• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Digital Foundry: PC Gaming on the Xbox Series X CPU

Bojji

Member



4800S is Xbox Series X APU made to work as normal PC platform by AMD (done much better than 4700S made from PS5), it looks like performance is lower than ANY Zen 2 CPU available in PC space and comparable to Zen 1 chips.
 
Last edited:

Closer

Member
I was about to make a new thread about this fascinating video. You beat me to it but thanks for removing the burden lol
 
Last edited:

Dr.D00p

Gold Member
Not surprising. When the specs first appeared for both consoles they thought these CPUs were just transplanted desktop Zen 2s, not bothering to read the small print in seeing what's been stripped out to get them working in an APU with such a low power budget
 

Bojji

Member
But the PS5 is a beast, how can you out-beast the beast? How dare you even hint at such possibility!

For consoles just as Richard said in the video devs can target those CPUs specifically. Other things that can help are GPUGPU (just like PS4 days) and for PS5 in particulal all the loading can be done on IO hardware.

Console CPUs are much better than jaguars that's for sure but pale in comparison with modern processors available for pc.
 

SmokSmog

Member
Console CPUs are handicapped by low L3 cache and high GDDR6 latency. Nothing new.
PS5 and XSX CPUs = Zen1 in gaming.

This is why desktop PCs are using low latency dedicated system RAM for CPUs paired with a lot of cache. GPUs have their own high bandwidth version with higher latency called GDDR.
 
Last edited:

Skifi28

Member
I always thought that the whole point of console CPUs is that freed from windows and driver overheads they punch above their weight. Installing windows on them and saying "here's the actual performance" is going a bit backwards. I'm not opposed to the testing, that's actually quite interesting in a vacuum. But the platforms wars this will be used on? Oh boy.
 
It’s really curious why people are that desperate for a PS5 Pro. Are they not happy with Cerny’s magic this fast?

To be fair neither system is maxing out games with resolution or framerate. Some might be interested in systems that allow a stable 60FPs or maxed out DRS.
 

SlimySnake

Flashless at the Golden Globes
4800S is Xbox Series X APU made to work as normal PC platform by AMD (done much better than 4700S made from PS5), it looks like performance is lower than ANY Zen 2 CPU available in PC space and comparable to Zen 1 chips.

Jurassic Park Ian Malcom GIF


This explains the 60 fps issues on consoles. Guardians made no sense because it was doing 100 fps on console equivalent GPUs.

Lets hope they fix this with mid gen refreshes. im guessing the 8MB L2 cache is holding them back.
 

Kilau

Member
Zen 2 mobile is just like Zen 1 desktop
And why are they saying the Xbox is 4800s and Ps5 4700s when PS5 has better frames in most games?
I think AMD numbering is more concerned with the whole board package versus caring about a PS5/Xbox comparison. These chips in this setting are interesting only for a PC bench not in relation to how they perform in their designed console environment.
 
I really don't understand the complaints to the PS5 and Series X hardware, both machines were great for a price of $499 in 2020 with better balanced hardware than the PS4-ONE generation with that pretty bad jaguar CPU, 8 zen2 cores, 10 -12 Tflops, 16 gigabytes of ram and a 2.4Gb-5.5Gb SSD, etc... there should be little complaint.
 

winjer

Gold Member
I noticed that when they tested the latency on these CPUs, that the 3600 had 90ns. This is really bad for this CPU.
It takes really bad memory and low speeds to get this bad.
But this also explains why in so many benchmarks that DF made with the 3600, it's performance seemed lower than what it should be.
Why did they screw up latency on this CPU so badly?
 

rofif

Can’t Git Gud
Which bespoke features? What are consoles doing that PC can't?
a lot of things. Consoles are not a PC despite "seemingly similar x86 architecture"
There are A TON of differences. more than similarities.
Consoles have way different IO with separate bespoke IO processors/accelerators, direct level hardware access, obviously no directx on ps5, apu(custom rdna), one shared pool of ram+vram. you access assets in entirely different way. Shaders can be precompiled, that's also one thing not to worry on console.
Then there are features differences like the kraken decompressions... and more. I don't even remember now.
point being - IT IS NOT a pc. There are very little similarities. You can't get a game and just put it on ps5. Rename .exe to whatever linux uses and go.

listen from 14:00


Or some interesting tidbits in this video:


Why do you think that teams of MOST experienced developers on the planet (like naughty dog or nixxies) are having trouble porting games to pc?
Games take months or up to 2 years (god of war) to port to pc. It's not that simple.

Or listen to this pdocasts. it highlights how differen pc and ps5 are:
19:50 onwards


edit:
sorry for adding to many links!!! This is also very insightful with some dev:
 
Last edited:

SmokSmog

Member
I noticed that when they tested the latency on these CPUs, that the 3600 had 90ns. This is really bad for this CPU.
It takes really bad memory and low speeds to get this bad.
But this also explains why in so many benchmarks that DF made with the 3600, it's performance seemed lower than what it should be.
Why did they screw up latency on this CPU so badly?
Yep, average Zen2 with normal decent memory should be at 75ns and manually tuned at low 60s.
 

SlimySnake

Flashless at the Golden Globes
I really don't understand the complaints to the PS5 and Series X hardware, both machines were great for a price of $499 in 2020 with better balanced hardware than the PS4-ONE generation with that pretty bad jaguar CPU, 8 zen2 cores, 10 -12 Tflops, 16 gigabytes of ram and a 2.4Gb-5.5Gb SSD, etc... there should be little complaint.
This is simply pointing out the bottlenecks in the hardware. Console gaming is an exercise in sacrifices made to hit an affordable pricepoint but you still have to be smart about it. There is no point including an 8 core 16 thread cpu running at 3.5 Ghz if you are going to gut it by removing the cache that was needed to take full advantage of those 16 threads.

These videos help articulate the bottlenecks and hopefully ensure that console manufacturers learn from their mistakes going forward. maybe few cores with more cache is the answer. maybe dedicated ddr ram for the cpu is the way to go.

yes, these consoles are great for $499 but if their GPUs are being held back then surely there is room for improvement.
 

winjer

Gold Member
Yep, average Zen2 with normal decent memory should be at 75ns and manually tuned at low 60s.

Exactly. And Zen2 is very depend on memory latency for performance. When I was tunning memory on a 3600X, I saw some games gaining up to 25% performance. And this was compared to memory with latency that was not on the 90ns.
This is a gigantic screw up on DF´s part, that skews results big a big margin.
 

Ozriel

M$FT
no. devs need to stop using console as budget pc and start using it's bespoke features and optimizations

Devs have been doing that for ages.
That said, they also have to develop games with PC in mind too, since Steam can be a massive outlet for their games.
 

Dream-Knife

Banned
I think it's funny how everyone said the PS4 and xone had bad cpus, then when this gen came out it was amazing hardware, and now we're back to saying the consoles have bad cpus.

Nothing really changed in the last 3 years, except for hype levels.

Pro consoles come out next year, and in 2 years we will need 10th gen since this hardware is so limited.
 

ThisIsMyDog

Member
I think it's funny how everyone said the PS4 and xone had bad cpus, then when this gen came out it was amazing hardware, and now we're back to saying the consoles have bad cpus.

Nothing really changed in the last 3 years, except for hype levels.

Pro consoles come out next year, and in 2 years we will need 10th gen since this hardware is so limited.
The difference is that Jaguar was a complete crap even at the PS4/XONE release day, where's Zen1/2 from PS5/SERIES was in much better situation compared to PC hardware.
 
Damn, this puts into perspective some technical performances we've seen then. It would even, IMO, explain why Series X might struggle with framerates more often on 3P games than PS5. The PS5 has a lot of custom ASIC hardware for offloading pretty much all of the I/O and decompression routines off the CPU.

While Series X (and S) have some I/O to handle these things, it's nowhere near as robust as on the PS5, so the CPU has to do more of the heavy lifting there. A 100 MHz advantage seemingly isn't enough to make up for this, not to mention I don't think the Series systems have dedicated I/O ASICs for handling cache coherency. Then there's the lack of cache scrubbers, so cache line flushes are more frequent with the Xbox systems, which take up CPU cycle times.

All of that would impact the rate at which draw calls for the GPU could be issued. Hence the 5-10 (sometimes more) lower FPS we see in a lot of Series X 3P games compared to PS5 ones at similar settings. This is just me giving a possible explanation for some of the performance we've seen in games on both platforms; the fact the CPUs in neither the Xbox Series or PS5 systems are as robust as first thought, just shows the increased likelihood of this all being true.

Guess this might make it two console gens in a row where the CPUs were rather weak-sauce? We need another Cell moment; the industry is ready this time.

CPU is crippled by GDDR6, so nothing new really. Hopefully in next cycle people won't praise the shared memory.

You think it's a latency issue? That's the only thing I could think of.

IMO then, unified memory isn't the problem. They just need lower-latency memory. HBM is the future for console memory.
 
Last edited:

ReBurn

Gold Member
a lot of things. Consoles are not a PC despite "seemingly similar x86 architecture"
There are A TON of differences. more than similarities.
Consoles have way different IO with separate bespoke IO processors/accelerators, direct level hardware access, obviously no directx on ps5, apu(custom rdna), one shared pool of ram+vram. you access assets in entirely different way. Shaders can be precompiled, that's also one thing not to worry on console.
Then there are features differences like the kraken decompressions... and more. I don't even remember now.
point being - IT IS NOT a pc. There are very little similarities. You can't get a game and just put it on ps5. Rename .exe to whatever linux uses and go.

listen from 14:00


Or some interesting tidbits in this video:


Why do you think that teams of MOST experienced developers on the planet (like naughty dog or nixxies) are having trouble porting games to pc?
Games take months or up to 2 years (god of war) to port to pc. It's not that simple.

Or listen to this pdocasts. it highlights how differen pc and ps5 are:
19:50 onwards


edit:
sorry for adding to many links!!! This is also very insightful with some dev:

Oh ok
 

M1chl

Currently Gif and Meme Champion
Damn, this puts into perspective some technical performances we've seen then. It would even, IMO, explain why Series X might struggle with framerates more often on 3P games than PS5. The PS5 has a lot of custom ASIC hardware for offloading pretty much all of the I/O and decompression routines off the CPU.

While Series X (and S) have some I/O to handle these things, it's nowhere near as robust as on the PS5, so the CPU has to do more of the heavy lifting there. A 100 MHz advantage seemingly isn't enough to make up for this, not to mention I don't think the Series systems have dedicated I/O ASICs for handling cache coherency. Then there's the lack of cache scrubbers, so cache line flushes are more frequent with the Xbox systems, which take up CPU cycle times.

All of that would impact the rate at which draw calls for the GPU could be issued. Hence the 5-10 (sometimes more) lower FPS we see in a lot of Series X 3P games compared to PS5 ones at similar settings. This is just me giving a possible explanation for some of the performance we've seen in games on both platforms; the fact the CPUs in neither the Xbox Series or PS5 systems are as robust as first thought, just shows the increased likelihood of this all being true.

Guess this might make it two console gens in a row where the CPUs were rather weak-sauce? We need another Cell moment; the industry is ready this time.



You think it's a latency issue? That's the only thing I could think of.

IMO then, unified memory isn't the problem. They just need lower-latency memory. HBM is the future for console memory.
It is very much a problem, since there is low L3 cache, the values are mostly store in RAM, which is a highway with few entry/exits, thus you are waiting way longer than you should on return values and so on.
 

winjer

Gold Member
You think it's a latency issue? That's the only thing I could think of.

IMO then, unified memory isn't the problem. They just need lower-latency memory. HBM is the future for console memory.

HBM is way too expensive for a console. Even for PC GPUs and that's why we only see it in enterprise class products.
Consoles have 2 solutions. One is to go for a dual pool of memory. DDR4 for the CPU and GDDR6 for the GPU.
The other is to add more cache. This helps a lot with scheduling. But consoles cut on this, as both the PS5 and Series X, only have 4MB per CCX.
Having the full 16MB per CCX that Zen2 has on PC, would alleviate a lot the issues with memory latency. Or even better, adding some 3DVcache.
 
This is so pathetic and ultimately pointless. Consoles are bottled necked, the shock """
What next, putting in an OG Xbox CPU in Alex's old PC and getting sh8t results?

Consoles are different to a PC and that's why one is called a console and the other the PC and a console user can't happily change their CPU or GPUs
 

Sleepwalker

Member
Hopefully the PRO versions of the consoles (xbox needs one too) get a bump in CPU spec as well and not only a modest gpu bump. Compatibility shouldn't be too much of an issue nowadays.
 
It is very much a problem, since there is low L3 cache, the values are mostly store in RAM, which is a highway with few entry/exits, thus you are waiting way longer than you should on return values and so on.

Yeah, and as SmokSmog SmokSmog said earlier, the 8 MB L3$ is really 2x 4 MB chunks so you have 4 cores/8 threads accessing 4 MB of L3$ in these systems.

There wasn't really anything Sony or Microsoft could've done there, though; that setup was all on AMD. A PS5 Pro actually has a more realistic chance of having Zen 4 than first thought, and since Zen 4 is BC with Zen 2 code, it might make the change there more acceptable from a design POV. IIRC the reason the PS4 Pro and One X used the same Jaguar cores was because AMD's new CPUs at the time were not BC in microcode with the Jaguar CPUs, so too much work would've been required in recompiling games.

At least, that's my understanding of things. But for next-gen systems, I genuinely think they need to move away from GDDR and towards HBM3 (maybe ben HBM-PIM but that depends on what can be done with AMD in GPU/CPU architecture by then), if they want a unified memory solution. Which is still preferable to a non-unified memory pool, especially in console, plus generally being cheaper due to more volume allowing better economies of scale.

HBM is way too expensive for a console. Even for PC GPUs and that's why we only see it in enterprise class products.
Consoles have 2 solutions. One is to go for a dual pool of memory. DDR4 for the CPU and GDDR6 for the GPU.
The other is to add more cache. This helps a lot with scheduling. But consoles cut on this, as both the PS5 and Series X, only have 4MB per CCX.
Having the full 16MB per CCX that Zen2 has on PC, would alleviate a lot the issues with memory latency. Or even better, adding some 3DVcache.

HBM prices have come down a lot in the past few years, and there's supposed to be a design that simplifies (or even removes) the interposer, which is where a lot of the cost comes from. But also, the main reason HBM prices have been high compared to DDR and GDDR, IMO, is because none of the customers for it order in large enough volumes to make the memory manufacturers price per-chip costs lowers. The volumes they get ordered are simply too small for economies of scale, so they make up for it by keeping the prices high for the clients willing to pay for the privilege.

If Microsoft or especially Sony were ordering HBM memories for their next consoles, they'd be ordering in volumes so large that prices would naturally come down because the manufacturers of the memory make more through the sheer bulk of order than they ever did with companies ordering way smaller quantities, so it'd not matter if the profit margin on each chip were lower for massive clients like Sony & Microsoft. That's my theory on HBM pricing, anyway.

I really don't see non-unified memory ever returning in consoles, it's just not worth the headaches. PC does have SAM/BAR but it's still not preferable to actual unified memory pool. There's a reason the industry in embedded systems has been trending towards hUMA: the benefits simply far outweigh the alternative. But, if next-gen consoles want to keep that while avoiding these issues, and don't address cache, then they have to move away from GDDR and towards something like HBM.

Cache option could provide a nice balance and would help a lot, though the consoles wouldn't want too much cache, not without ways of smart managing the cache data. In that respect I think the PS5 has a good idea with its GPU cache scrubbers, but yes they should also increase the amount of cache and fully unify the cache as well.
 
Last edited:

M1chl

Currently Gif and Meme Champion
Yeah, and as SmokSmog SmokSmog said earlier, the 8 MB L3$ is really 2x 4 MB chunks so you have 4 cores/8 threads accessing 4 MB of L3$ in these systems.

There wasn't really anything Sony or Microsoft could've done there, though; that setup was all on AMD. A PS5 Pro actually has a more realistic chance of having Zen 4 than first thought, and since Zen 4 is BC with Zen 2 code, it might make the change there more acceptable from a design POV. IIRC the reason the PS4 Pro and One X used the same Jaguar cores was because AMD's new CPUs at the time were not BC in microcode with the Jaguar CPUs, so too much work would've been required in recompiling games.

At least, that's my understanding of things. But for next-gen systems, I genuinely think they need to move away from GDDR and towards HBM3 (maybe ben HBM-PIM but that depends on what can be done with AMD in GPU/CPU architecture by then), if they want a unified memory solution. Which is still preferable to a non-unified memory pool, especially in console, plus generally being cheaper due to more volume allowing better economies of scale.
CPUs are very sensitive for fast memory, that's why you see massive uplifts in performance, because the developing process is small enough to fit a huge memory there, they are basically could get lazy for few years and just add more memory and it would be seen, since with multicore programs there is more overhead and you always take some "logical cell", which could be 512kb while you storing just a few bytes (depends on language, compiler) and that way you aren't ever have enough memory directly on-die.
 

hinch7

Member
Pretty much on par with a 3600 is what I estimated a while ago while this chip launched. A cut down Zen 2 with less cache and at a low 3.5Ghz with 8 cores/16 threads works okay for modern titles but falls apart compared to newer generation CPU's with considerably higher IPC and cache. That we had theoretical benchmarks of this chip that can be compared to others. Nice to see actual real world comparisons done here with the games tested.

Much better than we got with last last generation with Jaguar but its beggining to drag its feet a bit for newer releases. Though I guess a match for the GPU(s) in both XSX/S and PS5, it won't be that much that much detriment to the performance overall. Unless devs want to push really high FPS in their games at reasonable resolutions.

Assuming the PS5 Pro is coming, an APU with Zen 4 would bring a huge leap in CPU performance over current consoles.
 
Last edited:

Dr.D00p

Gold Member
This is so pathetic and ultimately pointless. Consoles are bottled necked, the shock """
What next, putting in an OG Xbox CPU in Alex's old PC and getting sh8t results?

Consoles are different to a PC and that's why one is called a console and the other the PC and a console user can't happily change their CPU or GPUs

FFS.

This is the kind of stuff DF do, particularly Richard, it's not meant to mean anything profound, just an interesting theoretical exercise that alot us find fascinating.

If you don't, don't watch it...
 

Clear

CliffyB's Cock Holster
Console CPUs are handicapped by low L3 cache and high GDDR6 latency. Nothing new.
PS5 and XSX CPUs = Zen1 in gaming.

This is why desktop PCs are using low latency dedicated system RAM for CPUs paired with a lot of cache. GPUs have their own high bandwidth version with higher latency called GDDR.

Exactly. I didn't make it to the end of the video because it should be extremely obvious that ram latency is huge, especially for multi-threaded applications.
 

ChiefDada

Gold Member
This is so pathetic and ultimately pointless. Consoles are bottled necked, the shock """
What next, putting in an OG Xbox CPU in Alex's old PC and getting sh8t results?

Consoles are different to a PC and that's why one is called a console and the other the PC and a console user can't happily change their CPU or GPUs

I mean, if anything I'm even more impressed with current gen console performance compared to PC. Has it ever been this tight 3 years into the generation?
 
Top Bottom