i mentioned it early but i thought i was responeding to another thread. yesterday tomshardware put out some benchies of 9800gx2 in sli verses 3 ultra's. the systems varied a bit the 9800gx2 with a 790i was running on a qx9770 @4ghz and another machine with 3 ultra's on a 780iwith a qx9650 @ 4.17. thg claimed, though they since pulled the article, the clock differance made both systems comparible considering the chipset differance.
4.17 tri ultra set up only scored 22k. as i said in my prior post i am dang close to that with less cpu clock. i think your grasping. on orb it shows not many systems out their beat mine with similar hardware and if they do, its by a nominal amount. Not to say their aren't uber good enthusist's who can squeeze more, but keep in mind the 790i is still "broken" and i am not the only one having some trouble. while the PO-5 bios has helped many with the hdd corruption and the 1600 glitch somewhat...the boards linked and snyc'd at 1600 freeze so u have to drop the fsb/ram to 1599 to avoid a pummle of crashes every five minutes. i an clock my ram lower to keep the FSB stable at 1600 but i get more benefit link'd and snyc'd 1599. so when i say i am at 4 ghz in truth i am at 3.995. Nvidia has stumbled. end of story. not to say they can't find new footing but they need to do so quickly.
Last thing i say on my score for 3dmark...its a synthetic and real world games i do see a great deal of bene's. crysis on very high no filters i get 40fps @1080P...or damn close with fraps depending on the map. when i first had the machine set up i got more like 32fps but between patches and bios updates i squeezed more out. i think the numbers your expecting just don't exsist unless you start water cooling your video cards and over clocking them to hell. as for my gigabyte cards, if i oc the GPU i can pull another 200 marks...but the second i overclock the ram even a little the score is actually less than stock. Had i bought a OC version of the gtx like i have in times past, yes my score should be higher, but i didn't and you get what u pay for. sad as it is yes i paid alot for my grphix cards...but did so over time i started with one in an old fx60 setup and upgraded over the last year. now i get double my score since when compared to my old rig in sli with 2 gtx's. in the end though there is no question you pay out the nose for running tri sli. disporationately so...but i also spent an ass load on my last rig and while it ended with an fx60 it started with an fx52, i had the same ram/mobo for 3 years.
i get that kinda of milage out of this thing by adding grphx (when jusitifible which for nvida cards its not right now or even the near future) and faster CPU's. plus don't get me wrong for my high end system..i have few systems, none quad gpu cause the support just isn't their yet...tri-gpu is great for 1080p though. all scores at default setting 3dmark setting.
q9300 @3ghz with 3870x2 ans 3870 (scores in the 17k range) on a 50" 1080p hdtv
phenom 2.5ghz be again 3870x2 + 3870 (scores in the high 15k range) on a 24" 1200P monitor
5000+ (am2 so 2.6ghz) with two 8800gts (g92) (scoring high 11k 3dmarks) 1280x1024 monitor
...point being it shows my speeds are somewhat in line and proves your point that less money can get you a better price performance ratio...i just happened to be able to afford uber high end to run my "personal" PC and using low/mid to high end machines for laning and servers for folding@home.