JayCeeBee64 wrote:

feipoa, see if this walkthrough on Youtube helps:
https://www.youtube.com/watch?v=UJqjUvQKTCc

Not particularly - that player seems overly naughty. At any rate, I have completed what I set out to accomplish - Half-life, is almost playable with an Am5x86-160 using a Geforce2 MX400 if you can accept ~15 fps at 1280x960. The game felt marginally faster with the GF2 MX400 on a Am5x86-160 compared to a Voodoo3 on a Cyrix 5x86-133. If HF scales at all like Quake II, the POD-100 should yield another 2-3 fps.

feipoa wrote:

JayCeeBee64 wrote:

feipoa, see if this walkthrough on Youtube helps:
https://www.youtube.com/watch?v=UJqjUvQKTCc

Not particularly - that player seems overly naughty. At any rate, I have completed what I set out to accomplish - Half-life, is almost playable with an Am5x86-160 using a Geforce2 MX400 if you can accept ~15 fps at 1280x960. The game felt marginally faster with the GF2 MX400 on a Am5x86-160 compared to a Voodoo3 on a Cyrix 5x86-133. If HF scales at all like Quake II, the POD-100 should yield another 2-3 fps.

What if you dialed down the resolution? 1280x960 sounds kind of high for a 3D game running on a Socket 3 system, methinks. 🤣

The game speeds up some at lower resolutions, but by an incremental amount only. In Quake II, the difference between 1280x960 vs. 640x480 was approx. 2 fps. I personally do not mind Half-Life at its current speed on the Am5x86-160, but I do not normally play computer games, and as such, have not been spoiled by high frame-rate game play.

feipoa wrote:

The game speeds up some at lower resolutions, but by an incremental amount only. In Quake II, the difference between 1280x960 vs. 640x480 was approx. 2 fps. I personally do not mind Half-Life at its current speed on the Am5x86-160, but I do not normally play computer games, and as such, have not been spoiled by high frame-rate game play.

I envy you for being unspoiled 😉

Would using a modern-ish video card on a Socket 3 system make a difference for 2D games, like say, StarCraft?

386SX wrote:

Hi, just built this pc Pentium 100 and 98mb 72pin on Win98 and I have no words it is running Half life with 3dfx miniGl and the Voodoo1 with the S3 Virge 2Mb quiet well almost 15-25fps 640p all effects.. I would like to try a 486 100mhz but i dont have the mobo.

feipoa wrote:

Not particularly - that player seems overly naughty. At any rate, I have completed what I set out to accomplish - Half-life, is almost playable with an Am5x86-160 using a Geforce2 MX400 if you can accept ~15 fps at 1280x960. The game felt marginally faster with the GF2 MX400 on a Am5x86-160 compared to a Voodoo3 on a Cyrix 5x86-133. If HF scales at all like Quake II, the POD-100 should yield another 2-3 fps.

really have to doubt the numbers stated in this thread, valve's p133 minimum spec is a joke. to illustrate i dug out 2 videos of the game running on a 200mmx:

https://www.youtube.com/watch?v=CGZZkQ7NS6U
https://youtu.be/JE5mfNUmJEw?t=10m

as can be seen from the fps counter in the 2nd video, it can reach 30-60fps looking at a wall in an empty corridor but the framerate tanks hard in any even remotely complex area or any fight, where it will average in the teens and dip to single digits at worst (6fps in the lift crash scene in second video). and a p200mmx is substantially faster than a 160mhz 5x86, plus the game is probably leveraging the mmx instructions to speed up a few things. also, later in the game there are bigger areas with more scripting going than what is seen here, so judging the framerate even before even getting the HEV suit is a bad call.

i think the half-life: uplink demo is a good way to get a quick idea of the game's actual performance, that container area outside really destroys slow cpus.

edit: didn't think of this first, but maybe hw t&l is being used on a geforce when running under opengl? i could see this reducing cpu load compared to a voodoo, but still find it hard to imagine a 486 class cpu pulling off a 15fps average in this game.

As there doesn't seem to be a benchmark script for Half-life, a lot of the ~15 fps is user perception. I noted that game "felt" about as demanding as Quake II, which does have a timedemo and I was probably inferring with how that games feels and comparing it with the instantaneous frame-rate counter. Don't read too much into "~15 fps". I'm not spoiled by high frame-rate game play and tend to consider slower game play as playable.

Best thing is for you to obtain the hardware and test it for yourself and make draw your own conclusion. I consider GLQuake on an Am5x86-160 w/Voodoo2 exceptionally playable. I think the timedemo returns around 28 fps (from memory).

playability is subjective so i didn't touch on that, it's just that the 15fps number for a 486 struck me as odd given that the substantially more powerful 200mmx tends to hang around that number in decently busy scenes, as evidenced by those videos i posted. i'd still be curious to know if hw t&l can be leveraged in opengl here to help out weak cpus though.

coincidentally there is a video on youtube showing how a 5x86-160 with voodoo2 runs the glquake timedemo, it scores 16fps, although there might be a penalty for running under 98se: https://www.youtube.com/watch?v=lh9MBbECsNo

16 fps ? Something is probably not setup optimally. Even with sound enabled it should be faster than that. I get 26 fps with sound disabled. If you are interested in 3D games on a socket 3, you can check out these threads,

Voodoo 1 vs. Voodoo 2 on a 486
Re: Voodoo 1 vs. Voodoo 2 on a 486

Unless there is a timedemo, you cannot infer anything from the instantaneous fps values unless you are comparing with another system at the same instant.

I don't know if HW T&L was helping with the GF2 system. I no longer have that system setup. I swapped the guts out for a VLB/ISA-only Am5x86-160 system.

leileilol wrote:

Actually with a V2 there's multitexture optimization with the world and the 486 doesn't have to crunch the lightmap blending or span drivers (Quake's two big 486 offenders) leaving more breathing room, GLQuake is far faster than software Quake on one. (as long as you have r_dynamic 0 or gl_flashblend 1 🤣

Are you saying that the Voodoo2 has an advantage over the Voodoo1 on a 486 with Quake?

Half life does have a timedemo command, it just doesn't come with a default demo. You can either record your own or download one though. I uploaded the 3fingers blowout demo. The blowout demo only seems to work with version 1.0.0.5 of the game. I know there is a thread here where someone uploaded a demo recorded with a newer version of half life, I just can't find it atm.

Now I have a Geforce 256 in a system with a K6-3 450 and it gets 28fps in the blowout timedemo, with a TNT2 it gets 26fps.

Also, I did a comparison of a AMD K6-2 400, Pentium 2 450 and Cyrix MII (300/100) in half life with a pair of voodoo2 cards in my asus p5a-b:

Here is GLQuake on the same system:

Edit: I should add that I did all my testing with sound enabled, high quality sound and eax were disabled.

I'll try this on my 486+Voodoo2 setup. Any special instructions needed to get it running?

cxm717 wrote:

Also, I did a comparison of a AMD K6-2 400, Pentium 2 450 and Cyrix MII (300/100) in half life with a pair of voodoo2 cards in my asus p5a-b:

How come the frame rate is the same for various resolutions on the MII but the PII shows a consistent decrease in frame rate with increasing screen resolution?

feipoa wrote:

cxm717 wrote:

Also, I did a comparison of a AMD K6-2 400, Pentium 2 450 and Cyrix MII (300/100) in half life with a pair of voodoo2 cards in my asus p5a-b:

How come the frame rate is the same for various resolutions on the MII but the PII shows a consistent decrease in frame rate with increasing screen resolution?

I think it's because the game is CPU limited on the Cyrix and limited by the video card on the Pentium 2.

I have Half-Life version 1.0.0.9 on my 486 and can confirm that the 3 fingers timedemo is not working. Any luck finding the recorded timedemo with a newer half-life version?

The 3 fingers readme mentions that you can use this command string for use on newer versions of Half-Life, but it doesn't work. I receive a couldn't open error.

hl.exe -protocol 35 -console

I found that thread Half-Life Performance Issues. I haven't tried his demo. It needs patch 1108, I uploaded it here: https://megaupload.nz/120fS6b5n7/hl1108_exe

tpowell.ca wrote:

Are you saying that the Voodoo2 has an advantage over the Voodoo1 on a 486 with Quake?

Yes.

auron wrote:

edit: didn't think of this first, but maybe hw t&l is being used on a geforce when running under opengl? i could see this reducing cpu load compared to a voodoo, but still find it hard to imagine a 486 class cpu pulling off a 15fps average in this game.

No. All a Geforce2 can help in a 486's case is the fillrate.

cxm717 wrote:

I found that thread Half-Life Performance Issues. I haven't tried his demo. It needs patch 1108, I uploaded it here: https://megaupload.nz/120fS6b5n7/hl1108_exe

Are you able to see how the 3 finger's timedemo result compares with the result from that vogons thread? For example, are they off by 50%?
I'll download and try the demo from that link this evening.

I tried that demo on my K6-3 system and it just crashes the game. I tried with the Geforce card and the TNT2 with driver versions 2.08 and 12.41. I ended up finding an iso of the goty version of the game (I just googled it). I thought maybe there was something wrong with the patch I was using. It's same 1.1.0.8 that he used to record it but it still just crashes. I'll give it a go on a P3 system and see how it goes.

auron wrote:

really have to doubt the numbers stated in this thread

you should be, they are from non gamer collectors, its pretty much always best case scenario no sound looking at the wall I can totally play Quake on a 486 type of deal 😉

auron wrote:

a p200mmx is substantially faster than a 160mhz 5x86

probably ~4x faster

auron wrote:

, plus the game is probably leveraging the mmx instructions to speed up a few things

nah, mmx was useless for games

auron wrote:

edit: didn't think of this first, but maybe hw t&l is being used on a geforce when running under opengl?

AFAIR this was dependent on game using standard opengl calls for geometry manipulation/lighting vs its own internal custom routines