That's very little actually. You should have got 60 fps. By Voodoo 3 I guess you mean the 3000 model.
Install the drivers once more.

F2bnp wrote:

That's very little actually. You should have got 60 fps. By Voodoo 3 I guess you mean the 3000 model.
Install the drivers once more.

60 fps @ 1280x1024?
It's a Gateway model (BIOS sticker says: 2.15.07), so perhaps it's slower than a "regular" 3000 model? Perhaps this Gateway V3 card is a 2000 model?

Edit: I'm using AmigaMerlin driver version 2.0, and Win98SE.

Edit 2: The BIOS POST screen says it's a 3000G model. The letter G must surely stand for Gateway, but the number certainly says 3000 and not 2000.

Edit 3: Using the 3dfx tools utility, it says that the graphics clock speed is 166Mhz.

Edit 4: Using the resolution of 1024x768, I get about 55 FPS shown in the top left hand corner of the screen in "Practice mode".

I'm not sure if the FPS value in the top left hand corner of the screen can be taken seriously. If you watch this Average value carefully, it simply goes down and down and down, in a steady downward way, no matter what is happening inside the Practice game. It doesn't seem to make much sense. Perhaps I need the full version of UT to run and properly benchmark the UTBench.dem demo?

That'd probably be a good idea. The UT demo wasn't released in a GHz world (fastest CPU was 600MHz at that point), perhaps 436 will work out better with all that post-release code refinements and the like. Dunno if UTBench.dem is compatible though

Thanks for the info. RG100 visits ebay and puts "Unreal Anthology" DVD game in to watch list. 😉 BTW, I ran 3DMark Max 99:

133 FSB (1667 Mhz CPU) : 6832 / 28094
160 FSB (2 Ghz CPU) : 6822 / 33722

Both 3D scores are about 6800. I wonder how much you can decrease the CPU clock speed by, in order to eventually see that ~6800 score begin to fall?

stat fps in the console

I have removed the Voodoo3 card, and replaced it with an AGP Voodoo5 5500 card. Here are some 3DMark 99 Max 3D scores:

2200+ rated mobile barton (1667 Mhz)
133 fsb = 12422
140 fsb = 12772
145 fsb = 13026

2400+ rated t-bred (2000 Mhz)
133 fsb = 12434
140 fsb = 12783

What's really interesting here is that the mobile barton seems powerful, compared to the t-bred CPU. I suppose these tests are inconclusive, because as the fsb is increased, so too does the 3D score. I have not reached a "plateau" yet, whereby the 3D score "flattens out" and remains constant. I suppose I could continue to increase the fsb with the mobile barton, but I'm not keen on OC'ing the mobo to obtain slightly higher 3D scores. The next best option is to obtain a faster mobile barton. Unfortunately, I can't find my slightly faster 2400+ rated XP-M, but I'm thinking of getting a faster CPU on ebay. The question is - will it work in a desktop mobo such as the Epox 8KTA3+Pro? I was thinking of getting a 2600+ rated XP-M, which has a multiplier of 15x. I'm guessing it might work, because although the Epox BIOS does not specifically allow for a 15x multi in the BIOS set up options area, the 2400+ rated t-bred certainly POSTs OK, and that has a 15x multi, which is stated in the BIOS POST messages area.

Also, for every test above, I adjusted the FSAA to 2x and also 4x. I noticed that the 3D scores obtained never altered, and so I will not mention them.

Edit: I put the 2200+ rated (1667 Mhz) XP-M CPU back in to the board, and did one more FSB increase to 150. The 3D score went up again to 13232. Using an AGP V3 card, I can increase the FSB to 160 on this board and have no problems. However, using the V5 I could not increase the FSB to 155, and so I couldn't continue these tests using the 2200+ rated mobile barton with an AGP V5 card.

Edit 2: Just 2 minor observations: 1) Both the V5's fans spin at the time of boot up, but after a minute or two they both stop. Also, the 3DMark 99 Max benchmark called "Bump map (emboss 1 pass)" produces an odd "washed out" appearance.

I might be approaching this with the wrong mindset. It might not matter if the V5 is capable of producing faster and faster 3DMark scores, by using a faster CPU. What might be the most important things to think about is - FSAA speed, and Glide speed. For 2x and 4x FSAA speeds, the 3DMark scores did not increase as I increased the mobo's FSB. So for FSAA, I have already reached the maximum CPU required speed, and a faster CPU is unnecessary. I think! For Glide speed, I guess I need to get a full copy of Unreal Tournament and run a Glide mode benchmark. If the UTBench.dem file is not compatible with the fully updated and patched UT game, then is it capable of doing a "standard" timedemo benchmark, in Glide mode? I need to know this before buying a copy, otherwise it's a waste of money. Also, is there another way of doing a Glide benchmark? I have tried Unreal's "stat fps" console command, but unfortunately it's not really very accurate as it doesn't provide me with a proper timed demo score.

Edit: Risky purchase yes, but fun yes. I've just bought a 2600+ rated (2000 Mhz real speed) XP-M mobile barton, code AXMG2600FQQ4C. You can see this CPU here -

http://www.cpu-world.com/CPUs/K7/AMD-Mobile%2 … G2600FQQ4C.html

Also, I'll get a copy of UT as they are cheap on ebay. I'll post back next Friday, when I can resume testing.

Edit 2: Although I have bought the 2600+ rated XP-M (with a 15x multi), I've just realised it won't POST correctly. This is because my 2200+ rated XP-M POSTs at 600Mhz (100FSB), and I have to manually adjust the BIOS multi value to 13x and reboot, in order for the BIOS POST messages section to correctly report its proper clock speed. The problem is that the 2600+ rated CPU has a 15x multi, and the BIOS multi option values only go up to 14x. It might POST OK at 14x multi, but it won't run at top speed. However, there is a chance that I can increase the multi from the BIOS maximum of 14x to the CPU's maximum of 15x by using a Windows program such as CPUMSR.exe.

Edit 3: I have bought a cheap copy of UT. I've discovered another odd thing about the demo version. When I use it in D3D mode, and select 32-bit color, I get a general protection fault error message.

What early games (hopefully downloadable demo versions) can I test my V5 on, in order to produce excellent 4x FSAA appearance? Am I right in thinking that the best resolution to test this feature with smooth FPS in mind is no greater than 640x480? I guess I'm thinking of games that do not have a video options setting allowing the user to increase the resolution to high values, such as 1280x1024 and beyond. Because these games will probably look better at higher resolutions with no FSAA, than selecting FSAA at a lower resolution.

Edit: I set the V5 to use 4x FSAA and ran UT demo in D3D mode @ 640x480, but the visual quality was terrible! It looked weird and blurry. Extra edit: I set V-Sync to Enabled, and I think that improved the image a bit. End Extra edit. Perhaps that's another thing which is fixed in the fully patched version?

Edit 2: I'm mystified. I've run Quake 3 Arena demo, using the V5's FSAA @ 4x, with a resolution of 1024x768, and I am not impressed with the overall visual quality, or the FPS performance. I can see that when FSAA is used, your player's "mug shot" looses its jagged outline, and your gun's edges are also somewhat smoother, but the overall general image quality isn't great, IMHO. Inside Q3's Video set up area, I tried to adjust the setting to "Voodoo", but when I return to this set up area to double-check that this setting has been saved correctly, it just says "Default" and not "Voodoo". Perhaps I need to install a "Voodoo GL driver"? But anyway, the "Default" video setting with 4x FSAA enabled should have made the image quality look good. The internet reviews say that FSAA looks good. When I play Q3 with FSAA @ 4x, I just see ordinary visuals, nothing extraordinary.

Does anyone have either a PCI or AGP Voodoo5 5500 card? If so, do you have this card in a PC that has a CPU speed of greater than 2 Ghz? If so, can you please run some benchies? 😀 I just want to see what speed CPU is needed to achieve the maximum performance possible from a V5. Tests need to include Glide and FSAA, because I think these are of greater importance than D3D and OpenGL tests - the reason being that lots of cards are capable of doing these tests well. Thanks a lot if possible.

retro games 100 wrote:

Does anyone have either a PCI or AGP Voodoo5 5500 card? If so, do you have this card in a PC that has a CPU speed of greater than 2 Ghz? If so, can you please run some benchies? 😀 I just want to see what speed CPU is needed to achieve the maximum performance possible from a V5. Tests need to include Glide and FSAA, because I think these are of greater importance than D3D and OpenGL tests - the reason being that lots of cards are capable of doing these tests well. Thanks a lot if possible.

That would be difficult with an AGP card since the V5 is AGP 2x and 3.3 volts. I am pretty sure most motherboards from the P4 era were 4x/8x and 1.5 volts.

sliderider wrote:

retro games 100 wrote:

Does anyone have either a PCI or AGP Voodoo5 5500 card? If so, do you have this card in a PC that has a CPU speed of greater than 2 Ghz? If so, can you please run some benchies? 😀 I just want to see what speed CPU is needed to achieve the maximum performance possible from a V5. Tests need to include Glide and FSAA, because I think these are of greater importance than D3D and OpenGL tests - the reason being that lots of cards are capable of doing these tests well. Thanks a lot if possible.

That would be difficult with an AGP card since the V5 is AGP 2x and 3.3 volts. I am pretty sure most motherboards from the P4 era were 4x/8x and 1.5 volts.

I think that the very last mobo to support the AGP V5 was the KT 333 chipset based mobo. I believe you can put in a fairly powerful CPU inside this mobo. I think you can put a Barton 3000+ in it! 😀

retro games 100 wrote:

Does anyone have either a PCI or AGP Voodoo5 5500 card?

Yes.

retro games 100 wrote:

If so, do you have this card in a PC that has a CPU speed of greater than 2 Ghz?

No. And frankly, CPU speed doesn't matter when it goes to FSAA. Just like higher resolution, FSAA is GPU-bound instead of CPU-bound.

Kreshna Aryaguna Nurzaman wrote:

Just like higher resolution, FSAA is GPU-bound instead of CPU-bound.

OK.

I downloaded a couple of early Direct X racing car demos - "Daytona USA Deluxe" and "The Need for Speed 2". I experimented by switching the Voodoo5's FSAA 4x mode on and off, and I could not see any difference in visual quality, in terms of jagged lines being softened or not. I double-checked that the FSAA 4x on-off mechanism was working OK by testing it with "Quake 3 Arena" demo, and I could see the difference OK. I tried another demo - "Mysteries of the Sith" which is Direct X 5, and I could see the difference when FSAA 4x was enabled, but IMHO it wasn't impressive. The thing is, with this game, you are allowed to increase the resolution up to 1280x1024, and so this can improve the visual quality with any card. I would like to test an older game, that doesn't allow the user to select a higher resolution, and so in theory, the V5 FSAA can improve the visual quality. Can someone suggest the title of a suitable game (downloadable demo) please? Thanks a lot.

I have installed a utility to my Win98SE system called RefreshLock. It allows me to display games on my LCD monitor @ 60Hz, instead of the usual 75Mz it often uses. This improves the general image quality. Now, when I try 4x FSAA with games such as Jedi Knight, Mysteries..Sith, and Deus Ex, I can see an improved image quality. This is running at 1280x1024 in 16-bit color. The FPS is quite low however.

I wonder if I should change the V5 driver. I'm currently using AmigaMerlin version 2.0, inside Win98SE. I notice that if I run 3DMark 2001 SE, the display modes listed for AmigaMerlin are all 16-bit only. No 32-bit modes are present. That might explain why every game I try to run in 32-bit mode fails. However, the Win98SE desktop can be set to 32-bit true color. Strange.

Solved. I installed the 3dfx reference driver package, and now 3DMark 2001 SE tells me that 32-bit color modes are available. I wonder why Amiga Merlin was not working properly? I tried various versions: 2.0, 2.5 and 2.9. Perhaps it didn't install correctly. Ah, I've just realised something. Previously, I tested a V3 on this system with Amiga Merlin. That card is 16-bit only. I expect it was some registry problem. Perhaps when I reinstalled Amiga Merlin for the V5, after having removed the V3, the system got a bit confused and treated the V5 as if it was a V3. Maybe that's why my tests so far with the V5 haven't been so great.

Edit: I tried 4x FSAA with "Mysteries of the Sith" @ 1280x1024, 16-bit color. Using FRAPS version 1.9D, I get about 31 FPS if I run around in circles. It looked good. I mean, not me running about in circles, but the general image quality. I think 31 FPS is quite good. Now that I have a) got the V5 driver 100% correctly installed, and b) got the LCD monitor displaying at 60Hz, I think the image quality is now good.

Background information.
I received the "Unreal Anthology" games DVD today. It's got "Unreal Tournament" on it. Using this game, the utbench.dem demo benchmark works successfully, in Win98SE. Also, I received a mobile barton (XP-M) CPU. It's 2600+ rated, with a real clock speed of 2000 mhz. It works inside the KT133A chipset based Epox EP-8KTA3+Pro mobo. In order for it to run at top speed (using its 15x multiplier), this multiplier value must be set inside a Windows software utility called CPUMSR.exe. If you don't use this utility, you can only achieve a real clock speed of 1667 mhz (12.5x multiplier) using the BIOS set up area options.

Introduction.
The purpose of these tests is to ascertain whether or not a Voodoo5 5500 AGP card can run at its maximum performance, using the mobo and CPU mentioned in the paragraph above. My tests indicate that it does not, and therefore, in order to maximise the AGP V5 card's performance, you need a faster processor. At this moment in time, I don't know what that CPU speed is, but I am going to buy a faster CPU (2800+ rated XP-M), and rerun these tests a bit later.

Testing information.
For all of my tests, I set Unreal Tournament to Glide mode. I wanted to give the AGP V5 a good fighting chance! I ran every test twice to get the average score. Also, the column at the far right end is labelled "2252 mhz". This was achieved by overclocking the FSB from 133 to 150.

Test results.
Please look at the table. Have a look at the first and fourth rows. You can see that as the CPU speed is increased, so does the frames per second scores. In other words, the FPS scores do not "flatten out/level off". However, using 2x AA and 4x AA, these scores do "flatten out/level off". Please note that some of the table cells are empty. This is because I did not perform that test, because it seemed unnecessary.

It would be interesting to read any comments from other people who have the PCI version of this card, and can run the same tests on a PC with a CPU speed of greater than ~ 2.2 Ghz.