Xbox One (Durango) Technical hardware investigation

Status
Not open for further replies.
IMO those two statements (said back to back in the interview) are trying damn hard to imply their bump in clock to 12CUs is better than 18CUs.

No it's not, at all.

They're saying that for the Xbox One the clock speed bump gave them more than an additional two CUs. Which no doubt it did.

There nothing in there that claims what you are saying. They don't even compare the the clock bump to a hypothetical 18 CU Xbox One, never mind the PS4.
 
mind quoting the part about HSA?

I talked about it on the last page:

The interesting thing for me from these results is the 290X time of 1.4ms vs the PS4 GPU time of 1.6ms. The only explanation I have for that is that the PC GPU time also accounts for the latency of sending the data back and forth over PCI-E. This gives us a pretty good idea of what kind of benefits the HSA architecture of the consoles might offer over a discrete GPU.

We can assume that given the power disparity between the PS4 GPU and a 290x the actual GPU time would be around 0.5ms which means the PCI bus is adding about 0.9ms to the overall operation. I bet if they used a 280X for this the overall PC GPU time would be slower than the PS4. Although obviously faster overall since the PS4 is still held back by the CPU.
 
No it's not, at all.

They're saying that for the Xbox One the clock speed bump gave them more than an additional two CUs. Which no doubt it did.

There nothing in there that claims what you are saying. They don't even compare the the clock bump to a hypothetical 18 CU Xbox One, never mind the PS4.

Never-mind, do your own homework. People are endlessly finding you quotes and links just for you to ask for more.
 
Never-mind, do your own homework. People are endlessly finding you quotes and links just for you to ask for more.

The link you posted in no way backs up the claim you made. The architects did not claim that a 7% speed bump made the system better than a unit with 50% more CUs.

You made false claim, have been called out on it, and are retreating behind another spurious claim.

It's not my homework to do - it's yours. And you just flunked the assignment.
 
The numbers are clearly just educated speculation but the fact that the 290x is vastly slower than it should be clearly points to something else going on and since we know that PCI-E adds a great deal of latency its a fairly safe bet to assume thats the culprit, or at least one of them.

If nothing else its an interesting comparison to speculate on and discuss possible reasons for, hence my original assertion.
 
there are like 500 words in between the 2 quotes, I feel you are taking them out of context. Besides, isn't it a Sony leak that actually said 14CU at balance? no?

Its been months since I read it but my take on the article was MS saying they could have bumped to 14 CUs but our test showed a clock increase improved performance more and beyond 14 CUs there is a dramatic drop in how the architecture scales so you won't see the increase in performance that you would expect by simply comparing which system has more CUs.

However what we are seeing with titles doesn't match those comments, perhaps its due to incomplete tools or something else but so far the article seems to not match the results.

Time will tell who was right.... But anyone can go back and look at the forum discussions here and see that some people used that article to suggest that 18 CUs was a mistake.
 
Its been months since I read it but my take on the article was MS saying they could have bumped to 14 CUs but our test showed a clock increase improved performance more and beyond 14 CUs there is a dramatic drop in how the architecture scales so you won't see the increase in performance that you would expect by simply comparing which system has more CUs.

However what we are seeing with titles doesn't match those comments, perhaps its due to incomplete tools or something else but so far the article seems to not match the results.

Time will tell who was right.... But anyone can go back and look at the forum discussions here and see that some people used that article to suggest that 18 CUs was a mistake.

Except is XBO particularly CU limited? I'm not seeing major shader differences in PS4 multiplatform titles that are 1080P on both consoles (or any titles). I think ROP's or ESRAM is probably what's keeping X1 from 1080. And more CU's wouldn't have helped that.

Still it sounds stupid when MS said they tested the impact of clock/+2CU's on launch titles. It seems obvious titles targeted at 12 CU's will benefit more from a clock bump since they wouldn't be ALU limited anyway!

We dont know how results would have been from +2 CU's, but it doesn't seem so far like 53mhz is yielding good results. I think if Microsoft could go back in time they would have enabled the 2 CU's as well. They likely didn't anticipate what a big deal the power difference would be (they could have just asked me :) ), with each new multiplatform deficiency causing a mini-scandal.
 
The numbers are clearly just educated speculation but the fact that the 290x is vastly slower than it should be clearly points to something else going on and since we know that PCI-E adds a great deal of latency its a fairly safe bet to assume thats the culprit, or at least one of them.

If nothing else its an interesting comparison to speculate on and discuss possible reasons for, hence my original assertion.

I don't know, maybe it's something called the OS? Point is I don't see how you can imply this is from HSA.
 
I don't know, maybe it's something called the OS? Point is I don't see how you can imply this is from HSA.

Feel free to explain in detail why you think the OS would be causing a 3 fold performance decrease in this GPGPU operation on PC compared to consoles while the PCI-E Bus which we KNOW adds significant latency in GPGPU operations is having no impact.
 
Feel free to explain in detail why you think the OS would be causing a 3 fold performance decrease in this GPGPU operation on PC compared to consoles while the PCI-E Bus which we KNOW adds significant latency in GPGPU operations is having no impact.

What? So I having no explanation has any bearing on how credible your claim is? I simply have no claim, because I don't know, and I'm calling you our on the HSA because I feel it's far fetched, and you have yet to give anything to support it.
 
Can it simply be a case of apu versus discrete? Even Llano could outperform discrete systems with beefier gpus in some data intensive gpgpu workloads because the apu was simply more efficient in moving data between the GPU and CPU.
 
Except is XBO particularly CU limited? I'm not seeing major shader differences in PS4 multiplatform titles that are 1080P on both consoles (or any titles). I think ROP's or ESRAM is probably what's keeping X1 from 1080. And more CU's wouldn't have helped that.

MGS Ground Zeros has a pretty sizable gap in image quality IMO (Resolution + X1 missing dynamic sky).
 
Except is XBO particularly CU limited? I'm not seeing major shader differences in PS4 multiplatform titles that are 1080P on both consoles (or any titles). I think ROP's or ESRAM is probably what's keeping X1 from 1080. And more CU's wouldn't have helped that.

Still it sounds stupid when MS said they tested the impact of clock/+2CU's on launch titles. It seems obvious titles targeted at 12 CU's will benefit more from a clock bump since they wouldn't be ALU limited anyway!

We dont know how results would have been from +2 CU's, but it doesn't seem so far like 53mhz is yielding good results. I think if Microsoft could go back in time they would have enabled the 2 CU's as well. They likely didn't anticipate what a big deal the power difference would be (they could have just asked me :) ), with each new multiplatform deficiency causing a mini-scandal.

The balance thing always read as a bad excuse, and afaik the 14 cu vs 12 was more related to yield than performance. I fully understand the need for doing pr like this week have seen it for every console.

I think it's a given that mp games haven't exploited the esram to its fullest, but it's just as likely that the extra cu power on the ps4 is seeing the same challenge. As we go deeper into this generation graphics engines will be tailored around this and new pros and cons will show up :)
 
The sites don't disagree.

Where's the difference?

it can play 4K video frames (3840x2160) in 4 ms PCs and 11 ms PS4/Xbox One using the CPU only (or 1.4 ms PC and 2.3 ms PS4/Xbox using GPU acceleration)!
http://www.radgametools.com/bnkmain.htm

If you want to break down official site numbers it would be like this:

Red part ---> CPU | Blue part ---> GPU acceleration

PC : CPU 4 ms. GPU: 1.4 ms.
PS4: CPU 11 ms. GPU: 2.3 ms.
Xbox: CPU 11 ms. GPU: 2.3 ms.

Then compare like for like numbers with gamingbolt:

PC: CPU: 1.3 ms. GPU: 1.4 ms.
PS4: CPU 2.3 ms. GPU: 1.6 ms.
Xbox: CPU 2.3 ms. GPU: 2.3 ms.
http://gamingbolt.com/new-benchmark-results-show-ps4s-gpu-is-faster-than-xbox-one-can-play-4k-video-frames-faster#BJzorBArCDEz2k2V.99:

The difference is Obvious.
 
What? So I having no explanation has any bearing on how credible your claim is? I simply have no claim, because I don't know, and I'm calling you our on the HSA because I feel it's far fetched, and you have yet to give anything to support it.

You're previous comment of "I don't know, maybe it's something called the OS?" which was dripping in sarcasm suggested you have a very good reason for dismissing my specualation and presenting this apparently much better one of your own. So I'm quite justifiably asking for some reasoning behind that.

Suggesting the relative speedup in a GPGPU operation from a HSA system over a discrete one with a more more powerful GPU is a perfectly valid and logical suggestion. That's exactly the kind of benefit HSA is designed to bring and exactly what Sony has been saying the PS4 will offer for months. Yet you have some reason to think that isn't the case? I'm keen to hear it.

And for the record, I'm not claiming HSA is the definite cause, I'm suggesting it as a plausable explanation for discussion. If you have reasons for doubting it's validity then I'd appreciate you stating them. Afterall, that's what a technical forum is for - technical discussion.

dobwal said:
Can it simply be a case of apu versus discrete? Even Llano could outperform discrete systems with beefier gpus in some data intensive gpgpu workloads because the apu was simply more efficient in moving data between the GPU and CPU.

That's exactly what I'm saying. It's about the time it takes to transfer the data back and forth over PCI-E adding latency to the overall operation. I simply used HSA as a catch all term for that setup in the consoles. taisui's reasons for dismissing that theory are yet to be revealed.
 
http://www.radgametools.com/bnkmain.htm

If you want to break down official site numbers it would be like this:

Red part ---> CPU | Blue part ---> GPU acceleration

PC : CPU 4 ms. GPU: 1.4 ms.
PS4: CPU 11 ms. GPU: 2.3 ms.
Xbox: CPU 11 ms. GPU: 2.3 ms.

Then compare like for like numbers with gamingbolt:

http://gamingbolt.com/new-benchmark-results-show-ps4s-gpu-is-faster-than-xbox-one-can-play-4k-video-frames-faster#BJzorBArCDEz2k2V.99:

The difference is Obvious.

The proposed explanation for those numbers that is under discussion is that Gamingbolts numbers are the breakdown of the overall CPU+GPU time which consists of two components which run in parallel.

So from the official site you have a TOTAL CPU+GPU time of 2.3ms on both consoles.

Gamingbolt breaks those totals down into their two constituent parts which run parallel to each other (again, assuming that assumption is correct). So you have 2.3 seconds + 1.6 seconds for the PS4. The longest of those 2 operations being the 2.3 seconds attributed to the CPU and thus the PS4's total time is 2.3 seconds just like the official site says (the 1.6ms GPU operation runs in parallel to the 2.3ms CPU operation and thus finishes before it, hence not contributing to the overall duration of the operation). Repeat for the XB1 numbers.
 
What? So I having no explanation has any bearing on how credible your claim is? I simply have no claim, because I don't know, and I'm calling you our on the HSA because I feel it's far fetched, and you have yet to give anything to support it.
He's raised a theory for discussing. Please discuss, or ignore.
 
The proposed explanation for those numbers that is under discussion is that Gamingbolts numbers are the breakdown of the overall CPU+GPU time which consists of two components which run in parallel.

So from the official site you have a TOTAL CPU+GPU time of 2.3ms on both consoles.

Gamingbolt breaks those totals down into their two constituent parts which run parallel to each other (again, assuming that assumption is correct). So you have 2.3 seconds + 1.6 seconds for the PS4. The longest of those 2 operations being the 2.3 seconds attributed to the CPU and thus the PS4's total time is 2.3 seconds just like the official site says (the 1.6ms GPU operation runs in parallel to the 2.3ms CPU operation and thus finishes before it, hence not contributing to the overall duration of the operation). Repeat for the XB1 numbers.

Your theory can't be true. Once again I have to say that there is noting exclusive to gamingbolt report except PC's GPU & CPU model.

As you can see gamingbolt just reported the results from official site, at first place. You can read exact quote from official site below (at Original Story section on gamingbolt):

“multi-core scaling and SIMD design (up to 70% of the instructions executed on a frame are SIMD). It is really fast – it can play 4K video frames (3840×2160) in 4 ms on PCs and 11 ms on Sony PS4 or Xbox One using the CPU only (or 1.4 ms and 2.3 ms using GPU acceleration)!”
After some confusion (gamingbolt thought that 1.4 ms is referred to PS4 and 2.3 ms is referred to X1 respectively) someone from official site contacted to them and tried to make things ok. Then something strange happened during this process, 1.4 ms changed to 1.6 ms for unknown reasons on their Update section.

We only need to do some comparison between first and second quote from official site, before and after correction, then every thing would be ok.

before:

“multi-core scaling and SIMD design (up to 70% of the instructions executed on a frame are SIMD). It is really fast – it can play 4K video frames (3840×2160) in 4 ms on PCs and 11 ms on Sony PS4 or Xbox One using the CPU only (or 1.4 ms and 2.3 ms using GPU acceleration)!
after:

multi-core scaling and SIMD design (up to 70% of the instructions executed on a frame are SIMD). It is really fast - it can play 4K video frames (3840x2160) in 4 ms PCs and 11 ms PS4/Xbox One using the CPU only (or 1.4 ms PC and 2.3 ms PS4/Xbox using GPU acceleration)!
Please read the bolded sections carefully.
 
Last edited by a moderator:
Your theory can't be true. Once again I have to say that there is noting exclusive to gamingbolt report except PC's GPU & CPU model.

Okay, so where on the original site does it state 1.3ms for the CPU component of the PC CPU+GPU time?

And where does it state 1.6ms for the GPU component of the PS4 CPU+GPU time.

This is new information.

As you can see gamingbolt just reported the results from official site, at first place. You can read exact quote from official site below (at Original Story section on
Please read the bolded sections carefully.

I've already read and fully understood the sections you quoted. You haven't read and/or fully understood my previous post so please go and take another look.
 
Status
Not open for further replies.
Back
Top