Zen5 Gaming: Where's my 5%? Windows vs "Patch" Windows vs Linux & The "Lost" Performance Ramble

[@Level1Techs]

Hey look AMD made a blog post! community.amd.com/t5/gaming/ryzen-9000-series-community-update-gaming-performance/ba-p/704054 Huge thanks to our supporters on Floatplane, Patreon and here for helping make this video possible. Y'all the best

[@jefmes]

"...and parity in gaming using the most popular games included in the reviews." That says everything most of the gaming reviewers have been trying to say, no? All of AMD's statements appear to be correct in productivity testing from the majority of the reviews I've checked out, but it's pretty clear that this is intended to be a more general purpose CPU gen, and for us gaming-first buyers, we should focus on the X3D parts. I think it gets lost in some of the discussion that it's not a BAD gaming CPU by any stretch, it's just not performing any better than 7000 series. I'm sitting on an AM4 board still with a 5900X and it still feels plenty fast to me. But once the 9800X3D comes out, and if it DOES show an appreciable increase over the 7800X3D, I think I'll still be on board.

[@kevinerbs2778]

In all SMT/heavy core work loads zen 5 is miles ahead of zen 4. The problem seems to be only in single thread or light work loads. I don't know how Linux handles the dual predictor, dual pipelines and dual decoders differently from windows. In that the Chips snd Cheese review he show that even though in zen 5 a single thread is supposed to able to use all of dual predictor dual, pipelines, & dual decoders that front-end it just doesn do it in most of the time. It seems like certain instruction sets trigger it only but nothing else. Even on techpowerup w1zzard disable SMT which should've forced them to use the dual predictor, dual pipelines & daul decoder they're still not being used. I'm really hoping an AGESA microcode fixes that. But worried that it was changed at hardware level in the silicon level to be disabled or changed for this behavior.

[@ChuckNorris-lf6vo]

Wow such a good job you are doing. Amazing. Learn a little bit assembly you can replace unoptimized subroutines with handwritten optimized ones when you understand the underlying CPU "units" and logic compute unites better than the compiler. This is especially modern with AI ML optimizations because those chips are unknown to compilers don't quote me on this.

[@ChuckNorris-lf6vo]

You are not showing ethe exact windows build 1. 26100 is not insider its RTM 2. when you install the KBs you get more specific build number and this is also important during these test.

[@chrisblack6258]

"this Windows update will improve performance for "Zen 4" and "Zen 3" as well". I feel bad for Intel

[@Desturel]

"i'm being verbose so you don't take it out of context". Sir, this is the internet, you are guaranteed to be taken out of context no matter how verbose you are.

[@IRefuseToUseThisStupidFeature]

Even the comments questioning the discrepancies were taken out of context.

[@TheDemocrab]

In fact, being verbose just means there's more words to be taken out of context.

[@alejandroespalter2601]

"you cant really benchmark windows and linux side by side" this video is still gonna be looked at as a linux benchmarking flex

[@uncrunch398]

@@TheDemocrab The more important aspect that guarantees this is the audience is normal humans controlled by evil humans who want everyone outside their tribe to fall down and disappear.

[@Dwyriel]

yep, I'm taking it out of context right now and telling all my friends linux > windows and sending them those charts :d

[@dansanger5340]

Wendell's discoveries will kick off several days of discussion in the gamer community. I love the collaboration.

[@zagan1]

It's going to show a couple of channels don't know what they are doing and that's being nice about it

[@zagan1]

It's going to show a couple of channels don't know what they are doing and that's being nice about it

[@NegerKim]

@@zagan1 Oh, the irony

[@stolenlaptop]

The more people on Linux throwing money at steam and other driving forces the better honestly.

[@roccociccone597]

I am one of those :D

[@seeibe]

True. Thousands of dollars spent on Steam while gaming exclusively on Linux.

[@charlestwoo]

Amen brother

[@IRefuseToUseThisStupidFeature]

I just wish steam would be better about pushing consumer ownership rights, and cross store compatibility rather than slowly turning things more and more into the apple store.

[@kevinerbs2778]

I'm more disappointed and losing options on windows in dx12 like no one wants to suppose mGPU even though most UE5 just about melt a rtx 4090 for even close to 60fps in 4K. :(

[@sfalpha]

It's not about branch predictor directly, but I tihnk it's about TLB which may get flush in virtualization environment to mitigate some vulnerability. Windows may precautionary enable some vulnerability mitigation for *unknown* Zen CPU by default or somehow kernel team do something wrong on Zen 5 CPU detection. Basically if it what I think, Cache line may get flush whenever it switch from kernel mode to user mode on that particular CPU and that mostly the thread handle GPU things which will get hit hard, same as spectre mitigation which result between 5-15% depends on how frequent thread are switching from kernel to user mode.

[@Level1Techs]

this is exactly what I was thinking, up to and including doing undocumented registry stuff, but I never nailed down that particular ghost. good brain thought tho. I can see things are cache miss that should never be cache miss for sure.

[@kevinerbs2778]

​​@@Level1Techsin my opinion that sounds like spectre patch is completely bugging out on zen 5 thinking its other predictor inside of the front-end of the dual predictor, dual pipeline, dual decoder is some kind of torjin/virus injecting false instructions into the core on light threads or in single thread? Those extra intructions get flushed (possibly be windows seeing as that) when they shouldn't be? But with SMT enable those instruction don't get flagged as false, becuase they get to the rename and order stage of the cpu when needed for the SMT. Unless the are the "cache misses" the ones that shouldn't ever be cache miss normally are also still there with the SMT active? The its a bigger porblem Anyways I'm only speculating here.

[@mycosys]

@@kevinerbs2778 thats the real pity here - whats going on here is REALLY interesting and its getting lost. Theres clearly a colossal uplift in the doublewide decoder and branch predictor on cache effectivenesses rather than a just massive raw IPC uplift, and cache misses are clearly REALLY expensive. Which also makes me wonder just how fast those X3ds are gonna be when they get em sorted.

[@theexplosionist2019]

Delete the microcode dll's. Disable performance nerfs: [HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management] "FeatureSettingsOverrideMask"=dword:00000003 "FeatureSettingsOverride"=dword:00000003

[@MacGuyver85]

Very interesting! Thanks for sharing your thoughts.

[@carpentb17]

The year of Linux is coming keep up the good work Microsoft

[@NoneRain_]

Every year the same thing lol

[@andreaspatsalides1914]

​@@NoneRain_ and every year the marketshare on linux increases by 1% since the Steam Deck. 1% is A LOT

[@nuclearpcs2139]

people is saying that to me since I started building computers, YEAH, linux will rule gaming ........... THAT WAS 1996 ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ

[@potatoes5829]

@@andreaspatsalides1914 1 percentage point, but yea the steam deck is pretty epic

[@andreaspatsalides1914]

@@potatoes5829 Yes 1% INCREASE PER YEAR That means 10s of millions of people switching from windows to linux every year and this will grow exponentially as more software companies adopt it. (And Microsoft advertising it)

[@johanmaasing]

Life time Linux user but gamed on windows until this year. The trend is real and the work Valve and others have put in is amazing. In practice I don’t care about 5% faster, it used to be gaming didn’t work on Linux and now it generally does. These are good times.

[@Hugh_I]

Yep, can confirm the trend is real. Been on Linux for over 20 years, but still had a Windows partition around on my desktop for gaming. For years I only booted this occasionally for some new title that didn't run well with Wine or had a native port. Ever since proton got really good, those occasions rapidly became rarer. I don't remember when I last booted that windows thing, must've been 1-2 years at least.

[@_unknown_guy]

I tried all-in on Linux when the proton hype started. And went back to gaming on Win after half a year. Linux Desktop sometimes is a mess, one update and things break or there are some minor issues but I just wanted to play some games. Some games need a lot of tinkering and then have to spend half a day to get running nicely. Win for gaming is more "it just works".

[@clansome]

@johanmassing If it's still only "generally" then that's not good enough for me or the vast majority of gamers. Might be good enough for "tinkering" but not much else imo.

[@Summanis]

I would love an official SteamOS desktop distro. My list of "things I have in Windows that I haven't found alternatives for yet in Linux" is pretty much just software/driver level frame gen, HDR, and controller paddle support with SteamInput.

[@kozekistudio]

arch btw is basically what steamos is based on

[@kboussa]

It existed, but no one used it.

[@cnr_0778]

@@kboussa Are you talking about SteamOS 1 / 2?

[@bonkerbanker]

There nothing magic with Steam OS, it's just another Linux distro shipped with the steam deck. You can get the same experience with bazzite for a handheld or tv-pc, or nobara for the desktop. There no point in waiting for something the basically allready here, a modern gaming optimerad Linux distro.

[@Invid72]

@@bonkerbanker Currently running Bazzite desktop on my gaming rig and Bazzite handheld on my Legion Go. Pretty sweet honestly.

[@benjaminoechsli1941]

When there's an issue, you turn to the janitor. Thanks for your continued investigation into this stuff, Wendell. 🏅

[@kodiroo8540]

Much as I love the Steves, Wendell is the only KZbinr with deep enough technical skill and experience to debug the issue. He's the only one who even discussed the new memory subsystem.

[@coolcat23]

I unsubscribed from Hardware Unboxed because in comment replies they repeatedly doubled down on their technical ignorance. I have zero issue with someone not knowing something but to actively claim nonsense, is just next level. I don't have time for that. FWIW, I admire GamersNexus' ambition with respect to testing but Steve's humour is aimed at 10-12 year olds. Can't stand it. I applaud their stance regarding consumer rights, but it is often too much sensationalism, for my taste. Wendell is the real deal, AFAIC.

[@JerryBanks572]

@@coolcat23 I agree with you about GN, especially on the consumer rights. I admire Steve also but he starts some news segments with shit-talk sensationalism right off the bat. That's just not journalism.

[@chhandobhihbhushan2742]

Off-topic: The only reason I started using Linux in the first place was because it could run fast on HDD, unlike Windows which would literally choke to death unless it's on SSD and SSDs were very expensive back then.

[@ubergarm]

Yeah, always felt that Windows Entropy Death after a year or so and it and so much junk was in RAM it would start swapping to slow spinning rust drives and grind to a halt haha... Time to re-install xD

[@akosv96]

So true. Steam deck easily runs games from SD cards of all things which choke the bandwith so much. It's so efficient. Same with CPU idle power. On Windows it is very high, on linux it's barely 3-5 Watts.

[@allinaxford]

It is a bug in the RAM cache, and windows power settings, not allocating enough default space if you are below a certain threshold, and then trying to power save causing the disk to not be ready from being off. Once fixed it is not fast to load, but runs so much more smoothly.

[@thoriginalSpaceDiver]

Gaming kept me in Windows for years as it was a headache to keep everything stable and functional with Linux updates breaking things once and a while. But now with all of the annoyances of Windows 11 forced me back to Linux and I've been loving it ever since.

[@fernycl]

Which distro??

[@keyboard_g]

We’ve already seen it on the Asus handhelds with Bazzite Linux. It’s atomic, and just works. Linux experts are tailoring your gaming experience without an interest in pushing ads or ai.

[@foxx1337]

I have noticed this, related to VBS / hardware virtualization, both on Ryzen 3900X and on Ryzen 5950X. For me, as I use an external audio interface, it's noticeable as sound pops and cracks. More subtly, it's visible as increased deferred procedure call latency and a spikey DPS latency graph, with spikes happening minutes/tens of minutes apart. Games with VBS have the same throughput as without VBS. But some, especially more CPU bound ones (DX11 vs DX12 as a simple criterion), are more "hiccup-y" when VBS is on - hundreds of thousands of frames render at 4-5 ms per frame, then there's a segment of 1000 frames or so where 3-4 frames render in 30+ ms.

[@kevinerbs2778]

I have that issue with no man's sky with one the settings in the nvida control panel. Ot shows like 120fps vu like 30ms. I have rest the setting to in the nvidia control panel to stop it.

[@PowellCat745]

Microsoft said we don’t need to disable SVM; we just need to turn off Virtual Machine Platform and disable Memory Integrity.

[@Level1Techs]

My windows seemed to get weird toggling it off and on a lot, necessitating a reinstall. Svm on/off via bios seemed to be the safest. Also sometimes it still seemed to be on even when those features were disabled in the "windows features" gui as I explained in the video.

[@WarriorProphet]

​@Level1Techs if you really want to be confused look at core Temps and boost duration on both zen4 and zen5 with svm off in bios vs all on bios and windows...

[@kboussa]

@@Level1Techs As I said in my comment, I had a hard time with windows and virtualization during my Devops classes, but my classmate, with the exact same laptop, the exact same os (installed by the school sysadmin with ms own Deployment toolkit), the exact same dell pro laptop with i7 10tg gen and 32gb of ram, had no problem. So I think we are not done yet and I'm sure we will see strange behaviours in the future as zen 5 is adopted broadly and people start messing with it. I'm still not sure if it's a bug or a feature, but as GN'S Steve said, it changes nothing regarding benchmarks since all hardwares are tested the same way. Regarding performance, zen 4 i3D s still here if you need it.

[@ChuckNorris-lf6vo]

@@Level1Techs I also notices weird stuff on the NTFS when touching the security features there were weird Copy on Write splits per container on the NTFS.

[@worldwar208]

Ancient game plays tested virtualization months ago make a difference with zen 4

[@przemysawukawski4741]

For the first time I noticed that strange performance problems related to VBS when using NEsted virtualization on AMD Zen3 and Zen4. Every time I try to enable HYPER-V inside the level 1 guest on AMD chips, I get immediately 20-30% performance drop even in CPU-Z benchmark. I noticed that this is not the case since 25xxx insider builds of Windows 11. So I think this branch predicting issue which harms nested virtualization on AMD chips is directly affecting also the gaming performance on Zen 5 we see today. I can't wait until we finally get those higher builds in stable channel.

[@markcentral]

Thanks for the great and informative video Wendell. As an aside, I laughed a lot when you showed off the texture mapping problem in Linux Steam flatpak because I've been swearing at it over the last week.

[@adamgarlow5347]

I had a small dark ages where Fedora didn't have X11 and the Nvidia 555 driver was still in development, but besides that I have found gaming on Linux to be pretty straightforward and performant. I wish modding my bethesda games as easier, but everything I play just works. Elden Ring, Nier, FF7, Fallout, it all just works.

[@ptessier73]

Appreciate your work. And you are right on point with the Tribalism stuff, doesn't help at all. Last but not least, really like that you test stuff on Linux, like you I'm full on that HOPIUM, maybe one day!

[@PlayJasch]

Ah yes, having 3 settings which all do somehow the same thing is the theme for Windows 11. Winception.

[@exscape]

Which 3 settings? If you're referring to VBS, SVM and Hyper-V the three are not the same thing at all, except all are related to virtualization.

[@ellsworth1956]

A semi new architecture is going to take a couple of updates before all the wrinkles are ironed out.

[@kevinerbs2778]

I hope so. Hope its not silicon spin or hardware desgn problem.

[@Malicious2013]

​@kevinerbs2778 I assume that it's not an architectural or physical defect or failure simply because dicking around with the software and OS settings can have such an impact. There are also some programs that work amazingly, but nearly identical programs seem to do very badly for seemingly no reason. To me, it seems like a lot of software nonsense because there are a bajillion moving parts. The bits that play nicely seem to do well, and if we fiddle with the OS software, a lot of performance can be had. It's a little weird. If I had to guess, the biggest problem is that AMD made a lot of pipeline changes and whatnot in this microarchitecture, and there's probably a lot of new stuff that various pieces of software haven't had to deal with in previous chips. I'm just speculating for fun at this point. It's very interesting.

[@zorbakaput8537]

The tribals here will conveniently forget most (if not all) cpus released in the last 25 years have required BIOS updates to fix one issue or another. I don't remember a single cpu release where they didn't have an issue.

[@RANDOMNATION907]

@@zorbakaput8537 . . exactly, ty.

[@timmyjohns222]

It shouldn't take any updates as these cpu's were being made for at least 2 years.

[@johncheng9665]

Linux gaming has become so good

[@suprem1ty]

I was honestly so surprised. I've used Linux for over 10 years, but only on my laptops. Gave it a go on my desktop recently and every game I tried (and I tried a bunch of different ones) just... worked. Ray tracing and all! Pretty insane. Great work on Valve's part! (and the Wine devs, and all the other people working on drivers and such!)

[@timf8684]

The root kits are the final hurdle. Once anticheat crap is sorted, Linux will roll hard. The number of gamer friends complaining about windows is almost 100%

[@tedzards509]

​@@timf8684 Its not like gamers wouldn't and don't already complain about Linux

[@MaryannLynch-z9c]

Agreed, Switched fully over from dualboot to Nobara 39 then fresh installed 40. It's my daily driver for gaming. I remember years ago it wasn't this good, had to do a bunch of methods/commands to get running certain games/launchers etc.

[@fugitive6549]

@@timf8684 Well, your friends might just be stupid. Imagine complaining about gaming in windows, I would love to see them game on linux and break their PCs in frustration cause gamin gin Linux still sucks and will suck for a long time. So many AAA games don't even work.

[@Dan-Simms]

Always appreciate what you guys do, keep up the great work!

[@IcaniCorrono]

I did not read all the comments on other channels, but even if for sure they are doing a lot of work, what make people “mad” is how they present the results IMHO. Let me explain, your videos (and others), are calm, objective, talk about the data, talk about the issue, the strange stuff and so on from an unbiased point of view. You have never told AMD is the absolute best nor AMD is a sh*t/embarrassing etc. You tried to understand stuff, you understand your limit and you do not ride any horses for views. Others (without mention even if straightforward) usually just put their self as “gods”, and instead to make a review they just judge with extreme arrogance. This does not make any sense, a lot of people at AMD, Intel, nVidia, Microsoft etc work very hard and with passion. Claim that all their work is embarrassing because you see some issue in specifics situation is just wrong. See that, they also complain about the hate their self has generated is ridiculous and only made for other views, in my opinion those channels are pure entertainment and zero competence. Obviously is a work and the views are important, but you should accept the consequences. And honestly, there are also channel that make entertainment without any value in knowledge but at least they know it and keep all around fun without spreading ignorance and hate without any reason at all. This is just my two cents.

[@manuelthallinger7297]

Couldnt have said it better

[@MikeBob2023]

same

[@tomstech4390]

HUB isn't trying to diagnose the issue or give answers, they're just making drama standing on stage at the shakespeare globe theatre preaching to the crowd trying to get a reaction but any critism they get they dismiss as the peanut gallery. That's aside from the audacious titles like "Zen5 is bad for gaming" while conveniently ignoring that means a 14700k is bad for gaming. HUB tested Wukong (nvidia sponsored which is fine) after being given specific drivers from nvidia which have had months of development while conveniently not wanting to wait 4 days for AMD to have ANY kind of driver (even a half baked rushed one because they've had minimal access to the game), So the native framerate is super low as a result even before you tehn cripple it with RT and of course Framegen requires a high starting framerate to work properly all while the resolution scaling is broken. Given similar computational power (on average across many games) where both gpu's on average perform equal... -So Nvidia spends months making a good driver and they get 120fps native which becomes 60fps with RT which gets inflated to 120fps with frame gen. -AMD has a week to make a driver, but it doesn't get used anyway, so they only get 80fps which becomes 30fps with RT which gets inflated to 60fps with frame gen. That's insanely flawed testing and it's hard to do that by accident, imagine testing a gtx660 using default windows display driver vs a 270x with the latest drivers (both ~$210 msrp) in battlefield4 using the Mantle API. When a journalists makes such flawed statements or uses incomplete testing then they no longer bring any value. Edit: I STILL use steves Techspot writing as a reference to help others, but nowreally have to vet and scrutinise it all, the number of people buying a DDR4 board because "its cheaper than AM5 and performs the same" completely missing that TS testing is using DDR5-7200 which you're not putting with your i5-13400! Those people don't realise if you use lga1700 with ddr4.. it performs like Zen3 on DDR4 [sigh] Trust but verify.

[@EhNothing]

Thanks Wendell. I've had to drop some subs to other tech outlets lately, and it's nice to have one I can add and get more unbiased info.

[@jayjaytronics8358]

@Level1Techs I absolutely love these deep dives - feels like sitting in the lab with Wendell. And the findings are so weird... Love it!

[@paulthebeardedonedowning6820]

I cannot express how grateful I am that we have a Wendell in this world great work

[@Sopel997]

32:53, would that be like the mouse cursor being jerky, things being generally slower, large visible stuttering on desktop? Because that would explain some of the issues I had, where like once every 30 boots it will behave like that and a reboot fixes it. Was suspecting DDR5 RAM training (because I enabled faster/saved training in the BIOS) but was never able to find any information on that.

[@andre_ss6]

Same experience here, would like to know that as well.

[@sirmonkey1985]

Sometimes that can be memory stability related where you're right on the edge. Ddr4 hynix for example was notorious for showing symptoms like that if it was running at 1.35v instead of 1.36+v at 3000+

[@Level1Techs]

yeah, similar symptoms. no errors and machine was stable. theoretically with platform first error handling off you should see whea errors in windows event log. but mostly you don't. it just acts weird.

[@jouniosmala9921]

There's a regression in Anandtech benchmark on communication between threads, but due to bursty nature its waking up a thread execution issue, or a security feature preventing L1 and L2 cache side channel attacks for SMT cores. In older games, the block that gets scheduled per thread is way larger thus both of those issues gets mitigated. Another point seems from that benchmark is Communication between SMT threads in a same core is handled through L3 cache, which would make sense as a security measure against reading other threads memory . That could mean that in modern games with fine grained threading, you could be in a situation where a thread goes to sleep and another thread awakes tiny bit later, both threads use same underlying memory for a thing but the security feature makes it L3 read instead of L1 read. And then it suddenly became a lot more important to keep same worker thread waking up in the same core to keep using shared data between threads through L1 cache. So the weakness could just be with any application that has lots of worker threads waking up and getting sleep frequently with scheduler that moves those around within a socket. And the fix could be trying to keep worker threads in exactly same virtual core or keeping each worker thread running longer time.

[@kevinerbs2778]

Thats the spectre patch. I mentioned this in another comment, but that wake could also be seen as torjin injection instruction because zen 5 uses dual predictors, daul pipelines,dual decoders. This could easily explain premature the dumping of instrctuions by windows. Wendal mentioned that things that never end up as cache misses are showing up as cache misses

[@chris27gea58]

Please point out the particular benchmark result(s) in the Anandtech article because what I'm seeing is simply a lack of change with AMD still behind Intel in the multi-threaded performance pecking order (when considering roughly equivalent x86-64 parts by market positioning and price). Indeed, there hasn't been a big uplift in the multi-threaded performance of these highish spec Ryzen parts (disappointing those who expected that) but barely noticeable improvements on that score don't amount to a regression. The only regressions that I am aware of are these regressions in selective gaming benchmark results. We are yet to see how the newer Ryzen 9 parts perform on Windows machines. Some results are in for Linux, though, and they look good.

[@jouniosmala9921]

@@chris27gea58 The microbenchmark about 9950 communication latency between cores. And I went to Zen4 review to look same benchmarks, to get comparison value to check it matches the text of Zen5 review. Compared to previous generation the communication latency between two SMT threads was 3X as was latency to another chiplet, and for same CCX was 25% higher. And conditions which the microbenchmarks tests also hit is something modern game engines would hit, but for lower fraction of time. The practical benchmark numbers, are from Phoronix Linux review. It has plenty of benchmarks where problems that microbenchmark revealed do not matter and in those cases there's often a very large jump in terms of performance. For gaming performance where the communication latency doesn't matter it got 30% more FPS. And with single threaded Javascript it got 18% performance uplift. And in several cases it got higher uplift with AVX256 instead of AVX512 in AVX specific article at Phoronix. (AVX512 was still faster but percentage of improvement of generation was higher with AVX256bit.) Which just eliminated the excuse that core didn't improve for old AVX. Linux has better scheduler for zen5, but I think it's also about everyone in Windows review world choosing benchmarks that hit the weakness of Zen5 architecture. If you need good single threaded performance Zen5 is a nice jump, as is with running independent processes.

[@stephanematis]

Wendell, never change. Keep speaking out.

[@MikeBob2023]

It's not just what he says, but also how he says it. Unlike a *lot* of very popular techtubers I could name (but won't 😇), *Wendell of Tribe L1T* speaks with a modicum of _humility._ It's a rare thing to see on KZbin, it's refreshing, and it's inspiring.

[@tonybrown7408]

If Wendell is shrugging his shoulders and going 'I don't know what these BIOS settings do' ? ...... what change have I got !!!

[@blackasthesky]

$ 0.57

[@MrSmitheroons]

So, to me we are just going to have to wait for AMD and/or Microsoft to sort this out. If there are fixes coming, then they will come. The big confusion was caused by AMD internally seeing (or reporting) one thing and reviewers (and presumably consumers) seeing another thing. That's on AMD to resolve. Maybe on Microsoft (Windows team) as well. It's great reviewers are pulling out all the stops to try and figure this one out, where all the performance went, but it's not really on them (not reviewers' responsibility to provide R&D for a product launched with issues). I'm not seeing 100% clearly how we go forward from here, because reviewers *are* the trusted perspective we go to as enthusiasts to validate the performance or any fixes. But here's an alternative I think people haven't considered strongly enough: Wait until there's some new patch to test, report what you see, and admit we don't understand what's going on until AMD and/or Microsoft lay it out crystal clear. Consumers will have to live with it being really unclear what's going on with this product until then. If they still want to buy, so be it, but I have heard sales are low. tl;dr: AMD (and maybe Microsoft) need to explain this, reviewers and review watchers should take a breather and ignore this gen unless/until they like what they see reviewers consistently hitting as results. We can speculate what's going on, we can look under every rock to find the answer, or we can accept temporarily not knowing. As hard as it is, I advocate for "accept not knowing" for now, until the companies explain it crystal clear. And if you don't feel a burning need to upgrade, stick with what you have. If you need a new system, intel 12th gen and AMD 5000 and 7000 are really strong options, or you can gamble about whatever remaining unknowns there are with Ryzen 9000, since it's not actually worse than Ryzen 7000 overall, just more expensive and a bit up in the air right now... But yeah, I'd like to see things calm down in the discourse. As far as I'm concerned, some people screwed up, but it's not our job as consumers, nor should it be reviewers' jobs, to R&D or test their products for them in place of their internal testing being valid. That is a bridge too far, IMO, we need to let them handle fixing their launch or just ignore the product IMO, either way. Going this deep down the rabbit hole to find the answers is beyond the scope of what should be happening outside their own labs, IMO. OTOH, it's been kinda fascinating. Maybe we'll uncover something and greatly improve Windows. But it's also so exhausting. AMD should have tested better, should have communicated better. And if they wanted better gaming results, they should have tested properly for one thing, and then let it cook longer or set expectations lower if they couldn't get there. Until then, we're seeing a half-baked launch IMO. Indeed, I'm not a fan of that in game launches, sure as heck not a fan of starting to see it in CPUs. But for this launch it happened, can't be undone this time around. Only maybe fixed. Which will take months to happen properly, at minimum, I'm guessing. ANYWAY... thanks for your insights over the years. I hope I can skip this topic of the mysterious Ryzen 9000 launch until the companies responsible fix the weirdness, but I won't hold my breath. I need my sanity back, so I'm gonna stop following this until it gets answered more definitively.

[@danijelpavlovic9871]

Here's a list of the bios settings mentioned. Bios stuff: AMD CBS - CPU Common Options Prefetcher setting - Set all L1/L2 Stuff to enable ACPI_CST C1 Declaration - Enabled Platform first error handling - disabled Pause Delay - Wendell had it at disabled/0 (edit: corrected) CPU speculative stores - More/Less (experiment some games work better some worse) Latency under load - disabled SVM Enabled - Disabled Corrector Branch Predictor - Enabled

[@Level1Techs]

pause delay is also "experiment" but not sure it was making any difference. I mostly ran with disabled/0

@@Level1Techs Core watchdog Timer = Disable ( BSOD code) Streaming Stores Control = Enable REP_MOV/STOS Streaming = Enable

[@n8spL8]

I have been saying this for weeks, everyone is like: whatever, just flip the switch. NO, disable in bios and you get even more performance. It will show the difference in taskmanager even "virtualization" enabled/disabled. Not sure how all these 'experts' miss this? Not to be tribal, but it seems so straight forward to me. TY Big W for trying this out for all of us.

[@RAZGR1Z]

I understand what you're saying but these kind of work arounds shouldn't even be necessary.

[@mycosys]

@@RAZGR1Z they arent if you dont run bleeding edge...........

[@n8spL8]

@@mycosys quite the opposite. older CPUs take an even bigger hit. The unfortunate part is we have to gird our defenses against thieves and it costs all of us performance. That being said, I believe some SMT, and prefetch effeciences can yet be made in windows OS.

[@AndreGreeff]

random interesting point, if you pause the video at 11:37, while the in-game FPS difference here is very close to an "acceptable margin of error", check how different the "Load Time" values are... jumping from a total of `10.162sec` to `27.706sec` when VBS is turned on. I'm not going to pretend to know exactly what that means, nor the potential ramifications of it.. but I can't shake the feeling like there's something more relevant there with regards to the other games. I mean, perhaps it's even down to internal processes of the underlying game engines? possibly even dependent on what specific game engine functionality the devs decide to use.. if the resulting system calls do things in a way that require more "is this safe to run" checks, obviously that specific internal logic pipeline will take a performance hit... this could also be relevant in the case of comparing old and new games, where the older ones were created before this type of "virtualization based security" was relevant.. now many years later, some of those engines would incur a higher performance cost, simply because some things that they decided to do back then were later found to be potentially dangerous and subject to additional checks.

[@dotplan]

I am on linux both at work and home for years. It's marvelous, nowadays.

[@rodrigorras]

Best review Ever!!!

[@martyplaysgames]

I'm finding the Windows desktop snappier after swapping out a 7950X for a 9950X with nothing else changed so same OS version, GSKill DDR5 32GB, RTX 4090 and gen 4 NVME (I'd been running on the ASUS X670E Hero latest BIOS for a couple of weeks with most recent chipset drivers) - this is purely anecdotal of course, as I have no real way of measuring it - but bringing up explorer and launching apps is noticible faster. There was one game I was getting strange stutters with and that was Battlefield 2042 (others have reported the same thing with the game on 7950X and X3D CPUs), frame rates in that game would randomly drop to < 10 FPS for a couple of seconds and that has disappeared, I had just thought it was bad EA code, but I do wonder if my particular 7950X was not 100% great silicon. I also saw a couple of microstutters with Teknoparrot - these are not shader related, and these have also gone. It does seem to me that there are QoL improvements on these chips ( based on my sample of one :D ) that isn't being highlighted by the benchmarking.

[@MooresLawIsDead]

Amen brother: kzbin.info/www/bejne/Zpa8ZGearNWcpbMsi=wOjOPM1hWBOyaXN2&t=784 The mere act of trying to explain why a given company did something (even if it was a mistake), or investigating why a certain product is having issues - should never invite bizarre attacks and accusations of "coping" or "defending" said given company. Just trying to know the truth, or explain the situation, does not put you on any "side".

[@kevinerbs2778]

I just want to know why the dual predictor, dual pipelines, dual decoder, doesn't seem to be in using in a single thread in a core ?? It's supposed to be able to use it for single thread in a core.

[@richard-davies]

MS needs a massive kick up the ass, instead of including ads and useless extra software in Windows 11 how about focusing time on improving everything from stability, bug fixes and performance.

[@IvoPavlik]

That's not how MS managers deserve their bonuses 🤷🏻

[@beaureddington4039]

33:54 Wendell demonstrates how to follow.. *checks notes* “The G-Code”

[@inflatablemicrowave8187]

A bit off topic, but it's funny seeing linux higher up than windows in a chart about something gaming related. Let's keep that up hahha

[@Alexx_80]

As Wendell have said, you can't compare it. Because a particular game can look different on Linux than on Windows. Some game elements (like RT) may not be present on Linux and vice versa. It can impact CPU/GPU load and framerate. Keep in mind that almost all these games are Windows games, not native Linux and translation layer is involved.

[@ThylineTheGay]

@@Alexx_80 he is overstating it a bit, aside from outright obvious graphics bugs, the differences will not really be noticeable to a person, also, i think he means more how the computer _technically_ runs it, vulkan vs direct3d and all that RT is generally well-supported on linux these days to my knowledge, and wouldn't windows games running better on linux even through a translation layer be a _positive_ for linux?

[@inflatablemicrowave8187]

@@Alexx_80 it was a joke, don't over think it haha

[@Alexx_80]

@@ThylineTheGay Cyberpunk RT was not supported for months on Linux after it was introduced. HDR basically doesn't work till this day.

[@ThylineTheGay]

@@Alexx_80 note how i said "these days" also you mentioned nothing about HDR and anyways HDR is how it's _displayed_ and has minimal effect on the actual graphics pipelines

[@Didymuss1]

On gaming on Linux vs Windows, from the start I couldn't even fathom using Windows on a handheld given that, well, it's Windows. It is big and fat and hungry and not condusive for a good handheld, they struggle enough on Laptops. Gaming on Linux might not be as good as yet, but it's close enough now for me to want to move over, especially with Windows going in the way it is. The general usability for Linux has come leaps and bounds (without needing to be a mega nerd, which I can do but cannot be arsed with), combined with the gaming being as good as it is already is enough for me. I'd rather have a system which isn't hostile, doesn't bombard me with adverts or tracks my every move, and doesn't hide literally everything from fear of the normies.

[@spladam3845]

Glad we have you working on it Wendell.

[@ash98981]

IF consumers first start with your reviews (as I do) before looking at the other reviewers there would be less tribalism! Negative headlines from reviewers before proper investigation of issues lead to said “tribalism “! As a hardware tech junkie I know I’ll get a proper read of cpu/gpu activity than sites that are just game sites. There’s so much more to hardware than game benchmarks.

[@fujinshu]

I usually start with LTT for a general review of both gaming and productivity, as well as power consumption, spikes and temperature, and their overall verdict on whether I should spend my money on this. Then I to L1T for more benchmarks mixed in with possible use cases for the product with this new knowledge. For example, running an always-on, always-pegged server is more feasible for Zen 5 CPUs due to lower power consumption and thus greater overall power efficiency, while running a mostly-idling server like an archive NAS is better suited for monolithic chips such as Ryzen PRO G, Ryzen mobile or Intel's offerings before their switch to chiplets.

[@MacGuyver85]

To be fair, when a product is released and properly tested, and it shows no uplift or even regression, then it's fair to be negative. It's up to AMD to have all this sorted BEFORE they release the product, not after.

[@ash98981]

@@MacGuyver85 just sayin’ Gaming benchmarks are not the end all to cpu evals. This generation of CPUS also include NPUs, the designs of both Intel and AMD are now focused on AI, the new $$$ target, another dimension has been added specifically for that target, Nvidia completely abandoned gamers for AI $$$ , I suspect this too is part of the move to higher compute and compute dollars. Gamer should stick to specifically gaming chips until this all gets sorted out as I feel folk primarily into productivity will not concern themselves with gaming as the primary objective of this buy because of the other advantages including Linux

[@ash98981]

The 9000 line is NOT a gaming focused first CPU, I think it was designed that way- not to excuse AMD but for me gaming is adequate for me while the other advantages, including compute and more has much more value

[@fujinshu]

@@ash98981 I think it's wise to wait for the inevitable X3D variations to come out, so that we get both compute and gaming increases.

[@SalvaBarbus]

It can't be understated how far gaming on Linux has come. I used to dual-boot Windows 10 & Linux on my old computer, but when I built a new one last year, I decided to go for Linux only, and it has been amazing. Of course, I can't say that everything worked first try, but really, only World of Warcraft gave me a bit of a hassle with the installation and so on. After that, everything worked super smooth.

[@maxwellsmart3156]

I ran into a Virtualization Security problem with VirtualBox a year ago. The VirtualBox VM's performance are impacted when Core Isolation Memory Integrity isn't disabled.

[@joe--cool]

That's normal. As soon as HyperV or core isolation is enabled Virtualbox is also forced to use HyperV instead of its faster Hypervisor.

[@ZipplyZane]

@@joe--cool Which means it is Microsoft once again crippling the competition. This is why I don't like Wendall's unqualified statement that it's always a good thing (save for this bug). It impacts performance a lot more on older systems. Even if it is "only" 1-3%, that can be a lot on some systems. It messes up certain programs, including anything else using virtualization. It disables certain drivers and peripherals. And there are plenty of reports of it causing worse slowdowns than it is supposed to. It just is not something that everyone should turn on. Any more than a TPM is necessary for everyone.

[@andrewjmarx]

It's a longshot, but I wonder how the new full pipeline AVX-512 instructions benchmark under these different scenarios. Like, could something about Windows, like scheduler heuristics or thermal management, be tuned for the old "double pump" pipeline and not properly be adjusting for the new AMD models? Only more modern games would be compiled with support for it; so it could correspond with differences between old/new games

[@Henrik_Holst]

I don't think there is a single game out there that uses avx-512 since Intel dropped it ages ago.

[@yusky03]

I use and game on linux exclusively. I have been considering windows gaming benchmarks as generally being around 3% of linux. Guess i am going to need to starting looking for linux exclusive reviews from now on.

[@griffin1366]

8:41 - Good explanation. I have VBS and Core Isolation disabled. In the past I used custom ISO's with Administrator accounts but this got me scratching my head as it never had a performance difference. Tested it again recently and no difference again. Being those security things makes sense.

[@brianmccullough4578]

Well if Steve, Steve and Wendell are ever in a real grudge match. I got $20 on Wendell

[@evalangley3985]

At least he is the real tech guy of the 3. The Steves are a bunch of glorified DIY PC Gamers running cinebench and a bunch of games.

[@robertmyers6488]

Wendell is rational and not a sensationalist and that is the difference.

[@robertmyers6488]

​@evalangley3985 They have a problem thinking outside the box. 😂😂😂😂

[@mycosys]

GN Steve would just have his legions attack, like he did to that reddit mod.

[@robertmyers6488]

@@mycosys What was that?

[@gabest4]

I was benchmarking my code a lot in the past, way before application virtualization, and once there was a regression by a similar ~5%. It took me a while to find out that just by activating hyper-v in Windows, it made everything run slower.

[@Mattia_98]

SOTTR is an ancient game? Oh my

[@darransmith32]

It's also worth mentioning that Windows games that use DirectX when running under Linux (via Steam) are forced by Proton to use Vulcan instead of DirectX. Vulcan is typically faster than DirectX (on any OS) which is why Windows games tend to get a perfomance boost on Linux.

[@comguaf8093]

I am glad to know that there are more people who share this goal of iteration instead of mere conformism, consumerism and tribalism because they prefer to expect something better from others instead of forming it with their own criteria and using these tools of the future that we have done, I always find myself have encouraged the existence of a consistent, self-taught and self-sufficient community in my personal cycle. Rather than complaining, we are going to iterate on what to do to improve those aspects that do not add up, people without the ability to contrast what they have in front of them and opt for the best option besides the media, from a third world person perspective ''my eyes can't conceive a world where people do not grow'' and I really appreciate what you do wendell :3

[@aaronturvey8819]

Saw you as a guest on a few other youtubers videos since all the intel problems and you really intrigued me so added your channel..... very indepth and full of good information keep up the good work.

[@EthelbertCoyote]

With Ai a lot of companies have taken their eye off the ball for software performance for other applications AMD, Intel, and Microsoft at least for me feel like they are in software maintenance mode. Even the ARM launch seemed less baked as far a my perception. While we get new OOOOh numbers per year with hardware again it may just be me but I feel less stable software wise in general over the last 2-3 year than I have since the 90's.

[@Hugh_I]

We've been waiting so long, it was always just a year away - but finally it came true. This is the year of Linux on the ... handheld! Close enough, I'll take it.

[@fatweeb1545]

Hardware Unboxed use clickbait titles like "Zen 5 Sucks" and then go surprised pikachu when there's backlash. They're the ones stirring up drama for views.

[@MikeBob2023]

I noticed this too. 🙄

[@rodrigorras]

same

[@evildude109]

Luckily there's a convenient little unsubscribe button below every video.

[@MikeBob2023]

@@evildude109 I used mine after I'd had enough of the drama and showboating -- on a couple of channels (so far), actually. Life's too short. ⌛

[@VoldoronGaming]

Fascinating video Wendell. I am glad you’re taking a much more in depth look at this vs other channels who just trash the product. Most channels benchmark with newer games which is fine but it isn’t an accurate representation of what gamers are actually playing on a daily basis. Amd in their own internal testing used older games but probably overlooked, and I can see why it can be easily overlooked, VBS on or off. Great video Wendell.

[@Terran.Marine.2]

Does the rain in Bahrain fall mainly on the plain? (Could not resist, I apologize).

[@ChillBuilds]

I have been gaming almost exclusively on Linux for about 6 weeks and I've had about the same amount of problems as I had with Windows, which is a great thing.

[@oscarcharliezulu]

So software issues I reckon. Need a few bios revisions and windows updates to improve.

[@ll1l11ll]

I really appreciate that you spend so much time down all these rabbit holes.

[@MikeBob2023]

⭐⭐⭐⭐⭐

[@andreaspatsalides1914]

Showcasing benchmarks with an anomaly is not a problem, not their fault. Doubling down after the anomaly is pointed out, without testing on linux is the problem. As a linux user I only feel represented when a benchmark includes linux in scenarios where the OS matters a lot, so Zen 5 benchmarks

[@rheh2570]

spot on...

[@NJ-wb1cz]

Yep. And making predictions that nothing will improve, that it's a flop, and generally being weirdly nasty about it - this is about hardware unboxed. They looked at obviously weird numbers and projected them into the future without any kind of critical thinking or investigation, and then started doubling down and trying to defend their own initial narrative and minimize anything that goes against it. If they were just reported the numbers they got, there wouldn't have been the same reaction

[@Osaka2407]

You nailed it. I love many of the tech channels and the work people do there. But doubling down on an anomaly, ignoring obvious ways of testing for it and pretending anomaly is normal behavior isn't right.

[@notjustforhackers4252]

Nailed it. The early reviews were horrendous. If you're testing hardware then test it and test it properly..... or change your channel name to "Windows hardware unboxed". 🙄

[@andreaspatsalides1914]

@@notjustforhackers4252 Exactly! 🤣

[@RetroTinkerer]

Yey, thanks for keep banging on it Wendell!

[@yeghigmikaelian]

True Hero, much respect !!!!

[@eliotrulez]

wow, what an incredible video! Thank you Steve aehm Wendel

[@jugklutzz]

Thank you for educating me. No matter how much my brain hurts after learning from you, it's a much better feeling than watching Angry Gamer Reviews.

[@maxwellsmart3156]

The Steves' work is fine and I have complete trust in their benchmarks results. What the problem is, with the tribalism, is the infantile and hyperbolic thumbnails, these are the things that tribalism is generated from. I mean it's on purpose and if they didn't have that effect then the YT gods would not respond well.

[@ThaexakaMavro]

100%

[@Injektil_o]

In the case of HUB, I don't think the thumbnail was being infantile and hyperbolic, in the sense that they genuinely believe Zen 5 is a bad generation.

[@fujinshu]

@@Injektil_oZen 5 Ryzen 9000 so far is not great, but it’s probably going to be better on the server and mobile sectors, since that’s where both Intel and AMD make the most money. Desktop and high-performance mobile is usually where they throw the scraps of their latest work.

[@Level1Techs]

faustian bargain tho? If no one choses to click the reasonable thumbnail then their task is harder. I almost exclusively do more milquetoast stuff, so I know :D Like.. Subs Will Not See videos because of how this platform is designed.

[@maxwellsmart3156]

@@Injektil_o Sorry but they are used to trigger people because they need to make a day 1 impact. Unfortunately this behavior makes consumers less discerning. The product launched successfully and were distributed for sale. The numbers they produced just indicated that it probably wasn't a good upgrade for gamers from Zen 4. Nothing more, nothing less. Zen 5 isn't a "bad" generation, that's ludicrous. I mean if you were gaming a 1080p with a 4090 then maybe, if you had bought one you could be disappointed.

[@MaryannLynch-z9c]

Thanks for testing for us. Am maining Nobara 40 KDE, have one of my alt rigs on W10 LTSC.

[@Zicrixdoesart]

If I can get HDR to work fine enough on Linux, I think that's pretty much all I need and I'm gone

[@DJDocsVideos]

HDR room he other bullshit that people got convicted they need...

[@gorgnof]

if you have a good HDR monitor (not 400, even though that is still better than no HDR) there's a HUGE difference, it looks much better ​@@DJDocsVideos

[@fugitive6549]

@@DJDocsVideos Did you have a stroke?

[@Henrik_Holst]

On SteamOS this is fixed by default, in other distros you can install GameScope (the window manager that SteamOS uses) and it will support HDR with the proper flags, can be a bit to setup if you are not accustomed to working with the terminal. Other than that Gnome is currently working on full HDR support and AFAIK KDE support HDR already.

[@MatthewSmall]

It's not just AMD. I'm running an I7 12700k and in Microsoft Flight Simulator at 4k, I have an average of 15 fps difference with VBS on or off. At the frame rates that game chucks out, 15 fps is massive.

[@he1go2lik3it]

11:24 look at the MIN FPS. How smoothly a game runs without too many stutters also makes a big difference. Games usually feel smoother without VBS.

[@enlightendbel]

To be fair, especially in stuff like Cyberpunk 2077, if you have 56FPS on 1440P, getting to 60FPS flat, which is V-sync for a lot of cheaper monitors, doesn't matter a lot less. It matters a whole lot more than getting 200vs250fps. For benchmarks these peak FPS may matter, for playing, they don't, at all.

[@ADINSX3]

One way to test if Windows vs Linux benchmarks are fair (for games with built-in benchmark): Do a lossless screen capture of the benchmark run on each OS, but limit the fps to 60 (or whatever each system can sustain). Then subtract each frame from one video from each frame in the other video (ie. take the difference in the RGB values of each pixel of each frame). The closer the resultant video is to pure black to the human eye, the more fair the comparisons are.

[@memory_stick]

Honestly, Great video Wendel, that really goes in depth, no sugar coating and tries to investigate some of the (perceived) issues around Zen5. It all points towards the SW Nd OS ecosystem not being quite ready for all the new HW stuff ( especially on MS side). Also thanks for the shout-out against tribalism, it really is not helpful. At. All.

[@keyboard_g]

We don’t need tribalism and we don’t need thumbnails of grumpy faces fire emojis and all capitals “JUNK”. That KZbin nonsense invites it. It’s youtube, but you can’t rage bait to get clicks and then be upset at the rage.

[@mikafursjo4259]

Amazing work you did! Very interesting.

[@KisaJeweler]

About frame times on Linux vs Windows. I have a Ryzen 7 5700g(don't ask why) and an RX 6650 tx from the Gigabyte 3 fan version. I had a lot of problems with GPU, starting from deleted discrete drivers because of integrated graphics(classic Windows stuff). It is an easy problem to deal with. I just turned off the integrated GPU(or u can try to go to the registry coz other options just don't work). The second problem is micro stutters and micro-freezing. On the AMD driver's side, the only solution was to lock the minimal VRAM speed to 2300. It helped with freezes, but frame times in almost any modern game were not perfect. Neither of these problems exists on Linux for me. Games run a lot smoother and I don't need to beg Windows to stop deleting my drivers on updates.

[@chrisblack6258]

why?

[@KisaJeweler]

@@chrisblack6258 I bought my setup during the GPU shortage and didn't want to spend Giga money on GPU, so I thought that I would stick with igpu. but a month later, I bought 6650 for a really good price, basically MSRP. I thought later I would make it as a home station mini PC, but stuck with it for now 🤣 5700g is a pretty decent processor.

[@kwerboom]

Thank you for making this video. It was good to watch a sober video explaining all the issues that could be affecting day 1 Zen5 gaming reviews on the current version of Windows, the differences between Windows gaming and Linux gaming, and the advanced BIOS stuff. My problem with the day 1 Zen5 gaming reviews from my experience was that they boiled down to the following: "We didn't see any uplift over Zen4 our Windows test systems. Our benchmarks don't match AMD's review guide. AMD lied. Zen5 is a terrible product. Don't buy." Which would be all well and good, bad products happen. But then I saw reviews done on Linux and reviews looking at productivity going the following way: "Oh, My, God! Zen5 is exactly what AMD said it was. All the uplift is there. This is amazing!" This situation didn't make sense to me. Normally a new CPU architecture is consistent across all types of tests; either 'good' or 'bad'. To see this divide, IMHO, meant something was broken somewhere. It's clear part of the problem is Windows 11 23H2 not being out-of-the-box optimized for Zen5 (or even Zen3 and Zen4 if reports on the new branch predictor code are to be believed). It's also clear that part of the problem is that AMD released Zen5 with an under cooked BIOS that wasn't ready for prime time. I don't thing the reviewers doing Zen5 game reviews are bad or wrong. I just think in this case, they aren't set up to debug a more complicated problem involving OS code and microcode.

[@jpullen581]

I wonder how Windows 10 does with the SVM on/off

[@Level1Techs]

win10 has been abandoned for new processors sadly

[@MrSaywutnow]

19:52 When I look at those numbers, I can't help but think of the late, great Paul Harrell. "Yes, there is a difference, but is it enough of a difference to make a difference?"

[@Lambretta_G]

At what point it stops being a bug/issue and its just OS tuning for gaming? This reminds me of the days we would try to find "lite" versions of Windows XP to get an "up to" 5% performance uplift in games. Is this what hardware releases have come to be? So meaningless upgrades that we have to tinker with the OS to get 5% more performance to make it a whole 10-15% and justify the "generational upgrade"?

[@NJ-wb1cz]

At the point where one OS gets gains from zen4 to zen5 switch and the other doesn't match them, it's a bug in that second OS. But when some tweak in OS affects all or most architectures equally, it's a tweak

[@Lambretta_G]

@@NJ-wb1cz In my mind a "bug" is something serious. Like 30% performance degradation, or lots of stuttering. You can get 5% with debloated/tuned versions of Windows (LTSC or Atlas). Is the bloat considered a "bug" as well?

[@blackasthesky]

I think with how expensive gaming hardware is, it's totally fine for people to use their PC for six years and then do an upgrade, instead of switching every year.

[@NJ-wb1cz]

@@Lambretta_G no, a bug is when one architecture performs much worse than it should relative to other architectures on the same platform. Then it's a bug in how that architecture is treated. It's like how Windows performed badly on CPUs with e-cores while Linux handled them just fine - it was a bug in Windows, it was fixed. There are countless areas and apps where Windows is more than 30% behind Linux in performance and that's okay because it's behind regardless of hardware. It's just how Windows is.

[@gabsriel]

Would you be able to run Cachy OS, with the Cachy kernel and so on ? Maybe frame pacing would be better.

[@OhhCrapGuy]

Linux is so much more friendly to gaming than it used to be, this is coming from someone who used to run WoW almost 2 decades ago on Ubuntu. Or was it Debian...

[@Hugh_I]

Can confirm, it absolutely is. I don't know how many days of my life i spend fiddling around with Wine configurations to make windows apps and games work over the last 20+ years I've been on linux. But I haven't done that in a long time now, proton/steam just works in almost any case.

[@Shantara11]

I was planning to build a new PC in the near future, and one look at the BIOS settings is filling me with dread. It's so much more complex than the one on my old motherboard! And if Wendell can't fully figure it out, what hope do mere mortals like me have? It's probably not the big deal, and I am already feeling a premature FOMO over the potentially missed performance.

[@JacobAsmuth-jw8uc]

Wendel I'm surprised that you're not talking about these FPS difference in terms of actual milliseconds. It is so much more useful to say that "The time to render a frame with VBS enabled increased by 1ms from 4ms to 5ms" instead of "The FPS went from 250 to 200".

[@roccociccone597]

Linux gaming has come a long way... I have been a Linux user for nearly a decade now. When I first switched the most reliable way to get games working was in a Windows VM with a GPU passthrough. Now all the games I want to play run on par with Windows without much tinkering. And the fact it keeps getting better is amazing. I couldn't imagine using Windows again as my daily driver.

[@movdqa]

I hope Microsoft gives us another couple of years of Windows 10.

[@joeykeilholz925]

people aren't moving so they have to.

[@RedSntDK]

That's why I'm hoping Wayland is in a good state early 2026, assuming people are slowly moving from Win 10 to Linux.

[@Jkaninteangemittnamn]

they hired android and linux people and dont give crp about the old windows

[@sweetdreams5454]

​@@RedSntDK with the current state of AMD cards in price/performance (hopefully increasing purchases), I think their driver compatibility is going to be really nice for many newish penguin migrants!

[@RedSntDK]

@@sweetdreams5454 AMD users will definitely get the better transition. Nvidia has been doing positive things as well re: Wayland support. As someone that games a bit using X11, that VRR support in Wayland is looking reeeal tempting. Feels silly to game with a freesync, displayport monitor in a 3070 Ti and have tearing in 2024

[@JKHYT]

Great info Wendell, as always.

[@ralmslb]

This man is our prophet! Can't wait for the day where playing in Linux is equal or better than Windows.

[@NKG416]

it's been good for me, almost all in my library is straight out of the box playable

[@ThePlayerOfGames]

I hard switched to Linux 3 years ago after hopping between for 2 before Every game I care about runs just as well on Linux, with better functions and less configuration often Stop letting software push you around, I say

[@zx14jeff]

Thank you so much for calling calm! You are the smartest of all the reviewers

[@MikeBob2023]

/agree

[@preludelinux]

Been using Linux for gaming and everyday usage for a long time.

[@funkymoine821]

if i may, nvidia or adm gpu ? (linux main here , with a second rig for windows gaming, planing to to ditch second rig)

[@Linuxhippy2]

Incredible - all those games benchmarked have been developed on and tuned for Windows, they even are compiled & linked against microsoft's DLLs quite likely using microsofts compiler - yet WINE in conjunction with the linux kernel have become so good that actually some games are running faster despite all the runtime translation going on. Awesome!

[@v3xx3r]

I switch back and forth between intel and amd idk why people give a single shit about a company.

[@heyarno]

On the bright side, there is a much of potential for optimisations.

[@BenState]

HUB taking notes

[@gasracing5000]

F n a man, f n a. Good video full of valuable insights 🤜🤛🏾