Are they ALL-IN-WONDERs?

Well, AFAIR, all Radeon 9000 were 64-128 MB ones.
256 MB was supported on Radeon 9200.
Radeon 9000 is the cutted Radeon 8500 with half of vertex/texturing units and w/o Trueform. Plus (not sure) there is no DX7 Hardware TCL.
Radeon 9200 is the Radeon 9000 with the AGP 8x support.
And yes, you can't say "these cards are updated Radeon 7500"...

The whole description is:
New ATI Radeon 7500 64MB PCI Video Graphics Card DDR 64-bit DVI /VGA /S-Video. Memory Size : 64MB onboard (256mb utilizing TurboCache) Specifications:Graphics Processor: ATI Radeon90 00
So adding TurboCache, is it safe to say the seller himself is clueless and sold so much because he under-priced 9000 as 7500?

Sorry, forgot to post the links. I paid $7 shipped for one of these cards new, but I was hoping it was a 7500 card. I already have a 9250 PCI card. It sounds like the advertised "7500" card is really a "9000" card then? Is there any benefit to having the 9000 over the 9250? Will 7500 drivers work on this fake 7500 card?

feipoa wrote:

It sounds like the advertised "7500" card is really a "9000" card then?

Well soon you will find out.

feipoa wrote:

Is there any benefit to having the 9000 over the 9250?

Most probably not, considering this one is 64 bit.

feipoa wrote:

Will 7500 drivers work on this fake 7500 card?

No, 9000 is generation apart from 7500.

Niche - those might be useful for DVD playing on older machines without AGP (both 7500 and 9000).
Or maybe the 9000 can be cross-flashed for Mac, IIRC there was some Mac model were a 9000 is the fastest option.

Don't know if any of these would explain > 1500 sales, though.

Working drivers: Well, ATI/AMD drivers are universal anyway. But you probably can't use some driver versions between release of 7500 (8/2001) and 9000 (8/2002).

Gamecollector wrote:

Radeon 9000 is the cutted Radeon 8500 with half of vertex/texturing units and w/o Trueform. Plus (not sure) there is no DX7 Hardware TCL.

The DX7 hardware T&L will run via the vertex pipeline. So it is hardware-accelerated, but it's slightly slower than on the 8500/9100 because it has only vertex pipeline instead of two.

What makes these cards interesting imho is that they are the only family of cards ever produced that have true ps1.4.
nVidia skipped right from ps1.3 to ps2.0.
ps1.4 is considerably more powerful than ps1.3, because it allows much longer shaders, it has a much bigger and more flexible instructionset, and it has higher internal precision. ps1.4 is actually closer to ps2.0 than it is to ps1.3.

For games that take advantage of ps1.4 explicitly, such as Half Life 2, these cards give better image quality than their GeForce3 counterparts.
Performance-wise I'm not too sure with the 9000, as it is a slower, low-budget model. I only used a Radeon 8500 myself. That card is more or less comparable with a GeForce3 Ti500 in ps1.1 games. With ps1.4, it could render some things in a single pass that took a GeForce 2 passes. In such cases it could even keep up with the faster, newer GeForce4 series.

At least the photo in the offer shows a fan usually meaning it is a higher-clocked 9000 Pro.
I'm slightly confused about the 256MB RAM, those should go with a 9200 (not that this amount of RAM makes sense for the chip).

These places seem to sell the same card, a 9000 with PCI bus and 256MB memory:
http://www.geeks.com/details.asp?invtid=V9000-P256
http://www.amazon.co.uk/ATI-Radeon-9000-Video … t/dp/B008R7UJD6

Seems to be the exact same card.

Scali wrote:

These places seem to sell the same card, a 9000 with PCI bus and 256MB memory: http://www.geeks.com/details.asp?invtid=V9000-P25 […]
Show full quote

These places seem to sell the same card, a 9000 with PCI bus and 256MB memory:
http://www.geeks.com/details.asp?invtid=V9000-P256
http://www.amazon.co.uk/ATI-Radeon-9000-Video … t/dp/B008R7UJD6
Seems to be the exact same card.

Well, the GPU photo (with removed cooling unit) will be the final test... IMHO, it is the 9200 (RV280).

GeForce FX supports PS1.4. Yeah they are PS2.0 cards but that's more of a feature bullet than a practical reality. 😀

swaaye wrote:

GeForce FX supports PS1.4. Yeah they are PS2.0 cards but that's more of a feature bullet than a practical reality. 😀

All PS2.0+ cards support PS1.4. D3D enforces that you support all lower pixelshaders as well.
But yes, FX was so bad that it was often better to run a game in PS1.4 mode than in PS2.0 mode.
I'm not entirely sure whether PS1.4 ran on an actual integer pipeline, or if it ran on the PS2.0 units with reduced precision (they had FP16 support, which was much faster than FP32 support... But Radeon 9x00 only had FP24, and most software did not go for reduced precision, since it didn't do anything on Radeon 9x00... so when the FX came out, most software would default to FP32).

Scali wrote:

All PS2.0+ cards support PS1.4. D3D enforces that you support all lower pixelshaders as well.

Oh that makes sense.

By the way I've never seen a Radeon 8500 match up to a GF4 regardless of PS1.4's theoretical advantages. The only in depth info I recall seeing about the reason for this was a post by John Carmack regarding his work on Doom3 for the GeForce3/4 and 8500.

http://www.bluesnews.com/cgi-bin/finger.pl?id … =20020211165445

Scali wrote:

But yes, FX was so bad that it was often better to run a game in PS1.4 mode than in PS2.0 mode.
I'm not entirely sure whether PS1.4 ran on an actual integer pipeline, or if it ran on the PS2.0 units with reduced precision (they had FP16 support, which was much faster than FP32 support... But Radeon 9x00 only had FP24, and most software did not go for reduced precision, since it didn't do anything on Radeon 9x00... so when the FX came out, most software would default to FP32).

Yeah I read something about how GFFX may be emulating PS1.4 to some degree. I wonder if FP16 / fixed point conversion would cause any visual changes. Though frankly FX12 or whatever those shaders use already causes precision related uglies so who would notice slight changes.

swaaye wrote:

By the way I've never seen a Radeon 8500 match up to a GF4 regardless of PS1.4's theoretical advantages. The only in depth info I recall seeing about the reason for this was a post by John Carmack regarding his work on Doom3 for the GeForce3/4 and 8500.

http://www.bluesnews.com/cgi-bin/finger.pl?id … =20020211165445

Yup, he wrote a R200-specific renderpath to leverage the better shaders, so Radeon 8500 works relatively well in Doom3.
Half-Life2 also has a full ps1.4 path, but it mainly delivers much better image quality, it's not used for performance only.
Some of the tests in 3DMark03 also use ps1.4 on a Radeon 8500+, where they need a multipass ps1.1-approach on GeForce3/4.

8500 is a nice chip with interesting features. Its various unique anti-aliasing modes are also notable. The lack of MSAA was a disadvantage however...

The main problem with 8500 though is driver quality. It tends to be pretty quirky for games outside of the video card reviewers' bell curve. For old D3D5 games, you don't get fog table support, which is the same as with prior Radeons but hey NV supported it until GF7. The other problem is non-Quake OpenGL games like NWN and KOTOR which are biased toward NV OpenGL behavior. The anti-aliasing behavior also varies with drivers and sometimes doesn't work. Something else I've noticed, with my 128MB 8500, is the anti-aliasing modes have the same resolution limits as with 64MB card.

What's particularly amazing is how much better texture filtering quality is with the R300 boards. Trilinear looks better and you don't have the R100-RV280 limitation of anisotropic only with bilinear filtering.

swaaye wrote:

8500 is a nice chip with interesting features. Its various unique anti-aliasing modes are also notable. The lack of MSAA was a disadvantage however...
...
What's particularly amazing is how much better texture filtering quality is with the R300 boards. Trilinear looks better and you don't have the R100-RV280 limitation of anisotropic only with bilinear filtering.

The R300 was the first card with 'modern' MSAA and AF.
Earlier cards had pretty 'bruteforce' AA and AF approaches, which really cut into the bandwidth.
R300 was the first with modern MSAA, running shaders only once per pixel rather than once per sample (although allegedly GF3/GF4 had some shortcut so they only sampled textures once per pixel... I never quite got good info on whether or not they execute the shaders for all samples, but the poor performance would indicate that they are).
And R300 was the first with a good adaptive AF algorithm. And they worked great in combination together. So in most cases you could turn on 4xMSAA and 16xAF, and effectively you'd only lose about 30% of the total performance compared to no AA and just trilinear filtering.
On all earlier cards it was better to leave those features disabled... well, except for older games that ran way too fast anyway.

R100 and R200 actually have very fast AF. It's extremely angle dependent and so misses most pixels, but it certainly adds quality to the end result. A lot faster than GF3/4 AF too. But compared to GF3/4 and R300 it is certainly lower quality.

R200's AA modes are wild, though very inefficient yes. It's reminiscent of Voodoo5's rotated grid SSAA but you get something like 12 different modes to play with. 2x,3x,4x,5x,6X samples with performance/quality options on each. It is very quirky but it can look nice if you can find settings that work well with a particular game.

Radeon 8500 is one my favourite videocards... unfortunately the AF quality is very poor (still better than pure bilinear/trilinear though) and antialiasing is quite slow. GF3/4 AF quality is perfect (similar to G80 / R500 levels) but the performace hit is huge. GeForce FX image quality in HQ mode is actually very good (same as GF3/4) and reasonably fast as well.

Oh, one more thing - R200 is the only GPU supporting full-HW Truform. 😀