Wii U hardware discussion and investigation *rename

Status
Not open for further replies.
I wonder what they mean by "final format"?

Could changes to the WuPad be incoming? Or maybe a bigger case that allows for a real, man-size heatsink and fan (and higher clocks)?

Well there was this portion from a Game Watch article around E3.

http://translate.google.com/transla...s.co.jp/docs/series/3dcg/20110611_452478.html

By the way, say system RADEON HD4000, but does not have to RADEON HD4890 high-end RADEON HD4350 low-end AMD and it is beyond the 1TFLOPS performance arithmetic I had information from another system, in the specifications middle RADEON HD4000 series instead of the range class may be close to high-end systems based perhaps RADEON HD4800.

The only other GPU over 1TFLOP and not a 4890 is the 4870. Could explain the overheating dev kits and the need to underclock them.
 
I sure hope that they aren't really going to stick something similar to RV7x0 in their state of the art custom chip. What would be the sense of that when we've had two generations of technology evolution in the meantime? I will be amazed if it doesn't have cues from the GPUs launched late 2010 (Barts, etc) or from the Fusion IGPs.
 
I think the timing of first dev kit might be why they went with what they did. Apparently the dev kits started going out in January. When looking at their choice of GPU it seems their choices at the time were probably either the 4870 or 5770. I would assume they went with the 4870 since it edges out the 5770 in most benchmarks. The 6850 may have been too new for them to consider at the time since I'm sure they would be building the dev kits not long after the 6850 became commercially available.

I still feel that it will end up being VLIW4-based.
 
Why do the demos look so crappy if the thing is packing a 4870? That would blitz the current consoles a few times over.
 
Given the apparent delay and the "we've learned something from the 3DS" statement, what would be the chances of the Wii U getting a major hardware overhaul and coming up in Q3 2012 with a GPU performance comparable to at least a 600MHz RV770 (along with a non-bottlenecking memory configuration) and thus making B3D's speculation article horribly innaccurate?

They do say they'll show the console's "final format" during E3 2012, which means the dimensions will surely be different from what was shown in June, which in turn kind of nulls the "it can't take a 1TFLOPS GPU because it's too small" recurring argument.
 
Given the apparent delay and the "we've learned something from the 3DS" statement, what would be the chances of the Wii U getting a major hardware overhaul and coming up in Q3 2012 with a GPU performance comparable to

None to non-existent.
 
And it's still a VERY small box, comparable to a laptop in size, and those usually use heatpipes and whatnot to dissipate the heat effectively. And we all know that this console is not meant to be expensive to produce, thus heatpipes should not factor into this equation.

I really cannot see ANYTHING near a 4870, even at 45nm or lower, put into this box, especially not with high clocks. The mobile parts, on the other hand, like a 4650, might fit the bill.
 
The 360 and especially the PS3 would be smaller if you converted some of their internal parts to external and used slim optical drives. When looking at the MB of the PS3 Sllim, it could be smaller if it weren't accommodating that internal HDD.

I'm about 99% sure they won't be using a mobility GPU.

Also the early rumors were saying it would cost $350 to $400 to make.
 
I'm pretty confident that Nintendo is not using anything like the furnace that is Mobility 5870 (40nm 800SP). Not unless they want people returning consoles for melting in home theater cabinets and sounding like a hair dryer.

I really expect something like a 160-400SP GPU integrated with northbridge and other custom stuff. Hugely better than Wii and maybe better than 360/PS3 but they can still make money on it at $250 and put it in their goofy little white box. Of course it will be flattened by Sony and MS's next $400+ boxes in performance, but N is not after the same audience anymore IMO.
 
And it's still a VERY small box (...)

What are you talking about?
The old Wii U from E3 2011, reported to have heating problems or the future Wii U from E3 2012 that you probably haven't seen and therefore couldn't have the slightest idea on how big it'll be?


I'm pretty confident that Nintendo is not using anything like the furnace that is Mobility 5870 (40nm 800SP).

We both agree on that, but:

1) Mobility 5870 is Juniper @ 700MHz -> higher performance standards than the suggested 1 TFLOP (~600MHz), higher heat output too.

2) 1 billion transistor chip @ 40nm fabbed during ~Q1 2012 should have a substantially lower power consumption and heat output than an approximately sized chip fabbed during Q4 2009. Even more, 600MHz should allow for LP transistors AFAIK.

3) Transistor size for the GPU hasn't been disclosed yet.
 
I'm pretty confident that Nintendo is not using anything like the furnace that is Mobility 5870 (40nm 800SP). Not unless they want people returning consoles for melting in home theater cabinets and sounding like a hair dryer.

I really expect something like a 160-400SP GPU integrated with northbridge and other custom stuff. Hugely better than Wii and maybe better than 360/PS3 but they can still make money on it at $250 and put it in their goofy little white box. Of course it will be flattened by Sony and MS's next $400+ boxes in performance, but N is not after the same audience anymore IMO.

Iwata recently stated they are debating on whether or not they will sell Wii U at a loss on hardware. So I doubt it would be made to profit at $250.

I think the GPU will be a 12 or 14 cluster VLIW4 chip. Then when you consider the other components going into it I'm expecting it to sell for $349 with that amount possibly being a loss.
 
I think the GPU will be a 12 or 14 cluster VLIW4 chip. Then when you consider the other components going into it I'm expecting it to sell for $349 with that amount possibly being a loss.


Sigh, you know you're inevitably going to be proven wrong :LOL: Ahh well. No harm I guess. But I saw so many go through this whole denial routine with the Wii (huge firestorm directed at IGN when they revealed Wii was essentially an overclocked Gamecube).
 
Last edited by a moderator:
Yeah the Wii hype was disheartening. X1600 and a dual core. lol

Part of the deal with Wii is that they had the expensive control setup instead of a simple gamepad and were aiming at $250. That's going to be an obstacle again.
 
so my take on the new Wii CPU, based on today's mood of restraint. it has two cores with four threads per core, eDram L2, 2.3GHz clockspeed and a 32bit interface to ddr3 1600 memory, 512 or 1024MB.
it does chipset duties - it's a SoC.

GPU has a 64bit interface to 512MB of gddr5.
 
lherre already confirmed it is a 3-core processor with two threads per core.

Sigh, you know you're inevitably going to be proven wrong :LOL: Ahh well. No harm I guess. But I saw so many go through this whole denial routine with the Wii (huge firestorm directed at IGN when they revealed Wii was essentially an overclocked Gamecube).

If I'm wrong I'm wrong, but Wii won't be the indicator of that. You'll need more than that to say it won't happen though. Nintendo acknowledges the shortcomings of Wii and they are going to turn around and do the same thing? Nintendo is stubborn to a fault at times, but when they realize something is wrong they fix it.
 
Last edited by a moderator:
I think the GPU will be a 12 or 14 cluster VLIW4 chip.

They are likely aiming for getting a lot of ports of existing games from other platforms to provide initial content. (Which they can sell because they will look prettier than they did in the older consoles). This means it makes sense to make the system look technically as much like either of the older consoles as possible, only faster. This is likely the reason for the 3-core CPU. I think this is a good enough reason to rule out VLIW4 -- VLIW5 can run shaders hand-optimized for the 4simd+1scalar on 360 at full speed with minimal compiler trickery.

As for how fast they make it, remember that Nintendo actually has a very good reason to not make it too good: It is going to be the best console out there for a while. If it absolutely trounces the competitors, that might cause them to accelerate their schedules for the replacements, which shortens the time that WiiU will rule, and make it's best profits. 12 or 14 clusters sounds like 2 to 3 times too much -- it would be cool, but it would also be expensive, and wouldn't really buy them anything useful.
 
The first part is a very legitimate reason why they won't go that route. I still think Nintendo would be interested in the transistor reduction though.

The second part, not so much. If the competitors shortened their time, then do you think they'd still be able to make that huge jump over Wii U so many are expecting? That would be the best thing for Nintendo. An already established console that's close to the power of the other consoles released after it while most likely being cheaper.

Also I doubt it would be expensive since I'm essentially proposing a Lombok from the AMD 7000 series.
 
Also I doubt it would be expensive since I'm essentially proposing a Lombok from the AMD 7000 series.

I think so too. Probably not the GCN architexture but one from the performance segment which are said to be still based on VLIW4 but are also on 28nm, indeed Lombok Pro would be a good fit. Would also fit the time frame because they will start in December AFAIK.
 
What are you talking about?
The old Wii U from E3 2011, reported to have heating problems or the future Wii U from E3 2012 that you probably haven't seen and therefore couldn't have the slightest idea on how big it'll be?

So... you dismiss my speculation that the size of the box stays the same by speculating that the box will get way bigger? Kudos to such talk... really.
 
Status
Not open for further replies.
Back
Top