NVIDIA GT200 Rumours & Speculation Thread
- Thread starter Arun
- Start date
- Status
leoneazzurro
Regular
But remember that GT200 has 32ROPs and 512-bit MC which should give some nice boost especially in 1920x1200 resolutions or higher with AA enabled.
I believe it`s true because they were right about G80 and G92 numbers too
IMHO:
GT200 has not 2x the shading or texturing theoretical performance of G92 (9800 GTX). ROPS and bandwidth are 2x. Now, if we reckon that G92 is heavily bandwidth limited, and that GT200 is a only slightly revised architecture, GT200 should have 2x the performance of G92, and more only in the case of G92 being framebuffer limited.
So I think it's reasonable to think that GT200 is between 1.5 and 2x the performance of G92 in most cases, and it could be more in very high resolution scenarios, due to bigger framebuffer and bandwidth, and it could be also less than 1.5x in some cases where G92 is not so limited by the bandwidth.
Megadrive1988
Veteran
Hmmm if 933 GFLOPs is true, then they've missed on the 1TFLOP by a significant margin. That's where a GT200 Ultra might come in, and I think Nvidia will need it with AMD's X2 coming in the fall.
Still, sounds like GT200 is gonna be a good performer. It may have room to grow with a respin/tweak. Nvidia could really give it a boost next year with a redesigned version that uses GDDR5. Imagine the bandwidth on that.
The 240 SP, 32 ROPs and 512-bit bus are good foundations to build on.
Anyone agree?
Still, sounds like GT200 is gonna be a good performer. It may have room to grow with a respin/tweak. Nvidia could really give it a boost next year with a redesigned version that uses GDDR5. Imagine the bandwidth on that.
The 240 SP, 32 ROPs and 512-bit bus are good foundations to build on.
Anyone agree?
Does Vantage use PhysX?
What do you think about this guys?
"Just wait until June 17th when the GTX 200 series of GPU's launch, and you'll start asking yourself when you last witnessed such a dramatic technology improvement. If you thought the GeForce 8 series blew the 7-series out of the water, this is going to leave you in shock. That's not my own marketing spin... Benchmark Reviews is presently testing the new GeForce video card."
Does it mean we will see at least as big performance bump as we have seen with G80 against G71?
I say it sounds very promising.
This is from Benchmarkreviews.
"Just wait until June 17th when the GTX 200 series of GPU's launch, and you'll start asking yourself when you last witnessed such a dramatic technology improvement. If you thought the GeForce 8 series blew the 7-series out of the water, this is going to leave you in shock. That's not my own marketing spin... Benchmark Reviews is presently testing the new GeForce video card."
Does it mean we will see at least as big performance bump as we have seen with G80 against G71?
I say it sounds very promising.
This is from Benchmarkreviews.
Yes it does
What do you think about this guys?
"Just wait until June 17th when the GTX 200 series of GPU's launch, and you'll start asking yourself when you last witnessed such a dramatic technology improvement. If you thought the GeForce 8 series blew the 7-series out of the water, this is going to leave you in shock. That's not my own marketing spin... Benchmark Reviews is presently testing the new GeForce video card."
Does it mean we will se at least as big performance bump as we have seen with G80 against G71?
I say it sounds very promising.
This is from Benchmarkreviews.
Reviewing the site I had this comment to make-
OK but as you see not only this site is saying that GT200 will be a incredibly fast GPU. We will see.
And one more thing. Are these results from this site much different than the other sites? I mean if GF8800GT is about 20% faster than HD3870 then on this site it is about 40% faster?
We will see.And one more thing. Are these results from this site much different than the other sites? I mean if GF8800GT is about 20% faster than HD3870 then on this site it is about 40% faster?
Last edited by a moderator:
True, there are a number of things that have to happen for them to get there. Raw unit count and bandwidth increases are the first step. But there's always the possibility that GT200 is more balanced than G80/G92 or has additional improvements in triangle setup, bandwidth efficiency, memory usage, caching, HSR etc.
It does but even if Nvidia's CUDA implementation works for CPU test 2 it wouldn't affect the overall score at extreme settings too much - both CPU tests combined account for about 5% of the final score (although CPU test 2 has a lot more weight than CPU test 1).
It has close to 2 times the shader performance with the increased efficiency its actually closer to 2.5 times (2 flops vs 3 flops) since the mul was only used at most 25% -30% in real world apps. Texture performance is abundent in the g80 the g92 was just well over that, so I don't see how we can factor that in, it already has ample.
There were places where the g80/g92 architecture can have great improvements, one was the mul where the g80/g92 loose around 15-25% effeciency, and the second is vertex set up performance.
Having a PPU(or PhysX support through CUDA) doesn't affect only the CPU scores, AFAIR, it helps the gametests as well...all in all, Futuremark suggested 200 or 300 points improvement from adding a PPU, so nothing earthshattering.
leoneazzurro
Regular
Tex ability was abundant and now it's bigger, but it's not 2x now. Let's say that this is not a limit anyway.
MUL should be functional now, OK, but you are not getting anyway 2,5x the performance of a 9800 GTX in any case, best comparison is 930/432=2.15x, which is closer to 2 than to 2.5 AND that's even in an heavy MUL using scenario, because if you use a lot of ADD operations then suddendly the GT200 advantage is only 1.44x. I really have no data if actually in real applications G92 is limited by triangle setup, if you have some info about that, I will be happy
Of course the architecture can improve a lot of parts (front end, caches, and so on) but if G92 is already bandwidth limited in a lot of scenarios, then with GT200 you wll have the same limitation, only with a limit 2 times higher because still sticking with GDDR3. Of course Nvidia could have invented a new bandwidth saving technique, but if this is not the case I don't really see how GT200 can go over +50% than a 9800GX2 as someone hopes...
leoneazzurro
Regular
Of course. I already told I was referring to all cases where framebuffer is not a limitation.
If this chip hasn't the power to run my fav game at 1920x1200 with SSAA 2x & the new high res textures (1024x1024 & bigger), at 60fps min frame rate, i'm going to be seriously disappointed.
Personally i think that this card will be bounded by the memory bandwidth (for my needs), cause my fav game seems to love memory bw:
http://forum.beyond3d.com/showpost.php?p=1147916&postcount=60
Personally i think that this card will be bounded by the memory bandwidth (for my needs), cause my fav game seems to love memory bw:
http://forum.beyond3d.com/showpost.php?p=1147916&postcount=60
did theInq's article get plagiarized from here?
http://www.hardware-infos.com/news.php?news=2090
a lot of the info seems to be from the charts with some "stuff" just filled in by Charley
- why doesn't he stick to AMD speculations; and those are getting worse
[and you guys talk about me being speculative fluff]
my German is poor, who is quoting who[m]?
otoh, Benchmarkreviews, seems a lot closer
EDIT: my bad; i am sorry .. "quelle" is Source; the german site is quoting theInq
--messed up news travels fast
=(
ignore me, please
http://www.hardware-infos.com/news.php?news=2090
a lot of the info seems to be from the charts with some "stuff" just filled in by Charley
- why doesn't he stick to AMD speculations; and those are getting worse
[and you guys talk about me being speculative fluff]
my German is poor, who is quoting who[m]?
otoh, Benchmarkreviews, seems a lot closer
EDIT: my bad; i am sorry .. "quelle" is Source; the german site is quoting theInq
--messed up news travels fast
=(
ignore me, please
Last edited by a moderator:
- Status
