The Matrox Millenium G100 gets 9937 FPS in 3dbench2 but the G200 only gets 1034 FPS 😁 both have 8MB SDR

It seems 3dbench2 is not the best tool to measure perfromance when it comes to Pentium 3 systems 😜

edit

I ran 3Dbench2 again with the cards that got really strange scores. There was a dot that was invisible on my LCD screen. I am using a CRT now that can display low res better.

The result for the G100 was 993.7 FPS
My Riva TNT still gets a strange score, 000.0 FPS 😁 I have remounted the card 3 times without the issue going away. FASTVID wrongly identifies the card as 4MB so that is a possible reason.
The Geforce2 MX 400 and Riva TNT2 M64 get very very low 3Dbench2 scores compared to my other cards, 167.8 FPS and 145.5 FPS.
All other tests give good numbers for all cards.

/edit

3dbench2 is also not the tool to measure performance on the 486slc2 machine I am testing... when the internal 1kb L1 cache is disabled I get 21.4fps , with the 1kb enabled (that REALLY improves performance) I get 11.2fps... it's still buggy :\ (buy comparison with PCPBench: 1kb=OFF 1.1fps / 1kb=ON 3.3fps with IBM's onboard sloooow vga)

IMHO 3dbench needs to go , maybe find a pseudo-3d game with benchmark like Roadhog or something similar

You learn stuff from benching.

I am pretty sure the 2D engine in the Riva TNT2 M64 is identical to the one in the Geforce2 MX 400.
They score identical except in 3Dbench2 and that test seem to produce a bit strange results 😀

Just added an i7-3770 for the hell of it. 😁

Doom wasn't that much faster on it, nor were the 3d benchmarks. Quake ran in 3.0 seconds though. I'd try it with the Nvidia Quadro K600 video card in it too, but I'd probably mess up my triple monitor install in order to test it, and I assume the built in card on the i7 is probably not a bottleneck in DOS.

keropi wrote:

I am currently benching Quake on my IBM PS/1pro machine, it is a cacheless 386sx/20 wonder that is upgraded with a 486slc2/40mhz , i387 fpu and 16MB IBM ram...
Quake runs but I estimade 0.3fps 🤣 🤣 🤣 🤣

edit: it bugs me why they did not include any cache at all 😢 , wtf IBM , "pro" model....

Can you try that chip in maybe another 386 board that has Cache?

Also this CPU has L1 cache right? Make sure it is enabled, either in the BIOS or through software.

Mau1wurf1977 wrote:

Can you try that chip in maybe another 386 board that has Cache?

Also this CPU has L1 cache right? Make sure it is enabled, either in the BIOS or through software.

Can't try it in another mobo , it needs a 386sx to clip on and I don't have another 386sx mobo... (nor planning to get one, not much point IMHO 🤣 )
Cache on the benchmarks I submitted was enabled ofcourse, if it was disabled then it is not faster than the stock 386sx/20 under it. ...and Quake was running at 0.7fps 🤣 🤣 🤣 🤣

btw, I also made benchmarks with the stock onboard IBM vga (some TI with 256? 512? kb of vram, nothing detects it to see specs) with and without the 1kb of cache, is that interesting to add since this is a pre-made machine and all...

That's awesome 😀

We are getting close to 100 submissions 😀

Here a quick graph I put together based on some of the data.

SO MUCH DATA!!!

Just added my P3 on 100 and 112MHz. Got 180fps in Doom! 🤣

I was going to add 133MHz too, but suddenly it won't boot anymore at that speed...

As I ran the tests, Doom in particular, I was amused to see that AMD DX2 CPU, unlike Intel SX and DX2 CPUs, creates mesh distortions at the left side of the screen. It lookes like whenever the edge of the polygon reaches the end of the screen, it is calculated as if it was placed directly at the edge of the screen by X and Y axis, but eternally distant by Z axis.

Am I the only one experiencing this problems, or is it time to say "Intel. The way it's meant to be played"?

jwt27 wrote:

Just added my P3 on 100 and 112MHz. Got 180fps in Doom! 🤣

I was going to add 133MHz too, but suddenly it won't boot anymore at that speed...

Nice 😀 133 is a tough setting. Not al video cards take it. I had good success with Nvidia cards though.

RacoonRider wrote:

As I ran the tests, Doom in particular, I was amused to see that AMD DX2 CPU, unlike Intel SX and DX2 CPUs, creates mesh distortions at the left side of the screen. It lookes like whenever the edge of the polygon reaches the end of the screen, it is calculated as if it was placed directly at the edge of the screen by X and Y axis, but eternally distant by Z axis.

Am I the only one experiencing this problems, or is it time to say "Intel. The way it's meant to be played"?

Well the AMD 386-40 seem to run it fine 😀 I had an AMD 486DX-4 as a teenager and Doom ran fine. Could be a BIOS support issue? I'm sure someone else with an AMD DX2 will add a result soon and then you could contact that person.

Mau1wurf1977 wrote:

jwt27 wrote:

Just added my P3 on 100 and 112MHz. Got 180fps in Doom! 🤣

I was going to add 133MHz too, but suddenly it won't boot anymore at that speed...

Nice 😀 133 is a tough setting. Not al video cards take it. I had good success with Nvidia cards though.

Weird thing is, I was always able to boot into DOS at 133, just not in Windows. Now it doesn't even POST anymore and I have to reset the BIOS settings using the jumper, although I do get the 3dfx init screen at first.

Voodoo Banshee and ATI Rage 128 Pro score more or less identical.

Did 3DFX do some secret 2D engine deal with ATI? 😀

Perhaps both cards are very bottlenecked by the CPU.

Added a few results...

@Mau1wurf1977
I tried to add another sheet for underclocked results, but apparently the permissions prevent the proper copying of the table to another sheet (It seemed to work at first, but it went havoc with permission errors after a short time, so I removed it again).

So, if you think this is a good idea, could you please move my underclocked results to a new page or just delete them from the "stock settings" sheet? Thanks!

Can we clarify the meaning of "underclocking"?

Is it to simulate slower, but existing, CPUs. E.g. running a Pentium 200 at 66 x 2? That can go straight into the standard database.

Or are we talking about things like a 486DX-66 chip with a FSB set to 20 MHz? I think that setting fits in nicely in the overclocking section, which is quite empty at the moment anyway.

Haha I just benched a Hercules Dynamite ET4000 ISA 512kb or perhaps 1MB? It dosnt have any memory in the extra sockets.
No Fastvid with the ISA card because Fastvid cant auto detect an ISA cards memory adress.
But a P3 850 helps performance a bit anyhow.
42.7 FPS in quake! Smooth as butter but Doom looked lika a slideshow.

Also the Doom FPS dosnt show for the last 3 systems.

I notice that from #97 and onwards DOOM fps are not getting calculated... is this a google-docs thing? maybe the system is busy or something?
I just finished entry #99 btw 😀

I'll fix it ASAP 😀

EDIT: Fixed

Another total shock! The fastest PCI card I have tested so far is an ARK Logic ARK2000PV.
Only the S3 Trio V2/DX comes close to the ARK Logic card performance wise.

[Edit]

Things are back to normal, the world is not going to end. Matrox Millenium II is the fastest PCI card I have tested so far 😀.
The ARK Logic Card is still faster in DOOM though.

[/Edit]

That ARK2000 chipset is very fast in VGA mode but can't keep up with a Matrox card in SVGA, and can't match the beautiful image quality of the Matrox either!