Maximizing 3dMark06 Scores

C

corruptblitz

n00b
Joined
Jan 4, 2007
Messages
59
See sig for specs. Last night I flashed the video BIOS to increase the voltage from 1.25 to 1.35 and was able to get the card to run at 640core/740mem from stock 565/715. My original 3dMark06 score at stock CPU/video settings was 5374. After OCing the CPU to 3.2 and the video to these new settings, I'm up to 6234. I really want your help with suggestions to push it just 120 more points that I need to beat my buddy's score.

I'm running Vista, and I know that just switching to XP would probably do it, but I really like Vista. I do switch to Vista Basic theme before benching just in case it makes any difference. I also only load the essential programs at startup such as the nvidia control and a few windows programs.

I don't think OCing the CPU any higher is going to get me much, because I'm running 400x8 and anything higher is going to put me into the 1333 strap. The memory is linked now, and I'm thinking about trying to unlink it, and run it closer to the 1066 it's rated at with 5-5-5-15.

I've thought about trying to change my current PCIE frequency from 101 to 110 or so, but I'm reading very mixed opinions on whether that actually nets any performance gain.

Any ideas?
 
So wait.... this isn't a joke? Wow, even if you don't know it, I think this is gonna draw out the flame community. Let's watch and see if I called it.

EDIT: for the sake of not getting my post deleted, ill give some input. make the inside cooler. I took my computer outside on a 12 deg F day in winter and pushed the clocks higher and BAM!
 
ATIIsBetter said:
So wait.... this isn't a joke? Wow, even if you don't know it, I think this is gonna draw out the flame community. Let's watch and see if I called it.

EDIT: for the sake of not getting my post deleted, ill give some input. make the inside cooler. I took my computer outside on a 12 deg F day in winter and pushed the clocks higher and BAM!
Click to expand...

I don't see what's so funny about the post, but anyways. I think my temps are pretty decent and not the limiting factor of the OC. I have multiple fans blowing at the card and one pulling away. The idle temp of the card is about 44C and the load is around 69C as reported by ATItool. Rivatuner says they're even lower. Aren't these decent temps for a 7950?
 
Uninstall 3dmark06 before you break something.

Compete with your friends in something that actually requires talent and skill. A big 3dmark score doesn't make your penis any bigger.
 
Have you tried just running it again. A variance of 120 points is not unusual.
 
3D mark score doesn't bench your computer. It checks online and compares your PC against others EXACTLY like it. It's a piece of garbage program in my opinion.

I'm running :

PentiumD 915 on Intel G965 motherboard
2gigs DDR2 667
1 x 160gig SATA hdd / 1 x 250gig SATA hdd
EVGA GeForce 8800GTS 320meg
Windows Vista home premium
20" Widescreen Dell E2007 WFP


My 3D Mark is 5799 and my system isn't as hot as yours (except I run a 8800), I also didn't have anything tweaked. Don't trust 3D Mark until they start to bench it against a single system.

So unless you and your friend are running very similar rigs, don't put much faith in 3D Mark
 
have2p said:
Uninstall 3dmark06 before you break something.

Compete with your friends in something that actually requires talent and skill. A big 3dmark score doesn't make your penis any bigger.
Click to expand...

dude dont be like that... i used to overclock just for benchmarks back in the day and it was alot of fun its a good hobby as long as you know when youve reached you max safe limit...which usually means locking up.
 
have2p said:
Uninstall 3dmark06 before you break something.

Compete with your friends in something that actually requires talent and skill. A big 3dmark score doesn't make your penis any bigger.
Click to expand...

I'll break my stuff if I want. And actually, my penis has grown 2 inches now that I score higher in 3dMark06. kthx.

Snooble said:
3D mark score doesn't bench your computer. It checks online and compares your PC against others EXACTLY like it. It's a piece of garbage program in my opinion.

I'm running :

PentiumD 915 on Intel G965 motherboard
2gigs DDR2 667
1 x 160gig SATA hdd / 1 x 250gig SATA hdd
EVGA GeForce 8800GTS 320meg
Windows Vista home premium
20" Widescreen Dell E2007 WFP


My 3D Mark is 5799 and my system isn't as hot as yours (except I run a 8800), I also didn't have anything tweaked. Don't trust 3D Mark until they start to bench it against a single system.

So unless you and your friend are running very similar rigs, don't put much faith in 3D Mark
Click to expand...

3dMark06 certainly does bench your computer, it just ALSO happens to compare it to other systems so you can see that. Most people with any kind of 8800 score much closer to 10k than you are, I would guess that CPU is holding your score down pretty low.

Anyways, I got the points I wanted by increasing the video memory clock to 815. I've learned my lesson that this probably should have been posted in the OC/cooling forum. Thanks for the suggestions all.
 
King_Weaver said:
dude dont be like that... i used to overclock just for benchmarks back in the day and it was alot of fun its a good hobby as long as you know when youve reached you max safe limit...which usually means locking up.
Click to expand...

Damn straight. Ever since the [H] stopped using 3DMark in their benchmarking everyone here on the [H]ardForums thinks it's cool to hate on 3DMark and anyone who dare admit that they've run it.

That's a BULLSHIT attitude and it needs to stop. :mad:

3DMark is a synthetic benchmark and it has value as just that. There isn't a damned thing wrong with some friendly comparing of scores and competition amongst a group of people. It's fun to see how far you can take your rig in that spirit. Hell I remember a time when that's part of what being [H]ard was all about! :p


OP: You're on the right track so there really isn't much we can add. Try and increase the video card o/c a little more and definately try either running your memory unlinked at a higher speed, or leave it linked and try and tighten the timings a little.
 
Say what you want, but in real world benchmarks the C2Duo isn't all that and a bag of chips. Overall it's 15% or so faster than what I have (comparing the PentiumD 920 to the Core2 Duo 6400, couldn't find a Pentium D915 / C2D 6300 to compare).

Check out Tom's Hardware CPU Comparison.

Here's the 3D Mark section : http://www23.tomshardware.com/cpu.html?modelx=33&model1=440&model2=433&chart=174

Check out the Price/Performance index from the drop down also.

3D Mark is a piece of crap that crap sites use to benchmark and solely rely on to make their ep33ns harder. HardOCP, Tom's, Firingsquad, Anandtech and other reputable sites use actual real world games and programs to test and that's where it matters. I've never liked 3D Mark (except for the nice graphic demos) ever since it was shown that certain drivers were being optimized for it so the card manufacturers could put something out that wasn't too great yet make sales on it because it could run 3D Mark.

Say what you want, but fire up 4-5 games and check their FPS and compare them to your friends. You'll get a much better idea of what your rig is capable of in regards to real world applications.
 
3D mark score doesn't bench your computer. It checks online and compares your PC against others EXACTLY like it. It's a piece of garbage program in my opinion.
Click to expand...

Wait... What... :eek:

OP: That's a fairly nice score for a 7950 under Vista. I just don't see what else you can tweak at this point. Turn all the GPU settings down in the control panel?

PCIe frequency has no bearing on score.
 
If you want to beat your friend's score, install XP, or get your CPU up to 3.3 or so... or do both and really beat him.

and :LOL: at all the people bashing 3dmark. It does exactly what its supposed to do, its a synthetic benchmark. If people running a synthetic benchmark on their system actually upsets you perhaps going outside would be a fine, fine idea.
 
Absolutly nothing wrong with 3dmark. I put more faith in 3dmark scores then I would a game I don't even play in a resolution I will never play in!

Bashing 3dmark is about as smart as buying a 850watt PSU...
 
Bashing 3dmark is about as smart as buying a 850watt PSU...
Click to expand...

Well gee, maybe if nobody bought a pair of GTX's and actually needed an 850 watt power supply then that quote might have some validity.

As far as bashing 3D mark, you know it's kinda hard to play 3D mark 2006..... I'm still waiting for them to implement WSAD controls.
 
3D Mark gets more useless as time goes by, and the OP is going to damage his system if he keeps trying to get anywhere with the synthetic PoS benchmarking program. Look around on the internet on people who have ruined their cards, damaged CPUs and voided warranties by OC'ing to get them last few points from 3DMark.

I've already stated that he and his friend should try some real world benchmarks for gaming, encoding and the such. That was my answer. I updated it to show really how useless 3D Mark is in comparison to real world benchmarking.

In case your lazy and can't click on links, that link from right here on [H] shows how skewed 3D Mark is from real world benchmarking. Add in that is has been shown, multiple times, that both AMD/ATi and Nvidia have different optimizations specifically for 3D Mark in their drivers and well, you can see just how mundane it is. Relying on 3D Mark for a PC Performance review is lazy. Pure and simple.

Have fun defending it though, it's the same as defending the Miles Per Gallon rating on 90% of new cars, they are rated in a lab not in the real world.

So unless you're really lazy, haul out some Doom 3, UT 2004, Stalker and time how long it takes to encode a piece of video or audio on a few programs. That will give a real world idea of who's PC is a better performer.

As for 3D Mark? When they start to reprogram it so it's not biased with certain hardware in some rigs, drivers and their performance ratings are done across the board and include OS differences, driver releases and overclocking then it may be a bit more practical. Until then it's a synthetic piece of trash only used by people who don't know any better (who have been tricked into believing 3D Mark is a great benchmarking program that mirrors real world performance) or how to run benchmarks in a real world setting.
 
Let's keep it on topic. The pros and cons of 3DMark as a benchmark aren't what the OP is concerned about
 
3dmark06 is great for measuring your own results from a baseline point of view. All tests can be distorted to some degree by drivers, temporary cooling solutions and whatnot.

As long as your testing methods don't vary and you understand what components and applications are managed by the tests, you cannot say that it a useless program.

As for tip on getting good results, stop all processes/services that aren't needed (AV, network, plugins, 3rd party apps...) and install XP. Balance your RAM, CPU and FSB overclocks as well.

My sig has my results with a small overclock to [email protected] and RAM@1142.
 
I would try your mem @ 1066 and see what happens. Is the core/mem speeds the fastest she will go?

Do you have the latest drivers?

Is V-sync OFF
Everything set to performance in the drivers?
You have the latest chipset drivers?

Try this...

uninstall your vid drivers
reboot
reinstall the drivers
and run it again
 
Hey Fried, I stated that it was useless to use it versus other computers as gaming and encoding from actual games and programs would give a much better perspective.

3D Mark is fine if your tweaking your system, but often won't find the stability issues nor give you time to address them as real world benching will. I was suggesting that the OP use other things aside from 3D Mark so he didn't burn his PC out while competing with a friend. Often 3D Mark won't crash or show artifacts with any regularity if you've gone too high on the O/C.

Take an OpenGL game (Doom 3 / Quake 3) some DirectX 9 games (UT 2004, Oblivion) and a few encoding programs (Photoshop - take large pictures and downsize them, do a load of them, or encode video) and run them a few times. You can find how to run the FPS benchmark tests on a few fansites out there.

Don't forget to apply the exact settings your friend has running on his PC for video settings, VSync locks frames at 60FPS on an LCD (or whatever your refresh rate is) and Triple Buffering will slow down things tremendously.
 
Snooble said:
Hey Fried, I stated that it was useless to use it versus other computers as gaming and encoding from actual games and programs would give a much better perspective.
Click to expand...

Hey, we could be discussing Super PI instead of 3dmark... this is all about the OP ribbing his buddy- which I fully suport :)
 
You must log in or register to reply here.
Back
Top