davis.anthony
Veteran
Not it's not, it's CONSIDERABLY slower...
Current Generation Games Analysis Technical Discussion [2022] [XBSX|S, PS5, PC]
- Thread starter BRiT
- Start date
Not it's not, it's CONSIDERABLY slower...
64 in that scene. Runs a tad bit higher. I also put every setting DF said was Medium or Higher to High just to be sure. Weather Particles is set to Very High but I suspect it isn't much of a factor in this scene.
As I've said COUNTLESS times, NVIDIA driver needs %20-25 more resources compared to AMD GPUs, a cost which also does not exist on consoles. NVIDIA practically do their scheduling on software. This fact alone is enough to skew all comparisons. This fact alone warrants the use of a Zen 3 CPU on an Ampere GPU. If you get an Ampere GPU it is best that you use Zen 3 or Intel 11th gen as a minimum. A person that spent 500+ bucks for a 3060ti+ tier GPU should also do what is necessary and upgrade their CPU. There's really no logic. Console makers can cheap out on CPU, they did so before.
Zen/Zen+ CPUs on desktop have enormous latencies compared to Zen 2 / Zen 3 products. It is impossible to take a Zen+ CPU serious in comparison to consoles. You will still get similar results, but it means nothing. There are tons of factors that puts Zen/+ CPUs way below consoles due to this. Also, PS5 is not even bottlenecked by its CPU in 4K-VRR mode. A 2700x also does not bottleneck a 3070 at 4K. Its a really meaningless angle to chase. You can use an 12900k or a 1700x, 3070 will behave similarly at 4K native Ray Tracing, more so specifically when you're VRAM bottlenecked. I've, Playsaves3, shared you that a 2700x is able to lock to a perfect 60 with ray tracing. And you still question the legitimacy of 2070 dropping 25s and somehow bundled it with CPU, GPU or other unrelated things.
"It’s not even console optimization the construction of a lot of pc ports make them more demanding to do a similar task basically the 3070 has to bench 20 pounds while the ps5 benches 17 and they both get 30 reps in"
This is simply wrong. You still do not get the main culprit after all this time. It is quite an achivement actually. What can I say? We have literal users sharing their 2080tis being %35-40 above PS5 in that specific scene. It is only %35-40 because game is not using complex ray tracing. An almost all games that use rudimentary ray tracing, 2080ti/3070 performs similar to a 6700xt and %30 above PS5. Only in games with really complex ray tracing such as Control and Metro Exodus 2080ti will take the ball away and run at +%50 perf. difference. So don't try to chase that angle either.
Zen/Zen+ CPUs on desktop have enormous latencies compared to Zen 2 / Zen 3 products. It is impossible to take a Zen+ CPU serious in comparison to consoles. You will still get similar results, but it means nothing. There are tons of factors that puts Zen/+ CPUs way below consoles due to this. Also, PS5 is not even bottlenecked by its CPU in 4K-VRR mode. A 2700x also does not bottleneck a 3070 at 4K. Its a really meaningless angle to chase. You can use an 12900k or a 1700x, 3070 will behave similarly at 4K native Ray Tracing, more so specifically when you're VRAM bottlenecked. I've, Playsaves3, shared you that a 2700x is able to lock to a perfect 60 with ray tracing. And you still question the legitimacy of 2070 dropping 25s and somehow bundled it with CPU, GPU or other unrelated things.
"It’s not even console optimization the construction of a lot of pc ports make them more demanding to do a similar task basically the 3070 has to bench 20 pounds while the ps5 benches 17 and they both get 30 reps in"
This is simply wrong. You still do not get the main culprit after all this time. It is quite an achivement actually. What can I say? We have literal users sharing their 2080tis being %35-40 above PS5 in that specific scene. It is only %35-40 because game is not using complex ray tracing. An almost all games that use rudimentary ray tracing, 2080ti/3070 performs similar to a 6700xt and %30 above PS5. Only in games with really complex ray tracing such as Control and Metro Exodus 2080ti will take the ball away and run at +%50 perf. difference. So don't try to chase that angle either.
Last edited:
I absolutely do run into CPU bottlenecks even at 4K, especially during rapid traversal where I see my GPU usage tumble down to the 60% and fps to the high 40s and low 50s at High settings with RT while my CPU is maxed out.
I'm almost convinced the elephant in the room for that game isn't the GPU like so many are harping about but the CPU.
Sorry but my 2700x never drops to low 50s, let alone high 40s with RT enabled at 3.7 GHz (so an improvement could be had with 4.1 GHz)
I don't know if you would call the traversal I'm showing in game is rapid enough or not, but this is the pace I usually play.
Also there's nothing wrong with the game regarding GPUs. Its just 8 GB GPUs buckling under the weight of 4K/Ray Tracing with added artificial limitation on VRAM.
I'm not having this discussion over and over again. I provide solid and sound hard facts with evidences. There's not much to talk about.
If my frames go up to 60 from 38 when I use High textures at 4K, then it means it is VRAM bottlenec;k
Some people evenwent so far to claim textures themselves have a rendering cost, which I had to prove it not being the case at 1440p;
Game performs like it should when you have enough VRAM. Even then, you might be talking about a situational thing. What we talk about here is general game performance. 2070 dropping to 25s in the very same cutscene PS5 renders 42-43 frames are being misused by specific users and reviewers to claim that PS5 is overachiving, a case we're trying to dispute. In this case, we're not even in the room where the elephant you're talking about is located. We're located in the messy room of Peter Parker where CPU is completely out of the picture.
If DF did benchmark that room with a 2080ti and 12900k, some people like them would associate the performance with 12900k. If you can lend me that 2080ti and I get the same performance with 2080ti just the way like 12900k does, IN THAT SPECIFIC ROOM scene, then some of them would at least hopefully believe it to be the case. I only have a 3070 and 2700x, and get 57-61 frames that scene with high textures. Then some of them will still think that very high textures are more demanding on CPU. This can go on forever. Only way to properly rebuke their claims is for a 2700x+2080ti user to share that 62-63 FPS result in that room and scene, and finally we can have closure. But problematic part is that most high end GPU users already have high end CPUs. That specific scene renders upwards of 70 frames with my 2700x, when I have enough VRAM.
I don't know if you would call the traversal I'm showing in game is rapid enough or not, but this is the pace I usually play.
Also there's nothing wrong with the game regarding GPUs. Its just 8 GB GPUs buckling under the weight of 4K/Ray Tracing with added artificial limitation on VRAM.
I'm not having this discussion over and over again. I provide solid and sound hard facts with evidences. There's not much to talk about.
If my frames go up to 60 from 38 when I use High textures at 4K, then it means it is VRAM bottlenec;k
Some people evenwent so far to claim textures themselves have a rendering cost, which I had to prove it not being the case at 1440p;
Game performs like it should when you have enough VRAM. Even then, you might be talking about a situational thing. What we talk about here is general game performance. 2070 dropping to 25s in the very same cutscene PS5 renders 42-43 frames are being misused by specific users and reviewers to claim that PS5 is overachiving, a case we're trying to dispute. In this case, we're not even in the room where the elephant you're talking about is located. We're located in the messy room of Peter Parker where CPU is completely out of the picture.
If DF did benchmark that room with a 2080ti and 12900k, some people like them would associate the performance with 12900k. If you can lend me that 2080ti and I get the same performance with 2080ti just the way like 12900k does, IN THAT SPECIFIC ROOM scene, then some of them would at least hopefully believe it to be the case. I only have a 3070 and 2700x, and get 57-61 frames that scene with high textures. Then some of them will still think that very high textures are more demanding on CPU. This can go on forever. Only way to properly rebuke their claims is for a 2700x+2080ti user to share that 62-63 FPS result in that room and scene, and finally we can have closure. But problematic part is that most high end GPU users already have high end CPUs. That specific scene renders upwards of 70 frames with my 2700x, when I have enough VRAM.
Last edited:
While there is some merit to your argument that the PS5 may also be CPU bottlenecked to an extent, your assumption that it could be to a similar degree to NXG's PC with its 2700x is way off the mark. On a console, the software is always adapted to the hardware. So where the developers see a CPU bottleneck in play, they will reduce CPU impacting settings to the point where that's no longer the case, same on the GPU side so that both the CPU and GPU are as fully utilised at any given time as they can be. Obviously it's not possible to do that with perfect success on a frame by frame basis but for the most part the CPU won't be a major bottleneck, because settings can be tailored to avoid that. That's why you see such specific settings for things like RT object distance, crowd density and vehicle density on the PS5 which are so tuned to the consoles specific capabilities there is no direct equivalent to them on the PC despite in the formers case there being 10 separate selectable distance levels.
Ideally to match the PC as closely as possible to the PS5 you would test a number of CPU's alongside a very powerful GPU (to ensure no GPU bottleneck) until you find a CPU that provides a very similar performance level to the PS5 at the same settings. You could then pair that that CPU up with a number of GPU's until you find a GPU that also matches the PS5's performance. At that stage you will have a full system that is comparable to the PS5. On the GPU side that system is likely to contain something around 2070-2080 level in this game. But on the CPU side you can be certain it won't be a 2700x.
Since we don't have detailed CPU benchmarks to know the exact best match for the PS5's CPU, the next best option is to use a CPU that is definitely more powerful so as to remove all possibility of a CPU bottleneck - if you are trying to do a GPU to GPU comparison.
And that's the crux of it. NXG is doing direct head to head GPU performance comparisons in his video while it is clearly more CPU limited by a large degree on the PC side. And that's simply wrong.
No, the 2700x is a Zen+ CPU. the PS5 CPU is Zen2 based which is much faster. This game in particular seems to be highly dependant on inter CCX latency on the PC side (probably because it's designed to leverage the PS5 CPU's strengths which does not use separate CCX's given it's mobile heritage - one of it's few advantages over Zen2 on the PC). And the inter CCX latency on Zen+ is much worse than Zen2 on the PC (which is in turn worse than the PS5). NXG comments on this himself in the video. The game is also highly CPU memory bandwidth dependent and he's using quite sow memory with his 2700x while the PS5 obviously has access to the much higher bandwidth of the GDDR5
This is a silly statement and shows a complete lack of knowledge on this subject on your part. Pairing GPU's with a very high end CPU when doing GPU testing in order to remove any CPU bottleneck its decades old best practice and the only sensible way of testig GPU performance.
And he was quite wrong, as has been shown several times now in this thread by posters who have 3070's or similar. The 3070 is in fact, quite a bit faster than the PS5 in this game (and every other game where head to head testing has been done).
This applies on the CPU side where the PC CPU is having to deal with decompression and thicker API's but not so much on the GPU side where comparable GPU's generally perform in line with the console equivalent provided they aren't otherwise bottlenecked by the likes of CPU or VRAM.
davis.anthony
Veteran
I'm on a 3060ti with a Zen 2 based 3600 and I'm not even close to thinking my CPU is a problem in any game.
*When running ray tracing settings that are below what PS5 offers in its performance mode.
Last edited by a moderator:
Metal Snake Eater
Newcomer
Thank you
This is enough Data
i know have 3 GPUs in similar scenario
First here is my RTX3080 in Two Scenarios vs RTX3070TI
68fps - 81fps = 20%
64fps - 79fps = 25%
this is 25% which is INLINE to what Hardware Unboxed Benchmarked between the Two GPU's in their test run
while we factor in the RTX2080TI in the same scene the RTX3070TI is about 6% ahead of the RTX2080TI
and if we compare it to HUB benchmark guess what? it is exactly 6%! so we are good in business !
the RTX3080 in that Scene is 86% ahead of the PS5 of where Peter shows up and 88% in the other Scene
The RTX3070TI is 54% ahead of the PS5 in Peter Scene and 50% in the other
the RTX2080TI now been Shared in two Scene "even though the second scene is different" showing 41-45% ahead of the PS5
so where does that leave the PS5 actual performance? since we got 3 GPUs lining up nicely with HUB findings when you add up how the PS5 performs?
that puts the PS5 Ballpark of around the RTX3060 .. which is also Inline with Digital Foundry
Just RT object range being at 6 instead of 8 is not going to change a lot in terms of performance. I can re-capture the same video with RT object range set at 8 if you accept that you're going to retract what you've said. What a narrow angle to attack. I said it is BEST to use Zen 3. I never said it was necessary.
I also suggest avoiding the use of vulgar words and tones. It is not appreciated and should not be accepted.
davis.anthony
Veteran
"A person that spent 500+ bucks for a 3060ti+ tier GPU should also do what is necessary and upgrade their CPU"
Where did you say it's "Best to use Zen 3" in that sentence?
And object range at 6 or not, it's below PS5.
Put it on 8 and go and swing around Time Square.
And I'm not retracting anything.
"If you get an Ampere GPU it is best that you use Zen 3 or Intel 11th gen as a minimum. "
If you're not going to retract this asteriks; "*When running ray tracing settings that are below what PS5 offers in its performance mode." this line when I do what you want, then I simply do not feel the need of proving anything to you.
"Put it on 8 and go and swing around Time Square."
And now it become the Time Square. Nice job, moving goalposts. Irrevelant, since I still get 60 FPS locked there. But since you say you're not going to retract, I'm not going to bother with you any further. You're also free to provide PS5 RT Performance footage taken in Time's Square. I base my footage on the footage available on PS5.
If you're not going to retract this asteriks; "*When running ray tracing settings that are below what PS5 offers in its performance mode." this line when I do what you want, then I simply do not feel the need of proving anything to you.
"Put it on 8 and go and swing around Time Square."
And now it become the Time Square. Nice job, moving goalposts. Irrevelant, since I still get 60 FPS locked there. But since you say you're not going to retract, I'm not going to bother with you any further. You're also free to provide PS5 RT Performance footage taken in Time's Square. I base my footage on the footage available on PS5.
Last edited:
davis.anthony
Veteran
You failed to address the criticism here, just went on a tangent
I can get 60fps on a 10 year old Intel CPU at your settings while swinging around some of the easiest parts of the map like you are. It's noting special.
Show a demanding section like Time Square and then brag about your frame rate and 'optimised' settings.
But you can't get a locked 60fps when just swinging around some easy locations but then claim you can get a locked 60fps around Time Square, one if the hardest areas of the game in terms of CPU?
I'm calling that out as BS.
EDIT: I've just swung around Time Square with your ray tracing settings (Textures on High) at native 1080p and even I get dips below 60fps on a CPU that is considerably faster than yours in Spiderman.
Last edited:
No, If I get 60 FPS there with PS5 equivalent RT settings, will you retract the thing you've said in the asteriks?
If you openly claim you will do so, I will do a video.
"When running ray tracing settings that are below what PS5 offers in its performance mod"
As I said, we never seen a PS5 footage running there. But let's assume it gets a locked 60. I openly ask you, if I present you a 2700x footage where it gets 60+ framerate in Times Square and never drops below 50, and very "rarely" drops below 56, will you retract the asteriks?
Marvel's Spider-Man PC - a patch or two away from perfection
Digital Foundry's in-depth tech breakdown - along with best settings - for this brilliant PC title.
www.eurogamer.net
I will use these settings as a baseline. Low crowd density (PS5 is more closer to low than medium), Medium traffic density, High RT/RT and 8 object range. All other settings set to parity as well.
I already have one such video, but unless you openly say that you will retract the things you said about me and me "bullshitting" and I only get that performance when I set my settings below PS5, I won't share it. And yes, I did get locked 60 there. But you have to openly say that you will retract it. I will also use a stock 2700x, not underclocked or overclocked (in the video it is underclocked to 3.7 GHz)
"but then claim you can get a locked 60fps around Time Square" I never claimed so. It is so.
davis.anthony
Veteran
I've already tested for my myself 5 minutes ago, PS5 settings at native 1080p (but high textures) I get dips below 60fps on a CPU that's considerably faster than yours in Spiderman.
So if I can't get there with a 4.2Ghz 3600 you're definitely not getting there with a 3.7Ghz 2700x.
No I won't retract, but funny how your tune has changed from what you claimed below.
You clearly state it is a locked 60fps you get in Time Square and that's what I want to see.
Last edited:
Can you try with a higher crowd density? I set mine to High.
You're once again, moving goalposts. You actually manipulate what I've originally said. I NEVER mentioned anything about matching a PS5 in this case. Let me rewind for you;
The original post:
"I absolutely do run into CPU bottlenecks even at 4K, especially during rapid traversal where I see my GPU usage tumble down to the 60% and fps to the high 40s and low 50s at High settings with RT while my CPU is maxed out."
My reply:
"Sorry but my 2700x never drops to low 50s, let alone high 40s with RT enabled at 3.7 GHz (so an improvement could be had with 4.1 GHz)"
Your claim:
"*When running ray tracing settings that are below what PS5 offers in its performance mode."
I never claimed I was getting rock solid 60 all the time. I initially said that mine config never drop to low 50s let alone 40s. And it does not. It rarely drops to 56, NOT Below 56. Low 50s means a framerate target around 50-54, and you know this better than I do.
I just said to that user that my low end CPU was not getting framerate drips below 50, let alone 40s. YOU WERE THE ONE to claim that I was getting that performance with "lower than PS5". You're the one who shifted to the topic to PS5. The topic was never about matching PS5. I just answered to you that even with PS5 equivalent settings, I'd still not get drops below 50 and into the 40s. I also said "rarely", as in being very rare. I literally had to do instantenous erratic turns to force it below 60. You're free to accept that you will retract once I publish the video. Until then, have a good day.
Locked or almost locked, if it rarely drops, then it is mostly a locked 60 FPS. You're just playing petty semantics at this point. But okay. Let's say locked. If I share you a locked 60 FPS vid with the settings I shared above with stock 2700x, will you retract?
davis.anthony
Veteran
Please evidence your claim in the attached picture either at your settings or at PS5's settings.
Attachments
davis.anthony
Veteran
If I use PS5's settings but reflection resolution set to very high I get a locked 60fps with no issues.
If I set everything to max but keep LOD at 6 I can get a locked 60fps 85-90% of the time.
First you have to accept that you will retract the asteriks. Or else, no video.
Similar threads
- Locked