I think CPU-Z or GPU-Z will show if they are 64-bit or 128-bit.
Maybe also run Tom2D?

wouldnt the PCI bus castrate any card in 3rd to the point where its almost useless to go faster than a Geforce 2 MX PCI?

candle_86 wrote:

wouldnt the PCI bus castrate any card in 3rd to the point where its almost useless to go faster than a Geforce 2 MX PCI?

Well Dirkmirks Benchmarks would at least indicate a difference between 6200 Geforce and 9100 Radeon. I don't have either of those in a PCI. I've got a ton of Radeon 7000 32MB PCIs and will probably at least try one out once I get to a certain point. I have a bunch of other PCI cards as well but I have to dig through them to figure out exactly what they are. I will do this once I get some time.

I have a 6200 PCI, 5500 PCI, FX600 PCI, Parhelia PCI, and a 9250 PCI but my testbed is busy with 386 hardware at the moment. What driver version are you using? I might be able to test these on a P4.

I was making good progress until my system died....

The 9100 is no good at Doom 3 getting like 9.1fps in the timdemo (17.5fps for 6200) at 1024x768,i t also struggled big time with the battle of proxycon in 3dmark 2003 4.3fps.

Quake 3 timdemo four 190fps 1024x768 & 168fps @ 1280x1024, Max Payne high quality in 3dmark was 51fps suggest good performance for this directx 8 game.

I could build a tualatin 1.4 system but not sure if that's a good enough system to test pci cards...

Well, it just so happens I've finished up my testing! I will great a chart. I tested the ATI 9250, FX5500, 6200, FX600, Matrox Parhelia 256 - all PCI. To put things in prespective, I also compared a 6600GT (AGP). I was wanting to use my P3 Tualatin board, but the damn FX5500 didn't work on that board, so I ended up using a P4 Prescott 3.4 GHz.

Attached are the results for 3DMark2001SE, Quake III, and Tom2D. I have also added an AGP card to add some speed perspective to these PCI cards. For 3DMark2001SE, which tests DirectX 8.1 capabilities, the Quadro FX600 came out on top, and way ahead of the GeForce FX 5500. The FX600 also was quite a bit faster at Quake III (openGL) compared to the FX 5500. Is this entirely due to it running at faster clock speeds, or is there more to it?

The GeForce 6200 was just slightly behind the FX600 in 3DMark2001 and Quake III.

Any idea why the Radon 9250 scored so well in Quake III? That score caught me off guard.

In Tom2D, the Quadro FX600 won by leaps and bounds, and was 3x faster than the 6600GT AGP 8X. Any reason for this? I'm not sure how reliable the Tom2D results are. It seemed like some tests get stuck in an infinite loop and I have to force it out, however the program then proceeds to the next test, so maybe it is fine? What is interesting is that the Top2D score more than doubled when the FX600 was tested in the PIII Tualatin motherboard compared to the P4. I wonder what the reason for this was?

Also interesting was the Quake III score for the Quadro FX600 was faster on the PIII Tualatin configuration (78.3 fps) compared to the P4 Prescott configuration (73.5 fps). Ideas? Is the P4 architecture somehow holding the FX600 back? Or perhaps the PCI implemtation on the P4 motherboard? The P3 Tualatin board I used does not have AGP implementation. Perhaps that played a role?

Lastly, using the Quadro drivers for the FX600 did not increase, nor decrease, the scores of the FX600.

Pending an explanation of the Radeon 9250's amazing Quake III score, and perhaps some further testign, I would say that the Quadro FX600 was the fastest overall graphics card in these tests. The Matrox Parhelia was a little disappointing, even though it has twice the memory bandwidth of any of the PCI cards. Keep in mind that it is intended for a PCI-X bus at 64bit/66MHz rather than a PCI bus at 32-bit/33MHz. Although testing elsewhere seemed to indicate that placing it in a PCI-X slot did not improve scores. Reference to this thread: Matrox Parhelia 256 PCI-X Performance

Test System P4: Intel S875WP1-E motherboard with 3.4 GHz P4 Prescott and 4 GB RAM

Test System P3: Intel SAI2 motherboard with dual 1.4 GHz PIII-S Tualatin and 4 GB RAM

For data, click the first image below for an expanded view.

Thanks for the work you put into this feipoa. FX600's range betwen difficult and impossible to find so the other choices are pretty clear. I wonder what benchmarks would look like on a 100% tualatin setup rather then a P4. Will it scale similarly or the slower CPU change things up. Once I get mine built I'll certainly add my 2-cents.

I used Radeon HD2400 PRO PCI card on my dual Pentium III-S 1.4 GHz build with Intel SAI2 board. Got ~6200 points in 3DMark2001SE. But it didn't work in 66MHz/64bit slot, only in regular one (in PCI-X slot the PCI bridge chip threw error in Device Manager).

I was able to use a 6200 in the PCI-X slot on the SAI2 motherboard. It worked fine, then I upgraded the drivers and it didn't work anymore in that slot. Even using old drivers again, it didn't work in the PCI-X slot anymore.

That's a really nice score for a PIII Tualatin. Did you run 3DMark2001SE at 1280x1024?

There are some newer NVIDIA cards with a PCI bridge too as well. The Zotac 610 comes to mind, but it is not cheap. I wonder how that would stack up to the competition. It is interesting to see how far you can take an ordinary PCI bus. https://www.amazon.com/ZOTAC-NVIDIA-GeForce-5 … L/dp/B0083Y1YVE

Even more interesting would be to use one of those PCI-X to PCI Express adapters. I wonder if they are intended mainly for non-graphic cards? If they work with an SAI2 and a graphics card, that would be pretty unique. I guess you'd need to find drivers which did not need SSE2. http://www.staples.ca/en/StarTech-PCI-X-to-x4 … 00_1-CA_1_20001

Now that you mention it, I think I have another PCI graphics card from ATI that has a bridge chip. Maybe I'll look for that and add it to the graphics.

Here is the other PCI graphics card I have. It is an ATI FirePro 2260 multi-view. That website says it is RV620, runs at 500 MHz, 256-bit DDR2, with a bandwidth of 32.0 BG/s. Supports DirectX 10.1, OpenGL 3.3, and OpenCL 1.0. On paper, this is my highest end PCI card. I stuck it in a box, so I wonder if it did not work with the SAI2, or was worse performing compared to the FX600.

https://www.techpowerup.com/gpudb/575/firepro-2260-pci

Are you sure thats an FX5500? I thought it was only the 5200s with memory clocked as slow as that and thats surely the reason why the performance is so bad, you could download rivatuner and try overclock the memory to 200/400mhz.

Good to see the FX600 performing similar to the 6200 I dont see a reason to pick the 6200 over the FX600 going off those benchmarks, it might be faster for games like Half Life 2, Farcry and Doom3 but lets not kid ourselves you wont get satisfactory performance with any PCI card, It would be interetsing to see how your radeon card runs games like though 256mbit memory is unheard of in PCI video cards.

What resolution were you running the Quake 3 benchmark?

Just going back the radeon 9100 scores in 3Dmark 2003, it clearly outperformed the 6200 in the wings of fury(89fps vs 70, DX7) & Trolls lair benchmark(24fps vs 13fps DX8), it scored worse in the Battle of Proxycon(4fps vs 13fps) but also remember the card couldn't run all the tests because it doesn't have DX9 like the 6200 so obviously lowers the score substantially.

I'll have to dig out another system to keep testing these cards, one thing for certain the 6200/FX600 are similar in performance but the Radeon could be a spoiler yet for the games that matter!

I ran 3DMark2001SE in default 1024x768 mode.

Paadam, I ran my benchmarks at 1280x1024, so there is perhaps some difference in scores due to resolution.

dirkmirk, the card has FX5500 256M printed on the PCB. It is possible that GPU-Z v0.6.0 messed up the core/memory speeds. Does that program actually measure the clock or are these values being report via table lookup?

For Quake III, I used:

GL Driver: Default
GL Extensions: On
Video Mode: 1280x1024
Colour Depth: 32 bit
Fullscreen: On
Lighting: Lightmap
Geometric Detail: High
Texture Detail: Looks like the position just before the far right
Texture Quality: 32 bit
Texture Filter: Bilinear

"It would be interetsing to see how your radeon card runs games like" Like what? Your sentance got merged with another at this point.

I never realised how much these PCI Radeon 9000 series cards had going for them. Were they buggy in games?

feipoa wrote:

"It would be interetsing to see how your radeon card runs games like" Like what? Your sentance got merged with another at this point.

I never realised how much these PCI Radeon 9000 series cards had going for them. Were they buggy in games?

Its only the 9100 thats high end for PCI its the same card as the Radeon 8500LE, the 9000/9250 should'nt be in the same class as the 6200/FX600, I was curious for your modern DX10 card thats what I meant for the radeon comment, I got 168fps with Default graphic settings @ 1280x1024 on the 9100 with Quake 3 demo four, could their be a cpu factor involved with the phenom 2?

Maybe I edited my original post but the games I meant to say were Far Cry, Half Life 2/Doom3, just for curiosity you would'nt think PCI would have enough bandwidth to run acceptably regardless of the hardware behind it.

Well, the Phenom II is quite a bit more advanced that a P4 Prescott, so there must be some influence w.f.t. the benchmark score. Though I would assume that the bottleneck is the graphics card in both cases. maybe get yourself a 9250 and find out? Or at least P4. The 9100 PCI is not so easy to find. Do you think a 9100 is a better catch than an FX600?

Are there benchmarks or at least frame rate counters in Far Cry and Half Life 2?

feipoa wrote:

Do you think a 9100 is a better catch than an FX600?

I dont know my card is only 64meg and has a 250/250(500ddr) cpu/memory clock 8gb memory bandwidth, looking at the gpureview website the main advantage seems to be 8 texture units vs 4 on the Geforce cards which equates to texture fill rate of 2000 Mtexels/sec vs 1000, not sure how this equates to which games it advantages.

The 6200 has a big advantage in vertex operations 225Mvertices/sec vs 125 on the 9100, I presume this explains the extra performance in the newer games like Doom3 or the Battle of Proxycon in 3Dmark but still less texture fill rate (1200 vs 2000),

Interestingly my 6200 is the "OC" overclocked version of the bfg card, 350mhz cpu clock but only a 205(410mhz) memory clock...... Thats a weird one!

Are you using GPU-Z to grab the bandwidth and clock rate information? I should also mention that I am running the Quake III benchmark by typing:
`
\timedemo 1
\demo four

How is the 9100 with DOS compared to the FX series? I sorta remember cards newer than the Radeon 7000 aren't as ideal for DOS.

EDIT:
I have added the ATI FirePro 2260 results to the table and charts. The 2260 is the fastest in 3DMark2001SE (DirectX 8.1), the slowest in Tom2D, and between the FX600 and 9250 in Quake III. We really need to increase our benchmark sample size to see which cards are overall better. I also found a PNY 6200 PCI card. I benchmarked the PNY and added the results to the chart. It was a little faster than the eVGA card for some reason. GPU-Z shows the eVGA and PNY 6200's to have the same CPU/memory frequencies. Any explanation for why the PNY is faster than the eVGA?

What 6200 64 bit do you have' DDR2?

I also have a Geforce 6200 64 bit DDR2, but is AGP, also have a Prescott 3.4 Ghz if you want to bench it and compare 6200 AGP vs PCI

Also i'm getting a Radeon 9250 for my dual tualatin setup (no AGP slot as well)

FYI, I was going through my PCI cards today and found an HIS ATI Radeon HD 4350. PCI 512MB. No drivers for Windows 9x as far as I know but for those running Windows XP on a playform with PCI only this could be a good alternative. Probably not for my build I was originally asking about but would be a good option for some.