VOGONS


First post, by RETROKOMODO

User metadata
Rank Member
Rank
Member

Hi folks,

I'm just about to embark on a project that's going to take a while to complete - running benchmarks on the following cards:

AGP 9600xt from 2003,
AGP 9800 pro from 2003,
AGP 7300 GT from 2006,
AGP 7800 GS+ from 2006,
PCI-E 7950 GT from 2006,
AGP x1950 pro from 2006,
PCI-E HD 3850 from 2007,
AGP HD 3850 from 2008,
AGP HD 4650 from 2008,
PCI-E HD 4850 from 2008,
PCI-E HD 6870 from 2010

Some of them may not work, as not yet tested and just arrived from Ebay. These are the only cards I have to hand, so can't add more even if I wanted to! all will be running the last available drivers for each card having first DDU's the previous card's drivers off. Some of these cards support higher than DirectX9, but as some don’t, I am not using benchmarks that run on a DirectX version higher than 9 so i'm using Win XP SP3 as I normally do, and the system i'm using is:

ASROCK 4CoreDual-VSTA
E7600 Core 2 Duo @ stock 3.06ghz
About 3.35GB of 4GB DDR2 Hyper-X in dual channel
8xAGP & 4xPCI-E slots. Considering the speed of the cards i'm hoping there won't de too much of a throttle on the PCIE slot - but I guess i'll find out!

I'm using 3dmark 2001se, 3dmark03, 3dmark05, 3dmark06, aquamark3, and the Unengine sanctuary v2.3, tropics v1.3 and heaven v3 (v4 won't install) benchmarks. Game benchmarks will be far cry 2 set to optimal at 1080p and 720p, lost planet 2, stalker: call of pripyat (medium settings), world in conflict (medium settings), crysis (medium settings), Half Life 2: Lost Coast (maximum settings) and FEAR (medium settings).

Any suggestions on alterations to the lineup? The combination of cards and benchmarks is already going to take ages to get through as and when I can find the time, but suggestions welcome!

Last edited by RETROKOMODO on 2022-10-12, 22:24. Edited 1 time in total.

Reply 1 of 7, by RETROKOMODO

User metadata
Rank Member
Rank
Member

There are plenty more games I could include that have benchmarks like Metro 2033, but they all require Steam to be installed, and this being an XP machine I'm not putting it online and messing with workarounds. It might still be possible to do.. but I don't think I want to go down that particular rabbit hole at this time!

Reply 3 of 7, by Imperious

User metadata
Rank Oldbie
Rank
Oldbie

Radeon 9600 and 9800 does not support shader model 3 which 3dmark06 uses I think for the Deep Freeze benchmark. Probably will just result in a lower benchmark number hopefully.

Alternatively scrap using 3dmark06. The cpu benchmark is incredibly painful to sit through time after time.

Atari 2600, TI994a, Vic20, c64, ZX Spectrum 128, Amstrad CPC464, Atari 65XE, Commodore Plus/4, Amiga 500
PC's from XT 8088, 486, Pentium MMX, K6, Athlon, P3, P4, 775, to current Ryzen 5600x.

Reply 4 of 7, by Putas

User metadata
Rank Oldbie
Rank
Oldbie

My main gripe with benchmarks is how repetitive the test selection is. Don't use things like 3dmark unless you have a specific purpose for that. Such results are easy to find.
Among games (and game engines!) try to find those not benchmarked. Of course, it is better if the game is renowned, but for some reason wasn't tested.
Then genres. Note how with exception of World in conflict you have only FPS? Not representative.

Reply 5 of 7, by RETROKOMODO

User metadata
Rank Member
Rank
Member
Putas wrote on 2022-10-13, 03:47:

My main gripe with benchmarks is how repetitive the test selection is. Don't use things like 3dmark unless you have a specific purpose for that. Such results are easy to find.
Among games (and game engines!) try to find those not benchmarked. Of course, it is better if the game is renowned, but for some reason wasn't tested.
Then genres. Note how with exception of World in conflict you have only FPS? Not representative.

True, Dawn of War 2 has a benchmark. but like many other games - it requires Steam to be installed first which disqualifies it. I'll admit I only really play FPS or RTS!

Reply 6 of 7, by RandomStranger

User metadata
Rank Oldbie
Rank
Oldbie

I'd drop 3DMark, Unigine and the likes. I think the scores those give are a lot less comparable than real game benchmarking. I'd include more of those. Even if they are popular and frequently used. I don't think it's useless to have some results from a large number of cards in the same system, under the same circumstances.

I'd add Unreal Tournament 2004. The game has a tool named Umark which makes it easy to set up benchmarks.
Splinter Cell: Chaos Theory also has a timedemo you can benchmark with.
Supreme Commander also has an easy to set up benchmark. https://archive.techarp.com/showarticle5aa3.h … rtno=493&pgno=1

You also don't have any racing games. Framerate is very important with those for precise and smooth control. I'd consider Need for Speed: Underground 1 or 2 and Colin McRae Rally: Dirt or Race Driver: Grid (you can find a GoG version for the latter if you look at the right places).

Maybe you could add Oblivion or New Vegas too, though I find them difficult to benchmark. The gamebryo engine is really poor at managing frame time.

sreq.png retrogamer-s.png

Reply 7 of 7, by RETROKOMODO

User metadata
Rank Member
Rank
Member

Morning all,

Currently in is the 9800 Pro as my slowest card - so re-running benchies based on it's capabilities. So far I have dropped all tests down to 1024x768 to give the lower cards a chance and the faster ones opportunity to shine. It's also more representative of 'the age'. Splinter Cell: Chaos Theory and Colin McRae Dirt are on the way. I have UT2004 somewhere.. SM3 tests now removed from lineup. Dirt 2 demo fails to load so can't use that.

Any other DX9 benchmarks/games with benchmarks suggestions welcomed!

Dead cards (so far) that will no longer be in the lineup are:

AGP 9600xt - Fan spins up but no signal and no POST
AGP 7300 GT - Fan spins up & no signal but windows loads. Reference can be found to it in DDU next time in with a working card.
AGP x1950 pro - Fan spins up but no signal and no POST

A bit bummed about that, but I guess that's the chance we take buying old hardware as untested 😉