That's not too shabby at all. If you still have it all together you should try out some PCPbench to get a feel for the VESA subsystem. V3's are supposed to have decent 2D, I know when I tested V5 VESA in my rocket it was among the top, second only to the TNT based card I had. So are you going to try to run your 233 at 300, or leave it at stock speeds?

I wonder if something's wrong with my thrown-together rig? I just ran pcpbench inside windows 98 with no command line parameters, and it only gives me a score of about 21. I wonder whether the V3 differs from the SLI V2, in that the V3 needs a beefier CPU to go with it? The SLI V2 cards seem to be "content" with the fact that the CPU was a meagre P MMX 166. Perhaps the V3 isn't as happy with this low speed CPU?

>> So are you going to try to run your 233 at 300, or leave it at stock speeds?

My reason for doing this project is to crank up the board + P MMX 233 CPU until I smell burning! 😈 I want to see just how far it will go. I've never overclocked a CPU before, so when the 233 CPU arrives, I may need some advice as to what to do with the mobo jumpers. Please. 😁

retro games 100 wrote:

I just ran pcpbench inside windows 98 with no command line parameters, and it only gives me a score of about 21."

Not sure man, I always did my PCPbench tests by booting into DOS and applying the Scitech Display Doctor NOLFB fix.

retro games 100 wrote:

I may need some advice as to what to do with the mobo jumpers. Please. 😁

The stock settings are 66mhz bus x 3.5 multiplier, with 2.8v core. The supposed magic recipe is 100mhz bus x 3.0 multiplier, with a small voltage boost to 2.9v. You may need some decent cooling to keep it stable. The heatsink I selected is small, but made out of pure copper and 100% Socket 7 compatible.

Thanks a lot for the OC'ing tips - I'm eager to get the 233 CPU nice n toasty! 😀 (BTW, I had to grab a K6-2 400 & 500 CPU to satisfy my curiousity for some additional testing.)

Anyway, I put the AGP V3 in to another board, just for some more testing. This board is a Gigabyte AMD chipset-based board (GA-7IXE4), with a Duron 600mhz CPU in it. The mobo's dip-switches look like they are in an overclocked setting. I think it may be oc'ing the CPU to 800mhz.

Inside win98, I run pcpbench 103 (800x600x8), which gives me a score of about 71. That's more like it! However, glQuake (800x600x16) only does about 60 fps, and still can't beat SLI V2s in a P MMX 166 setup!

Next test: I remove the Duron 600mhz CPU, set the board's overclocking dip-switches to "normal mode", and put in an Athlon 1400 CPU. This CPU is designed for 133mhz bus speed boards, and my Gigabyte mobo is 100mhz bus speed, so the POST screen reports the CPU as 1050mhz. That'll do. But, running all my tests again (pcpbench & glQuake) don't provide any fps improvement at all.

BTW, is there a download link for Scitech's NOLFB utility? I DL'd the display doctor DOS packages, but it isn't included. BTW, I think running pcpbench inside win98 does the "no lfb thing" automatically. Although having said that, when I run pcpbench it says: LFB in the corner and not NOLFB.

Edit: Sorry, LFB is what I need to see in the pcpbench test screen, so I reckon that just by running pcpbench inside win98 should be all I need to do.

Additional test: I "play with fire" (well, a bit) and set the Gigabyte mobo's OC'ing dip-switches to 115mhz. Now, the BIOS POST screen reports the CPU to be 1208mhz. Then I run some more tests:

pcpbench 105 (1024x768x8) = 80fps. That's not bad!
pcpbench 107 (1280x1024x8) = 55.5fps. Not bad either!

What's interesting is that rerunning the Quake 1 & 2 benchies didn't show any sign of improvement.

retro games 100 wrote:

Inside win98, I run pcpbench 103 (800x600x8), which gives me a score of about 71. That's more like it! However, glQuake (800x600x16) only does about 60 fps, and still can't beat SLI V2s in a P MMX 166 setup!

60fps sounds an awful lot like a v-sync issue. Try to disable that or at least increase the refresh rate to confirm it (assuming it's currently set to 60hz).

You're right! I went to the excellent 3dfxzone.it website, and DL'd a tweak utility called "V Control". It's got lots of options - I chose to untick the "sync to vertical refresh" checkbox, and that improved the fps score. Also, I adjusted the V3's core & memory clock value, increasing it from 166 to 190.

I then reran glQuake in 1024x768x16 mode and got 78.1 fps! 1280x1024x16 gave a respectable score of 48.5 fps. Interestingly, pcpbench scores seemed to be unaffected by the core & memory clock increase. BTW, I propped up a meaty case fan next to the V3's heatsink! 😀

Note that there's no practical reason to disable vsync other than benchmarking (which is definitely useful in this case) or possibly reducing latency if you're sensitive to it. If you're getting a higher framerate than the refresh rate of your monitor then you're just working your video card extra hard with nothing to gain other than ugly tearing artifacts.

Yeah, the default refresh rate being 60hz, and getting almost exactly 60fps is a dead give away. Cool, I'm really glad that worked out for you. I think it is also interesting to note that that it took an OC'd V3 to catch up to the V2SLI, but I guess it all makes sense when you figure that the V3 was meant to be a single card solution for the V2SLI and 2D.

Time for another benchie! - Voodoo3 AGP & PCI, head to head!

Board: IWill AX100 plus, Super socket 7
CPU: P MMX 166, overclocked to 250 (FSB has been overclocked to 100.2mhz)
AGP: Voodoo3, the "standard" SD-RAM edition, overclocked to 190mhz

glQuake (800x600x16) = 98.1 fps (not bad!)

PCI: Voodoo3, the "special" SG-RAM edition, revision D, overclocked to 190mhz

glQuake (800x600x16) = 97.2 fps (ooh, nearly!)

Also, the postman just delivered a P MMX 233 (SL27S) CPU. I'm gonna fry it ASAP!

I've done some P MMX 233 CPU testing. To summarise, test 2) shows that I haven't been able to overclock the CPU to 300.

1) I set the mobo jumpers to "normal mode" = V Core 2.8, FSB 66, Multiplier 3.5. Switch on power, Everything OK.

2) I set the mobo jumpers to "magic 300 mode" = V Core 2.9, FSB 100, Multiplier 3.0. Switch on power, the general "power sound" then dies after about 2 seconds. No POST on screen. Possible cause: not enough V Core?

3) I set the mobo jumpers back to "normal mode", as in test 1) above. Then, inside the BIOS, I set the FSB to 75mhz. The POST screen then reports the CPU as 266. PC behaves OK, no problems.

4) Inside the BIOS, I increase the FSB to 83.3mhz. The POST screen reports the CPU as 290. The PC then behaves oddly - it reports a "windows protection error" - I then power off the mobo.

5) I rerun test 4), and after POST, the PC reports a problem with Configuration Manager, and then the mobo powers itself off automatically.

6) I change the mobo's Multiplier jumper from 3.5x to 3.0x. I switch on power, and POST reports the CPU to be 250 speed. The PC behaves normally, no problems.

7) Inside the BIOS, I increase the FSB to 100.2mhz. The general "power sound" then dies after about 2 seconds. No POST on screen.

I think I need some expert OC'ing assistance please! 😀

Edit: I just had a thought: I managed to OC a P MMX 166 (old brown flat CPU type) by 50% to reach 250. Perhaps I could get another old brown flat P MMX, at 200. A 50% increase on it would reach 300. Just a thought.

This isn't worth the pain...

It needed 3.2V to prevent it from crashing while doing nothing under MS-DOS. Temperatures were around 60° C under load.

PCPBench results (Mode 100, 640x400x8):
37.5 FPS LFB
35.2 FPS NoLFB

Quake results: None, because of instability.

Agh!!!! I have been out projected on my project!!!! 3.2 volts is pretty high, I'm glad to hear you haven't started a fire 😈. 60c? Nice, we just hit Thunderbird / P4 territory. I'm curious, was the chip you used a 200mhz or a 233mhz one? Ceramic or Organic Packaging? Is there any way to know if it was .35 or .28 process?

prophase_j wrote:

Agh!!!! I have been out projected on my project!!!!

Sorry, I just found the CPU while sorting through my stuff and decided to give it a spin... 😉

prophase_j wrote:

3.2 volts is pretty high, I'm glad to hear you haven't started a fire 😈. 60c? Nice, we just hit Thunderbird / P4 territory.

Maybe the heatsink I used was not up to the task. It is a 50x55x35mm aluminium HS with a 50mm fan.

prophase_j wrote:

I'm curious, was the chip you used a 200mhz or a 233mhz one? Ceramic or Organic Packaging? Is there any way to know if it was .35 or .28 process?

233 MHz, PPGA and 0.35 micron I guess.
(FV80503233, SL27S)

5u3, does your speedsys screenshot say that your P MMX 233 CPU is a 0.28 process chip? Am I reading this screenshot value correctly?

I'm not going to risk increasing the mobo's core voltage from 2.9 to 3.2. Instead, I might try 3.0. If that fails to boot the mobo, then I'm not going to continue with the 233 CPU. As an alternative test, I will try an old-style 'flat' P MMX 200 CPU, and see if increasing the mobo's FSB to 100 will yield a 300 speed CPU. I need to wait a few days for it to arrive from the mothership...

The first thread on this webpage looks interesting:

http://episteme.arstechnica.com/eve/forums/a/ … 74/m/4640902955

The O.P. appears to have a stable OC'd P-MMX-233 CPU, running at 291mhz. The settings used are -

83.3 FSB x 3.5 Multilplier, with a V Core of 3.2

5u3, would you be interested in briefly testing this? I mean, you've been able to see with your own eyes a 233 CPU set to 100FSB x 3 Multiplier @ 3.2V, and lived to tell the tale! 😉

If not, I'll give it a go. 😀

I myself am not even going to bother with the 75 and 83 bus speeds since that will mean that the PCI bus is running at 37.5 and 41.5, which is out of spec and possibly crating other problems. I wish there was a way to conclusivley know what process the chips are made on, surely well'll have better luck a .28 then with a .35.

retro games 100 wrote:

5u3, does your speedsys screenshot say that your P MMX 233 CPU is a 0.28 process chip? Am I reading this screenshot value correctly?

Nothing wrong with your eyesight, but I'm not sure whether the SpeedSys output is correct, as it lists 0.28 µm on all my P55C CPUs (one P233, one plastic P166 and one ceramic P166). I wanted to check a P54C as well, but my last remaining specimen decided not to appear on the show (turned out to be dead 🙄)

retro games 100 wrote:

83.3 FSB x 3.5 Multilplier, with a V Core of 3.2
5u3, would you be interested in briefly testing this?

Sure, will test tomorrow. If it turns out to be more stable at 83x3.5, I'm going to increase the I/O voltage and try for 100x3 again.

prophase_j wrote:

I myself am not even going to bother with the 75 and 83 bus speeds since that will mean that the PCI bus is running at 37.5 and 41.5, which is out of spec and possibly crating other problems.

Depends on the chipset. The P5A has standard AGP/PCI clocks (66/33 MHz) at 83 MHz FSB. There are two options at 75 MHz: 75/37.5 MHz for real men and 60/30 MHz for wimps 🤣
But basically I agree: If the board doesn't support the correct dividers or async mode for in-spec bus clocks, it's better to avoid those bus speeds.

5u3 wrote:

Depends on the chipset. The P5A has standard AGP/PCI clocks (66/33 MHz) at 83 MHz FSB. There are two options at 75 MHz: 75/37.5 MHz for real men and 60/30 MHz for wimps 🤣
But basically I agree: If the board doesn't support the correct dividers or async mode for in-spec bus clocks, it's better to avoid those bus speeds.

I happily stand corrected 😀 That is really awesome to know, that they took the time to put in a proper diver. I guess it makes sense, if you want to use the Cyrix chips.

On this webpage -

http://www.cpu-world.com/CPUs/Pentium/Intel-P … FV80503233.html

- On the Comments tab, the commentater says he ran a P-MMX-233 @ 300, with a Vcore @ 3.1. He also says he tried 100 x 3.5 @ 3.5v, and does not specifically mention any permanent damage to the board, which is very interesting. The commentater also "praises" a cheap K6-2 @ 400mhz running Quake2, which seems a bit strange.

I have now tried 83.3 x 3.5 @ 3v, and saw the win98 desktop - for a few moments before the PC rebooted. Nevermind. It's a fun project, regardless of its outcome. I will carry on expermenting...

Edit: I adjusted the mobo's I/O voltage jumper from 3.5 to 3.6 (it has no 3.3 setting), and adjusted the Vcore from 3.0 to 3.1, and now the win98 desktop seems stable. I then OC'd the Voodoo3 card to 190mhz, and ran glQuake (800x600x16), and got a fps of exactly 100.0!

This is running the board & CPU at 83.3f x 3.5m x 3.1cv x 3.6io = 290mhz. I then promptly switched the machine off! (I wish my CPU heatsink clip wasn't broken.)