That's basically due to the refresh cycle isn't it? When it's on-die it becomes a big deal, but when it's not, the memory controller and signalling time takes over. I doubt the benefits of 1T-SRAM are that dramatic. You simply aren't going to have 20 cycle main memory accesses like you would in L2 cache. It's still probably a hundred cycles or more on average and I can't see this improving dramatically ever.
WiiGeePeeYou (Hollywood) what IS it ?
- Thread starter Megadrive1988
- Start date
- Status
That's basically due to the refresh cycle isn't it? When it's on-die it becomes a big deal, but when it's not, the memory controller and signalling time takes over. I doubt the benefits of 1T-SRAM are that dramatic. You simply aren't going to have 20 cycle main memory accesses like you would in L2 cache. It's still probably a hundred cycles or more on average and I can't see this improving dramatically ever.
I was thinking the exact same thing about the 1T-SRAM and where it was.
http://gear.ign.com/articles/713/713254p1.html
I guess it could be just for the 3MB but last gen they provided more. I was thinking they would provide the 24MB for GC compatibility and it's there for devs to use for Wii games as well. Is it possible the advances they talk about with this new SRAM are it's smaller and cooler?
http://gear.ign.com/articles/713/713254p1.html
I guess it could be just for the 3MB but last gen they provided more. I was thinking they would provide the 24MB for GC compatibility and it's there for devs to use for Wii games as well. Is it possible the advances they talk about with this new SRAM are it's smaller and cooler?
I still dont understand who they can assure 100% BC with GC using only GDDR3 given that the whole architeture would change a lot.
On the other side if the GPU had the 24MB then it would be (compared to Gekko/flipper size diferences) even bigger than it is when compared to the CPU (unless, of curse, the CPU had also a big upgrade)
BTW is just me or those chips (compared to his hands) seems much bigger than ones would expect to 90nm versions of the GC ones, and the CPU (even with the heat spreders) seems to not be much larger than we saw in the IBM video (leading me o think that the actual die size is almost as big as what we see in here)?
On the other side if the GPU had the 24MB then it would be (compared to Gekko/flipper size diferences) even bigger than it is when compared to the CPU (unless, of curse, the CPU had also a big upgrade)
BTW is just me or those chips (compared to his hands) seems much bigger than ones would expect to 90nm versions of the GC ones, and the CPU (even with the heat spreders) seems to not be much larger than we saw in the IBM video (leading me o think that the actual die size is almost as big as what we see in here)?
Last edited by a moderator:
Interesting responses - it didn't occur to me that the 24MB memory could be off-die but on the same package.
Alternatively, if the 24MB 1T is totally AWOL, will Wii be able to support widescreen progressive PAL with square pixels (approx 1024x576 pixels) in a 2MB frame buffer?
Alternatively, if the 24MB 1T is totally AWOL, will Wii be able to support widescreen progressive PAL with square pixels (approx 1024x576 pixels) in a 2MB frame buffer?
Size of the packaging is not directly related to die size.I still dont understand who they can assure 100% BC with GC using only GDDR3 given that the whole architeture would change a lot.
On the other side if the GPU had the 24MB then it would be (compared to Gekko/flipper size diferences) even bigger than it is when compared to the CPU (unless, of curse, the CPU had also a big upgrade)
On the Mosys Page there is TSMC 90nm Classic macro mentioned. It is 0.58 mm^2 big and has 1Mbit of memory. 24 Megabyte would thus be 111.36 mm^2 big. 1T-SRAM has a lot of build-in redudancy, so yields should be still pretty good. In their press release about 1T-SRAM-Q it mentions that it allows upto 256Mbit (32 MB) and that press release is from 2002 with 130nm. Flipper is supposed to be 120mm^2. At 90nm a combined Flipper and 24 MB 1T-SRAM Chip should be about 141 mm^2 big. Not small, but at 12mm x 12mm it should fit into that packaging and yields should be good because a huge amount of that die size is embeded 1T-SRAM with redundancy. 1/3 of the flipper die is 1T-SRAM too. A Hollywood chip like that would also explain why it took them such a long time to get it ready.
Not so fast. 1T-SRAM is only that dense on a dedicated process designed solely for 1t-SRAM. On embedded 1t-SRAM the density is about half that. http://www.cs.wisc.edu/~david/courses/cs838/notes/09-30-03.html
Wii draws 10W of power in standby mode and 17W when playing a game:
http://www.kotaku.com/gaming/wii/wiis-uses-about-onetenth-power-of-360-216051.php
X360 draws about 160W when running a game. So it does look like there's not much at all to that GPU.
http://www.kotaku.com/gaming/wii/wiis-uses-about-onetenth-power-of-360-216051.php
X360 draws about 160W when running a game. So it does look like there's not much at all to that GPU.
I don't think I've noticed dithering in Zelda yet. And it has been pretty dark. So maybe they fixed that.
I turned off WiiConnect24. Seems a bit pointless right now. There are basically no worthwhile channels yet. (Want Opera). So don't need the thing powered up 24/7.
I turned off WiiConnect24. Seems a bit pointless right now. There are basically no worthwhile channels yet. (Want Opera). So don't need the thing powered up 24/7.
Megadrive1988
Veteran
I cannot believe the 24 MB of 1T-SRAM is gone from Wii, no way. but I also don't believe it is actually on-die on the Hollywood as embedded 1T-SRAM. it seems it is just part of the package, and sitting right next to Hollywood. while a much smaller pool of 1T-SRAM is actually embedded, maybe the same 3.12 MB that was on Flipper.
so we have 24 + 64 MB of main memory split into two pools. And no slow auxillary DRAM.
so we have 24 + 64 MB of main memory split into two pools. And no slow auxillary DRAM.
Well, Mosys page about their Classic Macros doesn't mention anything about special mask steps required. http://www.mosysinc.com/classic_macros/
But it doesn't really matter, that additional mask can be used within a regular cmos logic process and very likely it already was used for Flipper, as that chip already contains quite a lot of 1T-SRAM. The modified CMOS process isn't like a dedicated DRAM process that is useful for DRAM but really bad for logic. 1T-SRAM is mainly supposed to be integrated into SoC applications, and it would hardly be useful for that task if you can't put serious amounts of logic onto the same die.
17w?!
What happened to the 50w Iwata talked about?.. Well I think now there's very little doubt that the system could have been much more powerful even in the same tiny casing.Also given another drop in process size soon this really makes me wonder what they could do with a new handheld with GC hardware, how much power does PSP use?
17w?!
Did Iwata mention a case size when he mentioned that wattage? Either way, it's would still be nothing but speculation. It's just like Shogmaster mentions; small, cheap, powerful, you can have two but not all three.
It is small, but it isnt anyhere near powerfull or cheap. Anyway they could easly put more 64MB or chips with twice the die size that wouldnt go anyhere near expensive or hot (meaning that could still small).
For what what we see here the console is so overpriced that it is even ridiculos.
http://www.winsupersite.com/showcase/xbox360_vs_ps3.asp
According to this side by comparison the Wii has 64+24+3 memory. Now they could be recycling rumoured specs but it's worth a look.
729 MHz IBM Broadway processor with 5 execution units
24 MB of system RAM (486 MHz) plus 3 MB of embedded DRAM (eDRAM)
64 MB GDDR3 RAM
According to this side by comparison the Wii has 64+24+3 memory. Now they could be recycling rumoured specs but it's worth a look.
729 MHz IBM Broadway processor with 5 execution units
24 MB of system RAM (486 MHz) plus 3 MB of embedded DRAM (eDRAM)
64 MB GDDR3 RAM
My comment on power consumption and case size was seperate from my question about Iwata's comments. Also there's not much point in that little catch phrase if you don't actually define the terms small, cheap and powerful. Not that it applies to my comments anyway considering I didn't mention anything about cost..
Last edited by a moderator:
Megadrive1988
Veteran
it makes absolutely no sense for the 24 MB 1T-SRAM to be gone. would that not destroy GameCube backwards compatibility ? even though GDDR3 memory can have more bandwidth, its latency is higher.
it makes no sense.
it makes no sense.
Well from a consumer standpoint in terms of cost vs. "power", you're certainly better off getting a 360 or maybe even a PS3 so I can't disagree with you there. But you have to remember, that both MS and Sony are subsidizing the true cost of their consoles.
Well that comment will always likely be true because it's relative to the other consoles out at the time. In other words, whatever technology MS has available, Sony & Nintendo have access to it as well. And of course vice versa. This is especially true since IBM/ATI did work for both MS & Nintendo. I.e. it's not as if Nintendo was going to get some sort of Alien technology or that MS was going to ask for an engineering disaster.
Not that it applies to my comments anyway considering I didn't mention anything about cost..
Fair enough. I just don't think it's safe to assume that Nintendo could have gotten a much more powerful console in that case.
- Status
