Wii U hardware discussion and investigation *rename

Status
Not open for further replies.
No need to do that, just briefly answer the dude's questions if you have the answers. This IS a discussion board after all.

The problem with that is that I've posted in response to someone else and then he started to argue with me, without doing his reading. I think I was more than reasonably patient with him.

Also, when I want to take part in a discussion, first thing I usually do is to see what it is about...
 
We have the GPU's approximate tech, some sort of die size, clock speed, and power draw. Wii U is basically achieving what PS360 are achieving in half the watts. That's what the technology and efficiencies are bringing to Wii U - not extra performance and abilities. We can be pretty certain of the limits because 1) devs have told us, 2) the games aren't showing any clear advantage, 3) the tech details tied with the laws of thermodynamics tell us there's only so much you can do with 40 watts of power, 4) It's been Nintendo's modus operandi due to its success to use simpler, cheaper tech in their products, 5) other aspects of the machine are poorly engineered and implemented, which doesn't indicate fabulous design within.

All of these points have been discussed. You can agree or disagree, but I for one aren't going to repeat the discussion. ;)

I agree what your saying completely about the Wii U except that a game studio named FrozenByte say that there is "a lot of untapped potential". Here is an article about it a summarized interview with this for the answer: http://gamersyndrome.com/2013/wii/wii-u/trine-2-developer-the-wii-us-untapped-potential/.

If this is true, then we might be able to see more out of the Wii U. If not, then your post still stands.
 
I don't see the point however in berating him for not backtracking through what can easily amount to dozens of pages of old posts. That's not the point of a discussion.
But surely it's the basis of an informed discussion? This is a technical board, and calling people with more technical experience or a different opinion than you 'biased' isn't ... productive. Netiquette is learn first, then talk. I don't consider even dozens of pages a big obstacle to learning. SoreSpoon is welcome to think WiiU is more powerful than PS360, but that doesn't require constantly making it the last word. Why not wait til we see more impressive games instead of bickering about untapped potential?

If the sticking point is whether "a game build from the ground-up exclusively for Wii U could match or exceed the best of current-gen without sacrifice," then what's the big deal? Just look at current ports to see that WiiU is roughly comparable at launch, so "matching" (insofar as exclusives aren't necessarily directly comparable in art style and technical choices [60 vs 30fps, etc]) shouldn't be a big ask. Precisely how much it can "exceed" might be interesting given actual game examples. I don't know how to approach "without sacrifice."

As far as ports being "the best they possibly could have been," I'm pretty sure compromise is a key word in porting among three somewhat distinct CPUs within a schedule.

SoreSpoon acknowledges WiiU's CPU has "bottlenecks" (without "truly knowing" enough to make a "definitive statement," curiously) but continues to antagonize Laa-Yosh over where exactly WiiU falls on the 1-5x PS360 scale. This seems pointless. Can't we agree it's in the ballpark and wait for some games to better glean where?
 
In that case they could solve it just by resorting to 1/4 resolution alpha buffers. Which is why I think it's more about the particles themselves, they could be quite expensive to calculate and writing a completely GPU-based system may not be a viable solution, either because of technical or budget restrictions.

Wouldn't a 1/4 resolution alpha buffer on a 870 x 720 game look awful though? It's not just alpha buffers, the Wii U also has at least two games that have removed trees and other foliage presumably in an effort to reduce transparent overdraw. This also fits with a ROP bound scenario but not with the CPU bound particle system scenario.

I'd bet big money (maybe as much as £5) that there's yet more hidden wiiksauce in the Wii U, and it's got something to do with ROPs / edram BW.
 
Wouldn't a 1/4 resolution alpha buffer on a 870 x 720 game look awful though? It's not just alpha buffers, the Wii U also has at least two games that have removed trees and other foliage presumably in an effort to reduce transparent overdraw. This also fits with a ROP bound scenario but not with the CPU bound particle system scenario.

I'd bet big money (maybe as much as £5) that there's yet more hidden wiiksauce in the Wii U, and it's got something to do with ROPs / edram BW.

Yes indeed, 1/4 resolution on 880x720 is pretty crappy, which is why all Black Ops 2 versions ran full resolution particles. There's a cost in downsampling the depth buffer and then upsampling the particles to the final buffer, which has to happen even with 1 particle on screen, so bottom line it ends up mostly a wash and the loss in visual quality is just not worth it.
 
Pardon my ignorance, I've read the past 20 pages but couldn't find an answer to these three questions.

The tear down revealed the Wii U's utilizes DDR3 1600 each on a 64bit bus, and it has 4 512 megabyte modules. The theoretical maximum bandwidth of each ram module is around 12GBps.

Could the Wii U's memory controller not also utilize dual or even quad channel? Pair the memory modules together to help improve bandwidth?


Also is it not possible for the memory controller to be reading from one module while writing to another?

Finally, is the Wii U's memory controller integrated into the GPU's die? I assume Nintendo have had AMD modify the GPU's memory controller to suit the Wii U's architecture.

Pardon my ignorance, I have a moderate understanding of memory latency, timings, and bus from experience overclocking. But my understanding of memory controllers is not that strong. Also apologies if these questions are stupid, or have been answered else where.



Finally, could the memory controller also not possibly be capable of reading from one module while writing to another?

I have a limited understanding of memory timings based on overclocking, but when it comes to memory controllers and the possibilities i'm rather uneducated.

The tear down revealed the Wii U's utilizes DDR3 1600 each on a 64bit bus, and it has 4 512 megabyte modules. The theoretical maximum bandwidth of each ram module is around 12GBps. .

Did a bit of re-reading and realised the bus for each module is only 16bit. With all 4 combined the Wii U has a 64bit memory bus.

I would re-edit my post but I don't have the option to do that....(wtf!)

None the less my questions regarding the memory controller still stand.
 
You don't have the option to edit your post because you're a new user, you gain the ability after 10 posts or so.

Yes, it's 12GBps for the 64bit configuration, this is the rating you see on a 64bit memory module for a PC : PC10600, PC12800 etc.
What we have is a single 64bit channel, more channels aren't physically there. i.e. this looks like an AMD E-350 or Atom, single 64bit channel whereas the higher end platforms (socket 1155, AM3, FM2 etc.) have two 64bit channels giving a 128bit total width.

The Wii U could have had 4GB ddr3 on 128bit, it would have been slightly more expensive and much better but instead Nintendo pursued low cost, size and power.
 
I know all of this. I'm more wondering about how we're so certain about the GPU's limits and how we know that the bottlenecks don't affect the GPU. Also, I'm wondering why the GPGPU and DSP can't possibly at least partially make up for the CPU bottleneck. I'm not expecting some kind of magic that makes the system 5x current-gen or something ridiculous like the fanboys; I'm just wondering if, possibly, a game build from the ground-up exclusively for Wii U could match or exceed the best of current-gen without sacrifice.


At a guess I'd say it's certainly possible, though we have not seen it yet obviously.

That is working on the assumption that the GPU is still better than PS360 GPU.
 
yeah after reading up I realise my question was stupid in the end.

4 memory modules with 16bit interfaces = total of 64bit bus

I was stupidly thinking this thing had 4 modules each with 64bit bus and the 12.8gbs was for each.

Hmm Nintendo must have pulled of someTHING pretty amazing to overcome this limited bandwidth configuration. They've traditionally been pretty crazy about ensuring their console's memory configurations have ample bandwidth and latency, but this would take quite an engineering feat.

I guess the eDRAM and CPU cache are meant to offset the limited mem1 bandwidth.
 
Yes indeed, 1/4 resolution on 880x720 is pretty crappy, which is why all Black Ops 2 versions ran full resolution particles. There's a cost in downsampling the depth buffer and then upsampling the particles to the final buffer, which has to happen even with 1 particle on screen, so bottom line it ends up mostly a wash and the loss in visual quality is just not worth it.

Thanks.

It's interesting to find that in addition to the loss of IQ that the performance gain from low res alpha isn't necessarily worth it. I did think that a flexible and high quality hardware scaler to work on buffers or textures during a copy or resolve in a next gen console might be a good idea.

Hmm Nintendo must have pulled of someTHING pretty amazing to overcome this limited bandwidth configuration. They've traditionally been pretty crazy about ensuring their console's memory configurations have ample bandwidth and latency, but this would take quite an engineering feat.

You're spinning this hard. AMD's APUs give the Wii U a kicking without any edram and only a 128-bit bus. I'd be singing their praises before I started on the Wii U as being a miracle of engineering.
 
Personally, I think the fact that the Wii manages to deliver something on par enough to the PS3 and 360 at half the power consumption impressive enough in its own right.

I don't think it makes any sense to then speculate how secretly Nintendo managed to also make it full of 'hidden power'. The texture compression options in the Wii are a nice enough bonus that should probably save at least some bandwidth, and make it even more abundant in available RAM. But in the end, that means it is quite similar to what the Wii was versus the PS2, and that is basically completely as expected, I would say, in terms of what they were going for.
 
You're spinning this hard. AMD's APUs give the Wii U a kicking without any edram and only a 128-bit bus. I'd be singing their praises before I started on the Wii U as being a miracle of engineering.

I'm not 'spinning' anything.

All im saying is I hope the Wii U has a way to compensate of the lack of its MEM1 bandwidth, be it eDRAM, cache, or any other factors. As with bandwidth that limited I'd say it's really going to struggle into the future.

Quite disappointed really. How much extra could it really have cost Nintendo to get 128 or 256bit bus going. Matter of dollars I would think. I don't get why the hell they would cheap out this much.
 
All im saying is I hope the Wii U has a way to compensate of the lack of its MEM1 bandwidth, be it eDRAM, cache, or any other factors. As with bandwidth that limited I'd say it's really going to struggle into the future.

I'm inclined to think that the main memory bandwidth is representative of the system as a whole rather than being a particular standout problem. Just going by the data on the chips, the Wii U has 57% of the 360's rated BW. But that's probably not the whole story, and any number of factors could potentially lessen the impact of this. Perhaps:

- The CPU simply can't eat through as much as much data
- The memory controller is more efficient
- The texture cache is more efficient
- There's no need to copy buffers into main memory...
- ... which would save some blending / compositing BW
- Scanout is done from edram
- etc

The Wii U is probably both more efficient and, in some ways, less powerful than the PS360 meaning it can make do okay with less.

Quite disappointed really. How much extra could it really have cost Nintendo to get 128 or 256bit bus going. Matter of dollars I would think. I don't get why the hell they would cheap out this much.

I think Nintendo decided that they needed a system that didn't look obviously worse than the PS360 and then set about building that system as cheaply as possible. And they seem to have achieved that pretty well (and at half the power consumption). If something is surplus to requirements, like a 128-bit bus, then I guess a hard nosed business man or woman will do without it even if it's cheap and seem like "good value" to us.
 
That's true. Nintendo ain't never shrunk nuthing, but there's always a first time.

I did think about suggesting a 96-bit bus last year around E3, but then decided not to say anything in case it was considered trolling. Little did I know that Nintendo were already planning to troll their hardcore fans with a 64-bit bus.
 
Personally, I think the fact that the Wii manages to deliver something on par enough to the PS3 and 360 at half the power consumption impressive enough in its own right.

I think that's a testament to how inefficient PS3 and XBox 360's CPUs are. The XBox 360 cores and Cell PPE run at this huge clock speed but get awful average utilization. A more conventional CPU at half the clock speed but with much lower taken branch penalty (on the PPE it's 8 fetch cycles!), branch misprediction penalty and ALU latency, wider issue-width, similar or better absolute L2 and main RAM latency, twice the SIMD width, and perhaps light OoOE could have easily performed as well but used much less power. Maybe it was believed that 8-wide SIMD was a lot harder to utilize than 4-wide, and everything else about the design was sacrificed to enable fast SIMD at high clocks. I don't know about efficiency/utilization on the SPEs, but those have their own problems.

These cores were designed around the same era Prescott was, and pushed pretty hard on frequency without much regard for efficiency..

With Wii U it currently looks like we have quite a bit less than half the clock speed and half as wide float SIMD (and no integer SIMD) leading us to the result we see.

- The CPU simply can't eat through as much as much data
- The memory controller is more efficient
- The texture cache is more efficient
- There's no need to copy buffers into main memory...
- ... which would save some blending / compositing BW
- Scanout is done from edram
- etc

You're missing a really big one: ability to texture from eDRAM. If the eDRAM is big enough and can both texture and input to the RAMDACs directly it could easily overcome the GPU's main RAM bandwidth limitations.
 
Status
Not open for further replies.
Back
Top