3Dmark2001 running orders and ATI

[BeardyMan]

I did some testing today and learned something more about 3Dmark2001 and running orders. I can say that running order = ~75% important but running it multiples times with reboot is as well. There is also no such thing as best running order, i tested Hipro's,Oppainters and Buttertoast (ocx) running order and they came not near my best run.

Setup:
ASUS P5K-E
E6850 @ stock 5-5-5-15
X1800XT 512MB cats 7.10 NO LOD

The winner is obvious DH-LL-DL-LH-N-CL-CH
You can see that car high gets the most points when running first.
You can see that nature get's untouched regardless it's place, yes this is ATI, nvidia shows otherwise.
You can see that almost every test that gets puts in front will have the boost. Minor boost wich at the end even doesn't matter.

My conclusion, running order comes in handy? yes only when brake into a new milestone.

[kiwi]

You know that there are 7!=5040 test order variations :)

Now if someone would do that :D

[BeardyMan]

You know that there are 7!=5040 test order variations :)

Now if someone would do that :D

an they will all produce the same end result scaling between X and Y .

[massman]

Interesting, though, I believe the run order will be influenced by the lods you use :)

[BeardyMan]

Interesting, though, I believe the run order will be influenced by the lods you use :)

could be yes, so far i tested nearly 16 possibilties lol :)

strange how nvidia and the 8series get such massive boost when running nature first.

[elmor]

How about LH-LL-DH-DL-CH-N-CL? ;)

[massman]

could be yes, so far i tested nearly 16 possibilties lol :)

strange how nvidia and the 8series get such massive boost when running nature first.

Only 5026 left to test :D

[BeardyMan]

How about LH-LL-DH-DL-CH-N-CL? ;)
I believe i have tested that one to, only the lows first and then the highs, the output was average compared to others. But i will give it a shot.

[eva2000]

nice.. wondering if you do try my run order to see how they stack up against the rest

my run order = lazy ass approach

1. DL + DH
2. LL + LH
3. CL + CH
4. N

then if need be rerun CL until get a boost in numbers

[BeardyMan]

Elmors run can as said be considerd as normal to good run order, eva's run is erhm..:D

[eva2000]

woah huge difference there... guess i better get off my lazy ass then LOL

thanks for the info :D

[metro.cl]

Nice testing Beardy, maybe you could do something similar with NVIDIA?

This is what i run on ATI:
CH-DH-LL-LH-DL-NT-CL

[G H Z]

All this info is invalid unless you run each test from a new boot ;)

[BeardyMan]

All this info is invalid unless you run each test from a new boot ;)

then i have some sad news for you it's valid.

Ofcourse i rebooted after every run, i spend nearly 2hours on this :rolleyes:

[G H Z]

Well why didn't you just say so in the first post? :lol:

[BeardyMan]

Well why didn't you just say so in the first post? :lol:

it's stated in my first post LOL :ROTF:

[elmor]

I tested your runorder BeardyMan and mine was better for my setup. Running 2x2900XT on X38.

[massman]

Because of the different bottlenecks, I presume

[G H Z]

it's stated in my first post LOL :ROTF:

Not really but thats OK. One more thing, you should throw out CL for a better comparison. This is a good thread, don't mistake my comments as a slam on what you've done here :up:

[BeardyMan]

I know, just kidding, anyways

I have found something interesthing :D

It's my secret but i will share it for 50% anyways.

There is a test that can be used to sort of insulate the next test from the previous, it's like running that test for the first time. It gives a nice advantage.
In practise it would give this

GT1 as the test to be used.

GT1-GT2-GT1-GT3-GT1-GT4-GT1-GT5
and so on...

Try it :D ;)

[massman]

I like the idea, Beardy. We all assumed that the ideal run order consisted of 7 tests where it could be even longer!

[GazC]

The problem is, Car Low will always throw a spanner in the works, that is why I always bench it last and multiple times.