PDA

View Full Version : PCPop GeForce FX review with benchmarks


undercover
02-05-03, 03:27 PM
http://www.pcpop.com

Well it's in japanesse but hey...

The benchmarks appear to favour the GFFX big time without AAAF, but even with 4xAA and 8xAF the GFFX looks like it wins in a good majority of cases.

discuss.

Evildeus
02-05-03, 04:15 PM
Yes i would say 12*10 4*AA and 8*AF is the best IQ situation for the NV30 ;)

MUYA
02-05-03, 05:03 PM
Originally posted by undercover
http://www.pcpop.com

Well it's in japanesse but hey...

Minor correction, it is in Chinese ....but on one of the benches, the gf4ti4600 they tested with, comes out faster at the resolution of 1024 by 768! Seems odd that.
http://www.pcpop.com/read.asp?id=627&page=16

Its the UT2k3 botmatch, CPu probably plays a major role here

MUYA

undercover
02-05-03, 05:44 PM
Originally posted by MUYA
Minor correction, it is in Chinese ....

But is that Mandarin or Cantonese ? ;)

As Evildeus says 1280x1024 4xAA 8xAF look very very fast.

The GFFX is 30-40% faster in Aquamark2.1 and the UT2003 botmatch. Nature bench is 33% faster at 1024x768.

The GFFX loses out in oldest benchmark, Q3A, and loses by 1% at the highest res (1600x1200) in the UT2003 Fly bench.

Perhaps the 3Ghz P4 being used is finding more headroom in the GFFX than the 9700P ?

Hellbinder
02-05-03, 06:28 PM
I'll tell you what i have to say.

1. See other reviews and compoare scores..

2. 4x FSAA is a joke compared to what you are getting with the 9700pro at 4x fsaa. Want a more fair comparrison? Set the GFFX to at LEAST 6xs.

undercover
02-05-03, 06:43 PM
Originally posted by Hellbinder
I'll tell you what i have to say.

1. See other reviews and compoare scores..

2. 4x FSAA is a joke compared to what you are getting with the 9700pro at 4x fsaa. Want a more fair comparrison? Set the GFFX to at LEAST 6xs.


1. I have. This is a new one :)

2. I would rather see for myself, or hear from some gamers who play rather than reviewers who review. That's not going to happen till the GFFX hits the shelves. Until then, I personally don't get much joy out of comparing screenshots of triangles blown up 2,4 ,8 times just to see is some jaggies are better than other.

It's just a personal preference but *any* AA looks so much better than none at all that I don't worry that at ATI has better AA. In fact I much prefer the way the GFFX handles fine structures - it AA's them without chopped up fine lines. It boils down to how much AA, at what resolution and at what FPS is acceptable for each game. Given 1280x1024, *some* AA and a cracking frame rate I would rather see a comparison on texture quality.

IMHO there's a lot of folks getting a bit retentive about AA ;)

of course the noise, heat and power are all still there to consider.

Smokey
02-05-03, 07:45 PM
Originally posted by Hellbinder
I'll tell you what i have to say.

1. See other reviews and compoare scores..

2. 4x FSAA is a joke compared to what you are getting with the 9700pro at 4x fsaa. Want a more fair comparrison? Set the GFFX to at LEAST 6xs.

How about 4xs and nvidias 4x isnt a joke at perfectly fine for everyone before the 9700, im not saying 9700 isnt better, just that nvidias 4xs is better than just 4x. I myself are not going to judge the gf-fx until at least a retail card, and some new drivers. ATI have made one good chipset, now everyones calling them king of the hill, nvidia sux and what not, ATI have alot to prove yet, nvidia have done a great job for at least 4 years, something ATI is yet to achieve. Im standing by the NV30, why? because since my TNT I havent had any problems, tnt1, geforceddr, gf3. No real problems with drivers, and the cards have always worked with every game I have played. Why would I start baggin on a company that with a good track record? The previews have been based on non retail cards and early drivers, at least when we bagged ATI cards, they were retail :p

Sazar
02-05-03, 07:53 PM
Originally posted by Smokey
How about 4xs and nvidias 4x isnt a joke at perfectly fine for everyone before the 9700, im not saying 9700 isnt better, just that nvidias 4xs is better than just 4x. I myself are not going to judge the gf-fx until at least a retail card, and some new drivers. ATI have made one good chipset, now everyones calling them king of the hill, nvidia sux and what not, ATI have alot to prove yet, nvidia have done a great job for at least 4 years, something ATI is yet to achieve. Im standing by the NV30, why? because since my TNT I havent had any problems, tnt1, geforceddr, gf3. No real problems with drivers, and the cards have always worked with every game I have played. Why would I start baggin on a company that with a good track record? The previews have been based on non retail cards and early drivers, at least when we bagged ATI cards, they were retail :p

ati has had many good video cards... their IQ has always (well quite often) been top notch... their drivers went downhill but after the launch of the 8500... they have been continuously improved to the extent that they have releases on a seemingly more frequent schedule than nvidia... (official drivers... WHQL)

you need to remember a lot of people KNEW the gf FX would have problems with drivers and that it was highly unlikely to take a massive performance increase off the bat... reason being that nvidia has routinely tweaked architecture in order to keep its fps high... and by using tweaked architecture has also had a fairly less-complicated driver issue when compared to ati with their new architectures...

the nv30 is a new architecture and nvidia will have to work out the bugs in THEIR architecture just as ati always does with their new cores...

its just the way it is...

perhaps with this more balanced playing field and both companies driver teams working diligently on their products... we will continue to see better performance being eked out and more optimizations as well :) and less fanboyism claiming one companies drivers are better than the other companies...

when looked @ in the whole context... specially over the past year or so... we know this not to be the case...

Bigus Dickus
02-05-03, 09:50 PM
Originally posted by Smokey
Im standing by the NV30, why? because since my TNT I havent had any problems, tnt1, geforceddr, gf3. No real problems with drivers, and the cards have always worked with every game I have played. Why would I start baggin on a company that with a good track record?

That's called fanboyism.

Shinri Hikari
02-05-03, 10:02 PM
THIS, from an on again off again fanATIc, Please spare me.:D

volt
02-05-03, 10:12 PM
Bigus Dickus: stop whining :eek: :)

Smokey
02-05-03, 10:36 PM
Originally posted by Bigus Dickus
That's called fanboyism.

No its not, most people will stick with the same brand of anything if they have never had problems. I never said ATI make bad cards, I just said they havent the same record. As for ATI releasing offical drivers more often than Nvidia, I do believe that may be true, but since the 9700 more than the 8500, and they were to fix problems more than anything. Things are getting better in the driver dept for ATI, and they now have a good chipset, but Im not turning my back on Nvidia just yet ;)

The_KELRaTH
02-05-03, 10:54 PM
The problem with many of these reviews is they just don't give info of the precise settings used.

If I run 3Dmark at 1024x768 using x8AF Performance with performance mipmap the score just drops 800 points. If I set Max Quality mipmap and texture I loose a further 498 points. (P4-2.53 9700Pro)

With regards AA quality, I think thats been covered enough and the only connection between them is the level numbers in the control panel. Ignoring the horizontal degree offsets is just ridiculous - even ATI's x2aa looks better than NV30's 8xSaa. Probably also reduces the performance cost too.
http://www.hardwareaccess.co.uk/NV30-ATIAA.jpg

As for the SpecviewPerf, a simple softmod and even the 9500 can fly past all of the results by a massive margin - which tells us it's purely a driver configuration.
Here's some examples with a 9500 with 64mb using an Athlon 1800+

ATI 9500 64MB - SpecviewPerf (http://www.nvworld.ru/docs/sfgl2_e.html)

Other things like ATI having hardware Gamma corrected AA and textures while NV30 has none. NV make a claim about gamma correction option in shaders but any card can do that and it's nothing to do with AA.
While the NV30 runs at 12bit Integer by default the R300 runs everything at 24bit floating point.

All these ATI features have a slight performance hit but improve image quality.
Does the fog being disabled on the NV30 offer an fps improvement - UT2003 is being used in practically every test and makes extreme use of fog effects. (among others).

Nutty
02-06-03, 08:39 AM
Why are you comparing completely different methods of AA?

IMO the Gf-Fx 2x AA looks better than the 9700 2x AA. I also think their texture quality is better.

I dont think the lack of fog will affect performance to any degree. Its a very minor detail that takes very little processing.

The_KELRaTH
02-06-03, 11:10 AM
Nutty:

The GeforceFX 8xS AA is obviously alot better than any x2aa but at the horizontal offset angles it seems to drop to the equivalent of x2aa. As, and just like the vertical offset, this is where you would see the most crawling when moving around ingame. The vertical offset is fine so why not the horizontal - is it just a driver flaw?

Sazar
02-06-03, 11:23 AM
Originally posted by Nutty
Why are you comparing completely different methods of AA?

IMO the Gf-Fx 2x AA looks better than the 9700 2x AA. I also think their texture quality is better.

I dont think the lack of fog will affect performance to any degree. Its a very minor detail that takes very little processing.

and hence with your last statement nvidia's driver team breathes a sigh of relief...

why render it if the consumer says no... I don't want it... it is not going to give a performance hit...

heck in that regard lets just have round blobs as targets in FPS and squares as castles in RPG's... I mean who will notice it and there will be no 'performance related decrease' in the frame rate :)

if you have looked @ the gf FX pictures you will see.. comparing side by side... gf FX no aa/fsaa to gf fx 2x AA that there is hardly ANYTHING being done... nothing... zilch... nada... zero...

now compare that to the 9700pro regular non AA/fsaa screenie and their 2x AA screenies... :) you WILL notice a difference in IQ... there is a discernible improvement and the base IQ is better on the 9700pro IMO than on the gf FX...

coming back to the fog issue... in a situation where the overall difference in most tests is +/- 5%... do you really think that not rendering the fog is NOT giving some advantage to the gf FX when it comes time to benchmark ? and that WITH fog enabled that the gf FX will not take a frame-rate hit ?



I have to go with Kelrath in his queries because I have been seeing the same thing...

Evildeus
02-06-03, 02:17 PM
Well, i don't know for you guys but i find the GFFX performance much better with time ... Driver optimisation? :confused:

oj213
02-06-03, 07:51 PM
:confused: Does anyone know any reviews where they checked out the TV-Out to see if Nvidia worked on this?