MHz, my first computer from circa 1996 is the only one clocked slower than that I ever had, and it had a turbo button

> The rest of the specs are not CPU

Fair enuf. I was really trying to comment on the ultra/arc chip performance.

> so 9 is fast?

seems to be - my previous laptop was a Core i7, and my new 9 Ultra one is definitely faster (for what I do with it). It goes into power saving only when it gets below about 30% of battery charge, and Asus give a lot of switches to fiddle about with that.

Not sure why this would be surprising.

Even if we assume theres a bit of a skew because a recycler has run a bunch of benchmarks on old hardware, it's something that felt pretty inevitable.

Average users dont need any more power than they have right now for their common activities (browsing and such), so havent been upgrading.

There's been a large increase in low power machines, be it the N100's, portables such as the steam deck, etc.

Until the next 'big thing' theres not going to be much of a need or desire to upgrade, you're not gaining anything at this point if your activities arent cpu constrained.

> Average users dont need any more power than they have right now for their common activities (browsing and such), so havent been upgrading.

Even on what could easily be called a "beefy" machine major websites on chrome make snails look like avatars of agility.

Do these graphs not include Macs at all? The text would suggest this to be true, but PassMark has existed in at least some form on Macs for a long time.

A linear scale is no fit for a performance graph. Can we switch to a logarithm scale ?

average reflects the population, of course. max graph shows no such fall - laptops going up even.

how about this interpretation: desktops are fast enough; laptops are getting faster but people tend to buy more usable laptops, not faster ones.

That happens every year after Christmas when old pcs hit the second hand market…

Lots of laptops marketed as AI powered but have lower CPU speeds and total ram?

most new software runs in the cloud. you only need the browser for many tasks. I even code in the cloud. for many people, why pay for something you dont need.

Because people with this point of view infest us with Electron apps which need 100000x the processing power that took men to the moon in order to run a chat app.

These charts should be on a log scale.

but we got more battery life

commoditization (i.e. enshittification) of computer hardware was taking place all through Moore's Reign, it was simply disguised and overwhelmed by all of Moore's "more".

> The average CPU performance of PCs and notebooks fell for the first time

Well, a lot of HN users care more about battery life than performance, so it shouldn't be any issue. /s

After computers dumbed down (Android, iOS, Windows), we now have computers that do nothing. /s

Moore's law crushed by capitalism? Does anyone else have a better explanation?

Most likely explanation is that we are comparing 1 months worth of data for 2025 to the full years worth on previous years. The laptops people will have over 2025 haven't been purchased yet. I'd expect the chart to be basically flat for the year to date, but a tiny drop isn't particularly surprising either.

I am super curious, if you’ll indulge me. What kind of economic policy would have let Moores law continue as it had been? Conversely, do you think Moores law would have been faster than 18? months if… someone else identified Moores law? Slower? Exist at all?

Intel spend was spending upwards of 15% a quarter on stock buybacks, while also being stalled out on 10nm with the yield improvements "right around the corner", every quarter, for at least 18 months. I don't blame capitalism, but I do blame lack of regulation regarding stock buybacks, especially talking about the United States only home fab.

Capitalism started Moore's law.

:shrug: CPU performance is a rounding error on your GPU and RAM bottlenecks.

The GPU being a bottleneck over the CPU is only in very specific scenarios.

Can you be more specific about what message you're trying to convey?

In the 90s and 00s paying more for CPU ment I could work faster. I was extreemly grateful for each breakthrough.

Even in the 10s this wasn't really true. Storage and bandwith took a bigger slice and CPU was in abundance.

We are now at the point I could easy serve the DB and App servers of a multi-million SaaS off my mobile phone.

The few remaining areas where compute is still the bottleneck at all: video encoding, 3d rendering, mining, and predominantly now AI; also tend not to focuse on the CPU.