• 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.

WiiU "Latte" GPU Die Photo - GPU Feature Set And Power Analysis

Status
Not open for further replies.

pulsemyne

Member
I just cannot see this being based off the RV770. It seems far closer to the HD5550 in every way. Prehaps the reason the RV770 was used in dev kits is because the deisgn of the WiiU GPU with its "special sauce" add on was quite simular.
 
Considering Marcan calls that pool "MEM0", that might actually be what it is in native mode. That would suggest that the pool is freely accessible.

Yep. They may have needed it for Wii BC and decided to just give devs access to it anyway, giving them a neat place to store their gamepad FB.

I agree it probably is accessible. It is, however, somewhat puzzling that they would necessitate its presence for BC when they still have 8 MB left over in the main pool. It's got to come down to latency, I suppose. Same with the SRAM. That would explain the 32 macros. Hey at least we get real SRAM this time around and not that phony bologna MoSys stuff.

(j/k the 1t-SRAM served admirably for many years)
 

IdeaMan

My source is my ass!
For what it's worth:

from what i've heard from a few developers, and the shin'en interview could hint at that also, it doesn't seems that the GPU is like 50 or 60% programmable shaders for a "meh raw power results" + 40 to 50% fixed functions that could add some magic into that. The hypothetic fixed functions should be in clear minority compare to the programmable part, maybe those are just for Wii BC.
 
For what it's worth:

from what i've heard from a few developers, and the shin'en interview could hint at that also, it doesn't seems that the GPU is like 50 or 60% programmable shaders for a "meh raw power results" + 40 to 50% fixed functions that could add some magic into that. The hypothetic fixed functions should be in clear minority compare to the programmable part, maybe those are just for Wii BC.

Thanks for this. Maybe Nintendo aren't as crazy as we all imagine. Seems alot of this chip does make sense taking into account the needs for Wii BC and streaming a separate image to the Gamepad.

I wonder what some of the other doubles in the diagram might be. Is a dual geometry engine too much wishful thinking? Dual tessellators?
 

Schnozberry

Member
Not completely, no. But I do think if his theory were true there would be a bit of a lack of space for "other stuff" that seems to be necessary for a GPU, as the RV770 relative size of components shows.

Do we know that the "uncore" stuff scales linearly? Is there more logic for certain core components than others?
 

Thraktor

Member
Ok, re GPU blocks/processors, the following can all be assumed to be in there:

  • Command Processor and Thread Scheduler (not necessarily the same block)
  • Trisetup and rasterizer (R800 dropped that and delegated the workload to SPs)
  • Global Data Share (traditionally not very large, and likely encased nicely by some of the numerous embedded pools, in a much larger size)
  • A bunch of caches (vertex, texture) which could be really tiny or not so much (again, memory pools ahoy)
  • DMA engines
  • Ring buses
  • Tessellator (likely still sitting in fixed-function silicon)

BTW, a quick google for ARM9 die area got me to this article discussing a Qualcomm broadband/app processor (yes, ARM9 stand-alone dies are a bit hard to track these days), which appears to be ~0.8mm2 @40nm (the original part is 90nm, so I've applied a squares rule).

Is a global data share really necessary when you're got 2 pools of eDRAM and a pool of SRAM on there? I assumed that it existed on GPUs to avoid a (long) round-trip through DDR, so I figured they might simply do away with it when there are other suitable memory pools. (Unless one of them, such as the SRAM, is the global data share).

Could V be vertex/texture caches? Lots of SRAM, and it's located by the DDR3 interfaces.
 

ozfunghi

Member
Yeah, but comparing to the size of the whole chip (including IO) doesn't make as much sense for our purposes as comparing to just the SIMD+TU block, which I did.

Sorry, hadn't refreshed my browser when i responded so i didn't see the last page and your images.

If like Ideaman & FourthStorm are saying, the "only/main" fixed function parts are mainly for WiiBC... how much can we realistically expect those to help out in Wii U games? I mean, Wii's GPU was +/- 1/20th as performant as the 360 GPU of which we know the Wii U is already 50% more potent. That would put the entire Wii GPU at 1/30th of WiiU's GPU... that doesn't really look like it will make much of a dent or a difference for new Wii U games? Simplistic logic, but you get the gist of what i'm saying.
 
Thanks for this. Maybe Nintendo aren't as crazy as we all imagine. Seems alot of this chip does make sense taking into account the needs for Wii BC and streaming a separate image to the Gamepad.

I wonder what some of the other doubles in the diagram might be. Is a dual geometry engine too much wishful thinking? Dual tessellators?

Dual tessellators? I like the sound of that.
 

Schnozberry

Member
Thanks for this. Maybe Nintendo aren't as crazy as we all imagine. Seems alot of this chip does make sense taking into account the needs for Wii BC and streaming a separate image to the Gamepad.

I wonder what some of the other doubles in the diagram might be. Is a dual geometry engine too much wishful thinking? Dual tessellators?

Shin'en said they would use tesselation in their next game, so it does exist in some form. I'm wondering if Thraktor might be onto something with the assymetric shader design. I'm wondering if Nintendo added some extra redundant hardware of lesser capacity, for gamepad streaming if nothing else.
 
Sorry, hadn't refreshed my browser when i responded so i didn't see the last page and your images.

If like Ideaman & FourthStorm are saying, the "only/main" fixed function parts are mainly for WiiBC... how much can we realistically expect those to help out in Wii U games? I mean, Wii's GPU was +/- 1/20th as performant as the 360 GPU of which we know the Wii U is already 50% more potent. That would put the entire Wii GPU at 1/30th of WiiU's GPU... that doesn't really look like it will make much of a dent or a difference for new Wii U games? Simplistic logic, but you get the gist of what i'm saying.

They could always assist in rendering a discreet image to the Gampad, I suppose....
 

Thraktor

Member
I've added this quote to the OP, as it's relevant to what Wii BC hardware might be in there:

Shiota said:
Yes. The designers were already incredibly familiar with the Wii, so without getting hung up on the two machines' completely different structures, they came up with ideas we would never have thought of. There were times when you would usually just incorporate both the Wii U and Wii circuits, like 1+1. But instead of just adding like that, they adjusted the new parts added to Wii U so they could be used for Wii as well.

(It's obviously not referring to the CPU, as that's a matter of adjusting the old parts so they could be used for Wii U, not vice versa)
 

blu

Wants the largest console games publisher to avoid Nintendo's platforms.
Is a global data share really necessary when you're got 2 pools of eDRAM and a pool of SRAM on there? I assumed that it was there on GPUs to avoid a (long) round-trip through DDR, so I figured they might simply do away with it when there's suitable other pools on there. (Unless one of them, such as the SRAM, is the global data share).
Definitely. Though it would likely have its requirements how far it sits from the shader cores.

Could V be vertex/texture caches? Lots of SRAM, and it's located by the DDR3 interfaces.
Absolutely.

Let me emphasize this once again: If anybody has any doubts that memory is fully used outside of BC mode - don't.
 

Thraktor

Member
Definitely. Though it would likely have its requirements how far it sits from the shader cores.

Yeah, that's what puzzles me about the SRAM pool. It's in a very awkward place for components to take advantage of the low latency. It's also on the opposite end of the die from the DDR3 interface, which indicates it's not being used as a cache.
 
Yeah, that's what puzzles me about the SRAM pool. It's in a very awkward place for components to take advantage of the low latency. It's also on the opposite end of the die from the DDR3 interface, which indicates it's not being used as a cache.

Yeah, that one gets me too. Actually, it's right across from the CPU on the MCM...

So any idea when we're going to get CPU shots? Though I suppose those wont be as exciting

Nope. Give it some time. We're not even close to finished here yet!
 

Thraktor

Member
A fairly notable update to the OP:

Major update: Some comments from Chipwork's Jim Morrison (ie a professional who analyses photos like this for a living):

Jim Morrison said:
Been reading some of the comments on your thread and have a few of my own to use as you wish.

1. This GPU is custom.
2. If it was based on ATI/AMD or a Radeon-like design, the chip would carry die marks to reflect that. Everybody has to recognize the licensing. It has none. Only Renesas name which is a former unit of NEC.
3. This chip is fabricated in a 40 nm advanced CMOS process at TSMC and is not low tech
4. For reference sake, the Apple A6 is fabricated in a 32 nm CMOS process and is also designed from scratch. It’s manufacturing costs, in volumes of 100k or more, about $26 - $30 a pop. Over 16 months degrade to about $15 each
a. Wii U only represents like 30M units per annum vs iPhone which is more like 100M units per annum. Put things in perspective.
5. This Wii U GPU costs more than that by about $20-$40 bucks each making it a very expensive piece of kit. Combine that with the IBM CPU and the Flash chip all on the same package and this whole thing is closer to $100 a piece when you add it all up
6. The Wii U main processor package is a very impressive piece of hardware when its said and done.

Trust me on this. It may not have water cooling and heat sinks the size of a brownie, but its one slick piece of silicon. eDRAM is not cheap to make. That is why not everybody does it. Cause its so dam expensive
 

Schnozberry

Member
I think the Chipworks guy makes a mess of good points there. Whether you agree or not with Nintendo's direction in terms of performance, there is nothing about the Wii U that screams design incompetence whatsoever, despite the GAF meme.
 

EloquentM

aka Mannny
That's an interesting perspective. It's pretty devoid of technical jargon and it should give others perspective into how much custom hardware costs a company. It's not cheap.
 

Lonely1

Unconfirmed Member
A fairly notable update to the OP:

Major update: Some comments from Chipwork's Jim Morrison (ie a professional who analyses photos like this for a living):

I have been saying this since the size of the package was revealed. Wii U is not low tech or Nintendo cheaping out. It would have been cheaper to just stick a 4770 in there.

However, I totally understand the people who state that a stationary console doesn't need such constrains and would have preferred a bigger, power-hungrier, noiser console that provides higher grunt power.

Just a question: Is that fixed function thing true or just speculations ?

Speculation, most of the stuff is still speculation, but with different degrees of likeliness. :)
 
I think the Chipworks guy makes a mess of good points there. Whether you agree or not with Nintendo's direction in terms of performance, there is nothing about the Wii U that screams design incompetence whatsoever, despite the GAF meme.

But GAF told me its 2005 tech!

Kidding aside, people who are that desperate to claim ols old tech are either stoll angry that Nintendo won last gen or that they got Bayo2.

Even though its not a flops monster, it will still produce very neat graphics, Period.
 
But GAF told me its 2005 tech!

Kidding aside, people who are that desperate to claim ols old tech are either stoll angry that Nintendo won last gen or that they got.

Even though its not a flops monster, it will still produce very neat graphics, Period.

It's definitely impressively engineered.
 

z0m3le

Banned
But, it is based on the R7XXX series...

Why do we still think this? did R700 ever group 40 ALUs in a SPU?
llano is the first chip that comes to mind when I think of 40 ALUs per SPU. That is R800, and not all R800s did this. As custom as it is I am not sure it matters to say anything beyond it is VLIW5 and it sounds like there has been customizations done there too if they made it compatible at the hardware level to run Gamecube/Wii software.
 

Kenka

Member
Who would have thought that trying to figure out a GPU real capabilities would take so much efforts and money ? I did not know that when I started taking part to the conversations, hot damn.
 
Btw, if it is really custom and not based on any previous Radeon design... Isn't it uncomparable to any Radeon GPUS? We even know what Radeons the Durango and Orbis GPUS were based on...
 

Kimawolf

Member
So basically, to simplify this down, Nintendo paid for a design that would be able to produce modern graphics a bit better than this generation, and instead of putting money into the pure overall power of the console like Sony/MS, they put money into making something that would be super efficient and low latency at the cost of pure oomph? Basically instead of going a full on V12 Viper that has that max speed of almost 200mph they went with a Super efficient Tesla with a top speed of 120 mph? if we're using car analogies?

Is that the gist of it? Would that signal they are looking more towards the future, cramming such parts into a hand held eventually?
 

Lonely1

Unconfirmed Member
So basically, to simplify this down, Nintendo paid for a design that would be able to produce modern graphics a bit better than this generation, and instead of putting money into the pure overall power of the console like Sony/MS, they put money into making something that would be super efficient and low latency at the cost of pure oomph? Basically instead of going a full on V12 Viper that has that max speed of almost 200mph they went with a Super efficient Tesla with a top speed of 120 mph? if we're using car analogies?

Is that the gist of it? Would that signal they are looking more towards the future, cramming such parts into a hand held eventually?
If that was the aim I would think they would have gone for an ARM solution. But people far more versed in hardware engineering can answer you better.
 

Meelow

Banned
So in other words when devs figure out the GPU, it will produce graphically pleasing games that's better than 7th gen consoles?
 
So basically, to simplify this down, Nintendo paid for a design that would be able to produce modern graphics a bit better than this generation, and instead of putting money into the pure overall power of the console like Sony/MS, they put money into making something that would be super efficient and low latency at the cost of pure oomph? Basically instead of going a full on V12 Viper that has that max speed of almost 200mph they went with a Super efficient Tesla with a top speed of 120 mph? if we're using car analogies?

Is that the gist of it? Would that signal they are looking more towards the future, cramming such parts into a hand held eventually?

Maybe. Or they just wanted to make a tiny console to spend less money possible in plastics. It's an investment, like starting a gen with the slim version.
 

z0m3le

Banned
Because it is the case.

Just a question to you, but what does that even mean? The big changes R800 did was in registry memory iirc, it had better GPGPU and a gen 3 tessellation engine from AMD. Both of these things could be different in Wii U, and in fact I would imagine the registry memory is a given.

I can understand if there is a real genuine reason to call it R700 still, but with the customizations made, I still think calling it VLIW5 instead of R700 would be proper. However if you have some interesting reason you could explain to us, ah nm. I doubt there is some sort of thing you or I could point to that says this is R700 or isn't. Since R800 is also based on R700 even a customized R800 chip would technically still be correct to call R700.

Matt is an insider/developer... he'll have his reasons i suppose.

I know, but if his reasoning is because he was told so, or he actually knows why that is. Well it is honestly a very big difference.
 

Pociask

Member
But GAF told me its 2005 tech!

Kidding aside, people who are that desperate to claim ols old tech are either stoll angry that Nintendo won last gen or that they got Bayo2.

Even though its not a flops monster, it will still produce very neat graphics, Period.

I have been very frustrated that the meme has taken place that Nintendo is using old tech. It has been clear since launch that they were actually using very impressive, new tech - both for their GPU and for their tablet streaming technology.

That being said - the design decisions Nintendo made resulted in less graphical punch in the Wii U than other potential avenues. Way too much time has put into the calling Nintendo cheap, when the real line of inquiry is why Nintendo made the choices it did.
 

ozfunghi

Member
Yes, it is. But apparently the IP is not AMD's.


Ok, to put things in perspective... if the chip is that expensive, why did Nintendo go that route? Wouldn't it have been cheaper to take a stock 4770 (RV740?) and a Wii GPU with?
Also, the guy from Chipworks says it's an impressive piece of kit... in what sence? Low TDP? Performance per watt? The eDRAM? And is this still cheaper or more performant (or both) than going with stock chips + GDDR5.

And does his comment mean we can expect some "bonusses"?
 

guek

Banned
At 352 gflops you probably won't notice much difference. Though it'll still be nice to see Nintendo IPs in HD.

Hey, if you put on those Nintendo glasses, polish your iwata4ever pin, and straighten that pikmin tie, you just might convince yourself of a difference :-D

I don't think we're ever going to get games that blow this gen completely out of the water but I think we're going to say "wow" when E3 comes around. This thread just gives me more confidence in that.
 
I have been very frustrated that the meme has taken place that Nintendo is using old tech. It has been clear since launch that they were actually using very impressive, new tech - both for their GPU and for their tablet streaming technology.

That being said - the design decisions Nintendo made resulted in less graphical punch in the Wii U than other potential avenues. Way too much time has put into the calling Nintendo cheap, when the real line of inquiry is why Nintendo made the choices it did.

Indeed. The 1 MB SRAM is impressive. That's quite a bit more than the L1/L2 caches that most (if not all) retail cards w/ a similar number of shaders/TMUs carry.

Again, how devs utilize this memory subsytem will make or break their results.
 
Status
Not open for further replies.
Top Bottom