NVIDIA 3-Way SLI and AMD Tri-Fire Redux @ [H]

_CH_Skyline_ said:
I wouldn't mind comparing some numbers with you, if you like.
Click to expand...

Sure loved to! these are my BC2 numbers:

Min
54

Max
121

Avg
89.2

I'm having issues with Warhead and Metro 2033 with the current 270.61 drivers, not sure what's up with that, those drivers have caused me a LOT of headaches. I don't have DA 2 or F1 2010. I have a pretty good inventory so throw out some ideas.

Thanks!
 
my setup is in portrait...not sure how that might affect things...was going to compare some #'s as well
 
heatlesssun said:
I'm having issues with Warhead and Metro 2033 with the current 270.61 drivers, not sure what's up with that, those drivers have caused me a LOT of headaches. I don't have DA 2 or F1 2010.
Click to expand...

The 270.61 drivers did one thing for me - let me actually play DA2 but it completly borked BFBC2 surround. So unhappy with these drivers - I'm tempted to buy a 6990 and sell the SLI 580's.
 
JayteeBates said:
The 270.61 drivers did one thing for me - let me actually play DA2 but it completly borked BFBC2 surround. So unhappy with these drivers - I'm tempted to buy a 6990 and sell the SLI 580's.
Click to expand...

I might be able to help yo with BC2. Are you having an issue with the in game main menu by any chance?

I do agree that the 270s are pretty rough, the worst nVidia driver release I've seen in some time, probably the worst since the first Vista drivers but as far as drivers go AMD's history isn't any better. The only thing that's really peeved me is that the 270s have borked 3D Blu Ray playback, they only play in 2D mode and unfortunately AMD is really a proven option for this type of thing yet.
 
heatlesssun said:
Sure loved to! these are my BC2 numbers:

Min
54

Max
121

Avg
89.2

I'm having issues with Warhead and Metro 2033 with the current 270.61 drivers, not sure what's up with that, those drivers have caused me a LOT of headaches. I don't have DA 2 or F1 2010. I have a pretty good inventory so throw out some ideas.

Thanks!
Click to expand...

heatlesssun said:
In the SP game, from the beginning for about a 15 minute run.
Click to expand...

Here are my numbers from the very first 15 minutes as well:

Min: 54

Avg: 115.369

Max: 199

i5-2500k @ 4.9Ghz and GTX580 Tri-SLi @ 850 / 1770 / 2004
Bezel-corrected Resolution is 5960x1200 and I'm using the Beta 270.51's.
 
_CH_Skyline_ said:
Here are my numbers from the very first 15 minutes as well:

Min: 54

Avg: 115.369

Max: 199

i5-2500k @ 4.9Ghz and GTX580 Tri-SLi @ 850 / 1770 / 2004
Click to expand...

Is that running everything at max in game HBAO on, 4xAA and 16xAF? Those numbers are substantially faster than [H]'s running a similar setup to yours.
 
heatlesssun said:
Is that running everything at max in game HBAO on, 4xAA and 16xAF? Those numbers are substantially faster than [H]'s running a similar setup to yours.
Click to expand...

Forgot about the 4xAA, all else is maxed. I'll do another.


EDIT: Yes, 4xAA makes a pretty big difference! The Max is from part of a cutscene, not game play, fyi.

BFBC2 4xAA Maxed 5960x1200:
Min: 44

Avg: 99.110

Max: 197

Also, here's 10-15 minutes of Crysis Warhead (Call Me Ishmael) All Enthusiast 16AF, 0xAA 5960x1200
Min: 25

Avg: 65.607

Max: 96

Again, my cards are overclocked and my board allows for x8, x16, x16 speeds and you must remember that I have no idea where (in the game) they did their review material compiling. These are just off the cuff.
I'd be happy to compare any game that I have in the exact same areas for more precise comparison results. :)
 
Last edited:
These tests by you 2 guys are showing that CPU frequency plays a HUGE role on nvidia, I wonder if the x4 in the X58 platform even played a role in the first run through at all. I'm starting to lean away from it little by little. BTW any of you guys play with the threaded optimization option in the driver, e.g.

auto is default, Can you guys run through the 15 minutes again in BC2 and Warhead with it set to off and on and see if there is a difference. It would be appreciated.
 
Lord_Exodia said:
These tests by you 2 guys are showing that CPU frequency plays a HUGE role on nvidia, I wonder if the x4 in the X58 platform even played a role in the first run through at all. I'm starting to lean away from it little by little. BTW any of you guys play with the threaded optimization option in the driver, e.g.

auto is default, Can you guys run through the 15 minutes again in BC2 and Warhead with it set to off and on and see if there is a difference. It would be appreciated.
Click to expand...

Sure, I'll do a few runs when I get home. I have it set to the default Auto in the NCP.


EDIT: Here is BFBC2 4xAA Maxed with Threaded Optimization set to OFF...

Min: 52

Avg: 98.764

Max: 197

And here is Crysis Warhead (Call Me Ishmael) 16AF 0xAA Enthusiast, also with Threaded Optimization set to OFF...

Min: 30

Avg: 63.592

Max: 106
 
Last edited:
Here's an interesting answer I got regarding that review directly from Asus:

''nVidia have some front end features in the NF200 that can be utilized for more efficient peer to peer writes across the PCIe bus (pw short perhaps??) during sli~tri~quad configurations. So it really comes down to what kind of optimisation AMD have in their drivers (hardware limitations withstanding). Given that the AMD cards have no way of utilizing any special features of the NF200 (and this can be down to hardware as well as software), as a result of this I’d have thought that the increased number of native PCIe lanes on the X58 platform is better suited to AMD multi GPU setups, too. ''


So the NF200 was really the problem here. I think it's clear. Directly from the makers of the motherboard: ''Given that the AMD cards have no way of utilizing any special features of the NF200 (and this can be down to hardware as well as software''. So what are those ''special features''' of the NF200 chip available ONLY to Nvidia GPUs? Typical Nvidia behavior.
 
Last edited:
Are you going to run the tests with Threaded Optimization ON, CH Skyline? just wondering since I would really like to see them :)
 
_CH_Skyline_ said:
Sure, I'll do a few runs when I get home. I have it set to the default Auto in the NCP.
Click to expand...

So on auto default the fps (especially minimum) are slightly higher than with threaded optimization off in BFBC2 but the exact opposite. Strange results. Try running with them ON even though technically auto should do the same but just to arouse curiosity it'll be cool to see all 3 numbers side by side.

However acknowledging the normal margin of error it seems that threaded optimization plays a huge role in very multithreaded games like bfbc2 but not so much in alot of the other games used in this article. It seems more and more that the frequency is to blame but I guess there is no way to know for sure how much running that 3rd card at 4x plays a role unless brent justice set the 2600k back to stock settings and ran with identical cpu frequency and settings to the initial x58 setup to see if there is a notable difference beyond the usual margin of error.
 
heatlesssun said:
Nothing that wasn't already mentioned.
Click to expand...

I've been following the forums for quite a while but I had to register just to say that I completely agree with this. All tests are amazingly tailored made for nVidia cards if the GPU isn't capable of certain settings H should just post it was incapable of running. very biased indeed!
 
Been a couple weeks since this thread has had any action, but my 3rd 3GB GTX 580 will here on Thursday so I'll rerun the benchies that I previously ran and post some results.
 
armartins said:
I'm looking forward seing those results too... I can't find anywhere info about how far unlocked 6950s can get, the maximum non bleeding point clocks.
Click to expand...

those should be interesting to see,the 6950's are not too far off 6970's and with the current scaling im sure a 3 way 6950 would be just ahead sli 580's
 
We experienced a 30% performance improvement on 3-Way SLI in this game, but a negative impact with Tri-Fire for some reason. This was the only game to perform this worse for us though.
Certainly we are hoping from feedback on AMD on this.
Our GUESS is that we have either some coding in the CAP that is not "Sandy Bridge friendly" or perhaps something in the game's code.
Hopefully we will find out.
Click to expand...

And Kyle / Brent, did you got any feedback from AMD, what the problem was F1 2010?
 
You must log in or register to reply here.
Back
Top