Benchmarking

Xterminator

I have just finished making some major upgrades to my system, and was wondering what is the best way to benchmark my system (free).
What do you use, and where can i download it from?
Oh yeh and whats your best score?

SickBoy

3DMark2001 & SiSoft Sandra are imo the best way to test a 3d GFX sub system and hdd performance.
3DMark can be gotten from http://gamershq.madonion.com/ and SiSoft Sandra can be gotten from http://www.sisoftware.demon.co.uk/sandra/

JustHalf

Remember to run these benchmarks several times. My hard disk's benchmarks can vary by over 10% depending on when I run it (almost certainly down to fragmentation). Then figure out the best, worst and average scores. Personally, I'd take 5 at least. For a good solid benchmark score, run everything 20 times (for luck)

elexes

20 times thats a lot of wasted time that could well be used for donkey porn research

---

what type of world will we live in when there is no world ?

BattlingCheese

You might want to use 3DMark 2000 for more accurate results as 2001 is optimized for DirectX 8 and graphics cards which can exploit it like the Geforce 3.

For example my 2000 score is 6012 or thereabouts compared to 2987 in 2001.
Thats using the same resolution, buffering etc. in both tests.




[This message has been edited by BattlingCheese (edited 09-08-2001).]

JustHalf

3DMark 2001 is putting the computer through some entirely different tests than 3DMark 2000. You claim that 3DMark 2000 is "more accurate" on the premise that 3DMark 2001 is designed for later cards.

It's a different benchmark! You can't, and shouldn't, compare the scores of the two. It is similar to comparing the 100m sprint to the 5000m... they're both about the same thing, but one is far more grueling than the other; although in this instance, the difference is not that severe.

You may be using the same resolution and buffering in both tests, but 3DMark 2000 isn't trying to do pixelshading.

Carnate

Personally i would just leave it in da gym and let it benchmark itself.. or is that brenchpress...arrrrrrrrghhhhhh!

dammit the drugs dont work!!

JustHalf

Personally, if I wanted to say card X or drive Y performs at a certain level, I'd want to confirm it. Such is the state of benchmark variance that it must be done several times.

However, running a benchmark, something which is designed to stress out the component, repeatedly is a very good way of testing out the stability of your component.