[Luminescent]

When do you guys believe the FX 3Dmark benchmarks will appear? If any of you already have the FX, would you mind benchmarking it with 3DMark03 and reporting the score, especially the synthetic PS 2.0 benchmark.

[LeStoffer]

Hey, I was just thinking the same!

In the words of Dave Baumann:
Brent - fancy doing a little more testing?

[Ichneumon]

http://www.hardocp.com/article.html?art=NDI4LDQ=

Interesting.

Remember those questions about how GFFX would perform on the PS2.0 test? Ouch. Granted it is running 128bit precision... but is that extra precision over 96bit on the 9700pro worth 1/3 the performance?

The new drivers make a big difference, but it still doesn't seem to paint that Ultra part as being terribly Ultra.

[BoddoZerg]

Why is the CPU score for the GFFX so much lower than for the Radeon9700Pro?

[Brent]

Originally Posted by BoddoZerg

Why is the CPU score for the GFFX so much lower than for the Radeon9700Pro?

the vid card still effects the cpu score, different vid cards will produce different cpu scores because they perform differently

[jvd]

what i want to know is what happens when you throw on aniso and fsaa . Since in game bench marks the radeon normaly lags behind and then over runs the geforce fx when eye candy is turned on (see anands benchmarks)

[Rodéric]

Now I just wait for the 9700 cost to lower...

[LeStoffer]

Brent, it's a massive jump on the GeForce FX with the new drivers (much more in line with where it should be). It all screams FP16 use instead of FP32 under PS 1.4. Can you verify this?

[Typedef Enum]

Another question to ask yourself:

How many other benchmarking titles (be it games, apps, etc) show any tangible increase in performance...OUTSIDE of FutureMark...in using the newer drivers?

[Doomtrooper]

I'd like to see some overclocked 9700 scores in comparison..my new OEM 9700NP came today and to my surprise the clock is already 324/310 (or I got a Pro instead) and with a simple Crystal orb running stable 375/323...not bad for $340 Canadian and no dustbuster.

[mboeller]

Originally Posted by LeStoffer

Brent, it's a massive jump on the GeForce FX with the new drivers (much more in line with where it should be). It all screams FP16 use instead of FP32 under PS 1.4. Can you verify this?

look at the Pixelshader 2.0 Feature Test !!



The R9700Pro is nearly 3times faster than the GFFx

[surfhurleydude]

Pixel Shader 2.0 for the GF FX seems VERY low... TOO low as a matter of fact... is PS 2.0 for the GF FX broken in the drivers or something? Hell, in almost every other previous PS test I've seen the GF FX SOLIDLY outperforms the Radeon 9700 Pro..

[Joe DeFuria]

Hell, in almost every other previous PS test I've seen the GF FX SOLIDLY outperforms the Radeon 9700 Pro..

Where have you been? Every test we've seen here has been pretty much the opposite. The 3D Mark PS 2.0 test is consistent with other DX9 shading benchmarks.

The big question is PS 1.4.

What changed in the GeForce drivers between revisions? Changed fp precision? Running in PS 1.1 (integer) vs. PS 1.4 (fp) mode?

[surfhurleydude]

Where have you been? Every test we've seen here has been pretty much the opposite. The 3D Mark PS 2.0 test is consistent with other DX9 shading benchmarks.

The big question is PS 1.4.

What changed in the GeForce drivers between revisions? Changed fp precision? Running in PS 1.1 (integer) vs. PS 1.4 (fp) mode?
_________________

Every test I've seen shows the GF FX lose badly in VERTEX shader performance, but not DX 8 Pixel Shader performance...

[Joe DeFuria]

but not DX 8 Pixel Shader performance...

DX 8.0 (Ps 1.1) or DX 8.1 (Ps 1.4) ?

[Ostsol]

Indeed. In ShaderMark, the GeforceFX got totally raped except in the fixed-function tests.

[surfhurleydude]

PS 1.1, sorry...

PS 1.4 3dmark test is horrible as well, scoring near Radeon 8500 levels. This leads me to believe that something is broken in the drivers...

[LeStoffer]

Originally Posted by Joe DeFuria

What changed in the GeForce drivers between revisions? Changed fp precision? Running in PS 1.1 (integer) vs. PS 1.4 (fp) mode?

My bet is that they changed fp precision. Since the game test #4 results are in line with game test #2 and #3, I don't think we are talking about running in PS 1.1 (integer) vs. PS 1.4 (fp) mode (although this could still be the case for GT 2 & 3 of course).

[Joe DeFuria]

In any case, it is still surprsing to me that in these GPU INTENSIVE tests that the GeForceFX with a 50% pixel rate advantage isn't absolutely clobbering the Radeon 9700 Pro.

That fact that the FX is "only" on Par in the DX8 benchmarks, and about 30-50% short of the Radeon 9700 in PS 2.0.....things don't look very good for the FX ARchitecture.

Either the FX architecture (or drivers?) is very inefficient, or Radeon 9700's larger bandwidth is having a bigger impact on shader performance than most of us anticipated.

[Joe DeFuria]

My bet is that they changed fp precision. Since the game test #4 results are in line with game test #2 and #3, I don't think we are talking about running in PS 1.1 (integer) vs. PS 1.4 (fp) mode (although this could still be the case for GT 2 & 3 of course).

Good observation!

[surfhurleydude]

Either the FX architecture (or drivers?) is very inefficient, or Radeon 9700's larger bandwidth is having a bigger impact on shader performance than most of us anticipated.

Even a Radeon 9700 vanilla scores much higher than the GF FX at roughly the same bandwidth... I'm not so sure that bandwidth has much to do with it... The differences in PS 1.4 speed between the Radeon 9700 Pro and Radeon 9700 are greater than the differences between the GeForce FX Ultra and the GeForce FX, so I'm thinking its perhaps the drivers or architecture that is holding it back. Considering the awesome performance of PS 1.1 tests on the GF FX, I'm hoping its the drivers.

[galperi1]

Originally Posted by surfhurleydude

Pixel Shader 2.0 for the GF FX seems VERY low... TOO low as a matter of fact... is PS 2.0 for the GF FX broken in the drivers or something? Hell, in almost every other previous PS test I've seen the GF FX SOLIDLY outperforms the Radeon 9700 Pro..

more like Nvidia is cheating somewhere in the drivers to get the performance increase from the original driver to the "updated" version

If you look at the tests (I'm doing Original Vs updated btw)


@ 1024 X 1048
-Game Test 4 - 18.7 vs 30.7 (64% increase)
-PS 2.0 - 12.0 vs 15.5 (29% increase)

@ 1280 X 1024
-Game Test 4 - 15.2 vs 21.2 (39% increase)
-PS 2.0 - 10.1 vs 12.6 (24% increase)

@ 1600 X 1200
-Game Test 4 - 12.3 vs 21.0 (71% increase)
-PS 2.0 - 8.5 vs 10.6 (25% increase)

While FPS for Game 4 blew up... PS 2.0 (Which is GT4) didn't vary by a huge degree. Where did they get these HUGE increases in the game but not in PS2.0 when both tests should be interrelated?

[Randell]

Originally Posted by surfhurleydude

Every test I've seen shows the GF FX lose badly in VERTEX shader performance, but not DX 8 Pixel Shader performance...

did you somehow manage to forget the APS result in Anands review?

http://www.anandtech.com/video/showd...i=1779&p=6

[surfhurleydude]

No, I did not...read my post in response to Joe asking whether I was talking about PS 1.1 or PS 1.4, then my post before this one

[Randell]

Originally Posted by surfhurleydude

No, I did not...read my post in response to Joe asking whether I was talking about PS 1.1 or PS 1.4, then my post before this one

aah yes sorry missed that - I was working on 'Geforce' in APS uses 1.1, doh! Brainfart