VOGONS


Quake III Arena Benchmark Thread

Topic actions

Reply 160 of 197, by SPBHM

User metadata
Rank Oldbie
Rank
Oldbie
rcarkk wrote:

Just to confirm the settings. Should we use the absolute maximum settings and 32bit colors except trilinear filtering? Or should we use high quality preset settings without trilinear filtering?

I'm using the settings from the OP
https://abload.de/img/q3a_settingsvcs4v.png

unless I specify on my post that I used something else.

it would've been easier to avoid mistakes if it was just about using one of the settings profile, well, I suppose we should just say the settings we are using when posting a result, be it "op settings" or "high quality" or whatever!

Reply 161 of 197, by rcarkk

User metadata
Rank Member
Rank
Member

Thanks. I did used those settings...except on the V2 SLI. I had to put the second highest texture setting, because they only have 8MB for textures. They only scored 11fps with max texture settings.

Baby AT socket7 - Pentium MMX 233MHz + 3Dfx Voodoo
Socket 8 build - Soyo 6FA + Pentium Pro 200MHz + 3Dfx Voodoo 2 12MB
PC Remake - Pentium III 450 + Matrox G400 16MB
The K6-III build

Reply 162 of 197, by cxm717

User metadata
Rank Member
Rank
Member
rcarkk wrote:
My results for a Matrox G400 32MB and 2x 3Dfx Voodoo 2 12MB in SLI on a Pentium III-S 1.26Ghz running on a QDI Advance 10T with […]
Show full quote

My results for a Matrox G400 32MB and 2x 3Dfx Voodoo 2 12MB in SLI on a Pentium III-S 1.26Ghz running on a QDI Advance 10T with Via Apollo Pro 133T and 256MB of PC133 at 2-2-2-2 timmings. Sound card is a Live! 5.1 digital and OS is Win98 SE.

6.23 drivers + G400 @125/166Mhz = 38.1fps @16bit
6.23 drivers + G400 @125/166Mhz = 25.4fps @32bit

Latest reference drivers + Voodoo 2 12MB SLI = 28.2fps @16bit (with 2nd highest texture resolution, because at max res, the framerate is very low due to only 12MB)

You should try the matrox TurboGL on your G400, you need to install a 5.xx driver though IIRC.

Reply 163 of 197, by rcarkk

User metadata
Rank Member
Rank
Member
cxm717 wrote:
rcarkk wrote:
My results for a Matrox G400 32MB and 2x 3Dfx Voodoo 2 12MB in SLI on a Pentium III-S 1.26Ghz running on a QDI Advance 10T with […]
Show full quote

My results for a Matrox G400 32MB and 2x 3Dfx Voodoo 2 12MB in SLI on a Pentium III-S 1.26Ghz running on a QDI Advance 10T with Via Apollo Pro 133T and 256MB of PC133 at 2-2-2-2 timmings. Sound card is a Live! 5.1 digital and OS is Win98 SE.

6.23 drivers + G400 @125/166Mhz = 38.1fps @16bit
6.23 drivers + G400 @125/166Mhz = 25.4fps @32bit

Latest reference drivers + Voodoo 2 12MB SLI = 28.2fps @16bit (with 2nd highest texture resolution, because at max res, the framerate is very low due to only 12MB)

You should try the matrox TurboGL on your G400, you need to install a 5.xx driver though IIRC.

I think i had tried it, and the OpenGL ICD of the 6.23 drivers, were faster than older drivers with TurboGL. But i need to check that again. One thing i did notice is that the G400 takes a huge hit with Trilinear filtering enabled in Quake 3. This benchmark requiring Bilinear filtering because of V2 is actually benefiting the G400 performance figures.

Baby AT socket7 - Pentium MMX 233MHz + 3Dfx Voodoo
Socket 8 build - Soyo 6FA + Pentium Pro 200MHz + 3Dfx Voodoo 2 12MB
PC Remake - Pentium III 450 + Matrox G400 16MB
The K6-III build

Reply 164 of 197, by rcarkk

User metadata
Rank Member
Rank
Member

My results for an Asus V3400TNT Riva TNT 16MB and one 3Dfx Voodoo 2 12MB on a Pentium III-S 1.26Ghz running on a QDI Advance 10T with Via Apollo Pro 133T and 256MB of PC133 at 2-2-2-2 timmings. Sound card is a Live! 5.1 digital and OS is Win98 SE.

21.83 + Riva TNT 16MB = 22.4fps @1024x768 16bit
21.83 + Riva TNT 16MB = 13.4fps @1024x768 32bit

Latest reference drivers + Voodoo 2 12MB = 31.8fps @800x600 16bit (with 2nd highest texture resolution, because at max res, the framerate is very low due to only 12MB)
21.83 + Riva TNT 16MB = 35.7fps @800x600 16bit (with 2nd highest texture resolution, to compare with Voodoo 2 12MB texture limitation)

I´m impressed. The Riva TNT is scoring higher all across the games i tested. This is true in Direct3D and OpenGL. Not much faster, but faster. Even UT at 800x600 16bit, runs at 47fps on both cards. And the TNT is using D3D vs the almighty Glide on the Voodoo 2.
I did notice that the image quality is better on the Riva TNT compared to Voodoo 2, even at 16bit. I never thought that the TNT was clearly superior to the Voodoo 2 12MB.

Baby AT socket7 - Pentium MMX 233MHz + 3Dfx Voodoo
Socket 8 build - Soyo 6FA + Pentium Pro 200MHz + 3Dfx Voodoo 2 12MB
PC Remake - Pentium III 450 + Matrox G400 16MB
The K6-III build

Reply 165 of 197, by SPBHM

User metadata
Rank Oldbie
Rank
Oldbie

I thought it was fairly interesting how close the results I got with the VIA P4M266 IGP (ProSavgeDDR, Savage 4 based IGP from 2001/2002) are to the TnT 16MB,
the CPU is not the same but I doubt it's having much of an impact with this level of IGP, (and the northwood Celeron is incredibly slow for how high clocked it is)

IGP is overclocked to 154Mhz and the ram is DDR 266

celeron21-q3-oc.JPG
Filename
celeron21-q3-oc.JPG
File size
302.46 KiB
Views
1620 views
File license
Fair use/fair dealing exception

OP settings
I also tested with 16 bits and it got 22.6FPS
and with 800x600 16bit and the texture quality reduced by 1 it got 36.5 (it actually made no difference to performance running with max or 1 step down)

Reply 166 of 197, by rcarkk

User metadata
Rank Member
Rank
Member
SPBHM wrote:
I thought it was fairly interesting how close the results I got with the VIA P4M266 IGP (ProSavgeDDR, Savage 4 based IGP from 20 […]
Show full quote

I thought it was fairly interesting how close the results I got with the VIA P4M266 IGP (ProSavgeDDR, Savage 4 based IGP from 2001/2002) are to the TnT 16MB,
the CPU is not the same but I doubt it's having much of an impact with this level of IGP, (and the northwood Celeron is incredibly slow for how high clocked it is)

IGP is overclocked to 154Mhz and the ram is DDR 266

celeron21-q3-oc.JPG

OP settings
I also tested with 16 bits and it got 22.6FPS
and with 800x600 16bit and the texture quality reduced by 1 it got 36.5 (it actually made no difference to performance running with max or 1 step down)

This means a standard S3 Savage 4 Pro 32MB is slower than a Riva TNT 16MB. The scores of the TNT i posted are with no OC. Weird, i thought that the Savage 4 Pro from 1999, was faster card than the TNT from 1998

Baby AT socket7 - Pentium MMX 233MHz + 3Dfx Voodoo
Socket 8 build - Soyo 6FA + Pentium Pro 200MHz + 3Dfx Voodoo 2 12MB
PC Remake - Pentium III 450 + Matrox G400 16MB
The K6-III build

Reply 167 of 197, by cxm717

User metadata
Rank Member
Rank
Member

I tried Q3 on a 32MB Rage 128pro and 16MB Savage4 in my OR840/P3800/512MB 800 RDRAM system.

First the Rage:

p3r128psys.png
Filename
p3r128psys.png
File size
65.88 KiB
Views
1582 views
File license
Fair use/fair dealing exception
q3rage128p6292.png
Filename
q3rage128p6292.png
File size
450.76 KiB
Views
1582 views
File license
Fair use/fair dealing exception

And the Savage4:

s3s4ep3sys.png
Filename
s3s4ep3sys.png
File size
73.26 KiB
Views
1582 views
File license
Fair use/fair dealing exception
s3s4eq3.png
Filename
s3s4eq3.png
File size
387.31 KiB
Views
1582 views
File license
Fair use/fair dealing exception

Reply 168 of 197, by SPBHM

User metadata
Rank Oldbie
Rank
Oldbie
rcarkk wrote:
SPBHM wrote:
I thought it was fairly interesting how close the results I got with the VIA P4M266 IGP (ProSavgeDDR, Savage 4 based IGP from 20 […]
Show full quote

I thought it was fairly interesting how close the results I got with the VIA P4M266 IGP (ProSavgeDDR, Savage 4 based IGP from 2001/2002) are to the TnT 16MB,
the CPU is not the same but I doubt it's having much of an impact with this level of IGP, (and the northwood Celeron is incredibly slow for how high clocked it is)

IGP is overclocked to 154Mhz and the ram is DDR 266

celeron21-q3-oc.JPG

OP settings
I also tested with 16 bits and it got 22.6FPS
and with 800x600 16bit and the texture quality reduced by 1 it got 36.5 (it actually made no difference to performance running with max or 1 step down)

This means a standard S3 Savage 4 Pro 32MB is slower than a Riva TNT 16MB. The scores of the TNT i posted are with no OC. Weird, i thought that the Savage 4 Pro from 1999, was faster card than the TNT from 1998

when I started testing this IGP I started looking around for savage 4 results and the IGP looks slower, it didn't look massively slower when comparing IGP with OC vs stock Savage 4, but still looked slower, but stock vs stock and OC vs OC the discrete card should be faster, as @cxm717 shows, with very close clocks the Savage 4 Pro is 50% faster than the ProSavageDDR IGP in this test!

the only thing is that I'm using windows XP, it might be possible that the IGP is better with 98 drivers,
because AFAIK the IGP has the same number of pipelines/tmus as Savage4 and the same the design, it might be that the shared memory is a significant performance penalty.

Reply 169 of 197, by rcarkk

User metadata
Rank Member
Rank
Member
cxm717 wrote:
I tried Q3 on a 32MB Rage 128pro and 16MB Savage4 in my OR840/P3800/512MB 800 RDRAM system. […]
Show full quote

I tried Q3 on a 32MB Rage 128pro and 16MB Savage4 in my OR840/P3800/512MB 800 RDRAM system.

First the Rage:

p3r128psys.png
q3rage128p6292.png

And the Savage4:

s3s4ep3sys.png
s3s4eq3.png

Great results for your Savage 4. Your card is running stock at 160/166Mhz on GPU/RAM. That´s very fast and much faster than reference Savage 4 Pro clocks of 125/143Mhz.
What card do you own?

Baby AT socket7 - Pentium MMX 233MHz + 3Dfx Voodoo
Socket 8 build - Soyo 6FA + Pentium Pro 200MHz + 3Dfx Voodoo 2 12MB
PC Remake - Pentium III 450 + Matrox G400 16MB
The K6-III build

Reply 170 of 197, by rcarkk

User metadata
Rank Member
Rank
Member
SPBHM wrote:
when I started testing this IGP I started looking around for savage 4 results and the IGP looks slower, it didn't look massively […]
Show full quote
rcarkk wrote:
SPBHM wrote:
I thought it was fairly interesting how close the results I got with the VIA P4M266 IGP (ProSavgeDDR, Savage 4 based IGP from 20 […]
Show full quote

I thought it was fairly interesting how close the results I got with the VIA P4M266 IGP (ProSavgeDDR, Savage 4 based IGP from 2001/2002) are to the TnT 16MB,
the CPU is not the same but I doubt it's having much of an impact with this level of IGP, (and the northwood Celeron is incredibly slow for how high clocked it is)

IGP is overclocked to 154Mhz and the ram is DDR 266

celeron21-q3-oc.JPG

OP settings
I also tested with 16 bits and it got 22.6FPS
and with 800x600 16bit and the texture quality reduced by 1 it got 36.5 (it actually made no difference to performance running with max or 1 step down)

This means a standard S3 Savage 4 Pro 32MB is slower than a Riva TNT 16MB. The scores of the TNT i posted are with no OC. Weird, i thought that the Savage 4 Pro from 1999, was faster card than the TNT from 1998

when I started testing this IGP I started looking around for savage 4 results and the IGP looks slower, it didn't look massively slower when comparing IGP with OC vs stock Savage 4, but still looked slower, but stock vs stock and OC vs OC the discrete card should be faster, as @cxm717 shows, with very close clocks the Savage 4 Pro is 50% faster than the ProSavageDDR IGP in this test!

the only thing is that I'm using windows XP, it might be possible that the IGP is better with 98 drivers,
because AFAIK the IGP has the same number of pipelines/tmus as Savage4 and the same the design, it might be that the shared memory is a significant performance penalty.

Maybe the clocks of the Savage 4 based IGP is running with low clocks. You need to open Powerstrip to see the clocks...
Edit: My mystaque. Yes, your Savage 4 IGP is OC to 152Mhz. Yes, maybe the shared system memory is slowing the IGP.

Baby AT socket7 - Pentium MMX 233MHz + 3Dfx Voodoo
Socket 8 build - Soyo 6FA + Pentium Pro 200MHz + 3Dfx Voodoo 2 12MB
PC Remake - Pentium III 450 + Matrox G400 16MB
The K6-III build

Reply 171 of 197, by cxm717

User metadata
Rank Member
Rank
Member
rcarkk wrote:
cxm717 wrote:
I tried Q3 on a 32MB Rage 128pro and 16MB Savage4 in my OR840/P3800/512MB 800 RDRAM system. […]
Show full quote

I tried Q3 on a 32MB Rage 128pro and 16MB Savage4 in my OR840/P3800/512MB 800 RDRAM system.

First the Rage:

p3r128psys.png
q3rage128p6292.png

And the Savage4:

s3s4ep3sys.png
s3s4eq3.png

Great results for your Savage 4. Your card is running stock at 160/166Mhz on GPU/RAM. That´s very fast and much faster than reference Savage 4 Pro clocks of 125/143Mhz.
What card do you own?

It's a number nine card. I really like it, it has much better quality output compared to my other S3 cards.

Reply 172 of 197, by cxm717

User metadata
Rank Member
Rank
Member

Here is my Geforce256 SDR on the same OR840 setup:

p3gfsys.png
Filename
p3gfsys.png
File size
68.16 KiB
Views
1553 views
File license
Fair use/fair dealing exception
q3gfsdr3082.png
Filename
q3gfsdr3082.png
File size
397.97 KiB
Views
1553 views
File license
Fair use/fair dealing exception

Reply 173 of 197, by RoyBatty

User metadata
Rank Oldbie
Rank
Oldbie

CPU-Z reports wrong CPU freq, it's actually 4.4ghz (same as north bridge). The elusive 1k FPS.

quake3demo.jpg
Filename
quake3demo.jpg
File size
608.58 KiB
Views
1540 views
File license
Fair use/fair dealing exception

Reply 174 of 197, by rcarkk

User metadata
Rank Member
Rank
Member
cxm717 wrote:

Here is my Geforce256 SDR on the same OR840 setup:

p3gfsys.png
q3gfsdr3082.png

Very good fps for the GF256 SDR. Can you do another benchmark with Trilinear filtering enabled?
My V5 5500 on a PIII Coop 866Mhz scored only 51.7fps @1024x768 with Trilinear filtering on, and the TNT2 Ultra is scoring 31fps in the same machine and same settings.
I always thought the V5 5500 is faster than GF256 SDR and the TNT2 Ultra was only 30% slower than GF256 SDR.

Baby AT socket7 - Pentium MMX 233MHz + 3Dfx Voodoo
Socket 8 build - Soyo 6FA + Pentium Pro 200MHz + 3Dfx Voodoo 2 12MB
PC Remake - Pentium III 450 + Matrox G400 16MB
The K6-III build

Reply 175 of 197, by rcarkk

User metadata
Rank Member
Rank
Member
cxm717 wrote:
rcarkk wrote:
cxm717 wrote:
I tried Q3 on a 32MB Rage 128pro and 16MB Savage4 in my OR840/P3800/512MB 800 RDRAM system. […]
Show full quote

I tried Q3 on a 32MB Rage 128pro and 16MB Savage4 in my OR840/P3800/512MB 800 RDRAM system.

First the Rage:

p3r128psys.png
q3rage128p6292.png

And the Savage4:

s3s4ep3sys.png
s3s4eq3.png

Great results for your Savage 4. Your card is running stock at 160/166Mhz on GPU/RAM. That´s very fast and much faster than reference Savage 4 Pro clocks of 125/143Mhz.
What card do you own?

It's a number nine card. I really like it, it has much better quality output compared to my other S3 cards.

Is the card overclocked with those 160/166Mhz settings?

Baby AT socket7 - Pentium MMX 233MHz + 3Dfx Voodoo
Socket 8 build - Soyo 6FA + Pentium Pro 200MHz + 3Dfx Voodoo 2 12MB
PC Remake - Pentium III 450 + Matrox G400 16MB
The K6-III build

Reply 177 of 197, by cxm717

User metadata
Rank Member
Rank
Member

Here are some stock and OC results on my G400Max. Both running the 555 driver with the 1.3 TurboGL.

The G400 at stock clocks:

g400sys.png
Filename
g400sys.png
File size
74.81 KiB
Views
1466 views
File license
Fair use/fair dealing exception
g40032.png
Filename
g40032.png
File size
392.86 KiB
Views
1460 views
File license
Fair use/fair dealing exception

and OCed:

g400ocsys.png
Filename
g400ocsys.png
File size
77.49 KiB
Views
1466 views
File license
Fair use/fair dealing exception
g400oc32.png
Filename
g400oc32.png
File size
395.35 KiB
Views
1460 views
File license
Fair use/fair dealing exception
Last edited by cxm717 on 2019-05-08, 18:16. Edited 3 times in total.

Reply 178 of 197, by cxm717

User metadata
Rank Member
Rank
Member
rcarkk wrote:
Very good fps for the GF256 SDR. Can you do another benchmark with Trilinear filtering enabled? My V5 5500 on a PIII Coop 866Mh […]
Show full quote
cxm717 wrote:

Here is my Geforce256 SDR on the same OR840 setup:

p3gfsys.png
q3gfsdr3082.png

Very good fps for the GF256 SDR. Can you do another benchmark with Trilinear filtering enabled?
My V5 5500 on a PIII Coop 866Mhz scored only 51.7fps @1024x768 with Trilinear filtering on, and the TNT2 Ultra is scoring 31fps in the same machine and same settings.
I always thought the V5 5500 is faster than GF256 SDR and the TNT2 Ultra was only 30% slower than GF256 SDR.

I can run more tests once I'm done with the G400. From what I remember the Geforce got about 55fps with trilinear. Keep in mind this version of Q3 defaults to s3tc enabled, that probably makes a difference (for the SDR version at least). New versions of Q3 seem to default to disabled.

Reply 179 of 197, by bofh.fromhell

User metadata
Rank Oldbie
Rank
Oldbie

My contribution to the 2000 GPU & CPU period correct list:

1l5bnnqm.png

Athlon 1200C.
ABIT KG7 with AMD 760 chipset (released nov 2000)
Gainward GeForce2 Ultra, 71.89 drivers.
1GB DDR memory, super tight 2-2-2-2-6 settings.
Soundblaster Live! (SB0060), funny enough I get better results with sound on.
Windows 2000 with Unofficial SP5.
Nothing overclocked.

Amazing machine IMO.
And rock solid since the recap, zero problems.
Can't belive that I never payed attention to the 760 chipset when it was current.
The Y2K machine is only missing a few parts now =)