Official 9800pro specs "leaked" by FIC
- Thread starter Hellbinder
- Start date
I'd like to see the Truform bugs worked out of the metal.
If R400 is delayed until next year, I'd also really like to see a .13 9900 Pro running at 500/500 for this fall.
If R400 is delayed until next year, I'd also really like to see a .13 9900 Pro running at 500/500 for this fall.
well we should have news from ati soon, so lets not waste too much time guessing whether or the press realease is correct or not. Ill believe the press realese until told its not true from more reliable sources. 
later,
later,
MuFu: Well, if it's DDR-II, then it's probably 460Mhz
You said earlier that there was a second card named Radeon 9800 Pro, with 256MB DDR-II and slightly different clocks.
Could it be that FIC is only selling that one, and not the "lesser" Radeon 9800 Pro?
It would make sense. After all, FIC is announcing that now, but we don't know when it'll be available. And we don't even know when ATI will announce it!
Of course, the Radeon 9800 Pro "regular" will probably have the +- same core clock, so about 400Mhz would also make sense.
Or it could be 360Mhz DDR-I.
I don't think it could be 400Mhz DDR-I, because that would give an unbalanced solution, unless ATI's pipelines have suddently become much more efficient. Sure, they're more flexbile, but we didn't hear *anything* about efficiency ( beside Doom 3 optimizations, of course - but it wouldn't warrant so much of a ratio change! )
Uttar
You said earlier that there was a second card named Radeon 9800 Pro, with 256MB DDR-II and slightly different clocks.
Could it be that FIC is only selling that one, and not the "lesser" Radeon 9800 Pro?
It would make sense. After all, FIC is announcing that now, but we don't know when it'll be available. And we don't even know when ATI will announce it!
Of course, the Radeon 9800 Pro "regular" will probably have the +- same core clock, so about 400Mhz would also make sense.
Or it could be 360Mhz DDR-I.
I don't think it could be 400Mhz DDR-I, because that would give an unbalanced solution, unless ATI's pipelines have suddently become much more efficient. Sure, they're more flexbile, but we didn't hear *anything* about efficiency ( beside Doom 3 optimizations, of course - but it wouldn't warrant so much of a ratio change! )
Uttar
Although a 460 MHz memory clock sounds high, it's not impossible for a high end product if keeping the performance crown is all ATI is thinking of.
Hynix should have it ready by now:
http://www.hea.com/PRESS/1209_02.htm
Hynix should have it ready by now:
http://www.hea.com/PRESS/1209_02.htm
Tagrineth said:I dunno, for all we know the texture samplers are trilinear-capable now, which would need a lot of extra BW.
You aren't serious, right? You can't be... It's not possible...
But damnit, then how many transistors does this beast have? I just can't imagine how a 400Mhz 0.15 beast with more than 115M transistors is gonna be cooled...
Are You Ready For More Noise & Heat?
Uttar
Nebuchadnezzar
Legend
Uttar said:Tagrineth said:I dunno, for all we know the texture samplers are trilinear-capable now, which would need a lot of extra BW.
You aren't serious, right? You can't be... It's not possible...
But damnit, then how many transistors does this beast have? I just can't imagine how a 400Mhz 0.15 beast with more than 115M transistors is gonna be cooled...
Are You Ready For More Noise & Heat?
Uttar
Why shouldn't they pull another rabbit out of their hats? Nothing is impossible.
Johnny Rotten
Regular
Uttar said:Tagrineth said:I dunno, for all we know the texture samplers are trilinear-capable now, which would need a lot of extra BW.
You aren't serious, right? You can't be... It's not possible...
But damnit, then how many transistors does this beast have? I just can't imagine how a 400Mhz 0.15 beast with more than 115M transistors is gonna be cooled...
Are You Ready For More Noise & Heat?
Uttar
Is the transistor count of the R350 substantially higher than that of the R300?
The cooling requirments are of issue to me. If ATI can keep a chip like this cool with a standard sink/fan combo, I'll pick one up.
LeStoffer said:
It is pretty much impossible due to RFTL issues, AFAIK. ODT is an absolute must at that kind of speed. Even a board using 360MHz DDR-I would be very tricky to qualify for mass production.
I don't think the R300-942 PCB (on which the R350-A075 board is very closely based) is capable of much more than ~350MHz, even with extensive tweaks. Just something I heard...
MuFu.
Isn't the Fx ultra a 12 layer design?
Maybe a (ultra) high-end card with 460Mhz/256 MB at 400+$.
Probably a typo but it would be fun to see the reaction of nVidia
P.S.- -Pixel shader 2.0+ (virtually unlimited instruction lengths).....I think this is the most importante feature
Maybe a (ultra) high-end card with 460Mhz/256 MB at 400+$.
Probably a typo but it would be fun to see the reaction of nVidia
P.S.- -Pixel shader 2.0+ (virtually unlimited instruction lengths).....I think this is the most importante feature
This is a PR document.
There are no new features mentioned apart from "512Kb of serial flash ROM", which fails to excite me hugely.
On the other hand, there has to be some worthwhile architectural differences, or there would be no need for an A98 at 325/310 while simultaneously offering an A97P at the same clock speeds which this document implies that FIC intends to.
Something doesn't quite click here.
Entropy
There are no new features mentioned apart from "512Kb of serial flash ROM", which fails to excite me hugely.
On the other hand, there has to be some worthwhile architectural differences, or there would be no need for an A98 at 325/310 while simultaneously offering an A97P at the same clock speeds which this document implies that FIC intends to.
Something doesn't quite click here.
Entropy
LeStoffer said:
Hmm... I don't know, maybe RussSchultz. He has "actual" industry knowledge though. Mine is just leeched off other people.
I wonder what their 256MB DDR-II board looks like. I suspect the layout will be a bit like that of the FireGL X1...
...except no DC, and probably more compact - using DDR-II saves on layout space a little.
The 256MB NV35 board is a bit different, sort of like this I think...
MuFu.
Joe DeFuria
Legend
MuFu said:The 256MB NV35 board is a bit different, sort of like this I think...
MuFu.
Man, nVidia has problems....the traces from the VPU don't even reach the memory chips!
Neutrality
Newcomer
Joe DeFuria said:MuFu said:The 256MB NV35 board is a bit different, sort of like this I think...
MuFu.
Man, nVidia has problems....the traces from the VPU don't even reach the memory chips!
ROFL!
-Neutrality-
Joe DeFuria said:Man, nVidia has problems....the traces from the VPU don't even reach the memory chips!
They are not traces, they're "Rays of Cinematic Shading Light". Part of this big PR deal, you see...
The smartass answer to that would be that the traces aren't on exposed layers, heh. Check out the NV30 PCB, closeup... OMG, SOME OF THE TRACES DON'T REACH THE CHIPS, NO WONDER THEY HAVE PROBLEMS!!!
MuFu.
Hellbinder
Banned
Its legit.
There are 3 R9800 comming. One with DDR-II.
Thus FIC is very likely talking about that one....
There are 3 R9800 comming. One with DDR-II.
Thus FIC is very likely talking about that one....
Typedef Enum
Regular
That's what I'm hearing as well...
Should that pan out...Man, if ever there was any reason to get a FX (the 9700 is already a darn good reason, IMHO), it's pretty much shot...
Heck...even if the memory was only clocked @ ~360 MHz., I think you would have to be a real "homer" to get an FX, based on what we can reasonably expect from R350...But add in even _more_ bandwidth to the mix...
Sheesh...
Should that pan out...Man, if ever there was any reason to get a FX (the 9700 is already a darn good reason, IMHO), it's pretty much shot...
Heck...even if the memory was only clocked @ ~360 MHz., I think you would have to be a real "homer" to get an FX, based on what we can reasonably expect from R350...But add in even _more_ bandwidth to the mix...
Sheesh...
