bgassassin
Member
Some nice new tid-bits of info and speculation, but bgassassin, seeing as you're the most reliable in my eyes, what are you looking at for the Wii U's final specs, taking on board the alleged alpha kit specs?
Also, analogue triggers eff yeaaah. Also, these FPS games mentioned... new IP's? Regurgitated current gen ports? Future gen cross-platfrom titles?
Pretty much my first speculation just with some tweaks.
CPU
Tri-core (POWER7-based cores) @ 3+Ghz
3MB L2 cache (split 1.5MB:768KB:768KB; maybe they'll increase this before launch)
GPU
600-800Mhz
640-800 SPUs
(I really think Nintendo is trying for a 28nm GPU)
Memory
32MB of embedded 1T-SRAM-Q
1.5GB of GDDR5 memory
Misc
6x-10x slim optical drive
8-16GB Flash memory
~1 USB 3.0 port (for HDD, hopefully)
Anybody working in retail know how this works?
The came out on the 19th of November.
Are there any other times in the year that retailers prefer to receive new products?
Im still thinking Nintendo will repeat what they did with the Wii and release in November. maybe capitalizing on Black Friday to create scarcity hype.
I would assume it deals with allocating product space for the Holidays. They would want to have something like that in order before the shopping period officially starts which would be Black Friday.
And we don't really all know what they are talking about. Their press release is vague. They say merely that the CPU is "Power-based" and they never explicitly mention L3 cache. Now, even if you choose to ignore the rumors, common sense dictates that the chip IBM make for Wii U is going to be a "bastardization" of Power7. But compared to a bastardization of a Power6, that ain't bad. Some of Power7's features apparently have little benefit for gaming anyway.
Anyway, I did a little searching and came up with this:
http://arstechnica.com/old/content/2007/02/8842.ars
This describes eDRAM being used as L2 cache rather than L3 (it also seems to align with a quote from wsippel's source concerning a "metric ton" of L2 cache. The rumored 3 MB is quite a bit compared to the 360's 1 MB.)
This link also provides an interesting tidbit in that IBM claims its eDRAM is almost as fast as SRAM. The latency is also super-low at 1.5ns (the 1t-SRAM on flipper, for a comparison, has a latency of 6.2ns). So perhaps the 1t-SRAM has been surpassed on some fronts.
The PowerPC A2 has 8MB of L2 cache and I believe that is the same eDRAM that makes up the L3 cache in POWER7. And I hope that 1T-SRAM has improved since Gamecube (for Mosys' sake).
Thanks for your input. I don't think it's unnecessary, however, if it's being used as an I/O processor. The Playstation 2 did this with the PS1's cpu in order to achieve perfect BC while also performing a necessary function for the current system. The Wii U will need something like Starlet but probably much stronger. So they could choose a shrunken Broadway, which for this purpose is pretty powerful and also familiar, or they licence something from ARM, which will also come with a cost. Having Broadway on the GPU die also might line up with the rumors about a SoC from wsippel's source and would eliminate the need for some kind of bus which connects the Wii U cpu to the theoretical pool of eDRAM/1t-SRAM on the GPU for BC. When the system goes into Wii mode, it could pretty much shutdown the Wii U tri-core CPU completely and run everything off the GPU SoC. That's sheer speculation right there, but it seems to be a logical possibility.
Wsippel was told an ARM is being used for I/O again and that's why I've been linking to that rumor about Marvell's ARM processor being in Wii U. And I'm sure blu would be "happy" to hear that I'm jumping on the software emulation bandwagon.

Also I think the SoC wsippel's contact mentioned will be a GPU LSI/MCM like with Wii.
Yes. GDDR3 on a 256-bit bus makes no sense at all.
I really wonder if that RV770LE is still present in the current devkits. If final hardware isn't there yet, it means the Wii U announcement could've been a greater rush than expected...
All things considered I still believe there is merit to it being a 4870 that was severely underclocked in the beginning. Lherre said early on every time they would push the kit the GPU would freeze. I have a tough time now believing a GPU that is normally clocked at 575Mhz and downclocked to 500Mhz would cause the system to freeze when being pushed. Not saying it can't happen, just harder to believe. Then you have this old quote from ShockingAlberto.
That probably helps cause the hard-locks, now that I think about it.
On the positive side, I have heard that (aside from the above mentioned issues), the new kits are a decent bump from the early summer ones.
Increasing a 4830 75Mhz doesn't sound like a decent bump since lherre said nothing has changed. Increasing a 4870 (or 4770) 250Mhz would be a decent bump IMO. So for me looking at the dev kit, if we're looking at 640 ALUs, then I give more credence to a 4770 than a 4830. If we're looking at beyond 1TFLOP then it's got to be a 4870.