so for me, a perfect 486 experience can´t be reached with a FIC 486-GVT-2 ?
the only way it could improve just a little bit is with a AMD 133mhz and using the TSENG videocard.

but in order to use the AMD133mhz i need to buy a voltage regulator socket3 addon, because the motherboard only has configuration for 5v cpus.

Cga.8086 wrote:

so for me, a perfect 486 experience can´t be reached with a FIC 486-GVT-2 ?
the only way it could improve just a little bit is with a AMD 133mhz and using the TSENG videocard.

but in order to use the AMD133mhz i need to buy a voltage regulator socket3 addon, because the motherboard only has configuration for 5v cpus.

You could try to overclock your DX4-100 to run at a FSB of 40 MHz (i.e., at 120 MHz) but, then you sometimes need to apply wait states on your VLB (usually via a jumper on the motherboard). At that speed, it should run much faster but, I would then add a fan on top of the heatsink. This is of course at your own risk.
As previously stated, check your CMOS settings to ensure that your RAM & cache timings are optimal.

Cga.8086 wrote:

so for me, a perfect 486 experience can´t be reached with a FIC 486-GVT-2 ?
the only way it could improve just a little bit is with a AMD 133mhz and using the TSENG videocard.

but in order to use the AMD133mhz i need to buy a voltage regulator socket3 addon, because the motherboard only has configuration for 5v cpus.

You could always use on of the upgrade chips that has a voltage regulator built onto it.

You could always use on of the upgrade chips that has a voltage regulator

...and horribly overpriced.

yea those chips sockets that go below a 3.4v processor are rare to find here and must be expensive

that was the reason why i used a pentium 486 overdrive. because it has a built in heatsink and regulator so i cound usr on this fic mothrboard

anyway i think i will just forget about VLB

today for example i got a pentium1 100mhz with a motherboard QDI P5i430TX/T1B+ and 16mb ram for 5 dollars and tried doom1 with a realmagic64 pci videocard and speed is perfect 3120 realtics in doom1 timedemo. And didnt see any slowdown when i played a couple levels

i used a pentium 486 overdrive

Once again, by your description you don't have Pentium Overdrive, which is overpriced too.

Cga.8086 wrote:

so for me, a perfect 486 experience can´t be reached with a FIC 486-GVT-2 ?
the only way it could improve just a little bit is with a AMD 133mhz and using the TSENG videocard.

but in order to use the AMD133mhz i need to buy a voltage regulator socket3 addon, because the motherboard only has configuration for 5v cpus.

Realistically, I think you've got a perfect 486 experience. 15-25fps is very normal in DOOM with a 486 class processor.

As you've now experienced, many things are just better on a Pentium.

My thoughts exactly. 😀 I know everyone has their own self-imposed goals around here, and maybe you just want to squeeze the most out of the 486 platform for the sheer thrill of it. I support that kind of nuttiness. It's why I'm not using DOSBox, or running my old boxen off flash media.

But if you're after performance, there were like... two or three generations of processor after the 486, and they do appear to be a little bit faster.

Only 2 or 3?

35 x gametics divide by realtics gives you FPS ( Frames per sercond ).
In Doom test.

trident vlb cxi : 3863 gametics in 4392 realtics
FPS = 35x3863 divide by 4392 = 30.7

tseng et4000 vlb: 3863 gametics in 3683 realtics
FPS = 35x3863 divide by 3683 = 36.7

diamond S3 vlb : 3863 gametics in 3803 realtics
FPS = 35x3863 divide by 3803 = 35.5

Wow ! that is low FPS.

This video may help.
https://www.youtube.com/watch?v=jlQXS1VrFUQ

DOSBench download. ( BURN to CD and then boot off DOS boot Disk and run dos tools from CD. )
https://www.philscomputerlab.com/dos-benchmark-pack.html

Also in MS-Windows I think you need to enable 32-bit access in System settings.

Cga.8086 wrote:

ok here are some doom1 timedemo demo3 […]
Show full quote

ok here are some doom1 timedemo demo3

trident vlb cxi : 3863 gametics in 4392 realtics

tseng et4000 vlb: 3863 gametics in 3683 realtics

diamond S3 vlb : 3863 gametics in 3803 realtics

i do still see a slight frame drop with tseng and diamond on open areas.

im starting to like more the amd 133mhz with s3 trio64 pci

For timedemo 3, usually gametics is 2134. Are you running timedemo 1? So you are getting between 31-35 fps? That seems about right, but a tad slow. An Intel DX4-100 with write-back cache (yours has write-through) gets about 41 fps with a fast PCI card. I don't think using a write-back i486DX4 will give you 41 fps. Its been commented that 486-based FIC boards run a little slow compared to others.

SirNickity wrote:

BeginnerGuy wrote:

How fast pixels can get blasted to the video memory will be your bottleneck

If only it were a Sega Genesis....

"Blasting pixels" was a common term used to discuss putpixel() and double buffer functions back then. The latency of the video memory which correlates to how fast words or dwords could be transferred is going to factor significantly into performance on a 2d card.

Faster bus, lower latency, faster doom, barring cpu bottlenecks or excessive wait states on memory. Hence the suggestion to compare the cards in top bench, two cards with the same chip may have greatly varied performance.

For timedemo 3, usually gametics is 2134

That's not shareware. Demo 3 is longer in registered version.

BeginnerGuy wrote:

"Blasting pixels" was a common term used to discuss putpixel() and double buffer functions back then.

I think he was referring to Sega's marketing term of "blast processing".

BeginnerGuy wrote:

Faster bus, lower latency, faster doom, barring cpu bottlenecks or excessive wait states on memory. Hence the suggestion to compare the cards in top bench, two cards with the same chip may have greatly varied performance.

DOOM has an added factor to performance though: it uses Mode X.
Most benchmarks just measure the raw bandwidth between CPU and video memory. DOOM however switches between bitplanes regularly during drawing, and renders per-byte.
So because of its byte-accesses, it will never reach the theoretical maximum bandwidth anyway. And certain VGA chips are notoriously slow in switching between bitplanes, which becomes an additional performance bottleneck.
In other words: benchmarks such as Topbench are not necessarily representative for DOOM performance.

Most likely benched Ultimate Doom's - "Unholy Cathedral", E3M5 - demo3.

As Scali pointed out, Doom is a way more interesting test for the gfx-hardware than "just filling the screen with x bytes" and measuring time elapsed.

I got about same amount of realtics as gametics with 100Mhz 486 and pretty much any pci card, so that matches quite well with those vlb results.

amadeus777999 wrote:

Most likely benched Ultimate Doom's - "Unholy Cathedral", E3M5 - demo3.

It is! I recognize the gametics number, this is the demo I use to benchmark for historical reasons.

Scali wrote:

I think he was referring to Sega's marketing term of "blast processing". […]
Show full quote

BeginnerGuy wrote:

"Blasting pixels" was a common term used to discuss putpixel() and double buffer functions back then.

I think he was referring to Sega's marketing term of "blast processing".

BeginnerGuy wrote:

Faster bus, lower latency, faster doom, barring cpu bottlenecks or excessive wait states on memory. Hence the suggestion to compare the cards in top bench, two cards with the same chip may have greatly varied performance.

DOOM has an added factor to performance though: it uses Mode X.
Most benchmarks just measure the raw bandwidth between CPU and video memory. DOOM however switches between bitplanes regularly during drawing, and renders per-byte.
So because of its byte-accesses, it will never reach the theoretical maximum bandwidth anyway. And certain VGA chips are notoriously slow in switching between bitplanes, which becomes an additional performance bottleneck.
In other words: benchmarks such as Topbench are not necessarily representative for DOOM performance.

Ahh good point scali, I was thinking doom was in mode13 using a double buffer from main memory for some reason