VOGONS


Modern graphics on a 486

Topic actions

Reply 20 of 371, by swaaye

User metadata
Rank l33t++
Rank
l33t++

When I tried a Voodoo3 2000 in my SiS 496 board there was a major performance jump when I moved from a Am5x86 160 to POD @ 100. Ironically, while the POD was clearly fastest for 3D games, it measured slower elsewhere because the L2 cache wasn't working right.

I don't see Voodoo3 offering a speed boost over a Voodoo2 though because of the CPUs here. There's no chance of Voodoo2 being a bottleneck. But I do prefer Voodoo3 because it only needs 1 PCI slot and you get unlimited resolution choice, higher image quality and its top notch 2D.

Voodoo1 would however be at a disadvantage because it does not have full triangle setup offload meaning the CPU is tasked with some more work. This is where Rendition hit 3Dfx in their advertising. A Verite V1000 might give more performance than a Voodoo1 on such slow CPUs. But then with a V1000 you pretty much must play Speedy3D/RRedline games so comparisons are difficult.

Reply 21 of 371, by feipoa

User metadata
Rank l33t++
Rank
l33t++
swaaye wrote:

When I tried a Voodoo3 2000 in my SiS 496 board there was a major performance jump when I moved from a Am5x86 160 to POD @ 100. Ironically, while the POD was clearly fastest for 3D games, it measured slower elsewhere because the L2 cache wasn't working right.

Which motherboard was this? When I tested the POD100 on UMC boards, L2 worked fine. Did the POD83 have L2 issues? Some motherboard/BIOS's don't tolerate the POD's L1 cache very well, but I was not aware of L2 issues.

Plan your life wisely, you'll be dead before you know it.

Reply 22 of 371, by numeriK

User metadata
Rank Newbie
Rank
Newbie

@ feipoa, to respond to your 640x480 curiosity:

default OpenGL:
Textures max, 8-bit ON: 7.8fps
Textures max, 8-bit OFF: 7.8fps
Textures minimum, 8-bit ON: 7.9fps
Textures minimum, 8-bit OFF: 8.0fps (ironically the same as my original 320x240 result)

software:
alpha stipple NO: 4.5fps
alpha stipple YES: 4.4fps

As soon as I can get my hands on the full version I'll be able to modify the cfg's accordingly. Similar to what you've noticed, I'm sure we can get near or over 10fps with some tweaking.

8433UUD v2 | AMD 5x86 @ 180MHz (60MHz x 3, 30MHz PCI) | 64MB EDO | TNT 16MB PCI | SB AWE64 ISA | Win98SE

Reply 24 of 371, by feipoa

User metadata
Rank l33t++
Rank
l33t++
numeriK wrote:

Textures max, 8-bit OFF: 7.8fps

Thanks for testing this. Very comparable results:

Cyrix 5x86-133 (2x66)
WinNT 4.0
Matrox Millennium G200 16MB
Quake II Timedemo1, 640x480, OpenGL
7.6 fps

AMD X5-180 (3x60)
Win98SE
RIVA TNT 16MB
Quake II Timedemo1, 640x480, OpenGL
7.8 fps

The logical next step now is to see how well the Voodoo3 does on a Cyrix 5x86-133 (2x66) and an AMD X5-180 then compare it with the POD83/100.

Plan your life wisely, you'll be dead before you know it.

Reply 25 of 371, by sliderider

User metadata
Rank l33t++
Rank
l33t++
swaaye wrote:

But I do prefer Voodoo3 because it only needs 1 PCI slot and you get unlimited resolution choice, higher image quality and its top notch 2D.

Can't you use a V2 with an ISA or VLB 2D card? You still use only 1 PCI slot that way.

Reply 26 of 371, by swaaye

User metadata
Rank l33t++
Rank
l33t++
feipoa wrote:

Which motherboard was this? When I tested the POD100 on UMC boards, L2 worked fine. Did the POD83 have L2 issues? Some motherboard/BIOS's don't tolerate the POD's L1 cache very well, but I was not aware of L2 issues.

It was a MSI MS4144 with SiS 496/497. Cachechk tests did not indicate any L2 present even though the L2 was enabled. It was years ago but I remember trying configuring the CPU jumpers in various ways for the POD but nothing restored the L2.

POD compatibility is spotty among motherboards and this is probably an example of that.

sliderider wrote:

Can't you use a V2 with an ISA or VLB 2D card? You still use only 1 PCI slot that way.

Sure you could do that.

Reply 27 of 371, by amstrad1640

User metadata
Rank Newbie
Rank
Newbie

Ati rage 128 pci run well in opengl.

For me i 've used these PCI graphic cards:
Voodoo 1
voodoo 2
voodoo 2 sli
voodoo 3 3000 (the best forme)
Kyro 1 64 mb (ok but slow, bad drivers?)
Ati Rage 128 (fast on opengl game and i don't know why i can't to use direct 3d?)
Geforce 2 mx 440 (geforce 2like) windows 98se ok, 3d gameand onpelgl not run
3d prohet hercules 9000 PCI (ATI) no post, no boot , nothing!

Reply 29 of 371, by feipoa

User metadata
Rank l33t++
Rank
l33t++
amstrad1640 wrote:
Ati rage 128 pci run well in opengl. […]
Show full quote

Ati rage 128 pci run well in opengl.

For me i 've used these PCI graphic cards:
Voodoo 1
voodoo 2
voodoo 2 sli
voodoo 3 3000 (the best forme)
Kyro 1 64 mb (ok but slow, bad drivers?)
Ati Rage 128 (fast on opengl game and i don't know why i can't to use direct 3d?)
Geforce 2 mx 440 (geforce 2like) windows 98se ok, 3d gameand onpelgl not run
3d prohet hercules 9000 PCI (ATI) no post, no boot , nothing!

You have used all these on a 486? What motherboard and chipset?

Plan your life wisely, you'll be dead before you know it.

Reply 30 of 371, by numeriK

User metadata
Rank Newbie
Rank
Newbie

Well, the full version of Q2 came in. 😈

FWIW, with all the tweaks in the Q2 book (in OpenGL basically everything is off) I received:

OpenGL demo1 640x480: 10.2fps
SW demo1 640x480: 4.9fps

OpenGL demo2 640x480: 10.5fps
SW demo2 640x480: 4.8fps

This is my baseq2\autoexec.cfg to get the 10fps+:
set cl_gun "0"
set cl_particles "0"
set gl_ext_palettedtexture "1"
set gl_modulate "1.0"
set gl_ztrick "0"
set gl_playermip "2.0"
set gl_picmip "2.0"
set gl_shadows "0"
set gl_dynamic "0"
set gl_lightmap "1"
set gl_texturemode "GL_LINEAR_MIPMAP_NEAREST"
set gl_triplebuffer "0"
set gl_polyblend "0"
set gl_flashblend "0"
set gl_cull "1"
set gl_round_down "1"
set gl_skymip "1"
set in_joystick "0"
set s_initsound "0"
set cd_nocd "1"
set cl_footsteps "0"
set intensity "2.0"

Returning these values makes it playable, while retaining decent fps: cl_gun "1", gl_lightmap "0", s_initsound "1"

With the above 3 settings changed (this is how I play the game) I receive:
OpenGL demo1 640x480: 8.5fps
OpenGL demo2 640x480: 8.9fps

8433UUD v2 | AMD 5x86 @ 180MHz (60MHz x 3, 30MHz PCI) | 64MB EDO | TNT 16MB PCI | SB AWE64 ISA | Win98SE

Reply 32 of 371, by numeriK

User metadata
Rank Newbie
Rank
Newbie
leileilol wrote:

With a Voodoo, how would a WickedGL minigl work with Q2? They're faster.

Yeah, TBH I'm considering a V3 2000/3000 to see if that'll work. However, I heard those may have issues with these UMC boards... ?

8433UUD v2 | AMD 5x86 @ 180MHz (60MHz x 3, 30MHz PCI) | 64MB EDO | TNT 16MB PCI | SB AWE64 ISA | Win98SE

Reply 33 of 371, by feipoa

User metadata
Rank l33t++
Rank
l33t++
numeriK wrote:

Returning these values makes it playable, while retaining decent fps: cl_gun "1", gl_lightmap "0", s_initsound "1"

I personally thought the game looked almost playable at 8.6 fps with these settings:
gl_flashblend 1
cl_gun 0

Interestingly, the POD83 only added 0.3 fps to the score, albiet at 33 Mhz vs. 66 Mhz fsb of the Cyrix 5x86.

@numeriK
I wonder what score you'd get if you added some L2 cache to that X5-180 setup? You'd probably need to run that L2 at 3-2-2.

Have you measured the CPU die cap temperature after prolonged use at 180 MHz? I see you are running it at 5 V. You might want to see what minimum voltage you can run it at stably to bring down the temperature and extend its life.

Plan your life wisely, you'll be dead before you know it.

Reply 34 of 371, by numeriK

User metadata
Rank Newbie
Rank
Newbie

@ feipoa, you're playing Q2 with no sound? I found that only gives me .5-1.0fps on average, so I kept it.

Yeah, I'll give the L2 cache a shot... even at grandma timings... 😵

8433UUD v2 | AMD 5x86 @ 180MHz (60MHz x 3, 30MHz PCI) | 64MB EDO | TNT 16MB PCI | SB AWE64 ISA | Win98SE

Reply 38 of 371, by feipoa

User metadata
Rank l33t++
Rank
l33t++
noshutdown wrote:
feipoa wrote:

With sound. Without is pretty dull.

but for benchmark, you should do it without sound to get a more accurate score.

I don't follow. For results to be accurate, they should be repeatable. I get the same fps score with the same sound card inserted each time. If you are refering to a comparison between the Matrox/Cyrix system and the NVIDIA/AMD system, then, yes, either the hardware would need to be the same or different hardware setup in such a way as to isolate the performance of the CPU and graphics. This would make for the best comparison. The Ultimate 686 Benchmark Comparison has been done with sound enabled, whereby the soundcard, graphics, RAM, motherboard, etc, are all the same.

@leileilol
Good suggestion. I'll try that out.

Plan your life wisely, you'll be dead before you know it.

Reply 39 of 371, by amstrad1640

User metadata
Rank Newbie
Rank
Newbie
amstrad1640 wrote: Ati rage 128 pci run well in opengl. […]
Show full quote

amstrad1640 wrote:
Ati rage 128 pci run well in opengl.

For me i 've used these PCI graphic cards:
Voodoo 1
voodoo 2
voodoo 2 sli
voodoo 3 3000 (the best forme)
Kyro 1 64 mb (ok but slow, bad drivers?)
Ati Rage 128 (fast on opengl game and i don't know why i can't to use direct 3d?)
Geforce 2 mx 440 (geforce 2like) windows 98se ok, 3d gameand onpelgl not run
3d prohet hercules 9000 PCI (ATI) no post, no boot , nothing!

You have used all these on a 486? What motherboard and chipset?

Yes i 've used theses cards on 2 motherboards: Abit PB4 et MSI 4144

Somes pics from french board Here i'm i386dx40