Counter-Strike 2 GPU Benchmarks & Comparison: NVIDIA, AMD, & Intel Tested
Vložit
- čas přidán 27. 07. 2024
- Sponsor: Thermaltake SWAFAN EX 120 3-Pack on Amazon geni.us/ardZR
This benchmark tests Counter-Strike 2, the successor to CSGO (Global Offensive), for GPU performance across a few maps and a lot of video cards. One of our primary areas of interest for this game has been frametime pacing. Although not too bad overall on our systems, the occasional hitches we did encounter were often large or noticeable even as one-offs. We spend a bit of time exploring those, but also talk about the difference of test environments with smoke grenades, incendiary grenades, or just firefights. We ultimately settled on a benchmark scene that involved smoke grenades for about half the benchmark, with a mix of traversal and firefights for the other half. This represents the average player's round pretty well. Testing included cards like the NVIDIA RTX 4090, RTX 4080, AMD RX 7900 XTX, 7800 XT, and plenty of prior generation cards, like the GTX 1650, GTX 1060, and so on. Intel Arc was also tested for the A750 and A770.
PATCH UPDATE & QUICK TEST: Valve has posted a lot of patches since this video. A couple days after we published, they posted a patch that's meant to address some of the shader compile or frametime spiking ("hitching") behavior on NVIDIA cards specifically. We ran some new numbers against the update and there's improvement in a few places; notably, in instances where we experienced massive stutters on NVIDIA due to (what we suspect) are shader compiling issues, it looks like they're improving. Our RTX 4070 was affected by this in our 1080p testing. AMD was also affected generally. We just ran the 4070 and 4090 through with the new patch and did not encounter the same stutters that hurt the average. The 4070 is now about 8% ahead of the 3070 (as this was effectively an RNG of if the game ran well, it was occasionally ahead and sometimes massively behind last data set, but now is consistently slightly ahead). That was either an error or a bug in the game, our data suggests a bug but we unfortunately don't seem to have a way to run the old version (that we know of - let us know if there is one) to validate against, so we just have to run new numbers and then make educated estimates against prior versions. Valve noted that the frametime pacing behavior is still present on AMD to at least some extent. They claim they're working with AMD, so we assume that means a potential driver update. We'll run everything fresh once Valve has some time to slow the daily updates - they are pushing them out so fast that we can't actually complete a round of tests before the next one, and each update could make the data incomparable, so we'll wait til they go to a slower cadence. Fast updates are good - they're fixing stuff. They're just not good for testing a lot of things.
Starting today and for a few weeks, we are giving away 10% of our store revenue to help our local cat shelter Cat Angels to fund their move to a new adoption center! Check our store here: store.gamersnexus.net/
Like our content? Please consider becoming our Patron to support us: / gamersnexus
TIMESTAMPS
00:00 - Counter-Strike 2 Benchmarks
01:47 - CS2 vs. CSGO Improvements
03:39 - Research: Maps & Smoke Grenades
05:56 - Frame Pacing Issues
07:08 - Frametimes (7900 XTX)
08:06 - Frametimes (RTX 4060)
08:43 - Frametimes (RTX 4070)
10:52 - 4K GPU Benchmarks in Counter-Strike 2
12:34 - 1440p GPU Benchmarks
13:35 - 1080p CS2 GPU Benchmarks
15:04 - Check Pinned Comment & Conclusion
** Please like, comment, and subscribe for more! **
Links to Amazon and Newegg are typically monetized on our channel (affiliate links) and may return a commission of sales to us from the retailer. This is unrelated to the product manufacturer. Any advertisements or sponsorships are disclosed within the video ("this video is brought to you by") and above the fold in the description. We do not ever produce paid content or "sponsored content" (meaning that the content is our idea and is not funded externally aside from whatever ad placement is in the beginning) and we do not ever charge manufacturers for coverage.
Follow us in these locations for more gaming and hardware updates:
t: / gamersnexus
f: / gamersnexus
w: www.gamersnexus.net/
Host, Writing, Testing, Editing: Steve Burke
Testing, Bench Course Match: Mike Gaglione
Testing: Jeremy Clayton - Hry
PATCH UPDATE & QUICK TEST: Valve has posted a lot of patches since this video. A couple days after we published, they posted a patch that's meant to address some of the shader compile or frametime spiking ("hitching") behavior on NVIDIA cards specifically. We ran some new numbers against the update and there's improvement in a few places; notably, in instances where we experienced massive stutters on NVIDIA due to (what we suspect) are shader compiling issues, it looks like they're improving. Our RTX 4070 was affected by this in our 1080p testing. AMD was also affected generally. We just ran the 4070 and 4090 through with the new patch and did not encounter the same stutters that hurt the average. The 4070 is now about 8% ahead of the 3070 (as this was effectively an RNG of if the game ran well, it was occasionally ahead and sometimes massively behind last data set, but now is consistently slightly ahead). That was either an error or a bug in the game, our data suggests a bug but we unfortunately don't seem to have a way to run the old version (that we know of - let us know if there is one) to validate against, so we just have to run new numbers and then make educated estimates against prior versions. Valve noted that the frametime pacing behavior is still present on AMD to at least some extent. They claim they're working with AMD, so we assume that means a potential driver update. We'll run everything fresh once Valve has some time to slow the daily updates - they are pushing them out so fast that we can't actually complete a round of tests before the next one, and each update could make the data incomparable, so we'll wait til they go to a slower cadence. Fast updates are good - they're fixing stuff. They're just not good for testing a lot of things.
Starting today and for a few weeks, we are giving away 10% of our store revenue to help our local cat shelter Cat Angels to fund their move to a new adoption center! Check our store here: store.gamersnexus.net/
It would be great to see a CPU test as well! Thanks Steve!
oh so you guys already received 14th gen for review? Hopefully this raptor lake refresh is better than the rumored 3% but I think I'm wrong :(
CS2 feels terrible, enormous amount of visual "lag"/high frame time on my rx6800. Maybe there's some sort of hidden setting that I am missing or maybe freesync/vsync is not working as it should
I don't like that they've replaced CS:GO with CS2. It bothers me from a game preservation standpoint, and it bothers me that a lot of people with old/low end hardware can no longer play the game. CS:GO ran comfortably on my i7 920/$20 1366 Xeon from ebay and an 4GB RX 480. It also bothers me that CS2 launched with a smaller map pool and they've taken away several of the game modes, and even community server support isn't there yet.
Games as a service is a nightmare for the end user.
Seems strange that most of the game modes and maps are missing, along with alot of settings like left hand are missing.
A friend with a FX 8370 notice that he got good FPS in actual matches, but barely around 40 with BOTs in practice, so... for people testing the game by themselves, be aware that bots are CPU heavy in low end hardware
Weil i have a 5900x.
And with Bots its still awfull But actuall gameplay is good
I'm on a i7 2600k and same stuff happening to me. offline bot matches and online dm and casual matches are a nightmare....premier and competitive on the otherhand is absolutely fine
yeah i go from 190 ish in an empty nuke to about 140 when i add a couple bots
Same here. I was "testing" with bots and i start crying hehe, gameplay runs OK in FX8350
same goes with dota 2,bots does reduce your fps
9 days without a video had me worried on if something happened to Steve, or someone finally fixed his sleep schedule .
Just trying to catch-up on a lot of background projects!
@@GamersNexusSteve was grieving at the death of CSGO 1
@@GamersNexustake whatever time you need. Not trying to push you guys, just genuinely impressed with how you have the energy to do so much, and I’m saying that as a guy in his 20’s.
@@LukeBroadhurstfilm Calling it "CSGO 1," hahaha
The latter will never happen
Worth noting that AMD gpu's seem to have shader caching issues in CS2, which might lead to the overwhelming stutters in game (and 1%/0.1% lows).
There are multiple reddit threads about it with many solutions, but the most popular one so far is disabling DXNAVI.
Edit: Seems like AMD has addressed the issue in recent driver update.
We'll check that out! Maybe worth a deeper look? We normally stop at benchmarks out of box, but sometimes tuning can be fun.
@@GamersNexus Wow, that was fast! I think that the review is great as it stands, because it shows that Valve clearly has a lot of work to do with the game's performance, but as I have been monitoring social media a bit, it seems to be a common issue among AMD gpu owners and highlighting the potential fix would be great for the community before it gets fixed by either AMD/Valve. Thanks for amazing analysis again!
@@GamersNexusThe big problem is once players are on the map, so benchmarks running around the maps won't trigger this.
@@GamersNexuscards with the dxnavi optimizations seem to be affected by the cache resetting and worse shader stutters from what I've gathered so far
It saves shader cache on cs2 exit and it seems like once it reaches 16mb+ and saves again then on the following launch it invalidates the shader cache and starts over again
Yap. I noticed that too. I have to rebuild my shader cache every 2 or 3 time I play cs2.
The issue is, in a game like CS where you have 1 life per round and the ttk is so low those spikes of almost one second really matter. I've had a few while playing casual on my 7900 XTX just as I was testing the game and it's enough to not make me go into ranked play yet. One of those in a ranked match can lose you a round and that would feel so bad in competitive. Probably gonna wait a while on this one.
yeah, also have a 7900xtx. Game is unplayable on highest settings. High still bad but okay for casuals
Especially with the netcode being wonky, this kinda inconsistency also makes it hard to understand whats exactly going on. Sometimes an enemy just did something 50 ms faster than you, and netcode/stuter/etc can easily hide that fact.
If you have the muscle memory, its not that bad. But building it up is kinda reliant on that, and it just "feels" more fun to have a stable and reactive game.
I've got a 3080ti and in any settings I experience absolutely terrible frame stuttering almost like a lag spike. But my latency remains low at 15ms all the time so this frame timing issue finally makes sense. It was driving me mad as it was only happening to cs2
@@Nawchos Against good players, the TTK in CSGO is lightning fast. Depends a lot on skill level.
I agree, stay away from Ranked until this bug is 100% banished to the point that nobody remembers it.
And I have the math to back that up:
Ranked matches in CS:GO could take up to 45 minutes to complete.
The game usually runs at or above 200 FPS.
And there's 60 seconds in a minute.
What does that all mean? Well, it tells you the minimum average frames in a match.
A single match, given those parameters, will on average contain 540k frames. Might as well just say half a million frames.
This problem with frame timing is happening far more frequently than "1 in a million odds of any frame being problematic in this way", but I can still use 1 in a million to prove the point, so what it's actually doing is much, much worse.
Even with 1 in a million odds of frame timings getting messed up on any given frame, you run past 1 million frames rendered in only TWO ranked matches, at an elapsed play time of 1h30m per encountered hung frame (on average).
But like I said, it's actually doing much, much worse than that. 2 runs of the same group of testing, in 2 cases, produced a total of 4 anomalous frames, which I'm gonna call "hung frames" because it's like a hang-fire with a firearm, it's gonna go off, you just don't know when.
To be honest, I think the odds are more like 1 in 50k that you get a hung frame.
Going off of those odds, it changes my warnings about CS2 from "Don't play ranked until this is fixed" to "Don't play the game at all until this is fixed".
And I still have the math to back that up:
1 in 50k odds, game running at an average of 200 fps, 60 seconds in a minute. A hung frame will be encountered on average once every 4 minutes 10 seconds.
I'll show my work: 50k frames / 200fps / 60s in a minute = 4 minutes with 10s left over, or 4m10s between hung frames, on average.
In any case, no matter how frequent the hung frames are, they can pivot the whole match from an easily handled victory to a crushing defeat, and it's 100% not the player's fault.
Going by that alone, even the 1 in 1 million odds are enough to scare me away from wanting to play ranked.
The 1 in 50k odds are enough to make me change my opinion from "Game's fun, just don't play ranked because of the bugs" to "This game is not fun to play because of the frequency of the bugs".
I've left negative steam reviews on games because of less severe issues than the 1 in 1 million odds case of things, and I have seen others leave negative reviews for far less than that.
So I can only see a flood of negative reviews coming the way of CS2 if they don't manage to fix this game post-haste.
CS2 has options for both DX11 and Vulkan renders (though at this point the Vulkan renderer just seems to be worse in terms of 1% lows and avg frames than DX11, and it is also being actively patched). That may be worth looking in to.
They just posted a patch talking about the Vulkan renderer last night, I think, so they're still working on it. We're hoping to run a comparison against both once Valve has some more time to patch-up the game. Seems like it's still changing a lot right now, but that's a good thing!
This is absolutely gonna absolutely suck for us 11 Linux users
edit: just played for a bit and i had absolutely no issues with stutters. i guess this _won't_ be a problem for us 11 linux cs players
@@Mizztvulkan works fine. I have yet to try non-vulkan on Linux in cs2 but on go my frame rate was minimum 120fps worse when not using vulkan. Considering I only got 170fps on vulkan, that's horrible.
@@Mizzt
> 11 Linux users
How many devices has steam deck sold? all linux users there.
@@XiaOmegaX I was just being sarcastic lmao.
Hey Steve/Crew,
Ordered a modmat, although I don't watch you much anymore, I remember when you first went to youtube from the forums and the many ways you have helped me in learning about computer technology. I mostly just watch your technical content/plant tours now. Thanks for all the amazing information over the years.
this man & the team are goat!
I hope you will follow this up with a compilation of results on Medium and Low settings: as you suggest, few people are going to seriously play this on the highest settings so those two sets of results will be meaningful.
yeah, I really doubt that many people play the game on high settings. 90% of fps-players immediately set everything to low, don't they? 😂
@@89gauna they dont
Even more importantly, nobody will play this @4K.
No. Why would I make the game needlessly ugly? Especially with CSGO having pretty good visual clarity, so you dont need low settings. Beautiful maps like Inferno are awesome, even in a competitive game.
And I doubt nobody cares about visuals. Like, theres a reason the game has more detail than CS 1.6, or people being obsessed with skins.
And with older lower-end cards, what most players use. Like 1050(Ti), RX 570
Thank you so much for putting together all of this data and bring attention to the stuttering. It was driving me crazy and I thought it was just my hardware and settings. Looks like Valve has some work to do on CS2. I am big fan of your detailed and scientific approach please keep up the good work.
This is my favorite game for well over 15 years now so please keep it in lineup and do some follow ups! Thanks for your time and effort!
Thanks again for all you do, guys.
Thank you so much for the support!
He pulled an LTT on this test... don't encourage him
@@mikeb3172 how
@@mikeb3172what are you smoking, dude?
@@mikeb3172 elaborate?
Thank you so much for taking a deep dive into the release performance and problems of my favourite game ❤
The frametime variance is definitely a large contributor to mouse aim feeling inconsistent and sluggish. It just doesn't feel the same as csgo on the exact same sensitivity, even if you match cs2's forced raw input.
It's kinda like the 1.6/Source strangeness
@@ZGtx 1.6 is the benchmark, and it's sounding like they failed =[
Sounds like a lot of people not good at CS making excuses they aren't good.
It's like when people used to write (cold hands) in their Xfire screen name
@@ZGtxwhat was weird about 1.6/source
CS GO has the same exact issue, but people can run it at way higher FPS so it's less noticable.
I would really want to see a CPU benchmark for CS2 as well. I have seen many threads in reddit from people with low/mid-tier GPUs that never even got close to being 100% utilized. I experience the same with my Ryzen 3600 + Rx 5700 xt and I also have some very low 1% lows. or does something else like RAM speed bottleneck many older systems? would be really interesting to know and I would love to see you dig into this topic a little more.
edit: btw, dropping a bunch of smokes around me is the only way I can push my GPU to 99-100% utilization. the fps are then higher than they are in normal gameplay... around 200-250. in competitive game, I'm more around 120-200, depending on the map (and with GPU usage of 40-60%). playing with mostly low settings on 1080p as most players probably do.
That's interesting. I was surprised on my setup that I appear to generate over 200 FPS average with the settings that Ropz posted recently on ESL's channel (so 1080p mid I would say). I have reflex on, even though it's supposed to be broken at the moment. I have a GTX1660 6G and a 5950X. Haven't checked utilisation and lows, but if feels smooth and like I said I checked FPS and saw it was running in the 200s.
Edit: I haven't tested in competitive play though.
Thank you Steve and your team to bring this test in Counter Strike 2 it’s was a must know from you guys point of view it’s this a game you guys will test with or is still not at that level Thank you in advance
Great video overall, and I watched until the end. Small note at 16:25, I believe there is a difference between "sub-tick" and "tick-less"
This is the 1st time i've noticed the monitor wall being used for more than mostly gn logos. I like it!
The spikes in my opinion, feels like a cache thing. I had a lot of spikes with my 6800xt the first match but never again after 2 matches played.
Same exact experience with the same card.
@@Trenchv1true im on a 3080 and 3700x and day 1 i was dipping to 58fps now im at 150-200 on max settings
This, I've not messed with S2 to see if they have something like Source 1's VProf, but I can bet it's unchached sounds
Same with RX6800@@Trenchv1
CS: GO is still better. XD
Woah! Awesome video, but what impressed me is how good the A-roll looks. Great job!
Let's GO!
Yay for more benchmarks.
I def will be testing it but I hope they fix the random stutters. 😊
Was watching long doors and I swear two guys just teleported in front of me. Thought i was losing my mind, looks like it was a stutter now after watching this.
Actually, they added a teleporter to the buy menu!
This studio background and lighting looks great!
Great work. Frame time pacing and consistency is exactly what you experience when playing in its current state.
But the video didn't show frame pacing issues, it showed stuttering. Overall the frame times were fairly consistent, which only a few ginormous spikes. Bad frame pacing would usually mean some frames being above, and others below the average frame time.
Last time I was this early, a mid range gpu was under $700
Under $400*
Edit: 1070 ti was $400 msrp. And it still is able to do 3440x1440p gaming 6 years later.
so you havent been early until now?
Would be interesting to see a Cpu comparison and whether the frame time issue is ralated to Intel CPU's as my 5800x3d + 2080ti doesn't seem to be producing those massive spikes, possibly smaller ones around 50 ms though.
I also have a 5800X3D + RTX 4070 and I definitely have these frametime issues.
5600 and a 3070, never had a single frametime spike since they opened the beta, played around 30 games thus far
its 100% shader cache issue, i am running it on linux with proton and shader cache pre-downloaded, and i haven't seen much spikes, even at unlocked framerates..
5800x3d + 4070ti, im pushing 300-500 in mm and the game still feels like shit bcs of random frametime spikes
Same 5800x3D + 2080ti, never seen huge frametime spikes aswell.
I am every time amazed on how you guys test things. 👍👍👍
I really love what you did with the decor and the lighting in your videos ! It really pops while looking comfy to my eyes for some reason ! Keep up the good work 💪 !
3060ti beating 4070 and 7700xt in 1080p is kinda weird.
almos tlike the 4070 is a scam
@@cadetsparklez3300 It is a scam.
@@cadetsparklez3300yes,in this case tho the game is clearly bad optimized for those new gpus
Legit most benchmarks are showing 3060 and 3060 TI rigs with the least issues. One of my friend has a 3050 and his game is running perfectly fine with high fps on medium settings. I'm wondering if Valve has first optimized this game according to the popular cards that people are using on Steam instead of higher end cards
Not really
Man, y'all's new set is amazing! The quality of the videos has gone through the roof!
Thank you! The new set is also a lot easier for us to set up!
Thanks for the analysis video
"over the years-- decades now" haha!
... yeah. That was a painful realization.
Thanks for adressing the frame pacing question. I was starting to get paranoid, having a 6900xt (with a 5800x3d) that has sporadic frame freezes out of the blue (especially on death match servers).
Sadly the player base of CS:GO is so heterogenious and huge that you don't find useful fixes because some players don't even properly recognize the problem and some "fixes" get just repeated because of too many beginners spreading them as solutions ...
Edit: The only real fix for AMD users is the DXNAVI driver mod (which is embarassing for AMD adn CSGO at the same time, that they didn't do anything by themselves about the shader loading problem).
They are fanatics.
4070 Ti getting 50% more FPS than 4070 at 1080p seems kind of insane, I wonder what's up with that result.
EDIT: Not to mention 3060 TI (!!!) and 3070 outperforming 4070. 13:35
And the 3070 beating the 4070 at 1440p seems weird to me.
@@dhm-2920 100% fake or just bad position, and writing fps meter. Its raddiculus. And crazy weird. I sell 3080 my and buy 3070 i think its fine for 300fps low 1280/960 stretched
could this do with vram and bus bandwidth?
The 4070 benchmark at 1080p is really weird. 4070 didn't improve when going from 1440p to 1080p. At 1440 its better than the 3070 and 3060ti.
@@melzyy 4070 has the same memory configuration as 4070Ti. It really looks more like they have an error in the chart.
I am so glad you called out the frametime issue. No one is talking about it and it goes away with V-sync on so, some people aren't even experiencing it.
i have fresh install of w10 22h2 with windows update disabled and i noticed that it is best optimisation possible. Before, I had all updates and junk and i had insane stutters every 2-3 seconds and today i had 1 small stutter over 3h of gameplay
The most important question for low end hardware in 2023: "Will it run CS2?"
surprisingly it not only ran better but the graphics looked better then csgso on my low end laptop. What im saying is csgo had worse graphics then cs2 and yet also had worse performance then cs2 for me which is a pleasent surprise.
@@HeadoN1337 you do realize that this was done on purpose to see the game's scaling? At the lowest settings it is highly likely that they will run into a CPU bottleneck at some point, which makes zero sense when you're doing a GPU benchmark
yes if you have dlss 3.5
@@HeadoN1337 Source Engine games were the type that you'd always put to maximum settings because it heralded from the time where the maximum settings actually meant something.
RTX 4070's performance at 1080p has to be a mistake in the chart
This is not LTT
@GamersNexus Can you share the replay you used and how to run it? I honestly can't say I've seen such spikes on my system (Ryzen 7950X and RTX 4080). I play at 1440p @ 144Hz. I've tried vsync on (caps at 139 fps for some reason) and vsync off (up to 380fps). Certainly I haven't seen spikes of nearly 1s!
I was waiting for this... :D
The scaling at 1080p maxed settings from 100fps on a GTX 1060 to well over 300fps on top GPUs before getting CPU bottlenecked suggests overall potentially good optimization as long as they quickly identify and fix those occasional frametime spikes..
I think it might be interesting to you (and valuable to millions of gamers) to test this game again a few months from now. valve is patching this game like there's no tomorrow. CPU scaling is a great theme for a video. also, since so many people who play csgo do it in low end machines, it'd be interesting to see 720p, 900p, 1080p LOW settings comparison video testing older graphics cards and/or cpus.
I still know some people who were playing csgo in a gtx750ti 😅😅
Id love to see you guys test that new AMD Fluid Motion Frames thing. That looks crazy for the amount of FPS that is given but I want to know what are the trade offs in quality.
This is not the game for that
I'm also very, very much waiting for GN to do a deep dive into FSR 3. I don't even own a Radeon, I just wanna see how it compares to DLSS 3 and how FSR 3's framegen performs on NVIDIA hardware, especially non-40-series cards.
Frame gen increases input latency. That’s the one thing u don’t want in CS
You mean FSR3? Its probably gonna be a slightly worse version of DLSS3, if its following the pattern.
Mind, it doesnt acutally give you "more frames", rather than generate smoothed frames for inbetween. So your game looks smoother, but you will not actually have the low lag or nice frametiming of high framerates. In fact, your input lag will be higher.
Motion will also look smoother, but have artifacts and could be a bit blurry.
Might be nice in a "cinematic" game or so, but not in a competitive shooter or action-based online games generally. No point for CS.
great video! im liking cs2 so far, they are working hard on it and i think they will update it much more frequently than csgo because of the code and engine. Im only missing some gamemodes and maps which i bet are gonna get released soon. I hope they fix the frame pacing and maybe work on vulkan more, sadly dota 2 and half life alyx (both valve source 2 games) had many issues on vulkan and still have in my experience
Great video 😊
Worth noting that the frame spikes shown here are in review mode, so even the enemies stop. But in an online match if you get a spike of 900ms, you're dead. Hope they fix this stuff (game needed more time in the oven, crazy move to just replace a working game out of the blue)
Valve Will never stop surprising us, abandoning perfectly updatable games like TF2, never releasing a 3rd sequel, never addressing important issues in the community such as cheating and scamming, now this fiasco with csgo....and thousands more that i feel bad for giving them money to do Jack shit 😢
Wonder if these large frametime spikes are only when viewing demos. Never got spikes like this in a real match. The replay system is pretty buggy and crashes often.
I've done some testing on the game's performance with dev mode on - nearly every frame stutter was caused by large chunks of the map suddenly being loaded at once.
For context:
Unlike in CS:GO, CS2 renders a large portion of the map, whether it's visible to you or not. Sometimes when moving your mouse, even when standing in the same spot, large chunks of the geometry that's invisible to you will load & unload, as I observed by testing with CS2's inbuilt dev mode & wireframe commands.
For example, large parts of B site overpass are still loaded when standing near toilets on A long, and nearly all of A site inferno is being rendered from bottom mid. The game is incredibly unoptimised in this regard.
When testing this with frametime graphs & gameplay footage, you could tell that this excess of geometry being loaded & unloaded happened at the *exact same time* the frametimes spiked on the graph, every single time as far as I could tell from my testing.
Which commands were you using? Would love to replicate this!
@@GamersNexus Update : i've found the solution!
Launch the game with -tools in the launch options, and enter *sv_cheats 1* and *mat_wireframe 2* in the asset browser window that launches with the game :)
I would also like to note that if you perform some testing with these settings, you should also check out the parts of the map being rendered when scoped in (especially with an AWP) - it seems to render most objects through walls no matter the distance when scoped in for some reason, which could explain why AWPing feels a lot less responsive than it did in cs:go
Why did valve do this? What CSGO player was asking for them to upgrade the engine? It was only non CS-GO players laughing at us playing an "old" game that had the most players ever consistently on steam for years.
@@MrChris00078bro GN responded to you and you didn’t answer the question they asked 😭😭
I am working on a review for a mini pc with an AMD 680m graphics chip, and the frame rate was usually fine, but like you said the frame times were horrible at times.
the other thing you should check out is the CRAZY difference between using intel cpus vs. amd x3d cpus...its a massive framerate difference.
Would have loved a test with the launch command -vulkan
Ahh the good ol internet cafe CS 1.5 days. Nothing like hearing 60 $5 speakers going off at once lol
Hi, thanks for testing. I run a 3080 at 1440p personally with similar settings, and haven't had any noticeable frame pacing issues myself (outside of when I was redlining my RAM, thanks chromium!). That being said though, I ONLY play on Inferno, which is a very largely segmented map. I know when @3kliksphilip was doing some earlier testing, moving to certain points on maps and looking at certain angles could load way more of the map in than is intended, leading to performance drops. I'm sure Valve will iron these out overtime. Overall though, I've been enjoying my time with the game, even if it does make me feel old now.
Great lighting on the set today, not sure what if anything changed but much better than normal
Why does 4070 have the same FPS values for 1440p and 1080p? Clearly a mistake has been made somewhere in a process
Csgo2 is still alpha build. No where near an actual working game. They forced us into it so they could make more money on skins.
I have a Lenovo Y700 (Y700-17ISK)with a Nvidia Geforce 960M and it does ok for this game. Yes I know my laptop is old, but I mainly use it now for music production (FL Studio) or to play older games. My favorite games to play are Ghost Recon Wildlands or Arma 3. It does handle Both of them fairly well with the graphics detect option in the options menu. Thanks for this review and Keep those videos coming guys!
Thank you for the video!! I believe the game deserves a deeper look, but I would also advise NOT TO do too much benchmarking, as there are multiple patches every WEEK ever since the release of the beta of CS2 few months ago.
From over 100 hours of CS2, spread across launch and Limited Test versions, I have never felt any of the second long stutters what so ever. I also have a 2080 Super so not quite top of the line anymore. Could it, instead of being a game issue, be a replay issue? Were the frametime problems present also in the pure gameplay tests?
Are you playing at 4k? These are all GPU bound tests.
@@Chopper153 Hm.. I play at 2560x1440 with around 200 FPS for the most part I think.
@@Chopper153 With that said, I believe I might have felt a stutter today of maybe around 400-500ms. However that was during Deathmatch which I haven't played much before. Still not a single case of it during normal play.
I thought my 3080 was going to be lower, gotta play with settings to keep FPS higher than 200 @ 1440 at all times but am happy to see its still very viable.
The 5800x3d isnt stressed with my 3080 in this game, I was expecting a little more CPU load.
I had some significant stutters at the beginning of a match on a map. I wasn't part of the limited tests. I figured it was some kind of asset loading/caching or something that shaders just didn't do well at the start. I would also get those 1/7 second drops in some action which disproved that.
I turned on FSR in AMD Adrenaline and turned off their overlay I forgot was on. Overall I think FSR did the trick. Been playing on a 4k tv at 1080p so not really overdriving my RX 6750 XT. Game set to Ultra Quality, rest on high or recommended defaults
Thank you for all the hard work on this wonderful first person shooter.
It’s great to see benchmarks for CS2 (shooter). Can we also expect GPU and CPU benchmarks for CS2 (city builder)?
Yeah I think you can expect that. Especially with the recent changes to recommended specs
I'm having a lot of stutterings in CS2. I have a 7900XTX and 7800x3D. Csgo ran smooth but in CS2 is being annoying. Looks like this happens more with AMD gpus. Looking forward for this fix, so I can really enjoy the game.
Yeah, it looks like Radeon 7000 series is underperforming across the board for some reason.
Definitely seems like something that'll change in the future with driver updates.
Google "disable dxnavi", worked for me with my 6800xt. Seems to be related to the shader cache, I won't claim to understand much more. I had terrible stuttering before the registry edit, and zero afterwards.
We need CPU benchmark video Steve!
i'm happy at the thought of steve and friends being able to play games instead of just benchmarking them in the making of some of these videos lately
I might be in the minority with this opinion since I am on a 4k monitor, but the stutters were unplayable. cs:go was the smoothest game and although i didn't play much cs:go in recent years I do have over 1000 hours. And anyway I was excited for cs2 to be a reason to come back to the franchise but it turns out I cannot play with the stutters I was getting (7900xt btw). fps for me was over 300 on low settings(I played on low back in cs:go as well) but with the stutters during fights I just cannot forgive valve for deleting cs:go instead of leaving it as an option.
This kind of post is super helpful for us because we can try to use your info/experience to look more into it. Great input to see others having these issues. Sounds like yours might be worse than what we saw. What CPU do you have? Did you also use this system for CSGO? No problems there? Were you at 4K? Any particular map/scenario it happened, or sporadic?
@@GamersNexus Not OP but I did have a similar experience on my 5600X + 1080Ti. Inferno is probably the worst map, but these stutters are observed on any map when smokes are disappearing, looking through molotovs and when smokes interact with HE grenades or molotovs.
@@GamersNexus I saw the stutters on all the high-end cards I tested the game with, the 7900XT, 7900XTX, and 4080. CPU is 7800X3D.
Im also on a 4k monitor but with a 13700k + xtx combo instead. I also found it unplayable and after fiddling a bit it turned out turning all settings to low, but keeping 4k works best. Still looks ok (way better than going 2k or even using upscaling) and no noticeable stutters. Iirc smokes and incendiaries didnt seem to do much either, but maybe i just didnt notice since i just played dust 2 and not too many were used
CSGO was super smooth for me. CS2 on the other hand is just unplayable. On a 5600x and 7800 XT.
Can you please do the same for CPU?
A cpu test would also be a nice addition :)
I am experiencing that frame time spikes. I tried with two different cards and it's like having a lag spike. It's so fecking annoying. I hope this gets fixed soon!
about frametime:
I have played like 5 games since the release and have not felt a single frame time problem, but it might explain why the game sometimes feels a bit weird.
To be honest, a lot of people cant really descsribe why a game feels or looks a certain way. Theres a lot of people that swear nothing is wrong with CS2 as well, despite mountains of evidence about tech and gameplay issues.
But how well the game performs will still affect how you play, even if you dont notice it.
it might be a bigger problem on higher settings and resolutions than lower ones
I have it set on 1440p at highest settings, with my 4080@@squidwardo7074
I saw your review of the RX 7800 XT, which was a GPU that I was planning to get, but after the disappointing benchmark results, I've opted for the Powercolor Red Dragon RX 6800 XT.
Great choice! The 6800 XT is a great value right now in a lot of places.
Hey, great video but would you guys consider doing a cpu comparison for this game? Also think it would be very useful to have some 4:3 testing as that is how me and all my friends play + a lot of players in the community/pros. Thanks.
Not sure that feasible - swapping out CPUs is a bit more labor intensive
Thanks for the great work ! I'd like the same but with CPU comparaison... 🙏
Love the benchmark but can you do 1 more test where most the settings are low/disabled as most people's going to play the game that way
One thing worth noting for anyone looking for more FPS. Reduce the Anti Aliasing setting. I have a 7800x3d and a 4080 and went from averaging 300 fps at 1440p to 450 when I went from 8x MSAA to 4x MSAA
Yeah mssa taxing af, 4xmssa should be enough for 1440p in any game
isnt msaa on 1440p kinda useless, msaa on lower resolutions would make more impact
Same HW as me
I mean MSAA has always been taxing for a good decade. lol
Anisotropic filtering also eats fps where there's loads of water like anubis CT spawn
Feeling hopeful seeing the 3070 FE float around the middle of these charts. Thanks for the information. :)
Nice job. Now please do a CPU benchmark as its even more crucial for CS2 (Crappy Sh..t 2).
I'm really upset about how they released CS2, completely removing CSGO was so anti-consumer (yes, I know using beta channel stuff you still can play -- NOT THE SAME)
Also, I wonder how huge GPU benchmarks look on Linux :P (performance seem fine to me)
100%
Anti-competitive isnt the right word for it. Anti-competitive is like AMD sponsoring Starfield and telling Bethesda to not have XeSS and DLSS in it (until the community rioted). Its more anti-consumer, as some people paid money for CS:GO and now Valve removed access to it, although CS2 is supposed to be a major update to it.
You can still play CSGO. For the BETA options from steam, choose the one that says csgo_demo_viewer 2023.
Valve only removed matchmaking.
@@__aceofspades you liked your own comment
@@__aceofspades thanks, was a typo, I'm the best proof reader :)
I appreciate the results.
Please review with low settings since competitive is usually played with highest posaible frame rate.
focus on old GPU, maybe from GTX 10 series and RX 400 series.
CS:GO was a game that was enjoyed by a lot of players. This game is pretty popular in 3rd world countries.
Would be interesting to see which GPU is the sweet spot for playing CS2.
My friend with old laptop which runs with i7 7th gen + GTX 1050 said his laptop cannot play CS anymore, at least competitively
We'll do Low or Medium for CPUs!
Uncapped framerate is stupid anyway. Diminishing returns after about 300 fps, higher GPU usage (which correlates to input lag), incosnistent frametimes / framerates depending on the area of the map (ie one area might run at 400 fps while another at 700 fps).
@@GamersNexusthank you! it seems to me like CS2 is still heavily CPU bound on low settings. especially on older CPUs, you get some very low 1% lows (not the stutters you describe, I think those are just shader caching).
thanks you! I thought I was going insane with so much frame drop
Are you ever going to get back to doing case and cooler reviews. That's the main reason i started watching this channel.
CS2 isn't tickless. It is still running at 64 ticks, but it has sub-tick implementation for aiming and shooting.
CPU benchmarks would be very helpful btw. As a competitive player I am much more interested in this as I am running the game at 1080p low settings anyway.
Those spikes are concerning, hopefully whatever is contributing gets figured out sooner rather than later by the devs.
ya
There is some weirdness with shader cache in this game. I was having frame pacing issues and wiping the DirectX Shader Cache on the Disk Clean Up tool solved it.
We've noticed within my friend group that the 6700XT / 7900XTX players were having significantly more and longer stutters than our Nvidia counterparts.
Our very basic testing seems to imply it's some sort of caching issue, we figured out a workaround by launching every single grenade type and using as many unique weapons as possible during warm-up
Though the latest bugfix update also seems to minimise issues.
Ie: the later you were into the match, the fewer of the crippling 100-900ms stutters you got, with grenades being the worst offenders.
very surprised by the 4070 vs 7800XT difference at 4K Very High...
We saw similar scaling to TechPowerUp and some others for that one.
@@GamersNexusyet in your video, 7800XT seems to beat the 6800XT all across the board, while in TechSpot, the 6800XT was better.
Similarly, your video suggests that 3070 is somehow better than 4070 at 1080p, yet that doesn’t happen at TechSpot
In their review the 4070 is right next to 3080 like it is in pretty much all of games, and yet in your video it performs on average much worse than even 3060ti. Must be a mistake, right?@@GamersNexus
yep, very weird.@@Fraudiola034
The RX580 8gb version seems to perform really well in CS2, maybe worth looking into.
yeah best value graphic card ever
with R5 3600XT at 1080p max settings It gives only around 60fps
Hey GN! Have you done a video on Nvidia reflex? And how it affects performance?
Thanks for the tests, Steve!
Quick question: Did you also experience more stuttering initially? I am on a AMD 5700 XT and it seems that the game suffers from shader compilation stuttering, but seems to run slightly less stuttery on another system with an Nvidia GTX 1070. playing for a while usually minimizes these short freezes, but occasionally, they still occur, just like you have shown in your frametime graphs. those spikes might not be caused by shader compilation then?
And if noticed these issues, did you ignore the first pass of your benches ("pass zero"), to make sure all shaders are compiled when measuring a run?
CS2!
So it's still perfectly playable on old hardware.
Yeah obviously. This game looks 10 years old.
@@username8644 no it doesn't...
What? 70fps?
@@Marco-cl9pb Bf4 looked better than this. That's a 10 year old game.
@@username8644 that's simply not true, huge improvement in the lighting and textures but ofc it's not Cyberpunk2077 visuals
Keep doing what you are perfect doing it. You are a legend all of you I will support you till my last minutes in my life.
I would love to see another counter strike 2 video. If possible, using some of the video cards represented from this video, but changing A-filtering anti-aliasing texture settings.
I have seen a lot of people commit to bi-linear texture filtering.... But my 3070TI does not seem to suffer any performance or input lag from running 16X, but I would love to see that compared across multiple vendors cards. Cheers
It would be nice to measure input lag latency, because in cs2 it is much higher than csgo even if you have 200+ fps
I'm definitely struggling on a 3090 and 5800x. It's fluctuating up and down from 120-160FPS on 3440x1440. Frametimes are messed up, I'm basically quitting until valve fixes it, not super fun lol.
🤡
@@sohiearth 🤡
There was one issue with antialiasing in beta i dont know if they fixed it or not but turning it to cmaa fixed some of the fps issues
@@sohiearth720p looks awful, i use 1440 x 1080.
I'm not sure if the frame time has been fixed since GN announced Valve patched some things, I feel like my post is after his update, I'll give it another shot to see if the frame pacing is fixed.@@OneDollaBill
Can you do an optimization guide for the best equilibrium of fps and video quality (and also the best settings for competitive advantage) like on Starfield
8:30 that weird frametime issue was rampant for me at launch. A couple days in and an update or two later, I haven't seen this issue come back and it appears to be fixed. I hope you guys would run the tests again after the initial updates to see if you can reproduce your results after the updates.
I'm on an r7 5800x, 3070 Ti, and 32 GB ram @ 3200MHz
It's weird because I think I had the frame pacing issues at the same time of your post, so I would not want to see it before your post. I still find the game unplayable on my 3090.
counter strike meh
~275fps in competitive matches matches with mixture of high/med settings (maxed textures) on 3090 and 5800x @4k, very happy with the performance.
Amazing video Steve and crew!
Personally I feel you should've waited for this video like a month or two more to see if there were more patches being put out by Valve (I think they will, hope so) considering the performance issues the game currently has.
I personally have "solved" the frametime and smoothness issues temporarily, after playing around with settings and resolutions for a while. My tradeoff, is that I have screen tearing. Or some, in some cases. I've tried anti-lag, vsync on, off, 16:9 full 1080p, lower settings, 16:10 and 4:3. Nothing truly gave me any setting I was comfortable with except 16:10 1680x1050, uncapped, all low. This was "the better" one for a while. And I've tried playing some weeks ago with vsync on. Reddit users who claimed the input lag was very low were correct. This was in favor of frame consistency. However, the input lag while log, is noticeable in certain things. In F1 terms, I'd describe the "mouse swings" as corners and say vsync on gives an artificial "slow" smooth feel to mid-cornering, which in this case, would be from the point you start swinging your crosshair to aim, if you have to track/follow up the movement to adjust, is where the "sluggishness" comes in.
My latest findings, in tradeoff for some screen tearing, was deactivating anti-lag in Adrenalin (My card is RX 580, sadge) which was what I've always used in CSGO. I also turned Vsync OFF. And with this card, I cannot double the frames at all times (double the frames to your RR means you can have vsync effect without the input lag). I've used (and this is a bit of a meme, I know) Radeon Chill instead to cap it to 75 (75hz is my RR). 75 alone makes you have a lot of screen tear (which is something I don't fully understand since the frames are capped to match the refresh rate but I guess these 2 are not the same thing working the same way therefore they don't match), so I set it to 75-85 and it was a bit better. In the meantime, I get the "raw input" let's say since vsync is off, and the frametimes and inconsistencies of going down to 70 and up to 180 doesn't happen. I don't know if this is like super logical or genius but I'm still pretty proud of myself for my own findings.
I think it's also important to note that the game is on my SSD (WD Green, sadge), my RAM rated at 2666mhz is at 2800mhz, 13-15-15-15-33, and the CPU which is a R5 2600 is at 3.9ghz (slight overclocks, stock cooler - is what I can do). If you have similar setup to mine or better, you should be able to replicate this or have better performance than me.
Also something else to note, bad internet such as mine, makes the game also be warpy and rubberbandy or have "stutters". Like if you're on an inconsistent ping, even if it's around the 50 to 80 range which is not that much, the game feels awful. When I get good network performance (My ping to the server would be around 27-40), it feels smooth.
I want to point out some things on the last question proposed at the end of the video: YES. The graphical overhaul is fantastic. In my case, I always liked the idea of graphics getting better (sadly, the industry decided that prettier meant shittier games but that's a different discussion and I digress, and no, I won't pay Jay for this) but even beyond that, it's 2023 and in CS this time having better graphics means actually better gameplay.
In the past, people came from 1.6 to Source or CSGO, and there was also the shift in monitor technology. Many were very used to the 4:3 AR and CSGO specifically, with its history of being first for mobile (or cross platform) and such, meant that the game itself no matter your graphics settings didn't change that much. In that sense, low settings were objectively better and not just for the frames in many instances.
In CS2 this changed and I think Valve knows this very well. Could you play it 4:3 and low settings? Yes, many people do. But having a crispier resolution makes models this time actually better to see. The maps lighting as a whole in the game got revamped and now you can actually see EVERYWHERE. You can distinguish models (Albeit with exceptions such as heaven in Overpass where the freaking shelf looks like a csgo ct and confuses our brains)- Another thing is reflections and shadows. Having global shadows turned on and if possible to high or very high, actually gives a gameplay advantage. In CSGO the only instance I am aware of this is in T spawn Dust 2 double doors. In CS2 this is a widespread thing and sometimes being actually key, such as in Ancient (mid square and A long).
So personally for me despite the performance issues (for now), Valve did great.
For the frametime issue, enable FPS_Max 0 in console and put this command in: engine_low_latency_sleep_after_client_tick true Thanks!