just how fast will the nv35 be?

[butter]

some people say it wont be faster than the r350 so that means its less than 15 percent faster than the gffx. the early benches at lower speeds say that its almost 200% faster. so whats the deal here?
i personally feel that it will be much faster than the r350. the r350 is really disappointing: reminds me of the gf3 to gf3ti500.

[Matt-AWD]

I think it will be faster than the R350, but not 200%. I'd give it the 20-30% that the nv30 was SUPPOSED to be.
As for R350, it was precisely what it was made out to be - an R300 with some slight optimizations, the F-buffer, and slightly higher speeds. I don't see why people are dissapointed with it, it wasn't meant to be anything revolutionary.

[Sazar]

Originally posted by Matt-AWD
I think it will be faster than the R350, but not 200%. I'd give it the 20-30% that the nv30 was SUPPOSED to be.
As for R350, it was precisely what it was made out to be - an R300 with some slight optimizations, the F-buffer, and slightly higher speeds. I don't see why people are dissapointed with it, it wasn't meant to be anything revolutionary.

its a revenge type philosophy more than practical statements that say the 9800pro sucks.. :)

coz everyone said the nv30 was crud... hence... the r350 is crud...

[Gar]

Well, we haven't seen anything about the "beefed-up" 256MB DDRII version yet have we :). I'm sure that will compete just enough with the NV35 with its 30+gb/s bandwidth.

[zakelwe]

The GF FX ( aka nv30 ) was pretty fast with no AA/AF and so they don't have to make that much faster in my opinion, rather they need to make AA/AF faster and if the 256bit bandwidth rumour is true then it should be.

But then they need to sort out the drivers first so there is a level of confidence in them, especially in regards to AA and AF again.

I'm looking forward to seeing how the nv35 does but as I have got a GF4 4400 I am not too worried about it not being fast enough, it will be compared to a GF4 4400:) , but to have the IQ that Ati have had for a while.

Ideally i'd like to see the 9800 Pro and nv35 both have roughly the same performance and both have similar IQ, this will make the prices hopefully move downward and provide lively debate between the forums.

There has too much incoming on nvnews forums of late, my hair almost got ruffled for a while back there ....

Once we can decide on 9800 or nv35 then the Canterwood / P4 or K8 / Opteron battle can add more variety.

Are we all feeling rich ?

$$$$$$$$$$

[borntosoul]

if the nv35 does have that much bandwidth look out.
what a year this year is going to be ,hehe with all these HW releases,nv35,athlon64,prescott,(maybeR400)ect

[Dazz]

the NV35 will most likly have 25GB/sec bandwith 800MHz 256bit DDR-1 memory so i would say it will be about 20~30 faster.

[Steppy]

It'll me like a GF3TI to a GF4 at best. GF3 to GF3TI at worst.

[Paul]

The biggest area of improvement will be when running games at high resolutions + AA/AF. Running at basic settings (800x600 / 1024x768, no AA/AF) will show minimal difference, since the G-FX is more than fast enough in those situations as it is.

[Dazz]

And hopefully a **** load of bug fixes.

[savyj]

Originally posted by Steppy
It'll me like a GF3TI to a GF4 at best. GF3 to GF3TI at worst.

It'll be like a GF4 Ti4600 to R 9700Pro at best. R 9700Pro to R 9800Pro at worst ;)

[Steppy]

Originally posted by savyj
It'll be like a GF4 Ti4600 to R 9700Pro at best. R 9700Pro to R 9800Pro at worst ;) I don't think it's possible for it to be a 4600-9700 type of upgrade...the timeframe between the two is too short, this is an evolutionary product, not a revolutionary one.

[surfhurleydude]

I don't think it's possible for it to be a 4600-9700 type of upgrade...the timeframe between the two is too short, this is an evolutionary product, not a revolutionary one.

The reason the 9700 Pro had such a big jump was because of two things: 256 bit bus and 8 pixel pipeline.

The NV35 will have the same thing over the GeForce FX... twice the bandwidth and twice the pipelines (at least in single textured situations). And of course there will be some more optimizations.

Sure I'm being fairly optimistic, but there's no reason to doubt that there will be utterly MASSIVE FPS improvements in the AA and AF arena with the NV35.

[gravioli]

Originally posted by surfhurleydude
Sure I'm being fairly optimistic, but there's no reason to doubt that there will be utterly MASSIVE FPS improvements in the AA and AF arena with the NV35.

LOL! There's no reason to doubt Nvidia's ability to deliver a great performer? The NV30 seems like a pretty good reason to me. ;)

Don't get me wrong, I hope the NV35 is a great product, but I have my doubts about "MASSIVE FPS improvements in the AA and AF arena."

[savyj]

Originally posted by gravioli
LOL! There's no reason to doubt Nvidia's ability to deliver a great performer? The NV30 seems like a pretty good reason to me. ;)

So the fact that they have delivered great performers time and time again (more often than any other GPU maker to date) is completely nullified by the NV30 mis-step? Or did I misunderstand you?

[savyj]

Originally posted by Steppy
I don't think it's possible for it to be a 4600-9700 type of upgrade...the timeframe between the two is too short, this is an evolutionary product, not a revolutionary one.

I would agree except that many consider the NV30 to be a flawed design and that the NV35 is what it "should" have been in the first place, therefore I do think it will be a decent jump in performance and will fall somewhere between my two mentioned points.

[Typedef Enum]

So the fact that they have delivered great performers time and time again (more often than any other GPU maker to date) is completely nullified by the NV30 mis-step? Or did I misunderstand you

Absolutely true. The past means nothing when it comes down to it. nVidia's CEO has said, time and again, that they will ALWAYS be one freakin' product cycle away from going out of business. Those aren't my words, but his.

Havins said that, I do anticipate NV35 being a very successfull product. nVidia is not completely idiotic, although this past year has clearly been one of the worst periods in their history.

Other than not having (likely) better AA methods than ATI, I think the writing is on the wall that they will be back come next month.

[StealthHawk]

Originally posted by gravioli
Don't get me wrong, I hope the NV35 is a great product, but I have my doubts about "MASSIVE FPS improvements in the AA and AF arena."

NV30 chokes with 4x FSAA because of....you guessed it.....bandwidth.

the bandwidth increase alone should make 4x FSAA scores much higher. and that was the NV30's biggest performance failure: benchmarks with 4x FSAA being used.

oh yeah...and shader performance...

[Uttar]

Originally posted by StealthHawk
oh yeah...and shader performance...

To be more precise, Pixel Shader peformance.
Vertex Shader performance is fine - it ain't amazing, but there's no reason to complain about it. Beside for branching maybe, but it still ain't bad performance.


Uttar

[Fotis]

Originally posted by Uttar
To be more precise, Pixel Shader peformance.
Vertex Shader performance is fine - it ain't amazing, but there's no reason to complain about it. Beside for branching maybe, but it still ain't bad performance.


Uttar
IMHO vertex shader performance is not good.
GFFX has a 500Mhz clock and radeon9700pro is still faster on vertex shader performance.

[gravioli]

Originally posted by StealthHawk
NV30 chokes with 4x FSAA because of....you guessed it.....bandwidth.

the bandwidth increase alone should make 4x FSAA scores much higher. and that was the NV30's biggest performance failure: benchmarks with 4x FSAA being used.

oh yeah...and shader performance...

Yes, you're right. I guess my perception of "utterly massive FPS improvements in AA and AF performance" is 2x NV30 performance in most situations. 2x NV30 is reasonable with regards to certain AA situations, but not AF. Or is it? How would Nvidia go about doubling AF performance? Not by increasing bandwidth.

[gravioli]

Originally posted by savyj
So the fact that they have delivered great performers time and time again (more often than any other GPU maker to date) is completely nullified by the NV30 mis-step? Or did I misunderstand you?

No misunderstanding. BTW, I didn't say it was completely nullified. However, I did give a very good reason to doubt their ability to deliver. Does doubting something completely nullify the possibility of it occurring? :confused:

[jbirney]

Originally posted by StealthHawk
NV30 chokes with 4x FSAA because of....you guessed it.....bandwidth.

the bandwidth increase alone should make 4x FSAA scores much higher. and that was the NV30's biggest performance failure: benchmarks with 4x FSAA being used.

oh yeah...and shader performance...

Faster but with lower quailty AA still is not a step in the right direction

[StealthHawk]

Originally posted by gravioli
Yes, you're right. I guess my perception of "utterly massive FPS improvements in AA and AF performance" is 2x NV30 performance in most situations. 2x NV30 is reasonable with regards to certain AA situations, but not AF. Or is it? How would Nvidia go about doubling AF performance? Not by increasing bandwidth.

well, AF performance isn't really bad per se. you could make the case that when you use Application AF performance is subpar, I suppose, and that would be a legitimate complaint.

I would think that going to a "truer" 8X1 configuration over a 4X2 configuration along with the extra bandwidth should help AF scores, maybe not.

Faster but with lower quailty AA still is not a step in the right direction

we weren't talking about quality, only speed. it is still unknown whether or not NV35 will have the same sample patterns as NV30 AFAIK.

[Steppy]

Originally posted by StealthHawk
NV30 chokes with 4x FSAA because of....you guessed it.....bandwidth.

the bandwidth increase alone should make 4x FSAA scores much higher. and that was the NV30's biggest performance failure: benchmarks with 4x FSAA being used.

oh yeah...and shader performance... I don't know about that...it's bandwidth isn't THAT far behind the 9700's...I'm still of the belief that there is more to the performance that just a lack of bandwidth. I'm guessing it'll be 30% faster than NV30 or so(that's what Nvidia's typical claims of 2-3 times faster usually translate to)...which should bring it a little ahead of a 9800 Pro. People expecting 2x performance of NV30 are gonna be left with a sour taste in their mouth and will probably end up bad mouthing it even though it will still be a good performer.