[H]ardocp Review for the Geforce FX

[gokickrocks]

well suprisingly im still up considering i need to get up early...
but i went back to look at the IQ pics...

and what i concluded...
9700 2xAA is similar to the FX 8xS
(look at the railings in the back for ut2003 and look at the cross beams at the top of the bridge for nfs2)

anyone see otherwise?

 

[Fred]

After cursory look at a few different benchmarks, I must admit i'm a little perplexed.

In various benchmarks on different sites, you see some benchmarks where the FX wins by 30% at some res/aa setting. Then on the next site, we see the exact opposite where the 9700 wins by a decent margin, on what appears to be the exact same bench/settings.

It could be an Athlon/Pentium issue, or some other registry setting thats being misused.

Until Anandtech does their benchmarks (and they tend to be the most reliable for pure benchmarking), im going to stick with Kyle Bennet's review (it seems to match Toms more closely than the other sites). One telling thing is the way they OC'd the FX. OC still shows gains at 1600*1200, indicating its not just a driver issue causing problems for the FX at that rez.

 

[Pete]

It may well be a platform issue--IIRC, nVidia cards work better with Athlons than ATi cards do. It may have something to do with the fact that they make Athlon motherboards.

I'd be more interested in comparing the 5800's 6xS to the 9700's 6x, as, according to B3D's NV30 preview, 6xS is skewed, while 8xS is ordered.

 

[Dave H]

After cursory look at a few different benchmarks, I must admit i'm a little perplexed.

Me too, although it's looking more and more as if the Extremetech benches are the odd ones out (too slow for GFFX).

One issue (dunno if this is causing some of the variance): do any of the sites say whether they're using "balanced" or "aggressive" AF on the GFFX? And, does anyone know whether the difference between the two is bilinear vs. trilinear ala R300's "performance"/"quality", or if it's turning on/off Nvidia's new more adaptive AF algorithm?

Anyways, could be a reason why Extremetech's AA/AF scores are so low...

EDIT: Having seen anand's clever mouseover AF comparison, the answer is "neither". Balanced looks way better than aggressive, and it's not because of funny polygon angles (the comparison zooms in on a distant piece of the floor), nor is it bilinear vs. trilinear (no mip-map boundaries). Instead it's simply "ugly" vs. "good-looking". Having seen that, I would guess (and certainly hope) everyone is using the balanced setting...

 

[Fred]

One oddity.

The villagemark scores show the fillrate advantage nicely on Toms site. It also refutes the most common belief that the bench is a test of bandwidth saving features. Not true, it tests fillrate and the ability of the card to discard pixels.

Its also one of the first time in recent memory that an Nvidia card has performed well on that bench.

 

[nggalai]

Well Toms has no pictures but sounds...gives a idea of fan noise.


The Fan

Sweet baby Jesus. The thing is louder and more annoying than my whole system with its case removed.

OK, let's sum up:

- draws 75W
- gets extremely hot
- is so large it has troubles fitting in medium towers
- gobbles up a PCI slot
- costs about €600
- sounds like a small-sized airport

Somehow, I don't think so . . .

ta,
-Sascha.rb

 

[Fuz]

I mean seriously..how can you do a video card review and not show ONE screen shot

I am trying to work this out my self. :?

 

[lwells]

Its Tom's -- need one say more?

Meanwhile both AnandTech and [H] review went to great length to highlight the IQ, which I think is a breath of fresh air...

LW.

 

[Nagorak]

Geforce FX AA is just no comparison to the R9700s. Who cares if the Geforce FX is faster in 4X AA than the R9700? In order to get a comparable image you'll have to run in 6X AA. Of course, I'm not a big AA user to begin with (starting to warm up to it a bit though). Overall the GF FX is faster than the R9700, but was there really any question of that? As soon as you saw the fan on it, you knew Nvidia was going to crank it up until it was faster.

As for max Image Quality, all I can say is OUCH. I hope that's just due to immature drivers? The IQ of the R9700 is just way higher in that shot too! After seeing that, it kind of seems like a very slanted review, since they never went beyond 4X AA while testing... Hrmm.....

Now the only question is when it will be available and when the R350 is going to hit the street...

 

[nggalai]

Surprisingly, I really liked the [H] review. Especially the "maximum quality" comparison shots are a good idea.

But something seems to be crooked with the GFFX's 2xMSAA mode. Sometimes it's there and then it's not? ???

ta,
-Sascha.rb

 

[Dave H]

Now the only question is when it will be available and when the R350 is going to hit the street...

Several of the reviews tonight are stating retail availability "in about a month" or "in February". These would appear to be official remarks (or at least official hints) from Nvidia.

As for R350, March seems thrown around a lot, but whether that means previews, reviews or availability, no one seems to say.

 

[tEd]

Surprisingly, I really liked the [H] review. Especially the "maximum quality" comparison shots are a good idea.

But something seems to be crooked with the GFFX's 2xMSAA mode. Sometimes it's there and then it's not? ???

ta,
-Sascha.rb

it doesn't really matter if it works or not it sucks either way IMO

 

[Nagorak]

When at its full speed in 3D under a heavy load of looping benchmarks at 4X AA and 8X AF the temp is 145 F on the back heatsink. We highly suggest letting it cool before removing it from your system.

 

[lwells]

When at its full speed in 3D under a heavy load of looping benchmarks at 4X AA and 8X AF the temp is 145 F on the back heatsink. We highly suggest letting it cool before removing it from your system.

If you read the forum thread that goes with the article....

Originally posted by FrgMstr
Brent went to bed....

You are speaking of a real possibility here considering they are right next to each other. Good point.

I was getting exhaust temps (I had the card here running for a few hours on Friday) at around 120F running the card in DOS.



__________________

Kyle "FrgMstr" Bennett
Webmonger @ [H]ard|OCP
Purveyor of Smoothness @ Ratpadz.Com

Ouch.

LW.

 

[Humus]

I'm surprised that Tom got technical enough about the Z-buffer issue, but not surprisingly he got somewhat confused. He blames the ChoosePixelFormat() problem on nVidia's drivers, while any problem with ChoosePixelFormat() should be directed at Microsoft who implements this function which is without control from the driver. It's way of selecting pixel formats are undocumented and have obvious problems. This is one of the reasons why the WGL_ARB_pixel_format extension was invented, which gives you a similar pixel format selection method as under X, and gives the driver control over the pixel format selection.

 

[Randell]

I'm surprised that Tom got technical enough about the Z-buffer issue, but not surprisingly he got somewhat confused. He blames the ChoosePixelFormat() problem on nVidia's drivers, while any problem with ChoosePixelFormat() should be directed at Microsoft who implements this function which is without control from the driver. It's way of selecting pixel formats are undocumented and have obvious problems. This is one of the reasons why the WGL_ARB_pixel_format extension was invented, which gives you a similar pixel format selection method as under X, and gives the driver control over the pixel format selection.

Humus - in laymans terms please how does this affect Toms conclusion about SS:SE results or is he just blaming the wrong cause?

 

[Dave Baumann]

Surprisingly, I really liked the [H] review. Especially the "maximum quality" comparison shots are a good idea.

Brent's ben a breath of fresh air for that site, in terms of 3D reviews. He's beginning to worry me a little...

 

[Luminescent]

Can someone explain if skewed grid is similar to jittered grid? It is superior to ordered grid, right? If so, the Geforce FX's skewed 6XS would be great for an AA image quality comparison with the 9700 (maybe against 4X on the radeon).

 

[Grall]

6xS is a mix of 4x ordered grid and 2x rotated grid, one being supersampling and the other multisampling. No, I don't know which one is which. I'm tempted to guess it's 2xMS and 4xSS since I don't think the hardware can do 2x rotated SS...

4xS was 2x ordered SS and 2x rotated MS after all, if the thing could rotate both varieties I'm sure they would have done so.

*G*

 

[Luminescent]

Does Nvidia use 4X rotated grid MS?