Gotta get a new video card

[GNEP]

bsd - having read round a bit it seems that the "cheating" goes much deeper than that, with NV and quite likely with ATI - I guess ALL review benchmarks, even "gaming" ones have a probabilty >0 of being inflated. It makes the situation very confusing if looking for a new card. Things like replacing shader code, application detection, bencmark-mode detection etc...

[The PIT]

Well dunno about ATI drivers getting 20% faster on each release. I haven't noticed much improvements over the last drivers releases.
Cheating on the drivers is getting a problem since both ATI and Nvidea want to be the fastest card out there. Much better if they concentrated more on bugs and IQ. Reviewers could help this by not concrentrating on FPS so much.

[bsdgeek]

Maybe this is a good time to announce it, I'm going to start a new review site soon. (www.cghq.com - you can visit, but everything there is basically a mockup) Do you guys have any suggestions on what I should do for reviewing?

Note: Use anything but IE, because IE does not support background-attachment: fixed; which I used in my CSS (which is a mess right now, needs to be cleaned/organized)) And you should have JavaScript on again because I had to use separate CSS for IE.

[The PIT]

Have probs with IE isn't a good idea since microshaft has got nearly everyone using it.

It would be interesting if you renamed the benchmarks and see if it effected the results.

[bsdgeek]

I know, it still works though, it just doesn't look as good.

I will try renaming.

[Gurm]

bsdgeek,

Some of us are already working on "The Flamethrower", an honest review site swimming in a sea of bought-and-paid-for reviewers.

- Gurm

[bsdgeek]

Good luck

[Kooldino]

Well, looks like it's about time for me to place an order. I'm thinking of going w/ the Powercolor...any last minute advice?

[Kooldino]

Uh oh...looks like a new competitor...Gigabyte...Hmmm...

http://www.newegg.com/app/viewproduct.asp?DEPA=&submit=Go&description=radeon+VIVO

[thop]

Did you say HardOCP?

[Kooldino]

Ok, I went w/ the Gigabyte. Can't beat that card for $90 shipped.

[GNEP]

[Indiana]

Sorry for the late reply, saw this thread only now. But there are some things said of which I feel that they definitely need to be adressed...

Originally posted by bsdgeek
Gurm, the 5200 has FSAA that is:
A) Faster.
B) Better Quality.
C) Actually usable (due to the supersampling on the 9200, it is very slow).

Nope. Sorry to disagree, but IMO Supersampling is still superior to multisampling from a sheer image quality standpoint for nearly all older games and even lots of the actual ones. Of course Supersampling is unusably slow, but so is the FSAA on the GF FX5200. Nearly none of the current games deliver framerates that are actually "playable"/"enjoyable" with FSAA even on the FX5200 - especially NOT if you take a look at the minimum framerates. Because of that the fact that the 9000/9100/9200 are even slower is not really important...

Without AA, they are pretty much equal on speed. (They both trade off victories in different games)

Well, that might be true for the 9000/9200, but the R9100 is a rebadged R8500LE and thus is only slightly below the performance of a GF4 Ti4200 - which is WAY FASTER than a GF FX5200Ultra, not to speak about the non-Ultra.

Gurm, I understand what you are saying. But, the cheating was about implementing static clip planes in 3dmark and lowering pixel shader FP precision, with NV dumbing down to 16FP or FX12 (integer). Nothing there about AA, and it is still in 3dmark. So look at the game testing.

I forgot to mention they did lower AF quality for a short time, but now it is fixed, and NV supposedly has a better method for AF than ATI as well. I'm not sure whether or not 9200 can run Trilinear + AF simultaneously either. (Because it is based on 8500 which could not.) [/B]

The driver cheating at nVidia goes MUCH deeper than this. The nVidia drivers detect the "common" Quake3 benchmarks and force a low texture LOD thus gaining speed and scrificing image quality. See benches with the "default" Q3 benchmark runs and custom made ones yourself and be astonished...
Same detection routines with "custom settings" are used for most if not all popular /benchmarkable games.
Other example: The nVidia drivers do force bilinear filtering in UT2003 regardless of the user settings, thus leading to review sites reporting wrong numbers for UT2003 with trilinear set.
At least most GF FX nVidia driver do force 16Bit FP rendering if not even 12 Bit integer on ALL DX9 shader-software, and not just 3DMark as some people like to tell.

For a more thorough overview about the MASSIVE and previously unparalleled cheating going on there have a look at the various threads at the excellent Beyond3D Forum.

(P.S.: nVidias newest Beta drivers now suddenly are encrypted after people began to search for custom game/benchmark detection routines in nVidia driver files and still haven't finished yet finding new examples of this sh*t going on. But no, they have nothing to hide....)


P.S.: For the prices you listed (>>100$!!) I would definitely go for a Radeon9600 today and the oc' the hell out of it...

[bsdgeek]

Originally posted by Indiana
Sorry for the late reply, saw this thread only now. But there are some things said of which I feel that they definitely need to be adressed...


Nope. Sorry to disagree, but IMO Supersampling is still superior to multisampling from a sheer image quality standpoint for nearly all older games and even lots of the actual ones. Of course Supersampling is unusably slow, but so is the FSAA on the GF FX5200. Nearly none of the current games deliver framerates that are actually "playable"/"enjoyable" with FSAA even on the FX5200 - especially NOT if you take a look at the minimum framerates. Because of that the fact that the 9000/9100/9200 are even slower is not really important...

It is true that Supersampling provides a sort of texture filtering, but you must also consider that the 9200 uses an ordered grid method compared to 5200's rotated grid (which provides better edge entialiasing).

Also, 2xAA on the 5200 is quite playable in current games at lower res (=/<1024x768).

Well, that might be true for the 9000/9200, but the R9100 is a rebadged R8500LE and thus is only slightly below the performance of a GF4 Ti4200 - which is WAY FASTER than a GF FX5200Ultra, not to speak about the non-Ultra.

And we were talking about the 9000/9200. Which is still slower than 5200 nonUltra.

The driver cheating at nVidia goes MUCH deeper than this. The nVidia drivers detect the "common" Quake3 benchmarks and force a low texture LOD thus gaining speed and scrificing image quality. See benches with the "default" Q3 benchmark runs and custom made ones yourself and be astonished...
Same detection routines with "custom settings" are used for most if not all popular /benchmarkable games.
Other example: The nVidia drivers do force bilinear filtering in UT2003 regardless of the user settings, thus leading to review sites reporting wrong numbers for UT2003 with trilinear set.
At least most GF FX nVidia driver do force 16Bit FP rendering if not even 12 Bit integer on ALL DX9 shader-software, and not just 3DMark as some people like to tell.

Maybe I should just sidestep this and say it doesn't matter. (as per [T])

Well I didn't know all of this at the time, so I will admit I am wrong on the cheating, but that wasn't exactly what I was trying to say. I was attempting to say that it did not effect FSAA directly.

For a more thorough overview about the MASSIVE and previously unparalleled cheating going on there have a look at the various threads at the excellent Beyond3D Forum.

(P.S.: nVidias newest Beta drivers now suddenly are encrypted after people began to search for custom game/benchmark detection routines in nVidia driver files and still haven't finished yet finding new examples of this sh*t going on. But no, they have nothing to hide....)


P.S.: For the prices you listed (>>100$!!) I would definitely go for a Radeon9600 today and the oc' the hell out of it... [/B]

Yes I have been reading there, and sometimes getting involved in the discussions.

The Gigabyte was still a good choice.

[leech]

Speaking of AA, Sampling, etc.... I have a simple question. WHAT THE HELL IS THE POINT? When you're running 1600x1200 resolution, what is the point in having AA on? Whenever I see a post that says "Oh my card xxx is so fast because I can play xxx at 1600x1200 with full AA and AF blah blah blah blah blah." Granted having the Trilinear filtering and Anisotropic filtering on DOES make a difference in how everything looks (especially Serious Sam!) But really, do people NEED to have 1600x1200 with AA? I understand having it on in lower resolutions from 1024x768 on down, but really it shouldn't be needed at the resolutions where all it does is slow down the game play. If I ran something like say Grand Theft Auto:Vice City at 1600x1200 with and without AA, the only difference I would notice is that with AA it'd be less playable.

I honestly want to know who can really notice any difference? The pixels are so small at that resolution that anti-aliasing things doesn't do any good (especially FSAA)

Just a quick rant.

Leech