Digital Foundry Article Technical Discussion [2024]

[Cyan]

1:13:43 Supporter Q2: Was Microsoft showing Series X or PC footage at their games showcase?

passionate speech from Alex here. In trailers, they can crank whatever settings they want up. On rigs like mine the settings are more like current gen consoles though. If you want those graphics this gen, well.....

Cue the 30fps.

 

[Charlietus]

At some point in the video Rich says that consoles CPU's perform more like a first generation Ryzen. But when there are comparisons in games it's almost never the case. There are games where it's true, like baldurs gate 3 where a PS5 is a bit below a 3600, but most times it's in line or even above. Recent examples that I remember are dragons dogma 2 and for the series x hellblade 2, where it doesn't get any traversal stutters.

This is all anecdotal, since I don't have all the comparisons in front of me, so maybe I'm remembering wrong.

 

[Remij]

I'm thrilled that John's looking to start a future "Retro Systems Comparison" series comparing PS1 and Saturn (+N64 potentially in the future) games against each other. I LOVE watching comparisons and retrospectives on games and development for past consoles. The 32/64bit generation is my favorite gen of all time, so I'm very excited for that!

There's so much potential for content for DF from the past.. it's staggering.. so I hope that they're able to get John the support and resources he needs to satisfy his ambitions for the project.

 

[arandomguy]

At some point in the video Rich says that consoles CPU's perform more like a first generation Ryzen. But when there are comparisons in games it's almost never the case. There are games where it's true, like baldurs gate 3 where a PS5 is a bit below a 3600, but most times it's in line or even above. Recent examples that I remember are dragons dogma 2 and for the series x hellblade 2, where it doesn't get any traversal stutters.

This is all anecdotal, since I don't have all the comparisons in front of me, so maybe I'm remembering wrong.

Depends on the context, I don't quite remember what it was in this case.

From a hardware context while the PS5 CPU is "Zen 2" it's an Zen 2 APU, and if you look at Zen 2 APUs they not performance equivalent to Zen 2 CPUs (especailly the ones with XMP memory) that desktop users associate with it. The APUs in general have perform closer to n-1 compared to the CPUs using the same generation CPU core due to uncore differences.

Now end result speaking due to software optimization difference you can't directly compare to the PC from an end result stand point, as a Zen 2 APU clocked to the same speeds (and I guess running very high latency for the PC memory) would likely run worse compared to the PS5. And this has always typically been the case to some extent.

As an aside this is worth remember in the context of future console CPUs and the ARM discussion (in another thread I think) in that it's worth remember that a console CPU, if it does go ARM, doesn't need to be performance equivalent hardware wise to the latest desktop enthuasist x86 offering AMD as that isn't what would go into consoles.

 

[Charlietus]

Depends on the context, I don't quite remember what it was in this case.

From a hardware context while the PS5 CPU is "Zen 2" it's an Zen 2 APU, and if you look at Zen 2 APUs they not performance equivalent to Zen 2 CPUs (especailly the ones with XMP memory) that desktop users associate with it. The APUs in general have perform closer to n-1 compared to the CPUs using the same generation CPU core due to uncore differences.

Now end result speaking due to software optimization difference you can't directly compare to the PC from an end result stand point, as a Zen 2 APU clocked to the same speeds (and I guess running very high latency for the PC memory) would likely run worse compared to the PS5. And this has always typically been the case to some extent.

As an aside this is worth remember in the context of future console CPUs and the ARM discussion (in another thread I think) in that it's worth remember that a console CPU, if it does go ARM, doesn't need to be performance equivalent hardware wise to the latest desktop enthuasist x86 offering AMD as that isn't what would go into consoles.

I understand, but if the end results are most times equal to third gen Ryzen, is it really fair to say that they are like first gen? I don't think that it matters how the PS5 CPU on PC performs since on a PS5 it apparently performs quite differently.

 

[pjbliverpool]

I understand, but if the end results are most times equal to third gen Ryzen, is it really fair to say that they are like first gen? I don't think that it matters how the PS5 CPU on PC performs since on a PS5 it apparently performs quite differently.

I think it depends in whether you are saying "this is how performant the PS5 CPU is, irrespective of the environment its operating in", or "this is how much performance you would need from a CPU in the PC environment to give equivalent to PS5 performance".

And remember it's game dependent as well since some games will put tasks onto the CPU in a PC where the CPU does not need to be involved on the console. If a game requires a lot of Streaming (with real-time data decompression), plus significant BVH rebuilds every frame, e.g. Spiderman, then something like a 3600x may not be enough.

 

[Charlietus]

I think it depends in whether you are saying "this is how performant the PS5 CPU is, irrespective of the environment its operating in", or "this is how much performance you would need from a CPU in the PC environment to give equivalent to PS5 performance".

And remember it's game dependent as well since some games will put tasks onto the CPU in a PC where the CPU does not need to be involved on the console. If a game requires a lot of Streaming (with real-time data decompression), plus significant BVH rebuilds every frame, e.g. Spiderman, then something like a 3600x may not be enough.

That's the thing, to the end user it doesn't matter if on PC the CPU overhead is higher, if on PS5 and series x some tasks are handled by coprocessors and more, that just means that on console the CPU is operating at a certain level. And that the console CPU on PC operates at a lower performance is curious, nothing more.

 

[davis.anthony]

DF did a comparison of the Ryzen 3600 Vs that Frankenstein PC (Basically XSX CPU)

And in a lot of the games the desktop 3600 smoked the Frankenstein PC.

The reduction of L3 cache and low clock speed really hurts the consoles in some games.

 

[Charlietus]

DF did a comparison of the Ryzen 3600 Vs that Frankenstein PC (Basically XSX CPU)

And in a lot of the games the desktop 3600 smoked the Frankenstein PC.

The reduction of L3 cache and low clock speed really hurts the consoles in some games.

Then again, it's the xsx CPU in a PC, so that doesn't tell us a lot about the performance in the console (especially PS5, that for a variety of reason performs better than the xsx, even with a slower clock speed and with the PS5 being cut down).

All of the above is still what I remember from various comparisons, so maybe I'm remembering wrong.

 

[davis.anthony]

Then again, it's the xsx CPU in a PC, so that doesn't tell us a lot about the performance in the console (especially PS5, that for a variety of reason performs better than the xsx, even with a slower clock speed and with the PS5 being cut down).

You're right, the faster, lower latency RAM used in the Frankenstein PC is likely producing better results than what the consoles have so shouldn't be used.

From the DF testing of 3600 vs 'Desktop' XSX CPU

CP2077 +37% faster
Metro Exodus Enhanced Edition +68% faster
Flight Simulator +44% faster
Far Cry 6 +44% faster
Hitman 3 +39% faster

Console optimisation will only get you so far, but the performance gulf can be substantial.

And the 3600 is considered a budget, low-end, entry CPU in the PC space these days.

 

[Charlietus]

You're right, the faster, lower latency RAM used in the Frankenstein PC is likely producing better results than what the consoles have so shouldn't be used.

From the DF testing of 3600 vs 'Desktop' XSX CPU

CP2077 +37% faster
Metro Exodus Enhanced Edition +68% faster
Flight Simulator +44% faster
Far Cry 6 +44% faster
Hitman 3 +39% faster

Console optimisation will only get you so far, but the performance gulf can be substantial.

And the 3600 is considered a budget, low-end, entry CPU in the PC space these days.

Again again, the xsx CPU in the PC is not rappresentative of the CPU in the console. And sadly there aren't many tests in this regard. I maybe remember some comparisons in some videos by DF but they were few and far between. And on PC I remember either comparable, lower or higher frame rates depending on the game and some bad frame times compared to the console. I would have to check for a thousand comparison videos but I don't think I have it in me

Some recent comparisons I can remember are BG3 with 3~ fps higher on the PC vs PS5, Hellblade with no traversal stutters on series x and warzone (?) with better frame rate on PS5.

I can be pretty sure that's much better performance than 1st gen Ryzen.

Ps, I also don't see ram slot on that motherboard, so DF is using the integrated gddr6 mem that's on the consoles. So there is no advantage there.

 

[pjbliverpool]

Again again, the xsx CPU in the PC is not rappresentative of the CPU in the console. And sadly there aren't many tests in this regard. I maybe remember some comparisons in some videos by DF but they were few and far between. And on PC I remember either comparable, lower or higher frame rates depending on the game and some bad frame times compared to the console. I would have to check for a thousand comparison videos but I don't think I have it in me

Some recent comparisons I can remember are BG3 with 3~ fps higher on the PC vs PS5, Hellblade with no traversal stutters on series x and warzone (?) with better frame rate on PS5.

I can be pretty sure that's much better performance than 1st gen Ryzen.

A lack of traversal stutters won't have anything to do with CPU performance.

 

[Charlietus]

A lack of traversal stutters won't have anything to do with CPU performance.

I disagree, if the API's are better on the console allowing for better CPU utilization, than the same CPU is allowing better performance.

 

[arandomguy]

I understand, but if the end results are most times equal to third gen Ryzen, is it really fair to say that they are like first gen? I don't think that it matters how the PS5 CPU on PC performs since on a PS5 it apparently performs quite differently.

By that reasoning how console games play vs a 3600 system is also just "curious" and nothing more since from an actual consumer perepctive I'm not sure how many people that would actually impact? Let's be blunt here the entire platform vs. platform hardware "normalized" performance comparison thing for the most part is just curiousity and not any sort of pragmatism.

Which is why the discussion depends on context (again I don't remember the context it was brought up). Practically speaking it also does matter depending on the context of the discussion. For example in terms of the future direction of console hardware.

 

[Charlietus]

By that reasoning how console games play vs a 3600 system is also just "curious" and nothing more since from an actual consumer perepctive I'm not sure how many people that would actually impact? Let's be blunt here the entire platform vs. platform hardware "normalized" performance comparison thing for the most part is just curiousity and not any sort of pragmatism.

Which is why the discussion depends on context (again I don't remember the context it was brought up). Practically speaking it also does matter depending on the context of the discussion. For example in terms of the future direction of console hardware.

One is how the CPU is operating on the console, the other is a CPU that is never going to get bought by the average consumer that is used to measure the ballpark performance of the console. And since the conditions for the tests are not equalized, it's not useful data. It's not the same platform, not the same API, not the same OS.

 

[arandomguy]

One is how the CPU is operating on the console, the other is a CPU that is never going to get bought by the average consumer that is used to measure the ballpark performance of the console. And since the conditions for the tests are not equalized, it's not useful data. It's not the same platform, not the same API, not the same OS.

Again you're looking at this contextually for some reason as a console vs PC comparison.

I'll give a more direct expect. The PS5 uses a Zen 2 APU. Let's say the PS5 was announced with a Zen 2 CPU (for whatever reason). Is that an upgrade? We know hardware wise that is, which is why the context matters. Let's say the PS6 goes ARM (for whatever reason) and an equivalent ARM CPU is available on the PC. Do we speculate the performance uplift vs. the 3600? No, because that isn't what the PS5 used hardware wise and we have no reason assume the PS6 would extract less out of the CPU.

 

[Charlietus]

Again you're looking at this contextually for some reason as a console vs PC comparison.

I'll give a more direct expect. The PS5 uses a Zen 2 APU. Let's say the PS5 was announced with a Zen 2 CPU (for whatever reason). Is that an upgrade? We know hardware wise that is, which is why the context matters. Let's say the PS6 goes ARM (for whatever reason) and an equivalent ARM CPU is available on the PC. Do we speculate the performance uplift vs. the 3600? No, because that isn't what the PS5 used hardware wise and we have no reason assume the PS6 would extract less out of the CPU.

The discussion started with me not agreeing with Rich (which I respect deeply, he is one of the best professionals in this industry) on defining the consoles CPU as closer to 1st gen Ryzen. And that is comparing consoles to PC. But as far as I can see, there aren't many "scientific" comparisons since we don't have the tools to measure with certainty If a console is 100% CPU limited. And the Frankenstein pc isn't good for precise comparisons. So there is no need for other hypotheticals.

 

[see colon]

Doesn't the 3600 basically perform like a 2700x, though? 2700x is Zen 1+, a "first generation" Zen processor. I don't even know what you guys are arguing about, what Rich said, broadly speaking, is correct. The console CPUs are performing more like a 2700x than a 3700x.

 

[davis.anthony]

I disagree, if the API's are better on the console allowing for better CPU utilization, than the same CPU is allowing better performance.

What better performance?

 

[Charlietus]

Doesn't the 3600 basically perform like a 2700x, though? 2700x is Zen 1+, a "first generation" Zen processor. I don't even know what you guys are arguing about, what Rich said, broadly speaking, is correct. The console CPUs are performing more like a 2700x than a 3700x.

The 3600 is usually faster than a 2700x in benchmarks. Also, DF always compared the consoles CPU to the 3600 in the directs and videos.