First few GFX benches

Status
Not open for further replies.
Qroach said:
How about performance and IQ. Doom, you almost sound like you think the FX will out perform the the 9700 or something...

I think the FX will outperform the 9700 (as we see it today) in fill rate tests just due to the higher clocked memory...as resolution and IQ enhancing features are increased it should even out.
I also hope reviews use a quality 9700 with some overclocking....like these monsters overclocked to 405/690..as you can see the card responds well and maybe a hint of something to come :)


http://www.ocsystem.com/ocenati97pro.html

ELSA_image014.png


Yet I believe ATI has a surprise coming too ;)
 
Qroach said:
How about performance and IQ. Doom, you almost sound like you think the FX will out perform the the 9700 or something...

With IQ then assuming the FX AA is the same as gf3/4 then i'd rate quality as 4x(fx)<4xs(fx)<6xs(fx)<4x(r300)<8x(fx)<6x(r300) and we end up with a similar situation as v5vgf2 where the v5 had its RGSS and the gf2 was only with ordered grid. What AA modes do you compare?
 
We will have to live through the usual BS reviews at Anand with small little thumbnails and/or no screen shots at all, different settings for each card, old game titles and the like...and of course the fan sites are always worth a laugh.

I certainly hope IQ is compared on both cards, including default LOD etc..

Yet Beyond3D will be of course the most indepth and interesting to look at...I hope they get a card at the same time as the bigboys.
 
It doesn't really matter for most people. There is a huge difference between no-AA and 2x/4x AA. But as you go higher and higher, the improvements in quality at each level aren't as dramatic. Some people may have hypersensitized themselves, but the vast majority of people won't really notice. Unless some vendor makes a huge leap to 16x or 64x sampling, I doubt we will get that "wow" factor again.

That's the problem with the IQ comparisons. Unless you can show a real stark contrast, like what we saw when we went from point sampling to bilinear filtering, or no-AA to AA, or no-AF to AF, it is enormously hard to market a product on minute differences that require microscopic investigation. Consumers are more likely to pay attention to the BOX design on the shelf. :)

Thus, we are going to go through another round of people 16x zooming into screenshots and arguing over minute aliasing details that 90% won't notice or care. And of course, depending on which side of the fence you are on, very minor IQ differences are going to be magnituded and hyped to the extreme.

But unlike performance, which is easily numerically recorded, IQ isn't, so whoever is the victor in the IQ battle won't really be able to translate this into increased sales IMHO.
 
Reverend said:
Pssst, here's a secret... never compare 4xAA performance between a R300 and a NV30...

Let me guess here: such a comparison would be unfair to the 9700 because its 4x is noticeably superior to Nvidia's ordered grid sampling pattern?
 
I certainly hope we get a baseline IQ setup..i.e even if it requires different settings on each card..then benchmark...using a popular game that allows easy screen shot comparisons.
 
And is there any real proof the GFFX uses the same ordered 4x as the GF4 :?:

ATI uses the same 4x patterns as the V5. nVidia acquired 3DFX and got over 100 ex-3DFX employees. I'd be surprised if they didn't realize V5 sampling patterns were better than GF4 ones :rolleyes:
With the GF4, nVidia had put the good patterns in 4xS, also including SuperSampling. But seeing how the GFFX got Color Compression and that might increase the performance hit quite significantly, I don't see why they'd do that anymore.


Uttar
 
There is no proof but unfortunately theres a lot of edvidence.

from the interview here:

The documentation made note of an 8X FSAA mode, how is this achieved and is it Multi-sampling?

The 6XS and 8X modes are a combination of Super-Sampling and Multi-sampling.

Are they still the sort of ordered grid sampling schemes that we've seen on GeForce4 or does GeForce FX have any jittered / programmable grids?

The 6XS is a skewed grid, the 8X mode is an ordered grid.

Als nvidias whitepaper says something along the lines of 'in addition there are 2 new modes, 6xS and 8x' suggesting that other modes are same as gf4.

Using gf3/4 2xRGMS with 1.5x2 and 2x2 supersampling gives a skewed grid and ordered grid for 6xS and 8x modes. :|
 
About benchmarks and NV30 being memory bandwidth limited. From the interview:

Is that full time colour compression, i.e. does it operate when AA is not enabled?

It pays dividends mostly for the AA modes because those are the ones that are memory bandwidth bound. If you're just doing aliased rendering, with the bounds of our pipeline and memory you are generally not memory bound, but it can colour compress without AA.

Marketing or not, ATI said that R300 is still memory bandwidth limited in most cases. Makes you wonder... :|
 
DemoCoder said:
But unlike performance, which is easily numerically recorded, IQ isn't, so whoever is the victor in the IQ battle won't really be able to translate this into increased sales IMHO.

If I ever saw Truth on these forums, this was it.

Entropy
 
Pssst, here's a secret... never compare 4xAA performance between a R300 and a NV30...

Knowing that you are a dedicated Nvidia Protagonist, I am confident you are saying that the Nv30 is faster... Right? I am also just as confident that you are saying Nv30 when you should really be saying the extremely OVERCLOCKED Ti 5800 Ultra with a Dustbuster Attachment. right? I wonder how the vanilla Ti 5800 will fare?
 
Rev an NVidia zealot? Clearly you weren't around in the V5 vs GF days on B3D. Rev was the leading proponent of benchmarking the GF @ 4X FSAA vs the V5 @ 2X FSAA (which makes GF performance look bad and V5 look good) because V5 2X ~= GF 4X IQ wise in his opinion.

Rev's opinion was to benchmark at settings where IQ is nearly equal. Adjusting LOD bias, AF, etc so that the cards are producing nearly the same image.

Of course, my opinion is, trying to market FSAA IQ to the general populace didn't work for 3dfx and it isn't likely to work for ATI either if that were their marketing thrust (fortunately it isn't. Currently, they tout performance and shaders). However, if ATI gets beat in performance and falls back to the ole "but our AA IQ is slightly better", they won't get any traction with most people.


Just try to imagine explaining gamma-corrected AA to the average-joe in an IHV ad for example.

To me this is like non-judged Olympic sports (track and field) vs judged (figure skating, etc) Non-judged sports are fairly objective. It takes 1 second to understand the scoreboard. Guy who crossed line first, wins. IQ is subjective. Takes more than the 5 seconds someone spends looking at an ad.
 
Hellbinder said:
Knowing that you are a dedicated Nvidia Protagonist, I am confident you are saying that the Nv30 is faster... Right? I am also just as confident that you are saying Nv30 when you should really be saying the extremely OVERCLOCKED Ti 5800 Ultra with a Dustbuster Attachment. right? I wonder how the vanilla Ti 5800 will fare?

One more unnecessary post like that and you'll be banned. . .again. Consider this fair warning.
 
DemoCoder said:
Rev an NVidia zealot? Clearly you weren't around in the V5 vs GF days on B3D. Rev was the leading proponent of benchmarking the GF @ 4X FSAA vs the V5 @ 2X FSAA (which makes GF performance look bad and V5 look good) because V5 2X ~= GF 4X IQ wise in his opinion.

Rev's opinion was to benchmark at settings where IQ is nearly equal. Adjusting LOD bias, AF, etc so that the cards are producing nearly the same image.

Of course, my opinion is, trying to market FSAA IQ to the general populace didn't work for 3dfx and it isn't likely to work for ATI either if that were their marketing thrust (fortunately it isn't. Currently, they tout performance and shaders). However, if ATI gets beat in performance and falls back to the ole "but our AA IQ is slightly better", they won't get any traction with most people.


Just try to imagine explaining gamma-corrected AA to the average-joe in an IHV ad for example.

To me this is like non-judged Olympic sports (track and field) vs judged (figure skating, etc) Non-judged sports are fairly objective. It takes 1 second to understand the scoreboard. Guy who crossed line first, wins. IQ is subjective. Takes more than the 5 seconds someone spends looking at an ad.

If one card can produce roughly equivalent IQ at a lesser setting, so be it. In fact, I can recall you saying that reviewer's should adjust tweakable settings to reach IQ as equal as possible (in their opinions, obviously) and then bench from there. Or you saying this isn't fair testing standards now or just that it tends to p*ss the fans off?

The truth, IMO, is that most people who post or even lurk on boards such as this aren't the general public and do care about such things. But you're right in saying that Rev used to get pegged as pro-3dfx and, bizarrely enough, he's now often labelled as pro-Nvidia. Sigh.
 
When benchmarking Speed sells, but people also don't lay down big money on top of the line cards to run with nothing on...and that is where Nvidia will win alot of races due to its fillrate, which ATI will counter with something of its own...
I Hope that people are smart enough to really look at the results for such a expensive piece and look at the benchmarks that show 4-6 x FSAA and AF as that is what we buy these cards for...the DX9 stuff is a added bonus but not a factor for at least 8-12 months.
 
John Reynolds said:
The truth, IMO, is that most people who post or even lurk on boards such as this aren't the general public and do care about such things. But you're right in saying that Rev used to get pegged as pro-3dfx and, bizarrely enough, he's now often labelled as pro-Nvidia. Sigh.

Or pegged as a Fair Weather Fan, riding whatever bandwagon that is current ??
I have had issues with Reverend in the past but try to move past that, there was legit reasons why he was pegged pro-Nvidia (even on Ve3d forums)...how many ATI reviews, interviews are done by him as their technology reviewer ??
If you constantly rant and rave over one IHV then the 'bed is made' for such a accusation.
Comments on this forum like:
ATI's dev relations SUCK
PR Suck
Doesn't bode well either...

I hope that these issues are resolved with him and ATI as he is far to 'smart' of a reviewer to be stuck with one IHV. :!:
 
I am no longer a member of Driverheaven, I've had some serious medical problems with a injury and did not have the time to be involved with a site or finish the article...
I may finish it if I can get the desire to do it, in reality though (effort vs. result) will mean little return personally..not to mention a few toes stepped on ( which doesn't bother me one bit).

If anything the net community is 'fair weather' in General, when things are going great they defend, fight tooth and nail for whatever they believe but generally if someone bests that company that same fan jumps the fence.
 
Reverend said:
Pssst, here's a secret... never compare 4xAA performance between a R300 and a NV30...

Right you are. Unless they have done something radically different with their AA methods, a fair comparison would be R300 Quality 4x vs. NV30 6x or even 8x (if such a thing actually exists.) And from these first benchmark leaks (assuming they are legit, which I somehow doubt), the lack of any > 2x benchmarking, coupled with the known NV30 memory architecture, to me clearly indicates a problem.

But then again, I always feel uncomfortable making statements based on very little fact, unlike many people on such forums.
 
Status
Not open for further replies.
Back
Top