Tables Turn Again (iXBT/Digit-Life Gainward 5900 GS review)

P

Pete

Moderate Nuisance
Moderator
Legend
Consumers get dizzy, fall off--news at 11.

Well, more trading places. At this point, it would appear the cards are still fairly even overall in current *games*, though each has its strong points (FX: AF, 9800: shaders). A good review, and better than their last (the MSI/3DM03 one).

As usual, it would be nice to see Doom III and HL2 demos finally released to put an end to some of this future performance speculation. I'll be sitting on the sidelines, cheering with my 9100. ;)
 
But it's actually a matter of taste: someone prefers shaders, another wants a beautiful card with a high performance in modern games. Besides, it's still unknown how future DX9 games will look like. Also one should remember that the NV3x is very flexible, which potentially gives a chance to the NV35 to look not worse than its competitor in such applications. I also wish the prices for the FX 5900 were adequate (lower than those of the RADEON 9800 PRO to increase demands).
Click to expand...

:oops: :oops: :oops:

Of course modern games will not use shaders. :rolleyes:
 
Their conslucion starts out o.k., but then really loses it.

In the tests where the anti-detect has no effect (i.e. where nvidia has no cheats, yet) the GFX get readily smoked by 9800, especially with AA&AF enabled. The difference is huge.

Referring to accessory packs is really stretching for straws, I think. Who buys highest end graphics cards for the accessory pack? Ok, throw in a good flat panel display and we can talk about it... ;)
 
Well, typical XBT review with it's obvious nVidia slant. However, even with their underlying leanings, even they cannot deny that the 9800Pro is equal with the 5900Ultra...... now they do seem to confuse the Ultra with the 5900 in this review, don't they? BTW, did anyone notice they compare - performancewise nVidias' 8X AF vs. ATIs' 16X? ;) And, while comparing nVidias' performance and quality modes to ATIs' quality mode only? :rolleyes: Of course they do say this about it:

The tests with the new demo benchmark show that the FX 5900 takes the lead only in the modes with the AA and anisotropy disabled and in the Performance mode with the anisotropic filtering used. But taking into account the way of realization of anisotropy in the RADEON 9800 PRO, the maximum level of the ATI's products can be compared to the Performance level (former Balanced) of the NVIDIA's solutions.
Click to expand...

Here's my fav quote:
I must say that we found no cheats in ATI's RADEON for this test! Probably, they are deeply hidden, but developer of Anti-Detect Aleksei Nikolaichuk considers that there is none.
Click to expand...

Overall, a pretty slanted review that still can't deny the real truth, but that does paint the 5900 in as positive a light as is possible.....
 
martrox said:
...
Overall, a pretty slanted review that still can't deny the real truth, but that does paint the 5900 in as positive a light as is possible.....
Click to expand...

Yes...What they are trying to do is to justify their many pre-release opinion pieces on nv35 and how much better it was going to be than already-shipping R3xx's, while admitting at the same time that nv35 isn't as good a chip... :D Quite a tall order and it's no wonder they fall all over themselves while doing it. It's a good reason why I usually ignore anything those sites generate and have for sometime. I don't think their goal is as much to promote bias as it is to try and wipe the egg from their faces...
 
Probably, they are deeply hidden
Click to expand...

I thought it was rather funny.

On one hand you still see posts stating that ATi drivers, & driver team, are rubbish & all of a sudden they are sooooo good they are better at hiding the cheats than nV! :? :LOL:
 
Re: Tables Turn Again (iXBT/Digit-Life Gainward 5900 GS revi

Pete said:
A good review, and better than their last (the MSI/3DM03 one)
Click to expand...

No it is not. Take a look at this:

Ixbt said:
However, some of NVIDIA's optimizations are useful: just compare the performance with the Anti-Detect and without it on our demo benchmark (not the standard one). Since it's impossible to make cheats for our demo version, the difference is caused by useful optimizations which were disabled by the Anti-Detect. Below are several pictures that prove that the quality is the same with the AD and without (but completely identical screenshots can't be attained for per-pixel comparison because the game uses random actions).
Click to expand...

The writer has completely misunderstood the issue with application detection. He goes to say that because he is using his own timedemo, he is immune to game engine-specific cheats. The image quality is not the same between the pictures he links! When you zoom the pictures, you can clearly see that the mipmap transitions are different, which is exactly what is going on with UT 2003 and Detonator drivers, the drivers not using full trilinear filtering even when the application asks for it.
 
Re: Tables Turn Again (iXBT/Digit-Life Gainward 5900 GS revi

Bolloxoid said:
The writer has completely misunderstood the issue with application detection. He goes to say that because he is using his own timedemo, he is immune to game engine-specific cheats. The image quality is not the same between the pictures he links! When you zoom the pictures, you can clearly see that the mipmap transitions are different, which is exactly what is going on with UT 2003 and Detonator drivers, the drivers not using full trilinear filtering even when the application asks for it.
Click to expand...

Not only that, there were no screenshots for the other benchmarks and games he accuses nVidia of cheating in. Judging cheats just by the numbers alone seems like a dangerous idea to me, surely once you see the discrepancy in the score you should start to look into exactly why that difference is there.
 
I'll let you in on a little secret: my standards aren't as high as I imagined them to be. ;) I'm not expecting the world from every review, just a decent core that I can add to other basically solid ones to piece together a reasonable picture of a card.

Puzzling commentary aside, the benches alone paint the picture for me. I didn't check the UT2K3 pics as I don't have that version of WinRAR, but we've already covered the issue here. I have both benchmark numbers, though, so the picture is kind of extraneous.

I'm curious about nV's Performance AF, though, as it seems to take a virtually nil performance hit. Is ATI's equally "inexpensive," or does nV hold the edge in both IQ and speed?
 
In the particular case of ut2k3 isnt it the case if nvidia is not using trilinear that any benchmarks done using it should be compared to ATis performance Af?

I'm curious about nV's Performance AF, though, as it seems to take a virtually nil performance hit. Is ATI's equally "inexpensive," or does nV hold the edge in both IQ and speed?
Click to expand...

Well Ati's bilinear performance af also has a pretty small performance hit. i.e. 16xaf perf was equal to 2xaf quality in some cases on the r9700
 
Re: Tables Turn Again (iXBT/Digit-Life Gainward 5900 GS revi

Bolloxoid said:
The writer has completely misunderstood the issue with application detection. He goes to say that because he is using his own timedemo, he is immune to game engine-specific cheats. The image quality is not the same between the pictures he links! When you zoom the pictures, you can clearly see that the mipmap transitions are different, which is exactly what is going on with UT 2003 and Detonator drivers, the drivers not using full trilinear filtering even when the application asks for it.
Click to expand...

i think there is nobody will zoom the content which in the framebuffer when play the ut2003, so there is almost not any visible mipmap level beam when play the ut003.

ATi also cheat in defaut , even set to the quality mode(turn on the trilinear filtering when turn on AF) in driver(most people do that).

And, ATi's AF on do at 45d, 90d, 180d, can we said it is not "defer to the applications" :)
 
Re: Tables Turn Again (iXBT/Digit-Life Gainward 5900 GS revi

cho said:
i think there is nobody will zoom the content which in the framebuffer when play the ut2003, so there is almost not any visible mipmap level beam when play the ut003.
Click to expand...
almost not != none
ATi also cheat in defaut , even set to the quality mode(turn on the trilinear filtering when turn on AF) in driver(most people do that).
Click to expand...
ATi offers full trilinear by default. ATi offers full trilinear with AF if the application enables it. Can you say the same about that other brand?
And, ATi's AF on do at 45d, 90d, 180d, can we said it is not "defer to the applications" :)
Click to expand...
I can't make sense of this.

Edit: Fix bad quote formatting.
 
cho said:
i think there is nobody will zoom the content which in the framebuffer when play the ut2003, so there is almost not any visible mipmap level beam when play the ut003.
Click to expand...

That is not the issue here. The issue is that Nvidia encourages people to check the filtering quality of their hardware with a filtering tester application and then (selectively, secretly and beyond users' control) does not apply the same level of filtering in the case of a popular benchmarking title, thus intentionally misleading people.
 
Re: Tables Turn Again (iXBT/Digit-Life Gainward 5900 GS revi

cho said:
i think there is nobody will zoom the content which in the framebuffer when play the ut2003, so there is almost not any visible mipmap level beam when play the ut003.
Click to expand...

You don't need to zoom the content - you can see that you are pretty much dropped to bilinear filtering (or at least their highest "Performance" more) on detail textures. These things become more evident in motion than on still screenshots.

ATi also cheat in defaut , even set to the quality mode(turn on the trilinear filtering when turn on AF) in driver(most people do that).
Click to expand...

"Default" is without AF on, and even then you do not get the correct IQ with NVIDIA's drivers. However, as has been pointed out, the issue at stake is that there is a clear inconsistency between the IQ modes that NVIDIA tell people to promote and what is actually offered under UT with their drivers.
 
Re: Tables Turn Again (iXBT/Digit-Life Gainward 5900 GS revi

DaveBaumann said:
You don't need to zoom the content - you can see that you are pretty much dropped to bilinear filtering (or at least their highest "Performance" more) on detail textures. These things become more evident in motion than on still screenshots.
Click to expand...

Indeed. I ran my old UT2003demo (build 2002-09-13_17.31) at default settings and I just started wondering what settings to use to get it to look that bad on my 9600pro (compared to iXbt:s screenshots) :)
 
I thot this was a pretty damning statement for NV:

"So, if the previous tests based on the standard demo benchmarks or games included in the standard packets showed that the GeForce FX 5900 had a good chance to beat the RADEON 9800 PRO, the elimination of the cheats by replacing the standard demo benchmarks with specially developed ones revealed that the FX 5900 is not that good."

If that kind of statement starts showing up in more reviews, NV has big troubles.
 
You must log in or register to reply here.
Back
Top