Current Generation Hardware Speculation with a Technical Spin [post GDC 2020] [XBSX, PS5]

Status
Not open for further replies.
There are a lot of assumptions here. You are assuming that they(Sony) had "trouble locking" the frequency when in reality what Mark Cerny said they had trouble with was none other than fan noise and by extension inconsistent power draw. This solution was done to fix those problems, not to "reach" a frequency number. That is unless you have intimate knowledge of their internal thought process. Variable clock/Continuous boost was enabled for a reason, true. But the reason was to keep fan noise down.

37:14 of the Deep Dive says you are mistaken.

 
Last edited:
Xbox Series X TeraFlops puts it above 2080 Super indeed, but the available memory bandwidth to it puts it under that mark, and in a significant way, Series X has at max 560GB for both the CPU and GPU, which means the GPU will access less than that amount, maybe even way less if we factor in CPU/GPU memory contention, on the other hand the 2080 has full 448GB all for itself. That's why don't be surprised if the Series X fell in between RTX 2070 Super and RTX 2080 in practice.

It gets worse for the PS5, it has 448GB for both the CPU and GPU, the 2070 and 2070Super have that same 448GB all for themselves, that's why I expect performance to fall in between those two.

Then there is the fact that NVIDIA reports TF differently, NVIDIA calculates TF based on the least stable boost clocks, but their GPUs often boost higher, a 2080 has 10TF @1700MHz, but it always boosts to 1900MHz, making it effectively a 11TF GPU. Same goes for other GPUs, whether 2080Ti, 2070, 5700XT .. etc.

Yeah i fully agree, i was using my example just as a reference :) XSX matches ballpark 2080 vanilla, PS5 around 5700XT level (close to a 2070).

That's the basic idea. And why a thermal solution works at those speeds but it was a problema with locked clocks event at 2 Ghz.

Exactly, and that's what we don't know, how much lower the clocks can go. Some say 3Ghz for CPU and 2Ghz for the GPU. Which makes sense per github.
 
They way I look at it Sony have built their solution around where the APU is likely to spend most of its time as opposed to where the peaks occur. But rather than throw away the possibility of peaking, developers can determine which peaks to forfeit and which to favour.

To not do this would have been to leave performance on the table that otherwise wouldn't have been possible in their given budget/design/power constraints. The caveat of course being that it requires some management, but given comprehensive profiling and dynamic resolution scaling, VRS etc. These pull backs should hopefully be easy to mitigate.

A lot of people, particularly on other forums seem to fail to realise that Clockspeed =/= Utilisation and this subsequently makes the entire concept a lot harder to wrap your head around.

I suspect that with Sony opting for a fast-narrow design their cooling system will likely have to be quite performant already; and having to over-engineer it further for edge cases would have led to an exponential increase in cost and complexity for very little real world results.
 
People tend to focus too much on the TF number, forgetting other parameters, chief among them is the available memory bandwidth for the GPU.

Xbox Series X TeraFlops puts it above 2080 Super indeed, but the available memory bandwidth to it puts it under that mark, and in a significant way, Series X has at max 560GB for both the CPU and GPU, which means the GPU will access less than that amount, maybe even way less if we factor in CPU/GPU memory contention, on the other hand the 2080 has full 448GB all for itself. That's why don't be surprised if the Series X fell in between RTX 2070 Super and RTX 2080 in practice.

It gets worse for the PS5, it has 448GB for both the CPU and GPU, the 2070 and 2070Super have that same 448GB all for themselves, that's why I expect performance to fall in between those two.

Then there is the fact that NVIDIA reports TF differently, NVIDIA calculates TF based on the least stable boost clocks, but their GPUs often boost higher, a 2080 has 10TF @1700MHz, but it always boosts to 1900MHz, making it effectively a 11TF GPU. Same goes for other GPUs, whether 2080Ti, 2070, 5700XT .. etc.
No, XSX wont lose ~120GB/s duo to sharing it with CPU. I think they really payed attention to it and mittigated that issue by giving as much access to 560GB/s to GPU as possible. Zen2 will live with ~50GB/s all day everyday, so I dont know where you get 2070S and 2080 range.

Its much more likely to end up exectly where the specs actually point at ~2080S.
 
No, XSX wont lose ~120GB/s duo to sharing it with CPU. I think they really payed attention to it and mittigated that issue by giving as much access to 560GB/s to GPU as possible. Zen2 will live with ~50GB/s all day everyday
CPU/GPU contention makes you lose way more than that 50GB of Zen2. As was evident in PS4.
 
First tests show XSX competing with a 2080 (DF), i think that's where the raw performance is, without taking console optimization into account etc. It makes sense to me that it would land about there. Future games are going to take more advantage of RDNA2, Turing, Ampere etc, instead of being optimized for 2013 architectures. So performance won't get worse to say the least.
 
CPU/GPU contention makes you lose way more than that 50GB of Zen2. As was evident in PS4.
Say you lose 100GB/s, which I think wont be the case because then PS5 would be starved of BW, there is still 460GB/s to play with.

We will have to wait and see how they did it, but its clear they prioritized GPU access.

Even in worst case scenario, XSX would only have deficit in BW. Everything else would be + on XSX side vs 2080S. This is why I put it in 2080S ballpark, because 2070S is relatively close to 5700XT. A card of previous AMD generation with almost 3TF less.

First tests show XSX competing with a 2080 (DF), i think that's where the raw performance is, without taking console optimization into account etc. It makes sense to me that it would land about there. Future games are going to take more advantage of RDNA2, Turing, Ampere etc, instead of being optimized for 2013 architectures. So performance won't get worse to say the least.
First test after developers saying its very quick port to XSX.

XSX literally cannot be at level of 2070S. Turing and Navi have pretty much TF parity, and XSX is RDNA2 with considerably more TFs.
 
Xsx is at least on the level of 2080 Super, based on what I know and have been told.

The memory contention issues isn't a new problem in consoles, you can code around it like we saw people do for the PS4, Pro, and X1x.

We saw the PS4, Pro, and X1x's GPU perform as well / better than their PC counterparts. If memory contention was an issue, that wouldn't have been the case.
 
Last edited:
We saw the PS4, Pro, and X1x's GPU perform as well / better than their PC counterparts
X1X has more memory bandwidth than RX 480/580.

X1X is 325GB/s
RX480/580 are 256GB/s

PS4 also has more memory bandwidth than any similar PC GPU.

PS4 is 176GB/s
HD 7850/7870 are 153GB/s

That's why X1X and PS4 can keep up with their PC counterparts.

The memory contention issues isn't a new problem in consoles, you can code around it like we saw people do for the PS4, Pro, and X1x.
We shall see, but I highly doubt that. The Jaguar CPU is not that demanding in terms of memory bandwidth, Zen 2 is a different beast altogether compared to that.
Say you lose 100GB/s, which I think wont be the case because then PS5 would be starved of BW, there is still 460GB/s to play with.
You can't eyeball the number like that. Fact is: a shared memory pool won't keep up with a dedicated one no matter what. Unless the difference in bandwidth is huge of course, which isn't the case for either consoles.
 
You can't eyeball the number like that. Fact is: a shared memory pool won't keep up with a dedicated one no matter what. Unless the difference in bandwidth is huge of course, which isn't the case for either consoles.
Sure it can, if you have enough of it. I dont even understand your point. Nothing we know of current Navi vs RTX2000 series, points to XSX GPU, which will be Navi 2, being somewhere between 2070-2080.

And besides, while yes, XBX has more BW then RX580 (70GB/s advantage), XSX also has more then 2070S and 2080S (64GB/s and 112GB/s respectively).
 
XSX and 2080 probably trade blow depending on the situation. It's the only test we have now so far, and the conclusion was 2080.

HD7870 is generally the better performer compared to base PS4.
 
I could be wrong here but I thought XSX had a different pool of memory
specifically set aside for the CPU ?

I'm not as knowledgeable on how a system should work but wouldn't that 560 Gb/s of Bandwidth be less
constrained if there's a path set at the hardware or software level specifically for the GPU & CPU?
As in you're more likely to hit your 560 Gb/s performance target because there's another path altogether set up for the CPU. Am I correct in assuming that ?

And if that is indeed the case wouldn't that put XSX above a 2080s, and not potentially at it's level
or lower as far as bandwidth is concerned?

I hope someone can clear this up for me !
 
Status
Not open for further replies.
Back
Top