NV33 Rumours, if anyone is interested
- Thread starter elroy
- Start date
BoddoZerg said:
Hmmm, I guess I shouldn't be so narrow minded when I think about people on this board/post this sort of stuff.
Back to topic - How many trannies to you think a 2x1, DX9 capable part would use? Do you think its feasible to use something like this in a tablet PC/PDA/Mobile Phone/GBA/Anything Handheld?
elroy said:Back to topic - How many trannies to you think a 2x1, DX9 capable part would use? Do you think its feasible to use something like this in a tablet PC/PDA/Mobile Phone/GBA/Anything Handheld?
whithout knowing any design specs or features its really hard to tell
i dont see why it wouldnt be feasible though
http://www.beyond3d.com/forum/viewtopic.php?t=3903&highlight=
She ( woah, didn't know it was a she... ) says directly in his first post starting the thread that both NV31 & NV34 are 4x1
Uttar
She ( woah, didn't know it was a she...
) says directly in his first post starting the thread that both NV31 & NV34 are 4x1Uttar
Uttar said:http://www.beyond3d.com/forum/viewtopic.php?t=3903&highlight=
She ( woah, didn't know it was a she...
Uttar
Yeah, like I said, neither did I. I hope none of this scares her off though. Her input to the board is invaluable. I think it is fairly safe to say that both NV31 and NV34 are 4x1, however, NV33 being 2x1 is a different story. I don't think we've ever heard this before......
Bigus Dickus
Regular
Uttar said:http://www.beyond3d.com/forum/viewtopic.php?t=3903&highlight=
She ( woah, didn't know it was a she...
Uttar
Yes, NV31 and NV34. This thread is about the NV33.
me too, i thought it was going to a joke about transsexuals.... was begginning to get very offended, good thing you didnt piss me offfMaverick said:elroy said:
For a fraction of a second, considering the subject of the previous few posts, I thought you meant something entirely different...
now then, anyone know when we could be actually seeing tablet PCs with PS2.0 hardware? Man I've been dreaming of a tablet PC with Photoshop for a long time and PS2.0 can be used to accelerate Photoshop 8)
Maverick said:elroy said:
For a fraction of a second, considering the subject of the previous few posts, I thought you meant something entirely different...
I didn't really think about this until I re-read this morning
. I'd had a few drinks when I typed that last night , and quite obviously couldn't be bothered writing the whole word. Anyone else on the board I can offend? I seem to be do a pretty good job of it so far (in this thread at least)JF_Aidan_Pryde
Regular
NV40 with eDRAM?
Highly doubt it.
Highly doubt it.
I think it's much more likely now than it was in the past. Of course, it couldn't have a huge amount of eDRAM, but even 2MB could be used to great effect as a cache. Anyway, you are probably right. But the time for large-scale memories embedded into processor cores is coming soon.JF_Aidan_Pryde said:
I could be wrong, but AFAIK texture caches are tiny compared to cache sizes cpus have. Around 3/4 of all tranistors on a cpu are for caches, but the number is much lower on gpus. I don't have an estimation, however, public information about gpu internals is not very detailed.Pete said:
eDRAM would be used for the frame buffer, not as cache (at least I think that was the idea on a bitboys design?). Of course, you can only put the whole frame buffer into this eDRAM if you have enough of it... 16MB might not be enough for FSAA, though you could probably only put parts of the frame buffer into this area.
arjan de lumens
Veteran
I don't think we will see large-cache GPUs anytime soon. For CPUs, it seems to me that the CPU makers are more or less running out of other ways to increase performance and therefore as a last resort just keep piling on more and more cache; until GPUs hit the memory 'bandwidth wall' (far off, as far as I can tell) you can always increase GPU performance by adding extra pipelines/processing units/memory controllers, helping performance more than large caches would.
AFAIK, GPU caches are mainly intended to buffer against memory latency and avoid reading data multiple times; as such, they don't need to be very large to be effective - after a few kilobytes, the added performance of extra cache tapers off too rapidly to be worth the added cost.
AFAIK, GPU caches are mainly intended to buffer against memory latency and avoid reading data multiple times; as such, they don't need to be very large to be effective - after a few kilobytes, the added performance of extra cache tapers off too rapidly to be worth the added cost.
