@d1stortion, it was mentioned in their faq section actually. My bad, if my post was ambiguous.
@swaaye, I think pci bus performance in general should be ok, though I may be dead wrong. probably it's more about the efficiency of the data transfers between the gpu to processor/memory via the reverse & host bridge. The closest I got to measure that was through Final reality bus transfer rate tests for 2d & 3d which seemed abnormally low to me. Or maybe the bus transfers are frequently getting interrupted resulting in poor efficiency. Changing the latency timer of host bridge affects the performance in both ways whereas changing it in the reverse bridge does nothing, which I don't understand.
I also came across a utility called pcitop, but afaik it is available only in Linux.

Some time ago I used an old monitoring utility to troubleshoot X-Fi. I think it showed PCIe (and possibly also PCI) bus usage or at least something very closely related to that. Can't remember the name right now though... 😒

Probably I should forget about half life, deus ex and other windows games and start playing dos games. I'm partial to half life as that was the time I got my first pc. It almost tempts me to wipe off windows and install dos and finally play some games. Though the problem may still haunt me in dos too, who knows. Atleast I would be able to contribute to Phil's vga benchmark project with scores at the bottom of the ladder. Getting too cynical, got to stop.

AFAIK Nvidia has broken VBE 3.0 compatibility at some point, so you can't get anything higher than 60 Hz in DOS games on those cards. I also recall a topic on palette issues or something like that, all making these cards not desirable for DOS gaming at all. Someone else should be able to elaborate on that.

I think with the introduction of unified shaders the classic vga core of those cards got removed.
The PCIe to PCI bridge may add additional penalties. It is probably faster to use Dosbox emulator and a fast CPU because the graphics card will be controlled by the driver software then.
It is questionable if the card shows the more custom modes of some games/demos at all.

Not much to discuss, probably nearly all old PCI graphics cards are better suited.

elianda wrote:

I think with the introduction of unified shaders the classic vga core of those cards got removed. The PCIe to PCI bridge may add […]
Show full quote

I think with the introduction of unified shaders the classic vga core of those cards got removed.
The PCIe to PCI bridge may add additional penalties. It is probably faster to use Dosbox emulator and a fast CPU because the graphics card will be controlled by the driver software then.
It is questionable if the card shows the more custom modes of some games/demos at all.

Not much to discuss, probably nearly all old PCI graphics cards are better suited.

The built in video could perhaps be an option in DOS.

swaaye wrote:

Does anyone know of a simple Windows graphics benchmark that will do a screen fill test and saturate the bus the graphics card is attached to? Something like vspeed for DOS maybe. I am wondering just what PCI performance he is getting.

i request anyone with pci or agp cards to run the final reality benchmark (ftp://193.219.28.2/vol/rzm1/demos/demos/group … emedy/fr101.exe) and share the results for the sake of comparison

@d1stortion, i'm not bothered about fps above 60, only the drastic variation in fps in comparison with similar systems. i will keep your advice in mind when i try dos games.

@elianda, though i don't have a fast cpu, i will try some dos games through dosbox just for the heck of it to find out where i'm placed.

@Skyscraper, the onboard video (trident 3d) performance is really pathetic. i will give it a try however in dos games.

i will try the non-acpi reinstallation of windows to rule out irq sharing though i didn't find any improvement even when i disable usb and sound from bios. still i will try it as i don't have anything else in mind.

Well, I have some old results from my Pentium MMX 200+Voodoo1 PC:
2D image processing: 1.92 RM
3D performance: 2.64 RM
Bus transfer rate: 2.61 RM
Overall: 2.42 RM

Here at 11:00 http://www.youtube.com/watch?v=SNiYRcm3NbE for S3 Trio3D
Number Nine Revolution IV: http://www.youtube.com/watch?v=ieUxUssxATk
and there are even more results: http://www.youtube.com/user/elianda/search?qu … y=final+reality

d1stortion wrote:

Well, I have some old results from my Pentium MMX 200+Voodoo1 PC: 2D image processing: 1.92 RM 3D performance: 2.64 RM Bus tran […]
Show full quote

Well, I have some old results from my Pentium MMX 200+Voodoo1 PC:
2D image processing: 1.92 RM
3D performance: 2.64 RM
Bus transfer rate: 2.61 RM
Overall: 2.42 RM

can you post detailed bus transfer rates? mine was,
Video card bus transfer, 5A, 15.34, MBps, 0.488, Rmark
Direct3D bus transfer, 5A, 19.28, MBps, 1.648, Rmark
Overall bus rate, 0.836, Rmark

elianda wrote:

Here at 11:00 http://www.youtube.com/watch?v=SNiYRcm3NbE for S3 Trio3D
Number Nine Revolution IV: http://www.youtube.com/watch?v=ieUxUssxATk
and there are even more results: http://www.youtube.com/user/elianda/search?qu … y=final+reality

how do such old cards like S3 Trio3D & #9 revn. have three digit transfer rates when mine is only in the lower 2 digits?
another thing was the display during the 2d image processing & 3d tests itself were garbled in my system. when i ran only the bus transfer tests i didn't get any display, don't know whether it was normal.

2D transfer rate: 62.79 MB/s, 2 RM
3D transfer rate: 42.67 MB/s, 3.65 RM
Overall: 2.49 RM

gandhig wrote:

can you post detailed bus transfer rates? mine was, Video card bus transfer, 5A, 15.34, MBps, 0.488, Rmark Direct3D bus transfer […]
Show full quote

can you post detailed bus transfer rates? mine was,
Video card bus transfer, 5A, 15.34, MBps, 0.488, Rmark
Direct3D bus transfer, 5A, 19.28, MBps, 1.648, Rmark
Overall bus rate, 0.836, Rmark

At least you outperform the ISA bus. 😉

A high performance PCI video card setup should break 80MB/s.

d1stortion wrote:

2D transfer rate: 62.79 MB/s, 2 RM
3D transfer rate: 42.67 MB/s, 3.65 RM
Overall: 2.49 RM

thanks d1stortion. mine is not even coming close

swaaye wrote:

gandhig wrote:

can you post detailed bus transfer rates? mine was, Video card bus transfer, 5A, 15.34, MBps, 0.488, Rmark Direct3D bus transfer […]
Show full quote

can you post detailed bus transfer rates? mine was,
Video card bus transfer, 5A, 15.34, MBps, 0.488, Rmark
Direct3D bus transfer, 5A, 19.28, MBps, 1.648, Rmark
Overall bus rate, 0.836, Rmark

At least you outperform the ISA bus. 😉

A high performance PCI video card setup should break 80MB/s.

i ran the benchmark with my onboard trident. guess what, it beat the discrete card with 40+ MBps transfer rate, overall: 2.23 RM. This is the first benchmark where my onboard graphics beat the zotac gt520 pci card.

i found out the reason for the garbled display. it was because my desktop resolution was 1280x1024. when i changed it to 800x600, it was normal.

Dell with P4 northwood HT @ 2.8 Ghz
Geforce Ultra 6800 with minor overclock.

http://www.outfoxed.net/final-reality/dell+68 … Performance.jpg
http://www.outfoxed.net/final-reality/dell+6800/3D_Tests.jpg
http://www.outfoxed.net/final-reality/dell+68 … ssing_Tests.jpg
http://www.outfoxed.net/final-reality/dell+68 … ansfer_Rate.jpg

I'll work on my Athlon with voodoo3 for comparison in a bit.

Edit: I know these are faster than your system and what you're looking for, but you said "Anyone with PCI or AGP cards" so..

Edit #2: Seems for some reason Final Reality is unable to detect a Hardware Direct3D Renderer on my AthlonXP system... even though it's using offical 3dfx voodoo drivers, and dxdiag completes all the Direct3D tests without error. Odd.. but guess I can't run it on that machine. Also on the dell above, I couldn't get it to run the AGP test for some reason, it errored out unable to read some file and crashed. It could of been because of WindowsXP it was run on, maybe. I don't know.

So how do we run this benchmark on a system with no Direct3D drivers for the video hardware? If you're running it in software then it's just 100% using the CPU and not the video card at all, so what does that test? I'm confused on that.

@kithylin, i have come to accept the fact that my card's results will be a fraction of expected normal values and really doesn't surprise me anymore whether the comparison is with PCI or AGP cards.

regarding your dell system, which service pack it is? i read somewhere that there may be a compatibility issue with sp3. i also could run the agp test on both trident onboard as well as gt520 pci(though i was warned it will be extremely slow, it was far from crawling).

i guess the bus transfer rate tests under software must be between cpu to gpu(frame buffer) through video bus.

Service pack 3 in XP X86 / 32-bit.

From my Voodoo5 rig:

2D image processing: 7.75 RM
3D performance: 7.29 RM
Overall: 7.69 RM

2D transfer rate: 169.37 MB/s, 5.39 RM
3D transfer rate: 204.51 MB/s, 17.48 RM
Overall transfer rate: 9.02 RM

AGP test (16 MB): 305.81 images/s

d1stortion wrote:

From my Voodoo5 rig: […]
Show full quote

From my Voodoo5 rig:

2D image processing: 7.75 RM
3D performance: 7.29 RM
Overall: 7.69 RM

2D transfer rate: 169.37 MB/s, 5.39 RM
3D transfer rate: 204.51 MB/s, 17.48 RM
Overall transfer rate: 9.02 RM

AGP test (16 MB): 305.81 images/s

thanks d1stortion. i reinstalled windows as standard pc today. the difference was insignifcant

Radial blur, 5N, 58.42, rips, 8.080, Rmark
Chaos zoomer, 5N, 91.91, rips, 4.466, Rmark
25 Pixel, 5N, 488.85, kpps, 15.623, Rmark
Robots, 5N, 35.93, rips, 9.307, Rmark
Fillrate, 5N, 17.70, MPps, 3.832, Rmark
City scene, 5N, 44.78, rips, 11.112, Rmark
Video card bus transfer, 5N, 22.49, MBps, 0.716, Rmark
Direct3D bus transfer, 5N, 28.31, MBps, 2.420, Rmark
-----------------------------------------------------------------------------
Visual appearance, 100.00, percent
-----------------------------------------------------------------------------
Overall 3D, 3.645, Rmark
Overall 2D, 6.273, Rmark
Overall bus rate, 1.227, Rmark
-----------------------------------------------------------------------------
OVERALL SCORE, 4.071, Rmark
-----------------------------------------------------------------------------