Tri 260 vs Dual 280... Again

[J hendy]

Hey folks, brief question here, in an SLI setup, which would be faster, three XFX Black Edition 260GTX's (666MHZ Core Clock, 2300 Memory Clock) or two BFG 280GTX OCX's (665MHz Core Clock, 2400 Memory Clock)

Conventional wisdom tells me that the TRI SLI setup would provide accelerated performance, at the cost of thinner driver support, primarily due to the fact that three 260's would have more combined memory than two 280's, i haven't ever really had a chance to familiarize myself with SLI as i've always used single card configs, so i'd appreciate some advice.

Thanks alot.

 

[L1qu1d]

how about just 1 4870 X2 it would cost less than all those setups and still pack about the same power, maybe 20-50% slower (depending on the game, sometimes it actually matches/surpases the 280 GTX SLI)...but again like 200$ cheaper than tri 260s, and 400$ cheaper than 2 280 GTXs.

 

[invisik]

tri sli is just overkill and it scales poorly. the difference between 2 gtx280 and 2gtx260 seems like only 10-15% difference. not worth the extra cost for a slight boost. best price/performance ratio would be 2 gtx260 or 4870x2.

 

[yonef]

Check this out: http://www.tomshardware.com/reviews/Blackbird-Falcon-Northwest-Vigor,2027.html

2x HD4870X2 is much faster than 3-SLI GTX280 (even with 2x CPU)
2x HD4870X2 system ~ $6000 while 3-SLI GTX280 ~$8000

 

[L1qu1d]

benchmarks have proven that the 280 GTX tri is still a tad over the 4870 X2 quad, since the quad doesn't scale properly yet.

It should follow the logics, 1 4870 X2 is 30% faster than the 280 GTX

4870 X2 quad is 30% faster than the 280 GTX sli

now the extra 280 GTX to make try only needs to cover 30% to make up for the loss, which shouldn't be all that hard.

either way the 4870 X2s would be a much better price but 4870 CrossX will not outperform the 280 GTx tri, just yet.

We'll have to see what the future holds. If that happens I'll just grab another 4870 X2 and do Quad fire

 

[yonef]

...I've corrected the link from my previous post. I've been posted the correct article now.

@L1qu1d: look again at the gpraphs:

Do not forget:
green bar: HD4870X2 QUAD XFire single CPU quadcore running @4.0Ghz
red bar: 3-SLI GTX280, 2X CPU quadcore runnign @4.0Ghz BOTH! (Two processors)

 

[L1qu1d]

haven't looked and trusted THG charts in a long time. How about you look at the cards tested on the exact same hardware??

Using Skulltrail

http://www.driverheaven.net/reviews.php?reviewid=609

There you are.

Just because a comp has 2 physical CPUs doesn't mean games can use it Remember how well SLI worked when games needed to be optimized first? 15 fps...

 

[J hendy]

Thanks for the replies, i'm not ruling out 2x X2's at all, but at current i'm just trying to figure out whether to go for three 260s or two 280s, once i've decided which setep performs the best i'll look into what ATI offers.

I'll be building this system in about a month and a half to coincide with the release of Nehalem.

 

[yonef]

haven't looked and trusted THG charts in a long time. How about you look at the cards tested on the exact same hardware??
Using Skulltrail
http://www.driverheaven.net/reviews.php?reviewid=609
There you are.

:lol: :lol: This looks completely fake :lol:
I don't know how they made it up, but lets start with something:
1) Drivers: "Catalyst 8.8.4 beta for X2" looks like driverheaven made this drivers by themselves.
2) They use reference radeons while on the other hand we have fully OC 3x GTX280 ...smell somthing!?
3) I do NOT know how they did this, here is a pic of driverheaven's setup of the radeons(left pic):

(right pic) is a good looking pic of what it should be:woooow, did you notive the difference? because I can smell it; Driverheaven somehow manage to "cut" the bandwith of the X2s; "CUT" the GPU clock to 507Mhz!!!, downclocked memory to 500Mhz ... what else...oh, Pixel fillrate! (looks like they win the downclocking tournament)

What is the lesson here: If you manage to cripple 2x HD4870X2s good enough, then you will get 3x GTX280s a bit faster than Radeons
Oh come on, even TH who are nvida fans didn't manage to fake the results that much.

 

[rangers]

i think one set is for the 3D clocks and one set for 2D

 

[yonef]

i think one set is for the 3D clocks and one set for 2D

No.
This shows current 3D clock for the GPU. AND memory is all the same in 2D and 3D mode it doesn't change freq, unles someone downclocks it. CCC shows current clock of the GPU and while it shows 500Mhz for 2D mode then GPU-Z still shows 750Mhz, if you switch to sensors tab then you'll see current clock and history of the clock change. (everyone with ati card can confirm that)
Plus (suspitio fact) 507 Mhz is not the clock for 2D mode. it is 500Mhz and user can NOT change 2D mode speed unless reflashing bios, which is possible.

P.S. if you look at the Crysis test you'll notice that the max FPS for Quad Xfire HD4870X2 is 59 ! this is very high settings and 19x12 res. Oh go to hell driverheaven :lol: - even single HD4870X2 can exceed 59 FPS and there is people that actually owe this card and can confirm that, isn't it?

 

[L1qu1d]

I see nothing wrong with it, seeing as most other sites show the 4870 X2 at a crippled 39 fps on Crysis. So I still don't see the problem.

Although driverheaven might not be as respectable of a site as lets say anandtech, it still holds some of the best benchmarks I've seen although you showed me 1 site comparing 2 completely different computers, doesn't say anything no matter how much more the other pc was

It would be nice if you could find another site that shows our dilemma, I'm not here to fight with u

 

[J hendy]

yonef, the thread really doesn't have anything to do with ATi, either atempt to answer my damn question or dont bother posting, i'm really tired of having threads like these hyjacked by ATi zealots, which seems to happen almost every time.

Sorry for the rant, but can anyone shine some light on my original dilema?

Performance wise Tri 260GTX vs Dual 280GTX?

Cheers

 

[L1qu1d]

260 GTX tri is better than the dual 280 GTX most of the time, in games that support sli properly eg COD4 Bioshock there are more

 

[0mg_1ts_m3]

dual gtx 280s perform worse than one single gtx 280

 

[L1qu1d]

LOL ^ dude honest why do you post? every benchmark says otherwise...LMAO!!!!!

that was funny it made my day lol....maybe in games that have no SLI support like warhammer LMAO!!!!!

 

[J hendy]

Thanks liquid, that's the first peice of actual advice i've recieved, i'll keep mining information.

 

[0mg_1ts_m3]

i was kidding. was hoping it would make him think he was a bit tarded.

 

[yonef]

All my posts is for just one reason, to get you focus on something else apart form GTX260/280s. I'm trying to tell you that there is better performing cards than theese two, or if you wish - better prize/performance cards. For example: 3x TGX260s ~ $1000 and 2x HD4870X2s ~ $1000 ! I bet that 2x HD4870X2s will perform much much better than 3x TGX260s for about the same prize (in fact will be faster than 3xGTX280).
BUT if you want to stick with GTXs then I have to appologize for the offtopic, and I have to agree with L1qu1d that 3xGTX260 will be better than 2xGTX280.

 

[J hendy]

As i said earlier, i just want to take a look at all my options, i'll be throwing a lot of money at this system so i need to choose the right hardware, if Nvidia really is the wrong choice then i'll go ATi, but not until i'm sure i'll get more performance out of an ATi solution.

And 0mg_1ts_m3, kindly gtfo of this thread you pea-brained douche

 

[L1qu1d]

:lol: :lol: This looks completely fake :lol:
I don't know how they made it up, but lets start with something:
1) Drivers: "Catalyst 8.8.4 beta for X2" looks like driverheaven made this drivers by themselves.
2) They use reference radeons while on the other hand we have fully OC 3x GTX280 ...smell somthing!?
3) I do NOT know how they did this, here is a pic of driverheaven's setup of the radeons(left pic):
http://www.driverheaven.net/reviewimages/skulltrail2804870/pictures/X2/gpuzCFXsmall.jpg http://xtreview.com/images/hd4870z2cf/gpu-z HD 487x2 cf.png
(right pic) is a good looking pic of what it should be:woooow, did you notive the difference? because I can smell it; Driverheaven somehow manage to "cut" the bandwith of the X2s; "CUT" the GPU clock to 507Mhz!!!, downclocked memory to 500Mhz ... what else...oh, Pixel fillrate! (looks like they win the downclocking tournament)

What is the lesson here: If you manage to cripple 2x HD4870X2s good enough, then you will get 3x GTX280s a bit faster than Radeons
Oh come on, even TH who are nvida fans didn't manage to fake the results that much.

BTW if you stop and read the link they give you in the review you will see that 500 mhz was when it was in low power mode, and it clocks up when 3D games start

The reason they used OC cards is because they used top of the food chain at the time for the 280s and top of the food chain for the 4870 X2s.

both 4870 X2 and 280s have received heavy optimization during the month after, knowing THGs style of testing weird combination, I didn't see what drivers they used for either video cards

Also you'll not that octo CPU doesn't help gaming at all seeing as we barely have games that use quad core (age of Conan) and second u'll notice the CPUs are the same clock, the RAM is DDR3 @ 1800 in CF computer and also RAID 0 Western digital VelocioRaptor Hard drives, while the tri is only ddr2 clocked @ 800 and 2 none raid old Raptor harddrives.

Moving on, the vigor uses the Skulltrail board, which has its own drivers and isn't support by Nvidia.

So either test both systems on skulltrail or get 2 boards that have similar performances and then test em

I'm not trying to defend tri-sli I really don't care which is slightly better since they are both about the same price ( I beleive the 280s are slightly cheaper now, 100$ or so because of the 374.99 price tag after MIR)