GeforceFX - what will be the best way to benchmark it?

[Geeforcer]

Nvidia's path seems to take several months longer and quite a few degrees hotter.

For now, there is no way (for web-reviewers) to benchmark the NV30 card. It may as well still be nonexistent/vaporware. Most disappointing <i>launch</i> ever.

--|BRiT|

How is this launch different from any other hardware-less launches of the past?

 

[BRiT]

How is this launch different from any other hardware-less launches of the past?

The launches that I like have benchmark numbers and more information that was revealed. Perhaps it's the fact that this NV30 launch added very little to the pool of information. I'm hoping that there's quite a bit to the NV30 that we still don't know. Even on the AMD/nforce2 launches where the product wasn't readily available until month(s) later, we had some solid numbers from external parties - hardware review sites.

ATI's latest launches at least said the hardware would be available in 30 days or less. This launch 3 months before availability is utter crap.

But yes, I suppose I'm bitching about it being hardware-less. But this one was even less than the other ones.

--|BRiT|

 

[antlers]

If they do manage to get boards to review sites in December and ramp up in January, it won't really be that different from ATI's 9700 launch.

 

[BRiT]

How's that Antlers4? From what I recall, I don't see the simularity at all. The review sites had ATI R300 boards at the time when they launched, and then the boards were available in retail 30 days later.

--|BRiT|

 

[Clashman]

Taken from anand's preview

It seems that there is at least 1 benchmark of the card. If you look closely, however, it appears that this is taken from a custom nvidia level. Has anyone seen this or checked it out? I wonder what the differences are between this nv demo and the actual game.

 

[ben6]

err I was at the 9700 launch , and the boards didn't get out to reviewers till August when the launch was in July. ATI did let us play with boards in San Francisco though

 

[Entropy]

It seems that there is at least 1 benchmark of the card. If you look closely, however, it appears that this is taken from a custom nvidia level. Has anyone seen this or checked it out? I wonder what the differences are between this nv demo and the actual game.

It favours the NV30.
Why else would they construct not only their own demo, but their own level to benchmark with? Which, incidentally, is the only comparative benchmark we have seen? Occams razor works wonders....

It does provide information though.
The NV30 is extremely unlikely to exceed this level of advantage vs the R9700. In fact, its relative performance in DOOM3 is pretty much guaranteed to be significantly below the level depicted in this graph.
Which is info too, of a sort.

Entropy

 

[Clashman]

It favours the NV30.
Why else would they construct not only their own demo, but their own level to benchmark with? Which, incidentally, is the only comparative benchmark we have seen? Occams razor works wonders....

Entropy

Yeah, I thought the same thing but kind of thought it would be redundant to mention. Still, I would like to see this nvdemo in comparison to the rest of the game.

 

[jvd]

well some sites might bench it with 2x fsaa and 4x aniso against the r9700 pro with 16x aniso and 6x fsaa...

sorry i allways wanted to use a

Serously I think we should find a setting that all review sites.. the major ones at least can agree on. A setting that makes both cards look equal. Then have everyone benchmark at those settings. So if a radeon 9700 looks better at 8x performance aniso compared to the 8x normal of the geforce fx it should be benchmarked that way. If the geforce fx 4x fsaa looks as good as the r9700 pro's 6x fsaa they should be compared at those... that would be fair but of course i can't see that happening.

 

[Ichneumon]

Taken from anand's preview

http://images.anandtech.com/reviews/video/nvidia/geforcefx/preview/doom3.gif

It seems that there is at least 1 benchmark of the card. If you look closely, however, it appears that this is taken from a custom nvidia level. Has anyone seen this or checked it out? I wonder what the differences are between this nv demo and the actual game.

That benchmark not only is of a custom nvidia level, but was provided to Anand by Nvidia... heh i'd take that one with a large grain of salt.

 

[antlers]

How's that Antlers4? From what I recall, I don't see the simularity at all. The review sites had ATI R300 boards at the time when they launched, and then the boards were available in retail 30 days later.
|

I may be mistaken, but I believe the 9700 launch was 30 days (it may have been shorter) before reviewers were allowed to post benchmarks (although they may have had the boards earlier ). Thus there was that interval where we had to debate the relative performance over the 4600 that was the only information available at launch.

Some boards were available at retail at about the same time benchmark numbers came out, but wide availability was about a month later.

Am I remembering wrong?

 

[Crusher]

NVIDIA had a custom level for Quake 3 back in the day as well, so this isn't much of a suprise. However, rather than make a simplistic demo that overestimates the typical in-game framerates, I would think they would make something more complex than the game itself, to show off how much the NV30 can do. No doubt that level is designed to emphasize NVIDIA's advantages over the competition and downplay their weaknesses, but I wouldn't expect it to show an extremely over-exaggerated framerate. If anything I would expect the actual game framerates to be higher than this. It's probably a safe assumption that the performance difference between the GeForce FX and Radeon 9700 Pro (if they are ever compared with this benchmark) will be greater on this demo than in the actual game.

 

[WaltC]

Nvidia's path seems to take several months longer and quite a few degrees hotter.

For now, there is no way (for web-reviewers) to benchmark the NV30 card. It may as well still be nonexistent/vaporware. Most disappointing launch ever.

--|BRiT|

*chuckle* Good answer... Heh-Heh..."several months and lots of C"...*chuckle*

I'm not as down on product announcements as some people, generally speaking. And I don't much like the term "paper launch" because announcing products months before they appear has really been a norm in the industry for most of the last 17 years that I can recall. But...

In this case I find it at least a little troubling that there were no demo cards available to circulate to web sites for reviews. In the case of an expensive, high-profit, high-margin product like GFFX I think waiting until you could ship demo cards to reviewers would have made for a far, far more interesting and effective product announcement, even if actual shipping products were yet a couple of months off.

I wonder why they did this? What they've done seems premature, and the amount of information they've left out is pretty substantial. I'm not sure whether this has been deliberately left out, or whether there are still some fundamental questions about the aspects of the final silicon that are far from being nailed down at present. I think one reason for this kind of superficial release is maybe to reassure business investors that all was well. I don't know--there are still so many unanswered questions. If not for investors, then why bother with something as weak as this?

It just strikes me as worrisome that nVidia is claiming the FXes will be available for consumers in January '03 (the official line is February), yet here we are a scant 8 weeks or so away and nVidia doesn't have any demo cards to lend out for reviews. Is something not right with this picture? Or does nVidia plan to go straight to market and bypass the review circuit completely until after the product begins shipping? (It's been done before.)

nVidia guys told Demo_Coder, for instance, that he could pick one up in January. OK, for that to be the case everything has got to be very near finalized right now--so why the lack of detailed information and the complete lack of review cards sent out to websites? I can't recall exactly, but *I think* what ATI did with the R300/9700 Pro was to announce the product and send out review samples almost immediately--even though shipping was still between 1-2 months away (Gladly be corrected if I'm wrong.) So if we are indeed this close to shipping it doesn't make a lot of sense for nVidia to be so close to the vest. As I can't think of any benefit for them in remaining secretive at this point, I think it must have something to do with nVidia's own internal doubts as to certain final aspects of the GFFX products. Ah, but if that's the case, then we can't be close to 8 weeks out from shipping. But I'm certainly not going to say that the GFFX won't show up in January. I'm disappointed by this, too, and have to confess that even though the bare outlines of nv30 are what I had expected, this poverty of information at the launch of the chip plus the absence of reviewable hardware makes for an interesting enigma (of sorts.)

 

[WaltC]

How's that Antlers4? From what I recall, I don't see the simularity at all. The review sites had ATI R300 boards at the time when they launched, and then the boards were available in retail 30 days later.
|

I may be mistaken, but I believe the 9700 launch was 30 days (it may have been shorter) before reviewers were allowed to post benchmarks (although they may have had the boards earlier ). Thus there was that interval where we had to debate the relative performance over the 4600 that was the only information available at launch.

Some boards were available at retail at about the same time benchmark numbers came out, but wide availability was about a month later.

Am I remembering wrong?

As I recall everything went ballistic with the 9700 Pro when the NDA expired--and people already had cards in hand, articles written, and stories ready to go. I wonder if there is a similar NDA afflicting nVidia people right now which might also be affecting nVidia's own ability to release more info. But, hey, if so then why did nVidia schedule a launch at this time??? They could have at least rescinded the NDA if one exists right now, I would think. Just seems strange to hog-tie your own product announcement in such a fashion.

 

[antlers]

I think any comparative DX9 benchmark should definitely include displacement mapping

 

[MikeC]

In this case I find it at least a little troubling that there were no demo cards available to circulate to web sites for reviews.

A couple of thoughts come to mind - more time to optimize drivers, which NVIDIA has control over, and waiting for a DirectX 9 benchmark, which they have less control over.

 

[demalion]

I wonder if there is a similar NDA afflicting nVidia people right now which might also be affecting nVidia's own ability to release more info. But, hey, if so then why did nVidia schedule a launch at this time??? They could have at least rescinded the NDA if one exists right now, I would think. Just seems strange to hog-tie your own product announcement in such a fashion.

What computer shows are there between now and February? How significant are they to investors and the computing world, if there are any?

 

[Glonk]

People buy cards for the moment, not the future, just compare what's out there and let the numbers speak for themselves: nothing complex about it.

Huh? Most people I know buy so they don't have to upgrade every couple weeks.

All you'd need right now is a GeForce 4 MX...

 

[cho]

3D Mark2003 already on the way and the alpha version was around in the labs of ATi , NVIDIA .... .

 

[Entropy]

It favours the NV30.
Why else would they construct not only their own demo, but their own level to benchmark with? Which, incidentally, is the only comparative benchmark we have seen? Occams razor works wonders....

Entropy

Yeah, I thought the same thing but kind of thought it would be redundant to mention. Still, I would like to see this nvdemo in comparison to the rest of the game.

Sorry.
It should be redundant to mention, true.
I cite lack-of-caffeine induced grumpiness in my defence.

Entropy