OS for next GPU Test System?

[W1zzard]

I will be switching to a Core i9-13900K for our GPU test system soon. Probably some time in December, after the NVIDIA and AMD reviews are done and I have 2-3 weeks to add new games and retest 30+ cards with 20-25 games. Also waiting for a couple of new titles to come out.

The big question is what OS to use?

I'm aware a lot of (majority?) our readers are using Windows 10 and are very happy with it. On the other hand, Windows 11 is the latest and greatest, and Intel claims that Windows 11 has better awareness of the Hybrid architecture on Raptor Lake. Is the difference significant? I haven't seen any data either way.

Your thoughts?

 

[64K]

I chose Win 10.

We still have almost 3 years of Extended Support for Win 10 so I don't plan to upgrade to Win 11 until then unless I see a good reason for it. If there is a substantial reason to upgrade my gaming rig to Win 11 then I would like to know it.

 

[ThrashZone]

Hi,
Both man inquiring minds want someone to confirm without waiting and doing it themselves just how bad or good 10 is from 11 on these chips.

Might help also to not use your gpu benchmark scripts too or compare both

 

[64K]

Since we are talking GPU reviews which are mostly used for playing video games the Steam Hardware Survey shows that 69% are still using Win 10 while only 23% are using Win 11.

There must be some reason for this.

 

[PuiuS]

Since this isn't about personal preference, you should do them on windows 11 (latest OS, latest drivers, etc). The benchmarks will use the recommended OS and they'll also be more relevant in the future (for both comparisons with future products and customers).

Since we are talking GPU reviews which are mostly used for playing video games the Steam Hardware Survey shows that 69% are still using Win 10 while only 23% are using Win 11.

There must be some reason for this.

Yeah, the biggest reason is that most gamers have old systems

 

[Yttersta]

That's a tough one.

As others have said, Win10 still covers more of the userbase but then staying on it would leave out possible new technologies that MS is implementing only on their newer platform solely because they want people to move over, such as the direct memory access implementation. Not that it matters, because the said newer techs aren't utilized by the game devs anyway. However, that's only a "yet" situation I believe; as sooner or later you'll need to update your setup again, and I'm fairly sure you'd rather that be later than sooner and the ability to just retest for a new tech instead of rebuild, reinstall and validate a complete new setup then test the new tech will be welcome

Through this train of thoughts I'm going to vote for Win11; the usage for it will only go higher and 10's usage will only ever go lower.

 

[ThrashZone]

Since this isn't about personal preference, you should do them on windows 11 (latest OS, latest drivers, etc). The benchmarks will use the recommended OS and they'll also be more relevant in the future (for both comparisons with future products and customers).


Yeah, the biggest reason is that most gamers have old systems

Hi,
Looking at steam hardware data I'd agree 6 core being most popular 4 not to far behind and is most popular on linux.

Steam Hardware & Software Survey

Which by the way would likely have to bootleg to install 11

 

[Dr. Dro]

11 is already counting with a yearly feature update and is considered stable by Microsoft, having shipped on new PCs for almost two years now.

11 is what represents the most realistic use case for a Raptor Lake system.

 

[dgianstefani]

I prefer 10, only use 11 on my laptop.

Unfortunately things like Directstorage, which will become more relevant, aren't fully supported on 10.

I think 11 is the most prudent choice, despite it's flaws.

 

[cowcatwithahat]

Windows 11 is the most relevant in my opinion.

 

[80-watt Hamster]

It'd be best to hedge your bets and go 11, IMO. With 10, there's non-zero risk (whether technical- or reader-driven) of needing to move the setup to W11 in the near future anyway, and you'd then have to go through the rebench process again. I don't want you to suffer like that.

 

[dirtyferret]

 

[dragontamer5788]

Core i9-13900K

Win10 scheduler doesn't work well for P + E cores though?

Win11 for sure if that's the CPU.

 

[L'Eliminateur]

Since this isn't about personal preference, you should do them on windows 11 (latest OS, latest drivers, etc). The benchmarks will use the recommended OS and they'll also be more relevant in the future (for both comparisons with future products and customers).


Yeah, the biggest reason is that most gamers have old systems

i'm sure that is the reason and not that win11 is utter trash, starting from the UI and being also slower than win10...

Win10 scheduler doesn't work well for P + E cores though?

Win11 for sure if that's the CPU.

exactly, and that's the point of testing it with win10.

Better yet, @W1zzard instead of committing to win10 to test, ¿why not do a "small" test of the 13900K in win10 vs win11 performance in productivity and a subset of games(no need to run the 53 games)? from dx11, dx12, vulkan eras and/or years/popularity, even only at 1440p/4k to save time. Ans also some kind of test to show the impact of a background processing task(not a high load one) at the same time a game is running to see how "smart" the OS schedulers are and if win11 really puts those processes in the e-cores.

 

[dragontamer5788]

exactly, and that's the point of testing it with win10.

No one buying a 13th gen Intel should use Win10. If they are, they're either misinformed, or purposefully deciding to gimp their own system.

We already know that Win10 is slower on 13th Gen Intel than Win11. Its reasonable to assume that a "rational person" would use Win11 in these circumstances, and that's what we should be testing for. Otherwise, we'd be testing all sorts of gimped single-channel RAM (or other such mistakes) / missing XMP-profiles / poor drivers / etc. etc. kinds of setups.

You setup test systems for the "best reasonable" computer. You never purposefully cut performance (ie: single channel RAM) unless you have bloody good reasons to.

 

[Athlonite]

Windows 11 22H2 for sure

 

[Dirt Chip]

I will go with 11, it is the future.
But many still use 10 so a small 10vs11 will be nice, just to see what we "miss" by using 11.
Also, 10 on RL might cripple it as time past and update stacks.

 

[Calenhad]

Windows 11 for sure. Windows 10 does not quite understand the whole P core E core setup Intel have these days. So, why add another unknown factor to the testing. And who are we kidding? Just like most of us are no longer using W7, we are all going to end up on W11 in the end. I doubt there will be a W12 before that happens

 

[hckngrtfakt]

Windows 10 LTSC, or Windows 11 Pro (for business) to prevent bloatware from interfering your benchmark results.

 

[Dr. Dro]

Windows 10 LTSC, or Windows 11 Pro (for business) to prevent bloatware from interfering your benchmark results.

That's just regular 11 Pro, it's not an special edition for businesses. There's also Pro for Workstations which is what I personally run but it just has some extras that aren't included in the regular Pro. All business editions can be quite customized.

 

[hckngrtfakt]

That's just regular 11 Pro, it's not an special edition for businesses. There's also Pro for Workstations which is what I personally run but it just has some extras that aren't included in the regular Pro. All business editions can be quite customized.

Windows 11 pro is windows 11 for workstations, not the regular "pro" that comes in the general ISO

Windows 11 Pro for Workstations | Microsoft

Set up your business for success using Microsoft’s Windows 11 Pro for Workstations. Empower your employees to work fast and accurately through heavy workloads.

www.microsoft.com

I too was using it for my AI/Machine Learning models and datasets, but Win10 in my experience, handles data blocks better, so I reverted back to it

(see my CPUZ signature link)

 

[Deleted member 211755]

Windows 10.
A lot of "features" on Windows 11 bog down the system. While it might be better with new Intel CPUs, it's still hit and miss with AMD ones, but since you'll be using the Intel CPU then, okay.
Plus, Windows 11 is filled with ads and bloatware, taking up resources and people still report stability problems like BSODs on the latest and greatest.
It would be best to test both how they fare then make a decision based on that.
Personally, I won't go near W11 up until I am forced to do so (after the support ends)...

 

[BSim500]

Windows 10. I can't use Windows 11 for health reasons (every time I look at the dumbed down UI my face physically hurts...)

 

[maxfly]

I'll defer to the folks that are actually running a RL proc on this one. I prefer 10 but if 11 treats p+e more effectively... well yeah.

I may or may not be going RL with my upcoming upgrade, so I'll be paying close attention to this thread.

 

[L'Eliminateur]

No one buying a 13th gen Intel should use Win10. If they are, they're either misinformed, or purposefully deciding to gimp their own system.

We already know that Win10 is slower on 13th Gen Intel than Win11. Its reasonable to assume that a "rational person" would use Win11 in these circumstances, and that's what we should be testing for. Otherwise, we'd be testing all sorts of gimped single-channel RAM (or other such mistakes) / missing XMP-profiles / poor drivers / etc. etc. kinds of setups.

You setup test systems for the "best reasonable" computer. You never purposefully cut performance (ie: single channel RAM) unless you have bloody good reasons to.

i'll put it differently based on what you say: Since that CPU (bad) design forces you to use a specific (horribly bad)OS, then it's not a viable CPU to buy in my book.
IF (and that's a big IF) i buy 12th or 13th gen or any hybrid CPU, i WILL use it on windows 10 and disable the "cinebench accelerators" e-core low-performance rubbish, and it will be faster than on win11.

The exact point of a hardware/review site IS to test all those "gimped" setups you mention so people know exactly what(if any) impact they have and to present the results since most people don't have all the resources to do that.

By testing with win11 you're testing on a OS very few use and even fewer like, this is not like win7 vs win10 which was better in every aspect and a real upgrade hence why the very fast win10 adoption back then. Heck, if your angle is "the best setup" then ¿why don't TPU move to popOS! or arch/clear?, they're faster(from near parity to noticeably faster) in absolutely everything compared to windows....

Windows 10.
A lot of "features" on Windows 11 bog down the system. While it might be better with new Intel CPUs, it's still hit and miss with AMD ones, but since you'll be using the Intel CPU then, okay.
Plus, Windows 11 is filled with ads and bloatware, taking up resources and people still report stability problems like BSODs on the latest and greatest.
It would be best to test both how they fare then make a decision based on that.
Personally, I won't go near W11 up until I am forced to do so (after the support ends)...

For me i won't go win11 at all, don't give a toss about support, it's not like your computer will stop working after 2025 when win10 support ends, nor companies will magically stop making drivers and games that work for it.
I'll stop using it when i load some game or app and it clearly says "this app does not support win10", and that will be long long long after support and extended LTSC support ends so i easily have 10+ more years of max-performance win10, maybe there will be a win10-2 by that time

And it's not like you're missing anything as back then with win7 you lost DX12 support as it was tied to win10, but with win11 there's not a single thing