Xbox One (Durango) Technical hardware investigation

[Pixel]

There are still many unknowns, some absence of information, although Digital Foundry is working into it. The people who worked on the console have been very open to questions, -we don't know what DF staff asked anyways, so we can speculate to infinity and beyond- except for one thing.

(thanks to Javisoft for the picture)

I think Misterxmedia will have this on his blog.

 

[Bigus Dickus]

Oh no. I am not stating one peak is real while the other is imaginary. I am stating both are real but are only possible under specific circumstances and are not sustainable especially over a full second.

And that the numbers in and of themselves tell you nothing about how much bandwidth will readily be exploitable over 30 or 60 frames.

Ah, figured I was misreading something. We are in agreement!

 

[Rangers]

I think Misterxmedia will have this on his blog.

I know. It must be proof of super dGPU

 

[RudeCurve]

This from the digital foundry vs microsoft article,

"Sony was actually agreeing with us. They said that their system was balanced for 14 CUs. They used that term: balance. Balance is so important in terms of your actual efficient design. Their additional four CUs are very beneficial for their additional GPGPU work. We've actually taken a very different tack on that. The experiments we did showed that we had headroom on CUs as well. In terms of balance, we did index more in terms of CUs than needed so we have CU overhead. There is room for our titles to grow over time in terms of CU utilisation."

What do you guys think they mean by having headroom on CUs as well?

Are they implying that they arent using all the ALU on each CU for graphics thus having extra for compute?
Or are they saying they have set aside a certain number of CUs from the 12 just for compute?

Remember PS2's "untapped power"? Same thing...

 

[Cyan]

As a summary of the past year.

Microsoft stated:


Maybe not literally, but this has been their focus until E3 at least.

The only thing that has changed, is that MS overclocked the cpu by 10% and the GPU by 5 or maybe 6%. No other changes have been made.
They discovered that the ERAM was (roughly) twice as fast as previously imagined.

Aside from this, nothing has really changed, hardware wise.
So either their initial reveal was incorrect, or there is some serious back-pedalling going on.

Nothing changed at all if you think about it, especially when Microsoft openly admitted themselves a couple of days ago that the GPU of the PS4 has a little more grunt. They also talk about the numbers of the eSRAM in the same interview, so I think they are being pretty candid on the subject, and their numbers are right --not like peaks of bandwidth but actual average numbers.


http://www.nowgamer.com/news/209288...h_a_benefit_to_developers_says_microsoft.html


I mean, that's not what I would define as PR talk. They are defending their system and that's about it, I think. Pretty much typical stuff, defending what's rightfully theirs. Sure, maybe it isn't the most powerful console ever but it doesn't mean that it can't reach the glory of silicon paradise.

Personally, I liked the design since the very first leaks, I still like it and I always will!, no matter what a good part of the critical mass say on the internet.

I got a bit tired of rumours though and I can't wait for any NDA to be lifted and have a "normal" generation again and focus on games.

It was bearable at first but for me this was the most convoluted pre-next gen time I've ever had. I've grown slightly tired of it, arguing, etc. I just want to start playing next generation games, for the most part. Although technology is always fascinating.

 

[Cyan]

I think Misterxmedia will have this on his blog.

ha ha ha ha, would you believe me if I told you that I visited his blog yesterday? I must admit that I remembered 3dilettante's words about MS Paint for some reason and it brought me a smile. I also read an article there where he talks to an insider and the interview goes like:

Me: /question about the Xbox One/

Insider: Reply.

Me: .....

Insider: ....

Insider is one of my favourite terms in console gaming. I've met a few over the years.

Rangers, Odd, isn't it? I have seen a dog passing by 15 minutes ago with a dGPU hanging on his mouth, it could be Xbox One's.

 

[Rockster]

For a specific subset of a frame in real game usage sure, this isn't what your going to get on average over the entire frame.

I'm not clear with what the average over an entire frame has to do with the price of tea in china. The only bandwidth measure of substance is the frequency at which the lack of bandwidth, or just as importantly access latency, has stalled the execution of some processing unit. The processing bottleneck over the life of a frame will move between different components depending on the operation. (ie. at some point you'll be CPU bound, or ALU bound, or other component, during which GB/s of bandwidth serve no purpose). The whole idea is to have enough bandwidth to minimize access contention, and distributing work as evenly as possible across the system throughout the frame to make the most effective use of your hardware resources. That's why it's said that XB1 has more available bandwidth per ROP and CU relative to its competition, and general memory bandwidth numbers are moot. Memory clients being equal, the amount of bandwidth and lack of contention per client, would be more relevant as it influences their potential efficiency.

 

[MarkoIt]

This from the digital foundry vs microsoft article,

"Sony was actually agreeing with us. They said that their system was balanced for 14 CUs. They used that term: balance. Balance is so important in terms of your actual efficient design. Their additional four CUs are very beneficial for their additional GPGPU work. We've actually taken a very different tack on that. The experiments we did showed that we had headroom on CUs as well. In terms of balance, we did index more in terms of CUs than needed so we have CU overhead. There is room for our titles to grow over time in terms of CU utilisation."

What do you guys think they mean by having headroom on CUs as well?

Are they implying that they arent using all the ALU on each CU for graphics thus having extra for compute?
Or are they saying they have set aside a certain number of CUs from the 12 just for compute?

It means that most workloads are not ALU bound, so there is always a certain amount of CUs which will idle. They think that esram could be the key to achieve an higher utilization of this 12 CUs (i.e less latency than DRAM).

 

[shredenvain]

It means that most workloads are not ALU bound, so there is always a certain amount of CUs which will idle. They think that esram could be the key to achieve an higher utilization of this 12 CUs (i.e less latency than DRAM).

Thank you fot your reply. I also dont feel the df article was pr and the info they shared made complete sense to me.

 

[baten]

On the first slide of the mantle presentation is states clearly "Works with all GNC GPUs" so probably it works on both XBO and PS4

 

[Shifty Geezer]

Mantle is a PC tech. Discussion here.

edit - fix link

 

[french toast]

Just noticed this article and test done by digital foundry.
Tries a rough comparison between two pc amd gpus, 7850 7870xt both clocked at 600mhz to try and replicate the 50% compute difference on paper between orbis and durango.
http://www.gamefront.com/balance-or-power-comparing-xbox-one-and-ps4-hardware/
The test someone validated some of microsofts claims that increase in CUs doesn't translate into raw game performance directly.

Of course this is a very rough and raw example, both console hardware is very different to the pc setup used in those examples, with cpu likely to be more of a factor in frame rate in consoles not to mention the very different memory setups and co procesors...also xbone has been clocked higher to both the cpu+gpu as well as other internal modifications not yet known for both. ..so its not directly comparable.
-still it does show that both consoles are going to be very close and comparable, certainly we wont be seeing anywhere near 50/40 or even 20% difference some of us (me) expected.

Note : this was not meant to be a verses post rather a continuation of the xbone digital foundry article.
Also I havnt kept up with pages on this thread so if this has been discussed already in depth just ignor.

 

[AlNom]

Well, you do have to account for other parts of the GPU since CU count isn't the only thing that's different or contributing. It's also just an isolation of CU count vs current workloads.

 

[french toast]

Well, you do have to account for other parts of the GPU since CU count isn't the only thing that's different or contributing. It's also just an isolation of CU count vs current workloads.

Yes boh cards have 32 rops where as durango has only 16 compared to liverpools 32.
Hd 7870 has more tmus although im unsure of the amount both durango and liverpool soc (both 48?)
Liverpool soc contains 2 x the ACEs...although durango gpu is marginally higher clocked.

Durango has according to the df interview. ..200gb/s of 'real world' bandwidth and much superior latency, with the proviso that only a miniscule amount of its main 8gb has that available bandwidth. ..requiring alot of fine tuning and messing about.
Liverpool counteracts this by being easier to program memory usage. .With the full 8gb of memory available to chuck in and out of for an approx 150-160gbs of real world bandwidth.

Thats just the main high level differences of the gpu and memory set up between the two...if durango contains the same amount of tmus as liverpool, it will have a slight advantage in fill rate due to the upclock, liverpool will hold raw processing advantage and ease of development.

It wl turn out to be very close im sure.

 

[3dilettante]

Yes boh cards have 32 rops where as durango has only 16 compared to liverpools 32.
Hd 7870 has more tmus although im unsure of the amount both durango and liverpool soc (both 48?)
Liverpool soc contains 2 x the ACEs...although durango gpu is marginally higher clocked.

TMU count scales directly with CU count, since they are part of them.

 

[french toast]

TMU count scales directly with CU count, since they are part of them.

Thanks.

 

[dragonelite]

How important are the rops actually these days?
My guess was that both consoles have enough pixel fillrate to do 1080p@60fps
Given that forza 5 runs at those specs and with some Msaa(?). And it has been a very long time that i heard pixel fillrate being named for causing a gpu to not render fast enough.

 

[AlNom]

Shadows/alphatest/depth-only passes etc. could use more depth fill at least. *shrug*

 

[Cyan]

Shadows/alphatest/depth-only passes etc. could use more depth fill at least. *shrug*

In that sense a technology that could help the Xbox One to handle better shadows than its 16 ROPs would initially allow for is Tiled Resources, I think.

 

[Cyan]

Phil Harrison said fairly recently that it is all about having the best games and not gigaflops and teraflops, and he also mentioned some interesting tidbits about the console, saying it is still unfinished, both the hardware and the OS.

Developers aren't working with the finished product, Harrison insists.

Developers have been complaining about memory issues in the next console generation, and making unfortunate comparisons between the "weaker" Xbox One and its PS4 rival. Naturally Microsoft isn't about to take this lying down, and Phil Harrison has stepped forward with an important message: it's not about who has the most gigaflops or teraflops. "It's all about having the best games and having the most impressive experiences, and clearly Xbox One has the best games," says Harrison, an argument which, you may recall, Harrison has made once before and no doubt will make again.

But is Harrison about to concede the memory argument, with PS4 lording it over the Xbox One? Not quite; it's all very nuanced, says he. The developers aren't getting the full picture because they aren't playing with the finished product. "We can increase CPU and GPU capability on Xbox One," says Harrison, "so I think it's impossible to draw any conclusions from that based on pre-launch unfinished hardware and unfinished operating systems." Even having said that, Harrison continues to beat the Games Games Games drum. Content is king, and as Xbox One has the best content - claims Harrison - it will win out in any consumer contest with the PS4.


Reuters would seem to disagree with that argument - according to a recent poll, US consumers prefer the PS4 - but it's early days yet, and the only way to know for sure is to sell the product. Sony's hoping for record breaking sales, and pre-orders for the PS4 are extremely high. Sony's Fergal Gara is very pleased with developer feedback so far. "The overwhelming thing is that developers are delighted with PlayStation 4," says he. "There's no glass ceiling in terms of what [developers] want to do, and it's easy to do it." November's looming. It's time to see what this console generation's really made of.

http://www.escapistmagazine.com/new...-Harrison-Its-Not-The-Gigaflops-Its-The-Games