Digital Foundry Article Technical Discussion [2023]

[SmooTh]

 

[Phantom88]

I love how Rich presents these tests. So the fastest desktop cpus we have can be more than 3 times faster than what consoles have and a ryzen 3600 is almost universally faster, by a gigantic margin.

 

[davis.anthony]

I love how Rich presents these tests. So the fastest desktop cpus we have can be more than 3 times faster than what consoles have and a ryzen 3600 is almost universally faster, by a gigantic margin.

You cannot conclude that from his testing or video.

All his video shows is how they perform when running Windows, not how they perform when running a console level OS along with a level of CPU optimisation desktop CPU's can only dream of.

 

[I'mDudditz!]

John is not talking about direct Storage here - here He is talking about why it runs on a hard drive at all on high end systems (aka why lot's of RAM and CPU make it somewhat work).

Yes and he is also discussing the huge impact of the decompression hardware, which is what I've been sayin for some time now. It shouldn't be an either or discussion between amount of RAM and decompression hardware. When thinking about the future of PC you should want the hardware to be maximum efficient and smart.

 

[I'mDudditz!]

It's as if DF is on a mission to eliminate false narratives this weekend.

An accurate GPU comparison where a clear CPU bottleneck can be seen?

His GPU load is fluctuating, it's CPU limited

 

[I'mDudditz!]

I love how Rich presents these tests. So the fastest desktop cpus we have can be more than 3 times faster than what consoles have and a ryzen 3600 is almost universally faster, by a gigantic margin.

Which suggests that the console GPUs are even more competitive against their PC counterparts than originally thought.

 

[pjbliverpool]

Which suggests that the console GPUs are even more competitive against their PC counterparts than originally thought.

Why?

 

[Dictator]

Which suggests that the console GPUs are even more competitive against their PC counterparts than originally thought.

Dude, Bro
Wut

Edit: the video is about CPUs

 

[davis.anthony]

It's as if DF is on a mission to eliminate false narratives this weekend.

I wasn't aware Ratchet & Clank was on Xbox to make the testing in the DF video relevant to our chat yesterday.

You seem hellbent on making yourself look foolish and so far you're doing a pretty good job of it.

 

[Riddlewire]

I wish he had tested Redfall.
If the 4800S can't reach 60fps no matter how low the settings, then that expected "eventual 60fps patch" for console can probably be considered an impossibility.

 

[I'mDudditz!]

Why?

Because DF Alex often runs comparisons between uncapped PS5 performance mode vs 2060S/Ryzen 3600, such as Spiderman. It's not off the table that there could be instances where PS5 cpu is hitting a much lower ceiling compared to the 3600 even though it was initially believed that these are comparable parts. This video suggests that the 3600 is actually much more capable than PS5 variant in most instances. This means the GPU to GPU comparison becomes flawed in favor of PC. Insomniac's own words further supports this at one time they mention they have GPU headroom to achieve 120fps, but the CPU is the bottleneck.

Dude, Bro
Wut

Edit: Didn't realize at first this was Dictator responding like this. I think you could have done a better job of asking clarification of what I'm getting at. Could've been as simple as the inquiry from pjliverpool. Nevertheless see above if you actually care, or not all fine either way.

 

[Shifty Geezer]

Which suggests that the console GPUs are even more competitive against their PC counterparts than originally thought.

1) You're pursuing a strange thought process, as if looking for a console advantage. This CPU investigation doesn't really talk about the GPUs at all.

2) We know what the GPUs are! There's no magic source there to make them 'even more competitive' that needs explaining. The console CPUs aren't particularly impressive. The GPU's aren't particularly impressive. The SSDs aren't particularly impressive now. The custom hardware, the decompression blocks and audio, offset shortcomings elsewhere, and the APIs and fixed hardware make them easier to tune games to. There's not really a lot to debate regards console vs PC hardware performance.

 

[I'mDudditz!]

1) You're pursuing a strange thought process, as if looking for a console advantage. This CPU investigation doesn't really talk about the GPUs at all.

I'm not looking for anything. I didn't wake up today knowing Richard was going to make this video. The video clearly communicates that the console cpu's are less capable than initially thought. So when you apply this newfound knowledge to DF previous PC vs console performance comparisons that attempts to isolate GPU performance, we can see that the PS5 cpu is likely handicapped in comparison to 3600, and in those instances GPU will be limited not by it's inherent limitations, but by the CPU.

If anyone can tell me, in a respectful manner please, where any of the logic above is incorrect or misplaced, I gladly welcome it.

2) We know what the GPUs are! There's no magic source there to make them 'even more competitive' that needs explaining. The console CPUs aren't particularly impressive. The GPU's aren't particularly impressive. The SSDs aren't particularly impressive now. The custom hardware, the decompression blocks and audio, offset shortcomings elsewhere, and the APIs and fixed hardware make them easier to tune games to. There's not really a lot to debate regards console vs PC hardware performance.

I didn't say anything about SSD and I/O so I have no idea why it's being brought up.

 

[pjbliverpool]

Because DF Alex often runs comparisons between uncapped PS5 performance mode vs 2060S/Ryzen 3600, such as Spiderman. It's not off the table that there could be instances where PS5 cpu is hitting a much lower ceiling compared to the 3600 even though it was initially believed that these are comparable parts. This video suggests that the 3600 is actually much more capable than PS5 variant in most instances. This means the GPU to GPU comparison becomes flawed in favor of PC. Insomniac's own words further supports this at one time they mention they have GPU headroom to achieve 120fps, but the CPU is the bottleneck.

That's not really how consoles work. As fixed platforms the developers know exactly how much CPU and GPU performance they have available and so can customise the load on each to balance things in such a way that neither component is a major bottleneck. For example if the game is CPU constrained, they can simply dial up resolution until the CPU and GPU are in balance again. That's exactly why DRS exists.

In addition to that we've seen numerous times now that despite it's raw performance advantage in PC code, the 3600x can be a bottleneck in the PC. Look at the RT view distance setting in Spiderman particularly for evidence of that.

I suspect that's because in a console the developer will specifically code around the CPU's weaknesses (less cache, high latency memory) which is coupled with the lower overhead OS and API's of the console.

 

[davis.anthony]

So because you've seen a video where a Series-X level CPU is bottlenecking a 61Tflop GPU when running under Windows with no console low level optimisation you've somehow deduced said CPU is also bottlenecking a 12Tflop and 9-10Flop GPU?

Do you not see where you're going wrong?

For the level of GPU both PS5 and Series-X have they have more enough CPU power.

 

[function]

Great video! Read the article too - found this bit interesting:

On the flipside, this PC version of the Xbox Series X CPU has some advantages over the console set-up. With SMT active - meaning eight cores and 16 threads - the console operates the CPU at a flat 3.6GHz, with one core held in reserve for OS level functions. On the AMD 4800S Desktop Kit, the CPU acts more like a standard Zen 2 processor in that clocks are variable. Typically, it seems to run all cores at around 4.0GHz like the other Zen 2 processors we've tested.

Whelp, at least on the CPU side the Series X silicone can turbo. Maybe the GPU can too?

Do it, MS, unleash the console Turbo warz!

Edit: perhaps PCIe 4X might be screwing with the results a bit more than DF are expecting? That's basically PCIe 2. That's .... going to be a bit of a problem for modern games?

@Dictator did you folks compensate by dropping PC PCIe down to second gen on the PCs?

Am a bit drunk, apologies if I've got the PCIe stuff wrong.

 

[davis.anthony]

Most of us, including Digital Foundry assumed the PS5 CPU and the Ryzen 3600 were close enough in performance to run head to head comparisons. The results of today's video analysis likely comes as a surprise to all of us. Labeling a reasonable hypothesis that prior comparisons may have understated the PS5 GPU performance as console warring is so insane.

You're correct, the PS5 CPU is bottlenecking its 61Tflop RDNA3 GPU.

 

[see colon]

I wish he had tested Redfall.
If the 4800S can't reach 60fps no matter how low the settings, then that expected "eventual 60fps patch" for console can probably be considered an impossibility.

Redfall's shortcomings have nothing to do with the hardware, though. There isn't a PC out there that can run it without performance issues, regardless of settings. That game is just doing something extremely inefficiently and if Arkane can sort that out, I'm sure the performance will be boosted accordingly.

 

[techuse]

Redfall's shortcomings have nothing to do with the hardware, though. There isn't a PC out there that can run it without performance issues, regardless of settings. That game is just doing something extremely inefficiently and if Arkane can sort that out, I'm sure the performance will be boosted accordingly.

I think I recall the issue being that it has almost no multithreading whatsoever.

 

[GhostofWar]

That video was intriguing, especially the contrast with batman and plaguetale.