Wii U hardware discussion and investigation *rename
- Thread starter AlNom
- Start date
- Status
That's probably the lowest hanging fruit, just re-using the base RV740 design (although it's more hairy than that, of course). However, I'm having trouble seeing them provide even remotely adequate bandwidth for such a wide beast, irrespective of what they'd do with regards to memory.
i wonder if it could handle something like that old fake fan made "tech" demo of the revolution in 2005, it gave me chills at the time.
http://www.youtube.com/watch?v=2cbdJeE-x6E
http://www.youtube.com/watch?v=2cbdJeE-x6E
hmm... 80W under load @ 750 core. What's the power consumption like at say 600 or 650MHz (+ accompanying voltage adjustment).
I'd almost prefer 480SPs so they can still keep higher clocks for say... geometry, but who knows what the goal is exactly.
As we all know, anything above 320SPs already significantly outclasses current gen in shading ability, but if the clocks are gimped, then geometry rate suffers.Just out of curiosity...
From the RV740 article:
They'd fix that, right?
I hope. >_>
500MHz is what I've been told. Also, DDR3. But like I said: No idea how reliable that information is.hmm... 80W under load @ 750 core. What's the power consumption like at say 600 or 650MHz (+ accompanying voltage adjustment).
I'd almost prefer 480SPs so they can still keep higher clocks for say... geometry, but who knows what the goal is exactly.
DDR3 would explain quite a lot at E3.
500MHz sounds a bit surprising, and I'm sure that's still one of the unsettled specs. What was worrying was the reports of overheating dev kits. Granted, an RV740 shoved into that chassis would have really awful cooling compared to the stock PC card, but that's why I wanted to know what the power characteristics were for lower clocked iterations.
500MHz sounds a bit surprising, and I'm sure that's still one of the unsettled specs. What was worrying was the reports of overheating dev kits. Granted, an RV740 shoved into that chassis would have really awful cooling compared to the stock PC card, but that's why I wanted to know what the power characteristics were for lower clocked iterations.
He was talking about the devkit before the current one, and the way I understand some of the statements earlier this month, there have been quite a few (significant?) changes with the latest revision. And I don't even think developers have target specs for the final hardware, either.
What about 32nm?
It should be quite mature for years end and ATI knows how to make a GPU on that process too.
I heard the revised Wii U kits are not more powerful though, they are basically just bug fix hardware.
Anyways if this info is real, and it has a plausible sound, seems like a lot of shader power drastically crippled by a lack of bandwidth.
Still guessing RV730 here. 500 mhz RV740 is still 2.5X shader ability of Xenos, so why dont we see that in the early games or hear of it from developers? Unless the bandwidth is just shutting it down that much...anyways the 500 mhz RV740+DDR3 does at least sound plausible to me.
Anyways if this info is real, and it has a plausible sound, seems like a lot of shader power drastically crippled by a lack of bandwidth.
Still guessing RV730 here. 500 mhz RV740 is still 2.5X shader ability of Xenos, so why dont we see that in the early games or hear of it from developers? Unless the bandwidth is just shutting it down that much...anyways the 500 mhz RV740+DDR3 does at least sound plausible to me.
Last edited by a moderator:
Well, it could be just as custom as RSX was compared to G71 in terms of hardware blocks within it, but I'd wager some of the customization has to do with their plans for tackling the bandwidth issue, whether that be waiting on some wacky CPU eDRAM communication (in which case, waiting for IBM to finish their work) or the GPU's own integrated eDRAM or SRAM. *shrug*
Yeah I don't understand the grasping at 3 year old discrete PCIe PC GPUs. Just consider all the extra hardware that is in RSX, Hollywood, Xenos, NV2A, Flipper, etc. It also makes no sense to me that they would jump on a chip from 2008 as the foundation for their future.
itsmydamnation
Veteran
when factoring in cooling whats going to be cheaper, a smaller, higher clocked chip ( higher TDP) or a bigger lower clocked chip ( lower TDP). whats going to cost them more 40-80mm sq @40nm (number pulled out of my ass) or better cooling?
That's not what I meant. I'm fully aware of RSX's GPU being G7x. I'm referring to how these console chips have many extras integrated. For one, they all serve as a northbridge. I don't know what else is in Xenos and RSX but Hollywood has an ARM CPU for system and security tasks and an audio DSP.
NV2A was a GeForce 4 Ti but it still was also a northbridge. A super nForce IGP if you will.
The only off the shelf choice I could see Nintendo using would be Llano because of its already high integration.
I don't know much about those things, but assuming the system is indeed using DDR3, and considering POWER7 has on die DDR3 memory controllers, would it possible to connect the RAM to the CPU and CPU and GPU via GX? And would that make sense? From what I understand, GX offers more bandwidth than any other bus, and POWER7's two integrated quad channel DDR3 memory controllers also achieve an impressive bandwidth of 100GB/s. Too complicated and expensive I guess?
There's an AMBA bus and an AMBA to AMBA bridge onthe Wii U GPU die, so there will certainly be an ARM core as well.
Last edited by a moderator:
What I don't get is why the GPU is not integrated as it is pretty tiny?
We know that the CPU will include "lot of EDRAM" (they did not state if it's all cache by the way) it could be a good idea to have the GPU use it (or most of it) instead of the CPU.
I don't bet on a powerful CPU for the same reason the article doesn't bet on a powerful GPU as no programmer in his right mind would say that a 2/4 cores power 7 CPU is in the same ballpark as say Xenon.
There is something fishy imho especially while taking in account the rumors about the RAM being DDR3. Nintendo will face bandwidth constrains (even with conservative hardware), so either they need two pools of RAMs or some extra EDRAM. I suppose that the IBM process allowing EDRAM is not their cheapest one, on top of it they would have either an external GPU with embedded EDRAM (which limit the possible providers to NEC, right?) and that's likely to come at a price too. Last choice like in Xenos they add an extra chip but I don't know to which extend this is practical assuming a RV7xx derivative, the RBEs are tied to the L2 right? So the link between the GPU and an hypothetical daughter die would have to support coherency, right? (either way they are redesigned...)
Honestly I hope we learn more soon as it is the whole thing as guesstimated currently looks like a pretty non optimal set-up even considering Nintendo performances targets (I guess slightly above ps360).
Last edited by a moderator:
TSMC can fab eDRAM as well. It was actually cheaper for Microsoft as NEC was charging more IIRC.
- Status
