25k+ in 06 - How do they do it?

Ok, so i was looking through orb's 3Dmark06 results for systems similar to mine and found some very impressive scores. I have an i5 oc'd to 3.8 stable and 5870 oc'd in CCC to 900/1300, my result is right around 25k with all the aa af etc off, vsync off. Although I found nothing with my exact configuration I did see a lot of i7s with 5870s and a few i5s with 4870x2s. The people with the i7s @ 3.8+ are scoring 28k to 36k max, the i5s @ 4.0+ with 4870x2 are around 28k -31k max.

How do they get such good results? I thought my system was functioning pretty well and i'm only looking to make sure things are about right, but going on those results it would seem mine should be doing much better ~ 26 to 27k. Maybe i have a bone GPU...

However comparing review benchmarks i'd say the scores i get are about right and there are some i7s with 5870s performing considerably worse. But how do the big guns get those scores...? What do you guys/girls think? Is it just a question of GPU overclocking?
16 answers Last reply
More about tomshardware
  1. 3dmark 06 is very very cpu dependent, so the i7 people obviously have faster CPUs, thus higher scores, same with 4ghz i5s. Try vantage to get a better idea of ur score.
  2. People do this kind of thing very competitively and it's reasonable to expect to not get "Top 10" results. Some people already have benchmarked i9s on 3DMark and the people with the i7s have them overclocked around 6-7GHz. There's no reason to expect comparable results.
  3. Ok thanks, that makes me feel better. With my CPU back at 3.6 i ran vantage and got these:

    P16390
    CPU: 16351
    GPU: 16403

    Is that respectable?
  4. at 3.8 an i7 doesn touch 30k in 3dmark06 you need least over 4 ghz with a high end gpu

    as you can see the current fisrt ranked uses some extreme methods

    3DMark06
    40257
    AndreYang
    United States
    INTEL Core i9 2.40 GHz @6ghz with ln2
    Radeon HD4890 (QuadCF) @ 950 / 1000
  5. Thanks for your replies guys. I think my system is performing about where it should be. So i feel a bit better.

    Hard facts are that i should have stretched to a 1366 setup with 5870, not only does the i7 920 oc more easily but it also seems to unlease a 5870 (or two) much more effectively (once oc'd a little). The even more annoying thing is, I'm on my second mobo, which basically negated the cost difference of an i7, damn it.
  6. Thats very impressive overclocking. I don't feel too bad now. Thanks for posting those up.
  7. I do 25K on my i7 920 @4GHz and 5870 at stock. 3DMark really isn't a good test for "performance" as each component tends to do better in certain areas and in certain games. But if you are comparing between upgrades and such it's helpful. I know my buddy hates my now that I beat him.
  8. I do around 27-28k with a pair of 4870x2s and a 4.2GHz i7. It's tremendously CPU dependent though - I only score ~17-18k with my CPU at 3.2GHz. Don't worry about it too much, as it isn't that representative of games (which tend to not depend as heavily on the CPU).
  9. Ok, i'll stop worrying and just get on with using it. I've been messing around for days trying to tweak this and that, thinking something isn't quite right. The only good to come out of these sessions: At 3.8 i can now get RAM stable running 1800Mhz instead of 1080. 2T though... A negligable performance increase, but nice enough.

    Thank you all for your replies, much appreciated.
  10. LowRes said:
    Ok, i'll stop worrying and just get on with using it. I've been messing around for days trying to tweak this and that, thinking something isn't quite right. The only good to come out of these sessions: At 3.8 i can now get RAM stable running 1800Mhz instead of 1080. 2T though... A negligable performance increase, but nice enough.

    Thank you all for your replies, much appreciated.


    If you're worried about gaming performance, I'd probably recommend tighter timings (CAS8 or CAS7) rather than a frequency over DDR31600.
  11. OK, I can get it stable at 9-9-9-24-2T @ 1810 1.6V or 6-7-7-18-1T @ 1086 1.6V. So i'll stick to the lower of the two for a while. Thanks for your help guys.
  12. LowRes said:
    OK, I can get it stable at 9-9-9-24-2T @ 1810 1.6V or 6-7-7-18-1T @ 1086 1.6V. So i'll stick to the lower of the two for a while. Thanks for your help guys.


    Well, 8-8-8-24 or so around 1600 1.6V would be about the sweet spot for an i7. I meant frequencies over 1600MHz aren't really worth it, not to use 1086, I would also recommend not using under 1333. :P
  13. Ah i get it 1333-1600. Interesting thats the sweet spot for an i7, because my oc at 3.6 (172x21) 1720 7-8-8-21 performs very close to 3.8 with 9-9-9-24, the LinX Gflops are 54-55 and the difference in Vantage score is small. Probably be around the same sweet spot for an i5? I couldn't get the 2:8 divider to boot with my bclk at 180ish, i'll try again tonight... My CPU seems so sensitive to ram settings, eg I can boot into windows at 3.8 with relaxed timings, run everything fine but fail LinX on first pass. Reboot with one latency change (tighter) and she'll do 50 passes. Took me ages to figure that out, adjusting Vcore with no effect.
  14. 3.8Ghz 1448 8-8-8-19-1T, was a matter of messing around with the MCH strap settings. Vantage score is ~P17-18k. Thanks for your help brockh.
  15. LowRes said:
    3.8Ghz 1448 8-8-8-19-1T, was a matter of messing around with the MCH strap settings. Vantage score is ~P17-18k. Thanks for your help brockh.


    You're very welcome. ;)
Ask a new question

Read More

Graphics Cards Graphics