3DMark Rules?

[Bravo]

I've got a question about several things.

1. Changing Graphic card clock speeds between tests, is this allowed/frowned upon?

2. LOD, lowering the quality detail allowed?

3. Does running the first 4 tests to gain a result really count as a result?

[calantak]

answer to
1: you can do that, lots of people do this
2: everybody does that...so that's okay
3.yep, coz that's the only tests that calculate the result.... pasting the others on it don't mean a thing cuz' the score won't change and you can downgrade your gpu OC to run them and still get the samer endresult, so I guess that's okay.


however... I must say I agree with the fact that it is bending the score a bit.

I would be quite happy if next time around the guys from Madonion release a benchmark they include a specific BENCHRUN for the Orb in which you
1.cannot pause between tests if you want to sbmit the result
2.they include a filter for lodbais, for prerender limits and all that stuff, so that the benchmark itself forces the normal settings

at that time we would have a clearer view of the results... and we would need the rig to be more stable than some rigs now to be able to run it...
plus it would give a more accurate view of the true power of a setup

just my very humble opinion

[DDTUNG]

I agree with you 100%. Right now the line between what is allowed and what is not is really not clear.

I used to always run my benchmark straight through, even though I skipped the non-essential tests because they were actually not stressing the system.

DDTUNG:cool:

[calantak]

maybe we should work out a must-have list for their next effort :)

wouldn't it be possible to send a mail to madonion HQ, just clarifying this?

it should be fairly simple for them to block the test per test runs in a future effort (making it clear that you cannot submit a result if it hasn't been done in one draw)

if you have to run it all in one streak, it'll stress more and eliminate 40% of all 'tweaks'

I know lots of us (including myself in the past) have done/do test per test tweaks, and I am in NO WAY critisizing anyone for doing it, we all have one or two dark and grim tweaks up our sleeve :)

But it would eliminate the fact that some people who know the tweaks and have an identical system compared to other people get 600 points more, just because they read more and not because their system is faster.

3dmark is all about
1. the extreme speed
2. and the stability at this speed

and that would be more adequately (if that word even exists ;) ) measured if we couldn't stop it between tests for a benchmark run...

[Adam Harvey]

So everyone lowers the quality detail? Is that how people get 14k? I am stuck at 9888 and I just want to get over 10k and I will be happy for now. I have a 1ghz Tbird at 1.4, and a GF4 4400 @ 320/635, Mushkin Level 2 2100 DDR. I have max settings on everything memory wise in the bios. Please give me some help please.

Adam

[calantak]

Originally posted by Adam Harvey
So everyone lowers the quality detail? Is that how people get 14k? I am stuck at 9888 and I just want to get over 10k and I will be happy for now. I have a 1ghz Tbird at 1.4, and a GF4 4400 @ 320/635, Mushkin Level 2 2100 DDR. I have max settings on everything memory wise in the bios. Please give me some help please.

Adam
no this is NOT how people get 14000...
people get 14000 having amd's over 2 ghz, or p4's at 3 ghz, extremely overclocked ti4600's...


you get an increase of max 400 points with decreased quality(lod bias and some other settings, like fog table and such)

[MrIcee]

Everyone makes such a fuss over LOD bias settings...to be quite honest..the difference in scoring is minimal with that tweak alone.
There are other tweaks in the drivers such as Anisotropy that will yield much higher results than the much over-ballyhooed LOD bias stuff. If you are clever with the good old NVidia drivers alone,without a 3rd party program you can increase your scores.:)

Randi:D

[majormav]

Originally posted by MrIcee


There are other tweaks in the drivers such as Anisotropy that will yield much higher results than the much over-ballyhooed LOD bias stuff. If you are clever with the good old NVidia drivers alone,without a 3rd party program you can increase your scores.:)

Randi:D how

[calantak]

Originally posted by majormav
how yes, how? :p

[DaGooch]

I would presume editing your registry if you knew the correct command and value.

Examples are:

Enabling coolbits in the registry is

[HKEY_LOCAL_MACHINE\SOFTWARE\NVIDIACorporation\Glob al\NvTweak]
"CoolBits"=dword:00000003

Enabling VIA agp 4X
[HKEY_LOCAL_MACHINE\Software\NVIDIACorporation\Glob al\System]
"EnableVia4X"=dword:00000001

I personally have not spent too much time but you can figure it out by looking at your registry.

***Caution*** If you do not know how to edit your registry do not even attempt the above! I am not responsible if you hose your Operating System!

[Marci]

Majormav: don't worry, they are tweaks that RivaTuner deals with under a different setting... I can put them on reg style for ya if ya want as I have them as mergeable .reg files. They were old tricks from the Geforce 256DDR days!

The main point for unifying the 3DMark results is down to the SYstemBuilders. More and more systems are being sold for gaming based on their 3DMark scores, and I know builders that have generated an overclocked result to sell the machine influencing the customer to beleive that that is the default result. When returning the system as it scored nowhere near as high as he claimed he just techno-baffled them and they were stuck with a what equated to a low end machine. People do abuse the service MadOnion provides. Perhaps they should create a separate section of the ORB for the Teams, where using our techniques are allowed and accepted as the norm...

All the UK PC Magazines bench all the PCs they review on Madonion products, and I've seen many specsheets that would in no way produce the resulting score they've claimed...

For the newbie PC Buyer or family buying for the kids etc (ya know, the ones that go out and buy a HP or a Compaq), a strictly default bench is what they wanna see... when we build our PCs under the MicroChoice brand locally, we provide 3dMark2001 on all systems and perform a full default bench, no tweaking.... our customers are always happy with the results they get as they are nearly always higher just from their own tweaking.... gives them a feeling of self-satisfaction.... the driving force for all of us...

*rant ends, Marci passes the J*

[Marci]

Oh yeah, and just FYI... my second to last published 3DMark score (http://service.madonion.com/compare?2k1=3668930) was done on a straight run thru the game tests without pauses between tests. My current score (click compare link under my sig) could only be achieved by pausing between tests

[Maranello]

i just started taking the 3dmark thing serioulsy not too long ago.... had no clue at all tho the amount of stuff people do...lol (still really dont).

the anisotropy set to level 0/use nearest point, antialiasing off and vsync off were the only "tweaks" i made to my gainward gf3 ti200 to get my 10k..... the rest was 260/590 (core/mem) and 201fsb...lol.

http://service.madonion.com/compare?2k1=3253455

[Bulldog]

What I would like to see is a actual comparsion useing a clocked cpu +gpu with and without the 2 big cheat programs and to really see how much a difference their is My feeling are 3d01 is a run what you brung bench, nothing should be out of bounds...or they should have different classes...uber clockers,gamers, oem or something on those lines...then they could filter lower classes to death and leave the masters alone.....