Niiiiiice!

The scaling is pretty similar between your 5900 and my 6800, you get extra ~30% and ~25% FPS respectively, when enablling SMP at 720p and 1080p. Pretty sweet! Couple of questions:
-are your results from a single run, or an average of various runs?
-did you touch the FOV setting?

I take an average of 3 runs + the first one that I disregard.

I did not touch the FOV setting, just did some quick and dirty tests as I didn't really have much time to waste :p. Wouldn't really make that much of a difference, I'm still mostly CPU bound.

You're right that it doesn't make much of a difference, but it does take a few frames away.

Thanks for running these!

This thread needed some love, so I updated the first post with a link to a web page with all the results uploaded so far. If people wanna keep contributing, please do so!

SMP results for voodoo5!!!

Compaq AP550: 2x PIII 1000EB, 768MB RDRAM, 3dfx Voodoo5 5500, SB Live!

Driver: AmigaMerlin V3.1 R11 (DX8 mode)
r_smp 0 ULTRA 1024x768 = 54.2fps
r_smp 0 FASTEST 640x480 = 77.9fps
r_smp 1 ULTRA 1024x768 = 64.3fps
r_smp 1 FASTEST 640x480 = 131.6fps

PD: OS is Windows 2000 SP4

Interesting.

KingPing wrote:

SMP results for voodoo5!!! […]
Show full quote

SMP results for voodoo5!!!

Compaq AP550: 2x PIII 1000EB, 768MB RDRAM, 3dfx Voodoo5 5500, SB Live!

Driver: AmigaMerlin V3.1 R11 (DX8 mode)
r_smp 0 ULTRA 1024x768 = 54.2fps
r_smp 0 FASTEST 640x480 = 77.9fps
r_smp 1 ULTRA 1024x768 = 64.3fps
r_smp 1 FASTEST 640x480 = 131.6fps

PD: OS is Windows 2000 SP4

I have added these to the global results, thanks!

Its pretty interesting to compare the Voodoo4 vs Voodoo5 in the results you have provided so far, and a ~20% boost at 1024x768 is really nice!

I ran Q3 1.32 after getting my OR840 up and running. I got 96.8 without smp enabled and 112.1 with. This is at 1024x768, max textures and bilinear filtering. I had 2xP3@933MHz, 512MB RDRAM@800MHz and a Radeon 8500.

Ok, I gave the Radeon x800 pro a try on the OR840 and it got 94.7 without smp and 103.2 with smp. I was going to do more testing but the x800pro was so unstable that I ended up just pulling it. I put in the Matrox Parhelia (128MB/207MHz) and without smp it got 77.3 and 118.1 with smp.

cxm717 wrote:

I ran Q3 1.32 after getting my OR840 up and running. I got 96.8 without smp enabled and 112.1 with. This is at 1024x768, max textures and bilinear filtering. I had 2xP3@933MHz, 512MB RDRAM@800MHz and a Radeon 8500.

cxm717 wrote:

Ok, I gave the Radeon x800 pro a try on the OR840 and it got 94.7 without smp and 103.2 with smp. I was going to do more testing but the x800pro was so unstable that I ended up just pulling it. I put in the Matrox Parhelia (128MB/207MHz) and without smp it got 77.3 and 118.1 with smp.

I have updated the results page with your numbers: the Parhelia results are impressive! Would you mind running tests at higher resolution and with trilinear filtering to see how it behaves?

I was honestly very surprised by the Parhelia with SMP. It's almost always slower than the 8500 in Q3. I'll run some some more tests and even try the newer Parhelia card in the board too, once I get the system back together. I was doing some testing with another system. I also need to install XP on it. I was being lazy and moved the XP install I already had from my d815eea board, I actually think that's what was causing the instability I was having. I'll try some of my other video cards in it too. I could also try these cards in my other SMP systems.

hrm...If I get a new case for it I might slap my dual Athlon MP rig together with a Geforce 2MX in it just to run Q3A.

misterjones wrote:

hrm...If I get a new case for it I might slap my dual Athlon MP rig together with a Geforce 2MX in it just to run Q3A.

Athlon MP sounds great, very curious about it. If you can run other video cards, please do so too--the more results, the better!

In general, non-Pentium3 results would be awesome if people want to contribute. Like early Pentium4 with HT maybe, or a BP6 with dual Celerons... or a PR440FX with dual Pentium Pros 😁

slivercr wrote:

misterjones wrote:

hrm...If I get a new case for it I might slap my dual Athlon MP rig together with a Geforce 2MX in it just to run Q3A.

Athlon MP sounds great, very curious about it. If you can run other video cards, please do so too--the more results, the better!

In general, non-Pentium3 results would be awesome if people want to contribute. Like early Pentium4 with HT maybe, or a BP6 with dual Celerons... or a PR440FX with dual Pentium Pros 😁

I have a Geforce FX5500 and 6200 AGP cards, but they both need to be re-capped. The 5500 is horribly unstable and the 6200 was originally in the machine I'm typing on right now but the caps started to bulge on it and I started getting random hard freezing. Right now this machine has an 8400GS PCI card in it.

I also have a few Matrox AGP cards (G200, G400, G450 DH.. also have a G400 DH but it's PCI) but I don't really think Matrox did much with SMP, though it would make sense if they did considering their Millenium cards with 32MB upgrades were often used as the low-end option in many SMP-capable 3D workstations in the late 90's.

I did some more testing on the OR840 and on my IBM intellistation m pro 6898 with a pair of 333MHz pentium 2s. Unfortunately I couldn't get many AGP cards to work in the IBM. It's an older board with the 440LX chipset and I don't think it can supply enough power for newer AGP cards. I did try the Geforce2 and everything was fine until I installed the drivers and then I got lots of artifacts and the monitor would lose the signal for a sec after opening a window. It was a pain to just get the drivers uninstalled. I actually do remember that I tried a Geforce256 in this system a while ago (sometime in 2000, I think) and it was the same way.

Anyways, I did more testing with the Parhelia at different resolutions and with trilinear filtering. Here is the link: https://docs.google.com/spreadsheets/d/1teBHo … dit?usp=sharing

I'm going to give my other IBM intellistation (6889, 2xP3@550) a go later and I'll try the Geforce4 in the OR840.

just for grins and giggles I ran Q3A on my DAW machine and came up with some interesting results.

Intel Xeon 3220 quad core processor, 4GB RAM, 2GB Geforce GT710 video card:

Without smp - 471fps
With smp on - 320fps

run at 1024x768x32 under Win7 Ultimate.

It runs significantly slower with smp enabled.

Time to expand the database a little.
Tested out my recently acquired Soyo MB (full recap and thorough clean).
Q3A is obviously CPU limited here with the GF4 TI 4200.
So its interesting to see the difference in gain with faster vs slower CPU's
Also Q3A doesn't seem to benefit from SSE, maby with sound enabled that comes into play?

Soyo SY-D6IBA (440BX), 1GB SDR CAS2, GeForce 4 TI 4200 driver ver. 71.89, Win 2K SP4.
Q3A v 1.32c "Demo FOUR"

Hope this looks OK when posted (ok it doesn't, pic added for easier viewing):

2xP3-600 Katmai SMP 0 SMP 1 Gain
Fastest 640x480 91 121 132,97%
High Q 640x480 81 104 128,40%
High Q 800x600 81 103 127,16%
High Q 1024x768 80,5 102,7 127,58%
High Q 1280x1024 78,5 99 126,11%

Average 82,40 105,94 128,44%


2xP3-450 SMP 0 SMP 1 Gain
Fastest 640x480 76 102,4 134,74%
High Q 640x480 66 88,5 134,09%
High Q 800x600 65,7 87,7 133,49%
High Q 1024x768 65,3 87,4 133,84%
High Q 1280x1024 64,4 85,6 132,92%

Average 67,48 90,32 133,82%


2xP2-400 SMP 0 SMP 1 Gain
Fastest 640x480 69,3 95,6 137,95%
High Q 640x480 60,3 82 135,99%
High Q 800x600 60 81,6 136,00%
High Q 1024x768 59,5 81,1 136,30%
High Q 1280x1024 58,8 79,3 134,86%

Average 61,58 83,92 136,22%


2xP2-350 SMP 0 SMP 1 Gain
Fastest 640x480 62,2 86,7 139,39%
High Q 640x480 53,9 73,8 136,92%
High Q 800x600 53,5 73,3 137,01%
High Q 1024x768 53,3 72,8 136,59%
High Q 1280x1024 52,5 71,6 136,38%

Average 55,08 75,64 137,26%

I did notice back in May when doing my compatibility testing that Quake 3 1.32e does work in Windows 2000. Haven't tested it on less than 2000 yet.
I tested quake3-1.32e-20200415 but it looks like Jan 31 2021 is the latest.
http://edawn-mod.org/forum/viewtopic.php?f=6&t=7
https://github.com/ec-/Quake3e/releases

Didn't bother to see if SMP worked
I do see that SSE2 and SSE3 are used.

Of course the changelog mentions the following but mabye something changed in 9+yrs. Even if no SMP may be worth comparing speed between original and 1.32e without it.

30-May-2011:
* more qvm optimizations - up to 30% speedup comparing to 1.32b/c 😀
* removed buggy \r_smp

Got a PR440FX, but only have a Matrox G450 or a VooDoo3 as 3D cards for it.
And both of em give negative scaling with SMP in Q3A, wich hints at lack of driver support or something else not right.

Could fire up the old A7M266-D tho!
That should have enough oomph for most AGP cards.
Even have a 3850 somewhere for that extra overkill feeling =)

I'm amazed how CPU sensitive Q3A is!

The AXP 1800+ Is slowly approaching what the TI4200 can do.
And even shows negative scaling in High Q @ 1280x1024, whats up with that? =)

The 6800GT however is seriously starved, and for some reason shows way better scaling then the 4200 across all resolutions.

My A7M266-D is a pain to use, passes all stability tests i throw at it.
But still has random lockups, not detecting KB/mouse etc. etc.
I did a partial recap of it, maby its time to do all the small buggers too.
Then again the 760 MP chipset (or rather the 768 B1 SB wich I ofc have) seemes to be kinda buggy overall.