• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

How do people hit 10K in 3DMark05 with a 7800GTX?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Revivalist

Member
Joined
Mar 27, 2005
Location
Fresno, CA
I see people hitting 10k in 3DMark05 with a single 7800GTX . . . but the best I can do is 8200, and that's with 575MHz GPU (+25MHz delta on the GD clock), 1500MHz memory. . . .

To achieve that I used 1.65V on GPU, 2.4V on memory, and a TEC that keeps load temps at 35C max. . . .

Is there some major setting/adjustment that I'm not aware of? . . . I just don't see how others are able to score so much higher? :eh?:

(By the way, I've done all the basic stuff like turn all graphic settings to performance, turn off all background programs, etc. . .)
 
Dual Core pushed really hard will make all the difference in the world.
 
Revivalist said:
I see people hitting 10k in 3DMark05 with a single 7800GTX . . . but the best I can do is 8200, and that's with 575MHz GPU (+25MHz delta on the GD clock), 1500MHz memory. . . .

To achieve that I used 1.65V on GPU, 2.4V on memory, and a TEC that keeps load temps at 35C max. . . .

Is there some major setting/adjustment that I'm not aware of? . . . I just don't see how others are able to score so much higher? :eh?:

(By the way, I've done all the basic stuff like turn all graphic settings to performance, turn off all background programs, etc. . .)

there is not just one magic thing todo to hit 10k

there is a hole load of things that add up

that said you must first find out why you cannot get 9200 this is the begining
or just into the 9000

i would start with reducing the ram oc to 1449
 
Thanks for the suggestions, guys. . . .

@ ziggo0 - I've actually seen some guys around here with single-core CPUs hitting around 10k. . . . So I think I should at least get close. . . .

@ bazx - I went ahead and tried lowering the memory to 1450MHz but unfortunately I got 8108 this time around. So I guess the problem isn't video corruption due to overheating the memory. (I have some pretty big heatsinks on them.) . . . . I wonder if there's something else I'm missing. . . .

By the way, I saw your old thread with the v-modded, phase-cooled 7800GTX. That was cool! . . . What did you manage to pull out of that one?
 
the last bench in 05 was 10709

but where you are now there is somthing wrong you should be in 9k for sure

i have a few bios you might want to try still
 
Drivers and OS tweeks make all the difference in the world. In addition did you bother checking if your card was actually stable at those speeds? If you push them too hard error-correction kicks in and they start to peform worse than they would 10mhz lower
 
Are you referring to thermal throttling? . . . Ok, I'll try doing a run with the RivaTuner speed monitor up and then check to make sure that the card stayed at the speed I set it for the entire benchmark run.
 
Well DCs are very very useful in 05. As are fairly new drivers. And all of that doesn't help out very much if you don't have a fresh windows install as well as a fresh defrag on your HDs. Finally hop on over to the 3dmark section and check out Guatams tweak thread.
 
@ Sentential - Ok, I just did a run with the RivaTuner monitor up the whole time and the clocks stayed right at the speeds I set them at without moving. So I guess it's not a thermal throttling problem. . . .

@ speed bump - Thanks for the info. . . I actually justed installed Windows on this system and defragged multiple times. So I'm good to go on that. . . If I remember correctly I saw few guys on here who benched with single cores and got up into the 10k level. . . .

For example, here's a link where Krag got 10,565 with 600/1480 on the same card. (He was using a FX-57 at the time before he got a FX-60.) Lower in the thread you can see that crimedog also broke 10k and from his sig it looks like he's running a single-core Intel. I think bazx also got his score on a single-core (please correct me if I'm wrong.) Maybe some of those guys will also chime in. ;)
 
I managed 10.3k bogomarks at default driver settings with 635(595+delta)/1500 settings 81.98 whqls. Obviously 'dirty' run. Don't remember voltage settings. Something is wrong with your setup though because this card will do ~8000 bogomarks on stock clocks/ 2.4 GHz single core a64 with no 'driver tweaking.' On my 24/7 settings I get about ~9500 bogomarks. Which is 554/1470 or 594/1470 depending how you measure clockspeed.

I would make sure your memory scores are in line with others who have similar setups and the same with spi runs. Other than that, having a million processes running while trying to run these programs may incur a slight performance penalty but nothing to that order. Disk fragmentation makes almost no difference in such programs.
 
Thanks for the input, SuperFarStucker. . . . Those numbers you mentioned help a lot. Now I can definitely see that something isn't right. I'm glad I discovered this problem before getting SLI working so I can focus on getting this card up to par first. . . .

I'm going to try the 81.98 whql drivers. (Right now I'm using the 84.21 whql drivers.) . . . As for my system memory, I've run Everest and I get 7,000 read / 2,500 write / 42.5 latency. . . . So I think the memory is doing ok. . .

The CPU is running 3.2GHz which is well above the 2.4GHz you mentioned. . .

I don't know what else it could be. I thought maybe it's video corruption hurting the score but dropping the video memory speed or GPU speed didn't help. It only hurts the score. . . .

By the way, does it matter what size monitor I run? I'm using a 21" widescreen with 1680x1050 native resolution. This causes quite a performance hit when running games. But I figured that the free version of 3DMark is locked to a certain resolution and that they setup the program so that it's not affected by the screen size. . . . Right?
 
I believe the default resolution 3dmark uses is 1024X768... you should be able to look at your monitor's OSD and see what resolution it is currently displaying.. check it in windows, then start a 3dmark bench and see what resolution it reports then...
I personally dont think u should worry about it... 3dmark is a synthetic benchmark, and 10K+ 3dmarks dont mean jack as far as real world performance. If you're happy with how ur games run, then its fast enough - and judging by ur sig : real-world performance should be killer :) Regardless of what some synthetic benching program says ;)
 
What you say is true, DareDevil_747. . . . I was just concerned that perhaps there is a problem with the way I am running my card that is limiting the performance of it somehow. . . .

By the way, I don't think my monitor reports the resolution it's running at in the OSD. (I'll have to check when I get home.) If it doesn't, then could I just take a screenshot during the benchmark and then see what resolution it comes out as? . . .

So is it possible that the size of the monitor is affecting the benchmark score afterall? . . . . I'd be really surprised because I thought they design those benchmarks so that they would be independent of the monitor used. :eh?:
 
i broke 10k with 600(+50)/1470 and a single core. i used my everday install (i was just playing around with my new gpu) but i tweaked it a little. get familiar with services.msc ;) i think i was also using the tweaked bios from the XS thread, but i doubt it made much of a difference. i barely broke 10k btw, it took a few tries.
 
Thanks for the input, crimedog. . . .

Wow, 600 +50 delta?! That's awesome! . . . I haven't been able to pass with more than 570MHz (even with delta set to zero). But I haven't gotten around to seeing how high of a delta I can run. I know that the GD clock can do 630MHz so I might try a delta of 60MHz . . .

About the tweaked bios, is it basically the memory timings that were tweaked? I was thinking about experimenting with that a little. But I'd like to wait until I get a second working card so that I can still post and flash the card back to stock if I end up giving it a bad flash.
 
Last edited:
yeah the modded bioses had tweaked memory timings. the GDC setting shouldn't really make much of a difference in your overclock (it doesnt work like 600+50 = 630+20) so you definitely want to bring it up. reading the massive 7800gtx bios modding thread at xs helped me to understand these cards a lot, i recommend you head over there. also, i've never heard of a bad gpu bios flash, and i've flashed probably 30x to 6800's and 7800's ;)
 
bazx said:
the last bench in 05 was 10709

but where you are now there is somthing wrong you should be in 9k for sure

i have a few bios you might want to try still

I'm only running 2.5GHZ with my 7900GTX, and got over 10K.
 
Back