VOGONS


Quake III Arena Benchmark Thread

Topic actions

Reply 140 of 197, by SPBHM

User metadata
Rank Oldbie
Rank
Oldbie
SPBHM wrote:

all stock apart from the e2140.
816.3 FPS, SPBHM, Radeon HD 5850, Core i5 2310, H61, 8GB DDR3 1333 DC, Win 10 x64
301.7 FPS, SPBHM, Radeon HD 4670 DDR2 (750/400) PCIe, Pentium E2140@2133MHz, VIA P4M900, 3GB DDR2 533 SC, Win 7 SP1 x86

I decided to test the same PC that I used for the 5850 but with the current card and I'm a little surprised...

618.9 FPS, SPBHM, Radeon R7 370 2GB, Core i5 2310, H61, 8GB DDR3 1333 DC, Win 10 x64

also decided to test the PC that currently has the same "HD4670" and again...

274.1, SPBHM, Radeon HD 4670 DDR2 (750/400) PCIe, Celeron E1400@2.54GHz, G31, 4GB DDR2 850 DC, Win 10 x86

the 4670 maybe it makes sense because perhaps it's slower on windows 10? is it really hurt by half the l2 cache? but G31 is much faster than the other chipset and there is over 400MHz more.

the 370 being so much slower than the 5850 I can't explain, only thing that I could think made the PC slower are the Spectre/Meltdown patches since that previous test.... or it's just that the drivers are far less optimized for this!?

edit: OK, cache is everything,
I removed the 2.54GHz Celeron dual core, and on the same board placed an e7300 (2.66GHz), and...

460.6, SPBHM, Radeon HD 4670 DDR2 (750/400) PCIe, C2D e7300 @ 2.66GHz, G31, 4GB DDR2 800 DC, Win 10 x64

Last edited by SPBHM on 2019-02-17, 22:55. Edited 1 time in total.

Reply 141 of 197, by Almoststew1990

User metadata
Rank Oldbie
Rank
Oldbie

I tried out my 4770K and 1070ti and got 997fps (1.3 seconds) at 1280*1024, the max that my screen supports :p

Ryzen 3700X | 16GB 3600MHz RAM | AMD 6800XT | 2Tb NVME SSD | Windows 10
AMD DX2-80 | 16MB RAM | STB LIghtspeed 128 | AWE32 CT3910
I have a vacancy for a main Windows 98 PC

Reply 142 of 197, by slivercr

User metadata
Rank Member
Rank
Member

Dear Q3 benchmark experts that have R200-based graphics cards and SMP configurations, please help me explain the following behavior I encountered while benchmarking. Here are the numbers...

1280x1024, r_ext_compressed_textures 0

r_smp 0
Run1 Run2 Run3 Run4 Run5
97.7 97.6 97.7 97.6 97.8

r_smp 1
Run1 Run2 Run3 Run4 Run5
114.9 107.2 107.1 107.3 107.2

1280x1024, r_ext_compressed_textures 1

r_smp 0
Run1 Run2 Run3 Run4 Run5
99.0 99.0 99.0 99.0 98.9

r_smp 1
Run1 Run2 Run3 Run4 Run5
115.9 115.9 115.7 115.8 115.9

Why do I lose frames when enabling SMP if I leave r_ext_compressed_textures at its default value of 0? Notes and observations:

  • I'm using version 1.32;
  • this happens to me with both a Radeon 8500 64 MB @ 275/275 and a Radeon 9100 128 MB @ 250/250;
  • when benchmarking with r_smp 1 and r_ext_compressed_textures 0, running vid_restart between benchmarks will prevent the FPS loss;
  • the performance loss in successive benchmarks happens at all resolutions, but the higher the resolution the more noticeable it gets;
  • in the same machine, no GeForce card incurs in a similar performance loss when r_ext_compressed_textures 0;
  • varying r_primitives has no effect, its default value of 0 yields the best results.

Any ideas?

Outrigger: an ongoing adventure with the OR840
QuForce FX 5800: turn your Quadro into a GeForce

Reply 143 of 197, by Kamerat

User metadata
Rank Oldbie
Rank
Oldbie

129.7 FPS, Kamerat, Intel GMA 900 @ 400MHz, Intel Pentium M 745A @ 2556MHz, Lenovo 3000 C100, 2GB DDR2 SDRAM DC @ 284MHz CL3-2-2-4, XP-SP3

DOS Sound Blaster compatibility: PCI sound cards vs. PCI chipsets
YouTube channel

Reply 144 of 197, by Standard Def Steve

User metadata
Rank Oldbie
Rank
Oldbie

I'm now running my PIII at 1628MHz, so I thought I'd retest my 9800 Pro. My previous score @ 1575MHz was 208.6 FPS.

PIII-S @ 1628MHz, 155MHz FSB
2GB DDR @ 310MHz, 2-2-2-5
9800 Pro @ stock, Catalyst 4.12
X-Fi Platinum
QDI Advance 12T mobo, Via Apollo Pro266T chipset
WinXP Media Center 2005 SP3

1024x768, all settings maxed = 220.2 FPS

PIII-1628, 9800Pro-Q3A.png
Filename
PIII-1628, 9800Pro-Q3A.png
File size
1.81 MiB
Views
1915 views
File license
Fair use/fair dealing exception

This machine is now EXACTLY twice as fast as a PIII-850.

P3-850-Q3A-9800Pro.png
Filename
P3-850-Q3A-9800Pro.png
File size
598.33 KiB
Views
1915 views
File license
Fair use/fair dealing exception

And 7 times faster than a P2-300!

P2-300-Q3A-9800Pro.png
Filename
P2-300-Q3A-9800Pro.png
File size
894.39 KiB
Views
1915 views
File license
Fair use/fair dealing exception

P6 chip. Triple the speed of the Pentium.
Tualatin: PIII-S @ 1628MHz | QDI Advance 12T | 2GB DDR-310 | 6800GT | X-Fi | 500GB HDD | 3DMark01: 14,059
Dothan: PM @ 2.9GHz | MSI Speedster FA4 | 2GB DDR2-580 | GTX 750Ti | X-Fi | 500GB SSD | 3DMark01: 43,190

Reply 145 of 197, by athlon-power

User metadata
Rank Member
Rank
Member

Now that I've got a graphics card that will perform acceptably, I'll go ahead and post my results here. Specs of the system are:

- Intel SE440BX-2 Motherboard
- Intel Pentium III Katmai 500MHz SL35E
- Single-stick 128MB PC133 RAM under-clocked by the motherboard to 100MHz.
- Creative Labs 3D Blaster 32MB (nVidia TNT2 32MB, 125MHz core, 150MHz memory, 128-bit memory bus)
- Cobra AW744L-II PCI (Yamaha YMF744)
- Samsung SC-148F 48x IDE CD-ROM
- Western Digital WD400 40GB IDE HDD
- 3Com EtherLink XL 3C905C-TX 10/100 PCI
- Mitsumi 1.44MB 3.5" Floppy Diskette Drive

I got 24.5 FPS.

When my system was using the 16MB TNT2 Pro I had (143MHz core, 166MHz memory, 128-bit memory bus), I got a whopping 9.8FPS average. I'm unsure as to why this is- the only edge that my current TNT2 has over my old TNT2 Pro is that it has 32MB of VRAM. I guess Quake III favors cards with more VRAM, I'm not sure.

IMG_20190415_014506.jpg
Filename
IMG_20190415_014506.jpg
File size
755.34 KiB
Views
1845 views
File license
Fair use/fair dealing exception
IMG_20190415_014702.jpg
Filename
IMG_20190415_014702.jpg
File size
717.45 KiB
Views
1845 views
File license
Fair use/fair dealing exception

Also, I noticed the extreme differences between the other systems used in the first post as compared to mine. Some people were getting ~38FPS average with a Pentium II 350! All of those systems had a 3DFX Voodoo of some kind in them, most of them used Voodoo 3 cards. I guess that's the power of Glide.

Where am I?

Reply 147 of 197, by slivercr

User metadata
Rank Member
Rank
Member
slivercr wrote:
Dear Q3 benchmark experts that have R200-based graphics cards and SMP configurations, please help me explain the following behav […]
Show full quote

Dear Q3 benchmark experts that have R200-based graphics cards and SMP configurations, please help me explain the following behavior I encountered while benchmarking. Here are the numbers...

1280x1024, r_ext_compressed_textures 0

r_smp 0
Run1 Run2 Run3 Run4 Run5
97.7 97.6 97.7 97.6 97.8

r_smp 1
Run1 Run2 Run3 Run4 Run5
114.9 107.2 107.1 107.3 107.2

1280x1024, r_ext_compressed_textures 1

r_smp 0
Run1 Run2 Run3 Run4 Run5
99.0 99.0 99.0 99.0 98.9

r_smp 1
Run1 Run2 Run3 Run4 Run5
115.9 115.9 115.7 115.8 115.9

Why do I lose frames when enabling SMP if I leave r_ext_compressed_textures at its default value of 0? Notes and observations:

  • I'm using version 1.32;
  • this happens to me with both a Radeon 8500 64 MB @ 275/275 and a Radeon 9100 128 MB @ 250/250;
  • when benchmarking with r_smp 1 and r_ext_compressed_textures 0, running vid_restart between benchmarks will prevent the FPS loss;
  • the performance loss in successive benchmarks happens at all resolutions, but the higher the resolution the more noticeable it gets;
  • in the same machine, no GeForce card incurs in a similar performance loss when r_ext_compressed_textures 0;
  • varying r_primitives has no effect, its default value of 0 yields the best results.

Any ideas?

Bump.

Any idea why r_ext_compressed_textures set to 0 malfunctions with R200 GPUs when SMP mode is used? It was pointed out before in this thread that it could be safely disabled if the GPU in use had 64 MB of memory or more. Should I just accept it as a lovable quirkiness of a bygone era with poor SMP support?

Outrigger: an ongoing adventure with the OR840
QuForce FX 5800: turn your Quadro into a GeForce

Reply 148 of 197, by athlon-power

User metadata
Rank Member
Rank
Member
havli wrote:
The requirements are: 1. demo001 2. 1024x768 resolution (windowed mode also allowed) 3. 32-bit colors and textures (except Voodo […]
Show full quote

The requirements are:
1. demo001
2. 1024x768 resolution (windowed mode also allowed)
3. 32-bit colors and textures (except Voodoo3 and possibly other 16-bit only capable HW), Lightmap lighting and full details

Garrett W wrote:

You are using 32bit color depth which probably kills performance on the poor TNT2. Pull it back to 16bit and performance will see a huge boost.

That's the official requirements of the benchmark, though- completely maxed out settings, save for the resolution. There's no way I'd ever actually run the game at these settings, it would probably be too slow, I just use the High Quality preset (even though I think that uses 32 bit color as well), and it's more than playable. I'll probably run it at the same settings, save for setting the color to 16-bit, when I get home later today, and see what results I get from that.

Where am I?

Reply 149 of 197, by matze79

User metadata
Rank l33t
Rank
l33t
slivercr wrote:

Dear Q3 benchmark experts that have R200-based graphics cards and SMP configurations, please help me explain the following behavior I encountered while benchmarking. Here are the numbers...

as far as i know enabling SMP only moves Audio to second processor.
So music and soundeffects are running on CPU #2.

https://www.retrokits.de - blog, retro projects, hdd clicker, diy soundcards etc
https://www.retroianer.de - german retro computer board

Reply 150 of 197, by slivercr

User metadata
Rank Member
Rank
Member
matze79 wrote:
slivercr wrote:

Dear Q3 benchmark experts that have R200-based graphics cards and SMP configurations, please help me explain the following behavior I encountered while benchmarking. Here are the numbers...

as far as i know enabling SMP only moves Audio to second processor.
So music and soundeffects are running on CPU #2.

Hmmm, I've had performance improvements when benchmarking with no soundcard installed, so what you're saying may be partially true, but its not the whole story.

But from your post can I guess you're hinting at a driver issue with ATI, compressed textures, and SMP?

Outrigger: an ongoing adventure with the OR840
QuForce FX 5800: turn your Quadro into a GeForce

Reply 151 of 197, by matze79

User metadata
Rank l33t
Rank
l33t

What about Nvidia ? is the Quadro Driver SMP aware ?

https://www.retrokits.de - blog, retro projects, hdd clicker, diy soundcards etc
https://www.retroianer.de - german retro computer board

Reply 152 of 197, by matze79

User metadata
Rank l33t
Rank
l33t

Found a description about SMP on Q3 Engine:

http://fabiensanglard.net/quake3/renderer.php

https://www.retrokits.de - blog, retro projects, hdd clicker, diy soundcards etc
https://www.retroianer.de - german retro computer board

Reply 153 of 197, by slivercr

User metadata
Rank Member
Rank
Member
matze79 wrote:

What about Nvidia ? is the Quadro Driver SMP aware ?

With GeForce cards I have no similar troubles, I get a performance boost and the performance of successive benchmarks doesn't deteriorate, no matter the setting of compressed textures.

EDIT:

matze79 wrote:

Found a description about SMP on Q3 Engine:

http://fabiensanglard.net/quake3/renderer.php

Beautiful! I'll give it a read.

Outrigger: an ongoing adventure with the OR840
QuForce FX 5800: turn your Quadro into a GeForce

Reply 154 of 197, by 65C02

User metadata
Rank Newbie
Rank
Newbie
Standard Def Steve wrote:
I'm now running my PIII at 1628MHz, so I thought I'd retest my 9800 Pro. My previous score @ 1575MHz was 208.6 FPS. […]
Show full quote

I'm now running my PIII at 1628MHz, so I thought I'd retest my 9800 Pro. My previous score @ 1575MHz was 208.6 FPS.

PIII-S @ 1628MHz, 155MHz FSB
2GB DDR @ 310MHz, 2-2-2-5
9800 Pro @ stock, Catalyst 4.12
X-Fi Platinum
QDI Advance 12T mobo, Via Apollo Pro266T chipset
WinXP Media Center 2005 SP3

1024x768, all settings maxed = 220.2 FPS

PIII-1628, 9800Pro-Q3A.png

This machine is now EXACTLY twice as fast as a PIII-850.

P3-850-Q3A-9800Pro.png

And 7 times faster than a P2-300!

P2-300-Q3A-9800Pro.png

Radeon 9800 is a pretty new card for the Pentium II/III. The video driver is probably leveraging the SSE instructions on the PIIIs making them much more efficient than the P2. As I recall, the PII wasn't much slower than PIII when SSE instructions weren't being used.

Running the timedemo on my Celeron 466 with GeForce 4MX-440 nets approximately 45 FPS. It's not great, but I would have been happy with it 20 years ago. 😊

Reply 155 of 197, by SPBHM

User metadata
Rank Oldbie
Rank
Oldbie
65C02 wrote:
Standard Def Steve wrote:
I'm now running my PIII at 1628MHz, so I thought I'd retest my 9800 Pro. My previous score @ 1575MHz was 208.6 FPS. […]
Show full quote

I'm now running my PIII at 1628MHz, so I thought I'd retest my 9800 Pro. My previous score @ 1575MHz was 208.6 FPS.

PIII-S @ 1628MHz, 155MHz FSB
2GB DDR @ 310MHz, 2-2-2-5
9800 Pro @ stock, Catalyst 4.12
X-Fi Platinum
QDI Advance 12T mobo, Via Apollo Pro266T chipset
WinXP Media Center 2005 SP3

1024x768, all settings maxed = 220.2 FPS

PIII-1628, 9800Pro-Q3A.png

This machine is now EXACTLY twice as fast as a PIII-850.

P3-850-Q3A-9800Pro.png

And 7 times faster than a P2-300!

P2-300-Q3A-9800Pro.png

Radeon 9800 is a pretty new card for the Pentium II/III. The video driver is probably leveraging the SSE instructions on the PIIIs making them much more efficient than the P2. As I recall, the PII wasn't much slower than PIII when SSE instructions weren't being used.

Running the timedemo on my Celeron 466 with GeForce 4MX-440 nets approximately 45 FPS. It's not great, but I would have been happy with it 20 years ago. 😊

it depends on the flavor of P3, Katmai or Coppermine, the 850 is Coppermine so it should be a good amount faster than PII even at the same clock and without SSE, if anything due to the much more effective l2 cache, and quake 3 seems to love some good l2 cache.

actually his scores kind of makes me wonder, how far a Coppermine could go with the same amount of love, max clock with that FSB and ram configuration,
because as cool as Tualatin is... I can help but always remind myself that it's newer than Pentium 4 🤣

Celeron 466 at least had faster l2 cache, the PII 300 runs its l2 cache at 150MHz

Reply 156 of 197, by 65C02

User metadata
Rank Newbie
Rank
Newbie
SPBHM wrote:
it depends on the flavor of P3, Katmai or Coppermine, the 850 is Coppermine so it should be a good amount faster than PII even a […]
Show full quote
65C02 wrote:
Standard Def Steve wrote:
I'm now running my PIII at 1628MHz, so I thought I'd retest my 9800 Pro. My previous score @ 1575MHz was 208.6 FPS. […]
Show full quote

I'm now running my PIII at 1628MHz, so I thought I'd retest my 9800 Pro. My previous score @ 1575MHz was 208.6 FPS.

PIII-S @ 1628MHz, 155MHz FSB
2GB DDR @ 310MHz, 2-2-2-5
9800 Pro @ stock, Catalyst 4.12
X-Fi Platinum
QDI Advance 12T mobo, Via Apollo Pro266T chipset
WinXP Media Center 2005 SP3

1024x768, all settings maxed = 220.2 FPS

PIII-1628, 9800Pro-Q3A.png

This machine is now EXACTLY twice as fast as a PIII-850.

P3-850-Q3A-9800Pro.png

And 7 times faster than a P2-300!

P2-300-Q3A-9800Pro.png

Radeon 9800 is a pretty new card for the Pentium II/III. The video driver is probably leveraging the SSE instructions on the PIIIs making them much more efficient than the P2. As I recall, the PII wasn't much slower than PIII when SSE instructions weren't being used.

Running the timedemo on my Celeron 466 with GeForce 4MX-440 nets approximately 45 FPS. It's not great, but I would have been happy with it 20 years ago. 😊

it depends on the flavor of P3, Katmai or Coppermine, the 850 is Coppermine so it should be a good amount faster than PII even at the same clock and without SSE, if anything due to the much more effective l2 cache, and quake 3 seems to love some good l2 cache.

actually his scores kind of makes me wonder, how far a Coppermine could go with the same amount of love, max clock with that FSB and ram configuration,
because as cool as Tualatin is... I can help but always remind myself that it's newer than Pentium 4 🤣

Celeron 466 at least had faster l2 cache, the PII 300 runs its l2 cache at 150MHz

Hmm, I was looking up some specs on these chips. The Pentium II and Celeron have 64-bit caches busses, whilst Coppermine and Tualatin have 256-bit busses. I suppose this would improve the gaming performance of these chips substantially! 😀

Reply 157 of 197, by SPBHM

User metadata
Rank Oldbie
Rank
Oldbie
SPBHM wrote:

well, still more than 1 frame per second!

K6-2 533, VIA MVP4, 128MB PC97, Trident Blade 3d IGP, ESS onboard, Win98SE: 1.7 fps

even the menu feels like it runs at 1FPS, if I lower everything it's kind of playable at low res, over 20FPS

same PC with a Voodoo 4 4500
K6-2 533, VIA MVP4, 128MB PC97, Voodoo 4 4500 PCI, ESS onboard, Win98SE: 31.0 fps

quake3-k62-v4500.jpg
Filename
quake3-k62-v4500.jpg
File size
334.1 KiB
Views
1678 views
File license
Fair use/fair dealing exception

Reply 158 of 197, by rcarkk

User metadata
Rank Member
Rank
Member

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?

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 159 of 197, by rcarkk

User metadata
Rank Member
Rank
Member

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)

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