Current Generation Games Analysis Technical Discussion [2022] [XBSX|S, PS5, PC]

PSman1700

Legend
True but you are comparing your result from Core i9 10900K with RTX 2070 Super to NXGamer Ryzen 2700X with 2070 (overclocked). You are not going to get the same results. A Core i9 10900K is almost twice the performance of Ryzen 7 2700X in games.

A desktop ryzen 7 2700X clocked at 3.8ghz will perform close to mobile ryzen 7 4800H give or take 2 to 4 percentage points in favor of 4700H but the consoles are clocked lower.

The point he was trying to make was that you shouldnt compare PS5 vs PC if that pc doesnt match the CPU prowess of said console. CPU's do make a difference.
Maybe test with a Zen2 product and 2070/S or 5700XT/RX6700.
 

Shortbread

Island Hopper
Legend
PC and console game comparisons are always going to be tricky, but still meaningful overall. I think both Dictator's and NXGamer's results are both good examples of how different PC configurations can net different results (which shouldn't be no surprise to anyone, or given into any temptations about oddball conspiracies on the reasons why *cough* Gaf *cough*).
 
Last edited:

function

None functional
Legend
Not sure how his 2070 (pseudo 2070s) results here are so much lower than what I measure on a real 2070 Super at 1512p original settings.

Yeah, that Zen+ / 2070 system of his is a bit of a mess. :nope:

My guess is that the heart of the problem is Nvidia's high driver overhead coupled with a relatively slow CPU particularly unsuited to handling that. But it's almost like there's more to it than that, because his PC consistently performs badly.

There's nothing wrong with only having the hardware you've got, and I appreciate he has limited resources. In the past though he's used his system to draw some rather dubious conclusions about the PC, PC vs console, and even Xbox vs PS5.

It is interesting to see just how differently a particular setup can represent the particular components that are in it though.

 

PSman1700

Legend
My guess is that the heart of the problem is Nvidia's high driver overhead coupled with a relatively slow CPU particularly unsuited to handling that. But it's almost like there's more to it than that, because his PC consistently performs badly.

There's nothing wrong with only having the hardware you've got, and I appreciate he has limited resources. In the past though he's used his system to draw some rather dubious conclusions about the PC, PC vs console, and even Xbox vs PS5.

I say it again, theres a reason NXG pc vs console comparisons (or even Xbox vs PS5) are best kept at a distance. They always stir up discussions like these above.
 

Globalisateur

Globby
Veteran
Supporter
+ using vsync/60fps cap as on console
Yes as the game is virtually locked on PS5 it could run at 80fps average for all we know. The only way to compare would be to find a spot on PS5, ideally GPU limited, that runs under 60fps and to compare on PC on the same spot. This is what NXGamer did on Death stranding and showed the PS5 game was performing very well compared to PC.
 

pjbliverpool

B3D Scallywag
Legend
13:50 rtx2070oc ps4 settings in 1512p can drop to 42fps where ps5 in same scane 60 on 4k cb, nice

I'm not sure I agree with the CPU being the reason for NXG's lower performance, particularly in this scene. Here's what NXG says elsewhere in the article:
  • At 11:10 NXG says the game isn't CPU bound at 60fps on his 2700X (at DLSS UP and Low settings)
  • At 11:54 he says a locked 60fps is possible on the same CPU and goes on to demonstrate that at Original settings @ 4K DLSS Q (love how he implies this is just "close enough" to the PS5's CB 4K in visual quality)
  • At 12:55 and 13:16 he states that the performance loss on the 2070 is regained (back to 60fps) simply by turning on FSR Q at 1512p
There could be a number of other reasons for it. e.g. he made a mistake in the settings, perhaps the settings didn't properly apply despite him setting them, some kind of background process running, incorrect CP settings, general windows craziness...

The only way to compare would be to find a spot on PS5, ideally GPU limited, that runs under 60fps and to compare on PC on the same spot. This is what NXGamer did on Death stranding and showed the PS5 game was performing very well compared to PC.

One has to wonder a little at the reliability of NXG's previous PC vs PS5 performance results in light of the above. In any case though, Death Stranding isn't a great comparison as the PS5 is running the Directors Cut vs the PC original which has received engine modifications which may result in higher performance.
 

snc

Veteran
I'm not sure I agree with the CPU being the reason for NXG's lower performance, particularly in this scene. Here's what NXG says elsewhere in the article:
  • At 11:10 NXG says the game isn't CPU bound at 60fps on his 2700X (at DLSS UP and Low settings)
  • At 11:54 he says a locked 60fps is possible on the same CPU and goes on to demonstrate that at Original settings @ 4K DLSS Q (love how he implies this is just "close enough" to the PS5's CB 4K in visual quality)
  • At 12:55 and 13:16 he states that the performance loss on the 2070 is regained (back to 60fps) simply by turning on FSR Q at 1512p
There could be a number of other reasons for it. e.g. he made a mistake in the settings, perhaps the settings didn't properly apply despite him setting them, some kind of background process running, incorrect CP settings, general windows craziness...
possible but still 2700x with 60fps frame cap is closer to console environement than 10900k with uncapped frames
 

iroboto

Daft Funk
Legend
Supporter
possible but still 2700x with 60fps frame cap is closer to console environement than 10900k with uncapped frames
the API differentials need to be considered here. GNM is smooth and code is optimized as butter compared to DX11; DX12 on PC still has issues as well, far from reliable performance. If you want to look at GPU performance, you need a CPU to push the bottleneck to the GPU.

Even though the machines aren't comparable, if the goal is to showcase PC performance, one must do so by removing potential CPU bottlenecks caused by poor API/driver performance.

I'm not blaming NXGamer here on it, hardware configurations are expensive, and very few sites have the funds and setups to run multiple configurations and benchmark PC GPUs properly. It's a nuance of benchmarking performance on PC, just like capped frame rates/floored resolutions are a nuance of measuring performance on console.
 

Remij

Veteran
the API differentials need to be considered here. GNM is smooth and code is optimized as butter compared to DX11; DX12 on PC still has issues as well, far from reliable performance. If you want to look at GPU performance, you need a CPU to push the bottleneck to the GPU.

Even though the machines aren't comparable, if the goal is to showcase PC performance, one must do so by removing potential CPU bottlenecks caused by poor API/driver performance.

I'm not blaming NXGamer here on it, hardware configurations are expensive, and very few sites have the funds and setups to run multiple configurations and benchmark PC GPUs properly. It's a nuance of benchmarking performance on PC, just like capped frame rates are a nuance of measuring performance on console.
I'm sure IGN could pony up to produce a video with reliable results.
 

snc

Veteran
the API differentials need to be considered here. GNM is smooth and code is optimized as butter compared to DX11; DX12 on PC still has issues as well, far from reliable performance. If you want to look at GPU performance, you need a CPU to push the bottleneck to the GPU.

Even though the machines aren't comparable, if the goal is to showcase PC performance, one must do so by removing potential CPU bottlenecks caused by poor API/driver performance.

I'm not blaming NXGamer here on it, hardware configurations are expensive, and very few sites have the funds and setups to run multiple configurations and benchmark PC GPUs properly. It's a nuance of benchmarking performance on PC, just like capped frame rates/floored resolutions are a nuance of measuring performance on console.
agree that on pc usually you need faster cpu too bruteforce api difference
 

PSman1700

Legend
I'm not blaming NXGamer here on it, hardware configurations are expensive, and very few sites have the funds and setups to run multiple configurations and benchmark PC GPUs properly.

If he cant afford to make valuable comparisons, he shouldnt make them. if he does, he can atleast make a disclaimer about what hardware he's using, to avoid nasty backfires like these.
 

PSman1700

Legend
both setups are literaly written on this video ;d

Aight, point taken and corrected there then. Up to each and one what to make of the PC vs PS5 comparison if the 2070/S can match what the PS5 is doing in this game. I'd say it easily does, aslong your not bottlenecking things with slower components, ofcourse.

Edit:

And to note, another good port from Sony studios. This is how you do PC ports done natively for one console (Playstation). Its something that has to be mentioned as its a problem for us pc gamers when its coming down to ports (cross platform isnt a problem, exclusives/ports are).
 
  • Like
Reactions: snc

Globalisateur

Globby
Veteran
Supporter
NXgamer doesn't do comparisons in order to show how such a platform (PC) is so much better than this console for the sake of showing how superior a $2000 PC is compared to a $399 console like if it was the only purpose of the comparison. How many people are owning that kind of PC?

He just compares a particular modest PC, his PC, against the PS5. But for most PC owners that comparison will be more pertinent because that's the kind of PC many own.
 

PSman1700

Legend
You dont need anything above a 2018 RTX 2070/S or a 5700XT/RX6700 regarding GPU's to play this at ballpark PS5 level for this game. Theres not many PS5 DE (399), if any available. Its usually the 499 dollar version if you can fetch one at mrsp.
You wouldnt need a 2000 dollar PC for this game at all, to match the PS5's performance in this one.
 

dobwal

Legend
the API differentials need to be considered here. GNM is smooth and code is optimized as butter compared to DX11; DX12 on PC still has issues as well, far from reliable performance. If you want to look at GPU performance, you need a CPU to push the bottleneck to the GPU.

Even though the machines aren't comparable, if the goal is to showcase PC performance, one must do so by removing potential CPU bottlenecks caused by poor API/driver performance.

I'm not blaming NXGamer here on it, hardware configurations are expensive, and very few sites have the funds and setups to run multiple configurations and benchmark PC GPUs properly. It's a nuance of benchmarking performance on PC, just like capped frame rates/floored resolutions are a nuance of measuring performance on console.

In the context of PC DX11/12 vs. GNM, the use "smooth" and "butter" against "poor" and "far from reliable" is harsh.

DX11/12 is designed to allow devs to avoid explicitly coding for specific hardware configurations. GNM isn't encumbered with such a monumental feature as it doesn't support multiple hardware archs in 1000s of configurations. There should be no expectation for DX11/12 to perform as well as an api with such a narrow focus.

The performance of a PS5 game on PS5 hardware should pretty much always trump a PS5 port on PC hardware similarly configured to a PS5. We don't really need videos to know this.
 
Last edited:

iroboto

Daft Funk
Legend
Supporter
In the context of PC DX11/12 vs. GNM, the use "smooth" and "butter" against "poor" and "far from reliable" is harsh.

DX11/12 is designed to allow devs to avoid explicitly coding for specific hardware configurations. GNM isn't encumbered with such a monumental feature as it doesn't support multiple hardware archs in 1000s of configurations. There should be no expectation for DX11/12 to perform as well as an api with such a narrow focus.

The performance of a PS5 game on PS5 hardware should pretty much always trump a PS5 port on PC hardware similarly configured to a PS5. We don't really need videos to know this.
Yup I agree. It’s an unfair assessment by me; but it’s one of the things we come to accept when it comes to PC and the variety of setups that are available.

I think a lot of folks just underestimate how much that can be on the CPU so like for like comparisons aren’t valid. A good example is that Halo Infinite uses GPU based dispatch for its draw calls, the CPU isn’t largely involved. I can assure that isn’t the case on PC where the CPU is doing all the work.

so I just felt that not using the best possible CPU isn’t doing anyone favours when it comes to benchmarking.
 

HolySmoke

Newcomer
On a different note, I just got an extra drive in my PS5 and did a bit of testing via the drive's SMART data and CrystalDiskInfo. I haven't seen anyone do this and figured it would be interesting to see how much data is flowing through since it's impossible to check on the console.

Here is where I started out.

I first tried R&C and loading into the main menu reads ~2GB from the disk. I then booted back in and started a new game and total amount of reads by the first checkpoint was ~18GB including the main menu and intro sequence. I then reloaded the checkpoint (bypassing the intro) and the total reads from that added another ~5GB. And since we already know that booting to the menu reads ~2GB then loading the checkpoint would have added ~3GB and with somewhere in the region of 13GB being streamed in during the intro.

I also did a bit of testing with Demon's Souls but not as granular. The long and short of it is that spawning into three different worlds read ~15GB from the drive. I then created a savegame at the start of World 1-1 and played until the Phalanx boss (which took me 30 minutes) and the drive had added another 100GB of reads to the counter. My reasoning was that this should give us an idea about the game's streaming rate.

I would test R&C some more but I haven't found time to start playing it yet. But I hope someone finds this interesting.
 
Top