RacoonRider:

Ok, here is new picture with current hardware (don't ask me why are HDD results so much different this time, no idea 😀

leileilol
You are right, Voodoo 2 is considerably faster! +3FPS on the same machine with the same settings 😀

https://www.youtube.com/watch?v=hwgiIK-hLLc

Your L2 cache isn't doing much here. I guess you had to set your L2 cache to the slowest for the system to be stable with a 50 MHz FSB?

feipoa wrote:

Your L2 cache isn't doing much here. I guess you had to set your L2 cache to the slowest for the system to be stable with a 50 MHz FSB?

Yep, 4-2-2-2. Machine will freeze right after RAM detection with faster settings.

Such is the downfall of using the 50 MHz bus. Many motherboards also need to have the cache slowed down quite a bit on the 40 MHz setting. It makes using them hardly beneficial. I wish AMD would have made a 5x Am5x86.

Would using 12ns (or 10ns) chips make a difference? Board is currently equipped with 15ns chips.

Not in my experience.

sledge wrote:

Would using 12ns (or 10ns) chips make a difference? Board is currently equipped with 15ns chips.

Does it crash when setting the tightest memory and cache timings in the BIOS?

That benchmark should take into account the maps have been 'vispatched' modified to display more than usual - through the water surfaces that is.

also gl_flashblend is 0 in the video. Turning on the orange balls helps the 486 a lot also.

dirkmirk wrote:

Good to see someone else try a 3D card on a 486 system! […]
Show full quote

Good to see someone else try a 3D card on a 486 system!

In my experience the best CPU to use is the Pentium overdrive 83mhz, I use a Voodoo3 on my machine but I dont think it would be any faster for this class of machine compared to the voodoo1.

For reference at 640x480 in Quake timedemo I got 27.0fps and 26.9fps at 800x600, cant remember the exact figures for the Cyrix 5x86-120 but its noticeably slower.

Did you run the quakedemo with the -nosound option?

Is Quake timedemo optimized for Pentium instructions? That might explain why it runs better with a PoD than a Cyrix despite the slower clock speed.

sledge wrote:

...and Descent with 3dfx patch (perfect - 25 FPS in 800x600!)

Descent has a timedemo? Could you provide instructions on how to run this?

sliderider wrote:

Is Quake timedemo optimized for Pentium instructions? That might explain why it runs better with a PoD than a Cyrix despite the slower clock speed.

dirkmirk wrote:

For reference at 640x480 in Quake timedemo I got 27.0fps and 26.9fps at 800x600, cant remember the exact figures for the Cyrix 5x86-120 but its noticeably slower.

I would be surprised if the Cyrix 5x86-120 was a whole lot slower than the POD83. I do get 27.6 fps (nosound) in GLQuake w/Voodoo3 on a Cyrix 5x86-133/4x (LSSER, FP_FAST, RSTK, LOOP, BWRT optimisations used). Usually the 40 MHz graphics brings the Cyrix 5x86-120 close to the Cx5x86-133/4x for graphic intensive jobs. Perhaps it would score around 25 fps. The Am5x86-160, with its crappy FPU, gets 27.5 fps and FPU of the Am5x86-160 is 20% slower than that of the Cx5x86-133/4x.

By contrast, using the Matrox G200 GLQuake wrapper yields a messily 16.3 fps.

Sliderider, did you ever try your Cyrix 5x86-120/4x chip at 133 MHz?

Preliminary test results indicate that the Cyrix 5x86-120/4x chip in my possession is funcational at 133/4x. I am running it at 3.7 V with a socket 7 cooler.

EDIT: There is likely one or more register feature which is not stable at 133 MHz, which is why it was marked for 120 MHz. With some trial and error, this feature can be singled out and disabled.

feipoa wrote:

Descent has a timedemo? Could you provide instructions on how to run this?

Not that I know of 😀 I played first level for a while and simply checked framerate counter from time to time. And of course my numbers are completely off - first level is "small", with narrow corridors and smal rooms. Hence high FPS. But in large areas framerate drops to single digit numbers. Sorry for misleading information 🙁

feipoa wrote:

I would be surprised if the Cyrix 5x86-120 was a whole lot slower than the POD83. I do get 27.6 fps (nosound) in GLQuake w/Vood […]
Show full quote

sliderider wrote:

Is Quake timedemo optimized for Pentium instructions? That might explain why it runs better with a PoD than a Cyrix despite the slower clock speed.

dirkmirk wrote:

For reference at 640x480 in Quake timedemo I got 27.0fps and 26.9fps at 800x600, cant remember the exact figures for the Cyrix 5x86-120 but its noticeably slower.

I would be surprised if the Cyrix 5x86-120 was a whole lot slower than the POD83. I do get 27.6 fps (nosound) in GLQuake w/Voodoo3 on a Cyrix 5x86-133/4x (LSSER, FP_FAST, RSTK, LOOP, BWRT optimisations used). Usually the 40 MHz graphics brings the Cyrix 5x86-120 close to the Cx5x86-133/4x for graphic intensive jobs. Perhaps it would score around 25 fps. The Am5x86-160, with its crappy FPU, gets 27.5 fps and FPU of the Am5x86-160 is 20% slower than that of the Cx5x86-133/4x.

By contrast, using the Matrox G200 GLQuake wrapper yields a messily 16.3 fps.

Sliderider, did you ever try your Cyrix 5x86-120/4x chip at 133 MHz?

Nope. Keeping it at 40x3.

You mean, you don't want to run it at 4x30 as intended?

Anonymous Coward wrote:

You mean, you don't want to run it at 4x30 as intended?

Find me a motherboard with a 30mhz bus. Beside 40x3 makes a faster system overall.

I disagree. I've ran my am5x86 at 120mhz for awhile and it makes me wish I ran it at 160MHz again. for my 59.79 speedsys and playable Quake score back.

Hello guys, Good that i found this topic. I have planned to build an Intel 486DX4 100MHz with a fast Voodoo 1 6MB graphics, but seems the CPU part would be a huge bottleneck. For the Voodoo I will rather choose a 233mhz mmx pentium cpu, but i unsure what graphics should i put into my 486 system?

Any S3 Trio64 or Matrox Millennium card will be fine for 486.