eight shader units for R520
- Thread starter kyetech
- Start date
Megadrive1988 said:now this conversation has moved beyond my ability to understand.
ah well, I'll just wait for ATI's R520 announcement, press release, interview, conference, demos, etc. it seems to me though, that ATI is likely to "own" the spring/summer and perhaps fall. while Nvidia works on its next-gen GPU ( fall 2005 unlikely, prolly spring 2006) as well as its contribution to PS3 GPU.
which part dont u understand, i can try to explain it in a non techincal way.
Hellbinder
Banned
With the added Complexity of Full SM3+ including pure FP32..
Try something more along the lines of This....
16 "pixel" pipelines
8 "vertex" Pipelines
And Something closer to 1ghz Core and maxed GDDR3.
Of course some internal improvements per clock. You are looking at something Quite a bit faster than the X800.
Thats just my opinion though 8)
Try something more along the lines of This....
16 "pixel" pipelines
8 "vertex" Pipelines
And Something closer to 1ghz Core and maxed GDDR3.
Of course some internal improvements per clock. You are looking at something Quite a bit faster than the X800.
Thats just my opinion though 8)
is it actually 4 RV410 cores combined into one!!?! 
The Baron said:is it actually 4 RV410 cores combined into one!!?!
HAHA.
What is the difference between a pixel pipeline and a "pixel" pipeline?
I don't think they'll leave their fate in the hands of mem manufacturers. Uber-fast memory will also require HQ on the board side and they've already had problems WRT this and R420.
So maybe, two full ALUs per pipe (x16) vs. a 24PS:16ROP NV47?
The Baron said:is it actually 4 RV410 cores combined into one!!?!
I think Hellbinder's is as good a guess as any, assuming "something closer to 1ghz Core and maxed GDDR3" means 650-750MHz and probably no higher.
Samsung have said 1GHz GDDR3 by year's end. All things considered, something clocked at 750/800MHz would consistently be ~40%+ faster than R420 and probably ~2x as fast in some bandwidth or shader-limited situations. I suppose the refresh will use GDDR4 and we might see quite a clock hike as they improve design for the 90nm node.
Didn't ATi say something like it'd take 25% more transistors just for FP24=>FP32 alone? ~250 million seems like a good guess, IMHO.
Samsung have said 1GHz GDDR3 by year's end. All things considered, something clocked at 750/800MHz would consistently be ~40%+ faster than R420 and probably ~2x as fast in some bandwidth or shader-limited situations. I suppose the refresh will use GDDR4 and we might see quite a clock hike as they improve design for the 90nm node.
Didn't ATi say something like it'd take 25% more transistors just for FP24=>FP32 alone? ~250 million seems like a good guess, IMHO.
I'll guess an even 300 mil, it has a nice PR-ring to it.
Sounds like it'll be damned quick though, will there be software out by then that can utilize it? (Hey, I'm having trouble finding things to max out my X800 still!)
Sounds like it'll be damned quick though, will there be software out by then that can utilize it? (Hey, I'm having trouble finding things to max out my X800 still!)
{Sniping}Waste
Regular
MuFu said:The Baron said:is it actually 4 RV410 cores combined into one!!?!
HAHA.
What is the difference between a pixel pipeline and a "pixel" pipeline?
I don't think they'll leave their fate in the hands of mem manufacturers. Uber-fast memory will also require HQ on the board side and they've already had problems WRT this and R420.
So maybe, two full ALUs per pipe (x16) vs. a 24PS:16ROP NV47?
Could we see a 512 bit bus to get the bandwith needed or better bandwith saving tech?
