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

[dobwal]

God I love this guy. Brings this forum to life.

I don't. He is like the technical video game analysis' version of The Onion.

 

[iroboto]

I don't. He is like the technical video game analysis' version of The Onion.

unintentionally

 

[PSman1700]

NXG stirring up the discussion again. This clown may only live in this small corner of the world.

 

[Flappy Pannus]

On top of that he's also saying that it's basically impossible to achieve 60fps on the PC

On his 2700x, mind you:

In that case, at least with the code that others have benchmarked with so far, it certainly looks to be the case. Zen2 CPU's atm are getting their ass kicked with this game. Computerbase tested it at 720p to find CPU bottlenecks, and a 2600X was getting 68 fps avg, but only 47fps with 1% lows - and that's without RT. With RT, not surprisingly it's far worse - just 48fps avg and 35fps1% lows.

So with RT, a 2700X, regardless of res, will likely not match a PS5 in RT performance mode which is pretty much 60fps locked, and could even struggle without RT to maintain 60. By comparison in computerbase's article, a $100 4-core, 8 thread 12100f gets 75fps avg and 54fps 1% lows with RT, and 113/85 respectively without RT.

So Zen2 just blows for it right now, or at least with the code NxG/Computerbase have tested with. To be as fair as possible however, even though it would be nice to see NxG finally retire that crusty 2700X (especially as mentioned $100 CPU's will likely beat it in most games), I also don't think it's unreasonable to expect better performance from a 2700X considering the PS5 is also basically a Zen2 arch - at least something far closer to a locked 60 without RT for starters.

Again though, patches are dropping seemingly by the hour and even though he benched it with 1.8 I have no idea where that is in terms of the latest releases, so we'll see.

whiel traversing due to "data transfer limitations" (its a PS4 game).

And yes, this is where we get classic NxG - extrapolating something that sounds all technical-like based on...? He does indeed say "..and largely because it can transfer data much better" which is why the PS5 can maintain 60fps solidly in RT mode. Like...what? He infers this is the cause of those 100ms spikes on the PC, but notes they are much rarer without RT - does he believe RT mode is streaming massively more amounts of data then? Perhaps he means CPU bottlenecks, but saying it's due to "transferring data much better" is a very odd way to put it.

Regardless, we've seen that the higher-end GPU's are not getting these spikes, so it's not even necessarily a CPU bottleneck (outside of Zen2), let along a 'data transfer' issue, whatever the heck that is. I mean when a quad-core can average 100+fps in a CPU bound situation, I don't think the lack of hardware texture decompression blocks on PC CPU/GPU's are the problem with this game right now.

 

[davis.anthony]

On top of that he's also saying that it's basically impossible to achieve 60fps on the PC while traversing due to "data transfer limitations" (its a PS4 game).

Other review sites have also shown that...... a locked 60fps with RT on is pretty much not possible on 99% of CPU's.

 

[pjbliverpool]

Doesn't he make the inability to hold 60fps claim for the 5950x too?

I think we need to be clear on what version of the code these tests are being run on as there are massive differences between patch versions according to DF. They should at least be using the day 1 patch which I'm certainly not convinced NXG is doing (he pretty much says he isn't at the start).

 

[Jay]

With the rate that patches were dropping for this, I wouldn't consider any a day one patch until game is unlocked and available to consumers.
So proper performance review can't happen until now.

 

[function]

On his 2700x, mind you:

In that case, at least with the code that others have benchmarked with so far, it certainly looks to be the case. Zen2 CPU's atm are getting their ass kicked with this game. Computerbase tested it at 720p to find CPU bottlenecks, and a 2600X was getting 68 fps avg, but only 47fps with 1% lows - and that's without RT. With RT, not surprisingly it's far worse - just 48fps avg and 35fps1% lows.

So with RT, a 2700X, regardless of res, will likely not match a PS5 in RT performance mode which is pretty much 60fps locked, and could even struggle without RT to maintain 60. By comparison in computerbase's article, a $100 4-core, 8 thread 12100f gets 75fps avg and 54fps 1% lows with RT, and 113/85 respectively without RT.

So Zen2 just blows for it right now, or at least with the code NxG/Computerbase have tested with. To be as fair as possible however, even though it would be nice to see NxG finally retire that crusty 2700X (especially as mentioned $100 CPU's will likely beat it in most games), I also don't think it's unreasonable to expect better performance from a 2700X considering the PS5 is also basically a Zen2 arch - at least something far closer to a locked 60 without RT for starters.

Again though, patches are dropping seemingly by the hour and even though he benched it with 1.8 I have no idea where that is in terms of the latest releases, so we'll see.

The 2700X isn't even Zen 2, it's Zen 1.x. So lower IPC, higher inter CCX latency, worse thread grouping for games (iirc) and stuff like that. Zen 2 arrived somewhat confusingly with the Ryzen 3xxx products, and it's a ton better as the Computerbase graphs show.

It's a bad choice of CPU to try and judge relative GPU performance with (or anything else), but he keeps doing it.

And yes, this is where we get classic NxG - extrapolating something that sounds all technical-like based on...? He does indeed say "..and largely because it can transfer data much better" which is why the PS5 can maintain 60fps solidly in RT mode. Like...what? He infers this is the cause of those 100ms spikes on the PC, but notes they are much rarer without RT - does he believe RT mode is streaming massively more amounts of data then? Perhaps he means CPU bottlenecks, but saying it's due to "transferring data much better" is a very odd way to put it.

Regardless, we've seen that the higher-end GPU's are not getting these spikes, so it's not even necessarily a CPU bottleneck (outside of Zen2), let along a 'data transfer' issue, whatever the heck that is. I mean when a quad-core can average 100+fps in a CPU bound situation, I don't think the lack of hardware texture decompression blocks on PC CPU/GPU's are the problem with this game right now.

Only PS5 can stream the rays off the SSD fast enough.

I think it's far more likely to be something processing related, like maintaining the RT acceleration structure as you move (removing and adding things). That might have a lot of CPU and/or GPU overhead, which increases the faster you move.

 

[snc]

The 2700X isn't even Zen 2, it's Zen 1.x. So lower IPC, higher inter CCX latency, worse thread grouping for games (iirc) and stuff like that. Zen 2 arrived somewhat confusingly with the Ryzen 3xxx products, and it's a ton better as the Computerbase graphs show.

It's a bad choice of CPU to try and judge relative GPU performance with (or anything else), but he keeps doing it.

We have ps5 cpu benchmarks and 2700x is faster.

 

[BRiT]

From the DF review, a bit of a shame that you can't have identical settings:

Optimised settings are on the page in an easy-to-digest table, but ultimately, we're looking at a variation of PlayStation 5's performance RT mode. It's not possible to match settings precisely, however, and I do wish that console equivalent presets were available to tweak from. Let's put it this way: the crowd LOD setting in the performance RT mode does not correspond to any of PC's offerings. Clearly, Insomniac made its selections carefully with a view to balancing cost vs fidelity, so why not offer PC users that setting?

 

[function]

We have ps5 cpu benchmarks and 2700x is faster.

That doesn't change the fact that a 2700x will gimp the PC it is powering, and make any comparison of e.g. PC GPUs Vs PS5 GPU flawed.

And yes there are different overheads on pc Vs console - we've always known this. What might not particularly hold back a console might gimp a PC. Look at last gen's Jaguar cores as an extreme example.

BTW, a 2700X is not faster across the board than the PS5 CPU. PS5 has weak FPUs, and loses out marginally there to the higher clocked 2700X (which also has weak FPUs). At other operations the PS5 CPU can pull ahead slightly despite the clock speed deficit and smaller cache. But this on its own is far from the full story, because when inside a PC vs a PS5 they are required to do different things by the various layers of software, and the impact of that can dwarf any small difference in performance.

Proportional to what's being asked of the processors, a 2700X is really not on the same level as the PS5 CPU. And that can be interesting to look at in it's own right, but it's also important bear that in mind when trying to draw conclusions about other parts of the system.

 

[snc]

That doesn't change the fact that a 2700x will gimp the PC it is powering, and make any comparison of e.g. PC GPUs Vs PS5 GPU flawed.

And yes there are different overheads on pc Vs console - we've always known this. What might not particularly hold back a console might gimp a PC. Look at last gen's Jaguar cores as an extreme example.

BTW, a 2700X is not faster across the board than the PS5 CPU. PS5 has weak FPUs, and loses out marginally there to the higher clocked 2700X (which also has weak FPUs). At other operations the PS5 CPU can pull ahead slightly despite the clock speed deficit and smaller cache. But this on its own is far from the full story, because when inside a PC vs a PS5 they are required to do different things by the various layers of software, and the impact of that can dwarf any small difference in performance.

Proportional to what's being asked of the processors, a 2700X is really not on the same level as the PS5 CPU. And that can be interesting to look at in it's own right, but it's also important bear that in mind when trying to draw conclusions about other parts of the system.

I mean we literaly have benchmarks and 2700x is winning by quite a margin

AMD 4700S Review: Defective PlayStation 5 Chips Resurrected

Break out the broken chips

www.tomshardware.com

 

[davis.anthony]

I mean we literaly have benchmarks and 2700x is winning by quite a margin

AMD 4700S Review: Defective PlayStation 5 Chips Resurrected

Break out the broken chips

www.tomshardware.com

Is that chip running PS5's OS and API in those benchmarks?

If it's not then it's a pointless comparison.

 

[snc]

Is that chip running PS5's OS and API in those benchmarks?

If it's not then it's a pointless comparison.

same hw, of course api and os is different, very interesting comparison for me, if for you pointless then dont read

 

[dobwal]

And yes, this is where we get classic NxG - extrapolating something that sounds all technical-like based on...? He does indeed say "..and largely because it can transfer data much better" which is why the PS5 can maintain 60fps solidly in RT mode. Like...what? He infers this is the cause of those 100ms spikes on the PC, but notes they are much rarer without RT - does he believe RT mode is streaming massively more amounts of data then? Perhaps he means CPU bottlenecks, but saying it's due to "transferring data much better" is a very odd way to put it.

Regardless, we've seen that the higher-end GPU's are not getting these spikes, so it's not even necessarily a CPU bottleneck (outside of Zen2), let along a 'data transfer' issue, whatever the heck that is. I mean when a quad-core can average 100+fps in a CPU bound situation, I don't think the lack of hardware texture decompression blocks on PC CPU/GPU's are the problem with this game right now.

Eurogamer's article with Nixxes explanation of the stuttering

"Running with frame-rate unlocked can also cause stutter in fast traversal: thankfully, today's update significantly improves consistency here - but it still needs work. Nixxes tells us that the Insomniac engine heavily utilises occlusion to reduce draw calls, which improves performance. However, when the camera is moving, the renderer lacks occlusion information for details that are suddenly in view, causing a spike in draw calls and a subsequent stuttering effect. Curiously, this situation tends to improve the better your GPU, but right now, limiting performance and not running flat out during city traversal makes the ride smoother overall."

 

[davis.anthony]

same hw, of course api and os is different, very interesting comparison for me, if for you pointless then dont read

When you're making ridiculous claims like this:

"We have ps5 cpu benchmarks and 2700x is faster"

And then produce an irrelevant article by Toms to some how prove 2700x >> PS5's CPU I will always call it out.

 

[snc]

When you're making ridiculous claims like this:

"We have ps5 cpu benchmarks and 2700x is faster"

And then produce an irrelevant article by Toms to some how prove 2700x >> PS5's CPU I will always call it out.

It terms of hw is faster, simple as that. I understand ps5 has low level api and system advantage (and even io system decrease cpu usage) but thats different subject. Btw quite hard to discuss with you, you have like this angry approach when you misunderstand somebody.

 

[davis.anthony]

it terms of hw is faster, simple as that

That's what you claimed though is it?

You claimed there are benchmarks that PROVE that a 2700x is faster than PS5's CPU.

So you share these benchmarks comparing what is essentially PS5's CPU running Windows and software that's not optimized to work around it's limits and get 1+1=3.

 

[snc]

That's what you claimed though is it?

You claimed there are benchmarks that PROVE that a 2700x is faster than PS5's CPU.

So you share these benchmarks comparing what is essentially PS5's CPU running Windows and software that's not optimized to work around it's limits and get 1+1=3.

Read my full post above.

 

[function]

I mean we literaly have benchmarks and 2700x is winning by quite a margin

AMD 4700S Review: Defective PlayStation 5 Chips Resurrected

Break out the broken chips

www.tomshardware.com

Dude, those tests are for a super fast discrete GPU connected connect over the 4700S PCI-e bus.

The 4700S desktop kit has a slow as fuck PCI-e 2.0 x 4 connection. Those tests are showing the massive bottleneck of a super, super, super slow PCI-e connection.

You can not infer anything about CPU performance from these graphs.

AMD 4700S + RTX 3090 (PCIe 2.0 x4): 1.33 GB/s <------ This is what you are looking at

Ryzen 7 3700X + Radeon 500 (PCIe 4.0): 25.5 GB/s <---------- This is what a normal PC has

This is a PCI-e bottleneck test.

"Here we can see that the 4700S' maximum attainable bandwidth weighs in at 1.33 GB/s with the GTX 3090, but dropping the same GPU into an X570 motherboard results in ~10X more throughput (13.43 GB/s) for a PCIe 3.0 connection and 19X more with PCIe 4.0 (25.5 GB/s)"

same hw, of course api and os is different, very interesting comparison for me, if for you pointless then dont read

It's not the same hardware being used, the PS5 is not connected to its GPU by a tiny, prehistoric, super slow PCI-e 2.0 (lmao) x4 (double lmao) bus. It's connected internally over something very fast, probably in the tens of GB/s.

You can infer absolutely nothing meaningful about the PS5 from this. This is not a configuration that would ever be used on a PS5.