What's your best bit of advice for the 3dmark enthusistic apprentice?

[zakelwe]

Mine would be:-

Try all permutations

but

Attack one limit at a time so you do not get confused with the data


What would be your advice?

Regards

Andy

[massman]

Don't hope for too much, benching = disappointment most of the time. But when you break WR's you just know that all the work was worth it ;)

[K404]

yea...take your time- test one variable at a time

[BeardyMan]

yea...take your time- test one variable at a time

can't agree more, taking your time for everything. the more you put in it, the more you will most likely get out .

[TheGoat Eater]

can't agree more, taking your time for everything. the more you put in it, the more you will most likely get out .

well said taking your time to get everything squared away usually yields better results than trying to do it fast - even after a few fast benches doesn't compare to a well thought out pass ... at least that is the way it is for me:up:

[cadaveca]

My advice? Get a ramdisk to run your benches from.

[justin_c]

set pcie frequency to 130mhz (thanks hipro for that tip, it made my 20k!)
dont worry about frying your hw by overvolting; painless :D

get snqtools (snq is a member of XS, here is my mirror: members.shaw.ca/justin_c/snqtools.zip ), make sure you take screenshots.

keep ethernet connected so you can upload your score to orb; forgotten many times.

[G H Z]

Look for the guys who's scores are often the best for a given setup, then pick their brains, or you can do it the other way, by clicking 'Benchmark' till your eyes bleed :up:

[1Day]

**Test then test again and then test some more**
That way you will know your system inside out. You can never do to much testing.


Clearly define what it is you are testing. In other words know your objective and stick to it. Otherwise it is too easy to get sidetracked. Say for example you are testing LOD settings do not start changing ram timings halfway through your LOD testing.

Write your findings down - each time you do a test record all the parameters. Basically record your entire bios settings, plus your windows based tweaking applications and of course the results of that particular test. And if air-cooling try and include ambient temperature as this will impact on both CPU performance and GPU performance when testing on the edge. No one want to re-invent the wheel and repeat a sequence of tests and thereby waste time. Plus having a record that you can examine at your leisure often provides clues as to what is the way forward for you think of it as a reference manual.

[massman]

keep ethernet connected so you can upload your score to orb; forgotten many times.

You can always save the 3DMark project and upload the results when you're done benching from another PC ;)

[BenchZowner]

My take on this is:

One small step at a time...cover all the aspects...when ready...find the max CPU benchable...then vga benchable...then combine & go for the real bench.

And as always, the more effort & time you put into it, the better the chances you've got ;)