Comparing 3DMark06 scores on currently available GFX cards

Discussion in 'Benchmark Mayhem' started by SickBoy254698, Jun 18, 2009.

  1. SickBoy254698

    SickBoy254698 Master Guru

    Messages:
    662
    Likes Received:
    0
    GPU:
    ASUS GTX 780 OC
    I wasn't going to make a whole thread about this but I changed my mind after I realized how long my post was gettting. I've seen a lot of posts here at Guru3D and elswhere online comparing 3DMark06 scores. I've even been flamed here at Guru3D for discussing the issue of 3DMark scores on certain cards and how (if at all) they relate to actual game performance.

    This issue has irritated me since the day my GTX 285 arrived in the mail. I bought my 285 to replace my 8800GTS 512MB. Of course the first thing I did was fire up 3DMark06 and Vantage to see what my new scores would be. I honestly can't remember what my 8800 scored in Vantage but I think the 285 was about 4000 points higher. Much to my surprise my score was almost the same in 3DMark06. Which brings me to my irritation.

    IMO 3DMark06 has become completely unreliable in it's scores. Despite this it is still widely referenced when comparing new GFX cards. In the past six months I've run 3DMark on a 4870 512MB, my 8800GTS 512MB, and of course my GTX 285.

    While not exactly a huge upgrade from an 8800GTS 512MB the 4870 scores just under 13,000 on the VGA charts. (I would reference my own tests but I hated the 4870 so much I didn't save any.) The lowest score I ever got with my 8800 was 13,596. The highest score I got after OC'ing my 8800 was 15,232. Since the 4870 didn't OC well it got destroyed despite the fact that we all know it's about 10-15% faster than an 8800GTS 512MB. After those tests I didn't really think much about it until I ran 3DMark06 on my GTX 285.

    My scores with the 285 demonstrate what I mean by 3DMark06 scores no longer being reliable. The highest I have been able to get with my 285 is 16,531. Despite the fact that when I tested my 8800 my CPU was clocked at 3.0Ghz and with the tests on the 285 it's clocked at 3.2Ghz. Anywho...

    While I understand why the GTX 285 scores are only 8% higher than my scores with the 8800, it's obvious from what I read in this forum and elswhere that the vast majority of people do not. There are many reasons why the scores come out this way. Without going into the issues too much the main problem lies with the default settings of 3DMark06 which almost everyone uses because you get the best score that way and then you can brag to all your nerdy friends. :nerd:

    Unfortunately the majority of us do not play games @ 1280x1024 resolution with AA and AF turned off. Since a card like the 285 doesn't really begin to shine until you run games with AFx16 and AA 8xQ the default settings of 3DMark06 make it look like a huge waste of money. That being said anything over AAx4 would make my 8800 slow down signifigantly while it takes a setting of 16xQ with Supersampling Transparency AA to slow down my 285.

    While I firmly believe that Futuremark Benchmarks are a great way to get an idea of how a card will perform I think they are relied on a lot more than they should be. This is especially true when comparing newer cards with older GPU's and GPU's with more than 512MB of memory. I just think we need to shift the conversation away from benchmarking programs and focus more on real world performance a.k.a in game FPS.
     
  2. Psytek

    Psytek Ancient Guru

    Messages:
    3,367
    Likes Received:
    0
    GPU:
    2x 260 GTX 216 SLI
    Yeah, I barely use it anymore... every time I see the results after upgrading or overclocking, all I can think is WTF future? ... how hard is it to count how many frames and tell me which cards are faster?
     
  3. benne

    benne Member

    Messages:
    20
    Likes Received:
    0
    GPU:
    evga 275 sli
    For 3d 06 has been completely unreliable in it's scores since it start to compare cards.
    I had a 4850 at same cpu clocks as my 88gt it was slower but a better gamer.
    I think this is just for the benchmark hobbiest now

    All thuo,I wish more guys with rigs like yours would run more benchmarks of all types to help the guru3d benchmark team.
     
  4. MisterWhippy

    MisterWhippy Guest

    Messages:
    806
    Likes Received:
    9
    GPU:
    Asus GTX 960
    QFT

    The base settings that "get the best score" are available in the demo version. Applying any settings to the program require purchase of it. This is probably the main reason that its set this way.
     

  5. MM10X

    MM10X Guest

    Messages:
    4,240
    Likes Received:
    1
    GPU:
    3080 FE
    its the CPU limitation in 06, vantage is a much better benchmarking tool for real performance.
     
  6. ricardonuno1980

    ricardonuno1980 Banned

    Messages:
    4,407
    Likes Received:
    0
    GPU:
    GTX 780Ti Classified :D
    ok :) but intel is very bad for single core (graphic tests 3dmark01~06 and older games) only since pentium D because single-core doesn't increase yet from 3.8 GHz without OC. it's bad!! :confused: :( I think the solution is 10ghz w/o OC for elimine cpu limitation. ;)

    next cpu's "simulation":
    3.8ghz - 2004
    10 ghz - 2007
    25 ghz - now :)
     
    Last edited: Jun 29, 2009
  7. SickBoy254698

    SickBoy254698 Master Guru

    Messages:
    662
    Likes Received:
    0
    GPU:
    ASUS GTX 780 OC
    I agree that Vantage is definitely a better all around benchmark. With the release of Windows 7 this fall I imagine 3Dmark06 scores will no longer be a topic of conversation when comparing gfx cards. Not to mention that ATI and Nvidia will be releasing DX11 cards by years end as well. I'm sure Futuremark has already prepared a new benchmarking program to be released shortly after the new DX11 cards hit the shelves.

    Anywho I stated above that Vantage is definitely better but I disagree with your reason why. I think for gfx cards it definitely hits the nail on the head and gives you an easy way to compare performance. I double checked just now and my 285 averages a GPU score of around 12,000 in Vantage. Using the VGA charts for a comparison we see the 8800GTS 512 scoring just under 5,000. In contrast to the results from 06 the 285's score reflects the actual performance difference much better with a score 240% higher than the 8800.

    Ideally any benchmarks scores should accurately reflect the actual difference in performance in it's numerical computation. Vantage does this beautifully with gfx cards and IMO horribly with CPU's. The only i7 benchmark reference in my version of Everest is a 965 @ 3.33Ghz. It's scores are about 50% faster thanmy Q6600 @ 3.0 and about 40-45% faster then my Q6600 at 3.2. I found some more realistic benchmarks online that show the 965 is actually about 200% faster than my Q6600 which is all well and good I expected that. Than what the hell is your point Sickboy we don't have all day?

    My point is that Vantage scores for my Q6600 @ 3.0, with PPU disabled, are just under 11,000. While scores for i7-965 and 975's are an astronomical 30,000+. Over 3x higher in most cases while actual performance gains are only about 2x higher. Not exactly an accurate reflection of the true difference in performance and I think (judging by the posts in the forum) this is confusing to non geeks who don't understand the role the CPU plays in gaming or in there daily PC use for that matter.

    While I understand the difficulty faced by the dev's at Futuremark when it comes to creating an accurate scoring system I would just like to see a little more accuracy the next time around. I also think PPU should disabled by default not because of innacurate scoring just because it's not an industry standard so it makes no sense to test it by default. One last thing...

    For the love of God please let Vantage's succesor have shorter loading times than Vantage. I can't freakin stand running Vantage because it's takes so long to load the first two tests.:bang:
     
    Last edited: Jul 1, 2009
  8. Grozer

    Grozer New Member

    Messages:
    1
    Likes Received:
    0
    GPU:
    Asus GTX260 Matrix
    Same thing here. Used to get around 15000 with my old 9600GT Top.
    When i got a GTX260 Matrix i couldn`t believe the 14000 score. Really
    bugged me and i couldn`t figure it out.
    So thanks for picking it up!
    Cheers
     

Share This Page