Intel Core i5 12400 "Alder Lake": A Great ~$200 CPU For Linux Users

[Markopolo]

I really like the format/layout of the article and the focus on geomean of sub-suites vs individual tests. Past articles sometime felt excessively editorialized because you can’t possibly put all 100+ tests and so which ones where shown exposed the article to accusations of bias.

the only question I have is how the final geomean is computed. Does it weigh each test equally or is it a geomean of geomeans? I can see arguments both ways, but I do not that if you have more individual tests for web benchmarks then those could be over represented in the final geomean vs a geomean of geomeans.

Jesus, too many means. You knowhwatimean?

[Sonadow]

Maybe it's a perfect match: an underdog OS and an underdog CPU company (even despite the fact that AMD has stopped competing price-wise since Zen 2).

Unfortunate but true.

And that's why it's so immensely satisfying to trigger such people by simply pointing out the blunt facts and watching the preposterous lies and misinformation they use to try and dispute the facts.

[MadCatX]

More like you get home, turn on your PC, your dGPU suddenly dies and then you remember that you don't have an iGPU.

And when you look on eBay to realize that your 3 years old GPU that just died sells for the same money it did when you bought it brand new.

[Sonadow]

More like you get home, turn on your PC, your dGPU suddenly dies and then you remember that you don't have an iGPU.

Have you ever had the on-die iGPU fail before?

Just because it's very unlikely to fail doesn't mean it won't fail.

I've once had to buy an entry level Nvidia dGPU to use because the on-die Intel iGPU started artefacting and triggering driver crashes and restarts in Windows and outright just locks Linux up (even until today Linux cannot recover from gpu driver crashes).

[user1]

Have you ever had the on-die iGPU fail before?

Just because it's very unlikely to fail doesn't mean it won't fail.

I've once had to buy an entry level Nvidia dGPU to use because the on-die Intel iGPU started artefacting and triggering driver crashes and restarts in Windows and outright just locks Linux up (even until today Linux cannot recover from gpu driver crashes).

I had the opposite case. In 2018 my year old RX 580 started artifacting and then just died completely, so it took a week until I got a replacement unit. That week I was using the iGPU on my i7 8700k, so I could use my PC at least somehow. iGPU's can be a neat backup option in such cases.
While of course it's likely that iGPU's may die, I think these cases are less common than dGPU's dying.

And yeah, it's a shame that Linux still can't recover from GPU crashes. I heard Windows Vista already could recover from GPU crashes.

[MadCatX]

Have you ever had the on-die iGPU fail before?

Just because it's very unlikely to fail doesn't mean it won't fail.

If you have it only as a backup GPU to your dGPU the likelihood of it failing when you suddenly need it is really low.

I've once had to buy an entry level Nvidia dGPU to use because the on-die Intel iGPU started artefacting and triggering driver crashes and restarts in Windows and outright just locks Linux up (even until today Linux cannot recover from gpu driver crashes).

My experience with GPU driver crashes is actually rather good. SNB HD3000 used to crash a lot back in the day and the kernel usually recovered fine. Intel fixed their driver since then (although a few years too late). AMD with the latest kernels also does pretty well. I remember having one or two hard crashes with Vega 8 when Ryzen 4000s were pretty new. Earlier days of Polaris were pretty bad; till this day it gives me the creeps when my monitor takes a bit longer to light up because I always think that amdgpu has flaked out again.

[Sonadow]

And yeah, it's a shame that Linux still can't recover from GPU crashes. I heard Windows Vista already could recover from GPU crashes.

That's correct. Vista from 2007 could do that. Multiple crashes and recoveries over a short span of time is usually a telltale sign that the GPU is starting to get messed up.

In Linux you couldn't even tell whether a system lockup was caused by a GPU driver crash unless you got lucky and the system managed to log the first few lines of a panic in dmesg before the system locked up. Most of the time a panic caused by a GPU driver crash that was triggered by faulty hardware doesn't even have time to write the kernel log before it locks up, so you can keep searching in var/log/kern.* and never find anything.

[user1]

That's correct. Vista from 2007 could do that. Multiple crashes and recoveries over a short span of time is usually a telltale sign that the GPU is starting to get messed up.

In Linux you couldn't even tell whether a system lockup was caused by a GPU driver crash unless you got lucky and the system managed to log the first few lines of a panic in dmesg before the system locked up. Most of the time a panic caused by a GPU driver crash that was triggered by faulty hardware doesn't even have time to write the kernel log before it locks up, so you can keep searching in var/log/kern.* and never find anything.

That's kinda what I experienced when my first RX 580 started showing its symptoms (usually during gameplay). On Windows I mostly had black screens for a few seconds, but then the Windows desktop showed again. On Linux on the other hand, I had some extreme slowdowns in games, until everything completely froze and I had no option but a hard reboot.

[MaxToTheMax]

I really like these geometric means of subcategories, that's a great way to present the information.

[Alpha64]

Hey Michael, I just found out my Core i5-12400 had AVX512 enabled by default in the BIOS that shipped with my motherboard (which was the latest BIOS until about a week ago). I posted about it in the Gentoo forums here. In your testing, did the Core i5-12400 have AVX512 enabled by the motherboard by default?