Real GF FX test at tecchannel

[Indiana]

Here:

http://www.tecchannel.de/hardware/1109/index.html

Unfortunately in german. The results don't look too good for the FX, though, only up to 10% faster in game benchmarks, sometimes even slower.

Especially interesting is the fact that the FX while a bit faster in Q3 standard lost against the R9700Pro in the FSAA+anisotropic filtering test. This for all those that claimed the GF FX wouldn't need 256bit wide RAM.... And I'm sure that if TecChannel had included even more FSAA and aniso tests, the GF FX would've had an even harder time, but then very likely NVidia wouldn't let them do this

It seems that this round has been won by ATI, the NVidia fans will have to wait for the supposedly much better performing NV35 refresh. We will have to wait for other reviews to confirm the tecchannels findings, though.

I think there would've been a place for a Pitou, looking at the not so bright GF FX numbers, and here especially considering the mediocre results in the tests with high quality (high FSAA and aniso levels).

[dZeus]

one question about the review:

why run SPECviewperf? They say the test strains the RAM and graphics core a lot, but isn't driver optimisation a big factor in the results? If so, this test is completely meaningless, since both manufacturers have a different range of cards for the workstation/CAD market.

Further more, isn't CPU speed a bigger bottleneck for the botmatches than the videocard's / driver's speed in most situations?

I can only see 1 FSAA/Anisotropic filtering test result, so basing any conclusions on just that might be premature. I do see the card winning nearly any other test, or at least being equally fast to the 9700 Pro. Pretty good for a card with just 128-bit memory bus, though perhaps not worth the cost savings on that front if the total production costs of the card are higher (are they?). I've read that the PCB was very complex, and DDR-II probably doesn't come that cheap either.

[BuddMan]

Let's not forget the R350 isn't far off. If the GeforceFX doesn't do too well against the 9700, just wait till ATI releases their refresh. Rumors are saying March.

[thop]

I smell another 3DFX in the industry.

[Indiana]

Originally posted by dZeus

why run SPECviewperf? They say the test strains the RAM and graphics core a lot, but isn't driver optimisation a big factor in the results? If so, this test is completely meaningless, since both manufacturers have a different range of cards for the workstation/CAD market.

....

I can only see 1 FSAA/Anisotropic filtering test result, so basing any conclusions on just that might be premature.

1. Yes, SPECviewperf has IMO nothing to do in a gamers card review and by the dramatically different scores between the GF2 and the Quadros based on the exact same chip (same goes for ATIs Radeon vs. FireGL) you can clearly see that this score is mainly dependent on driver optimization (or the lack thereof). But then this is only one test in the review.

2. I complained about them having only one FSAA and anisotropic filtering test in my post above. It's totally unimportant if you have 250 or 270 fps, so why not use these settings. Only reason for this I can imagine is that NVidia did give them some points on how to review the card thus preventing tests with HQ settings - this is especially considering the GF FX really doesn't look good in the single FSAA test, it loses to the R9700Pro by a considerable margin.

3. Looking at the availability and price tag of DDR-2 RAM clocked as high as 500MHz and the fact that ATI apparently even uses the 256 bit wide memory bus on its el-cheapo non-Pro cards, I'm quite sure that the 256 bit wide bus is cheaper than the extremely rare 500MHz DDR2 RAM. The 256 bit wide bus only rises the costs for the mainboard but you can use normal clocked cheap RAM.
NVidia as it seems needs a very complex mainboard to be able to grant troublefree operation at those excessive clockings.

[R.Carter]

Maximum PC Magazine has published some NV30 benchmarks as well.

You can find the results here.

[Guru]

580€ for the ultra version over here in Finland!