Hemlock*2: This is Quadfire, baby!

vjcsmoke

Supreme [H]ardness
Joined
Dec 5, 2006
Messages
4,511
Quick GPU-Z screenshot of the 4 5900 series GPUs. =)

gpuz5970Quadfire-1.jpg


As you may note, the PhysX card is out temporarily until I get my Quadfire setup stable. But the 5970s are all loaded up and hooked up to my TPQ-1200 and EVGA Classified. Just need to install some drivers. Then of course I'll have more pictures to show of these puppies installed and hopefully a quick 3dMark Vantage run to test the power of Quad!
 
impressive. i will eventually get a pair of 5870s and a new D0 stepping 920 and hopefully get close to the same results.
 
I'd play a game rather than use 3D Mark Vantage to test the performance and power of anything. 3D Mark in all it's most recent incarnations do not run an actual game engine. The numbers they generate are arbitrary at best. Factors that do not effect actual game performance drastically alter its results. You get huge gains from minor CPU clock speed increases and you get a nice point increase for having a PhysX capable card in your system, yet in most PhysX based titles, the performance you get with PhysX enabled drops. Usually it adds minor eye candy effects and that's it. Another example of this reaches back to the Geforce 8800GTX / Radeon 2900XT days. Back then the Radeon 2900XT was certainly the inferior card but you'd never know that judging by 3D Mark scores of the time. The 2900XT benched awesome in 3D Mark but in reality it was inferior to the 8800GTX in actual gaming performance.

I'm not saying 3D Mark is totally useless, but as a measurement of performance relative to different configurations it is worthless to say the least.
 
ctrl+left mouse then scroll wheel down. Uber rig btw. I want one. :D
 
Thats 2 5970s in crossfirex right?

It's a very sweet set up thats for sure, but pretty much useless at the moment unless you REALLY absolutely need a lot of AA in Crysis in 2560x1600, thats the only thing 1 5970 won't deal with.
 
I'd play a game rather than use 3D Mark Vantage to test the performance and power of anything. 3D Mark in all it's most recent incarnations do not run an actual game engine. The numbers they generate are arbitrary at best. Factors that do not effect actual game performance drastically alter its results. You get huge gains from minor CPU clock speed increases and you get a nice point increase for having a PhysX capable card in your system, yet in most PhysX based titles, the performance you get with PhysX enabled drops. Usually it adds minor eye candy effects and that's it. Another example of this reaches back to the Geforce 8800GTX / Radeon 2900XT days. Back then the Radeon 2900XT was certainly the inferior card but you'd never know that judging by 3D Mark scores of the time. The 2900XT benched awesome in 3D Mark but in reality it was inferior to the 8800GTX in actual gaming performance.

I'm not saying 3D Mark is totally useless, but as a measurement of performance relative to different configurations it is worthless to say the least.

Good advice, of course. From my long experience with 2x 4870 X2 playing games vs. benching in 3D Mark is an entirely different ballgame. Most often the extra horsepower was just sucking unused power. Many games seemed they were only taking advantage of 2 GPUs (or 1 in worst case) while the cards were a powerhouse in 3D Mark.

Still though, having theoretical 512 GB/s is pretty damn impressive.
 
Just a Furmark Bench.

Furmark5970Quad.jpg


I have Borderlands in hand. Just arrived from Amazon same day as the second 5970. So I think I'll install that and bench it next.

Here are just some physical pictures of the cards and the setup installed to tide you over until I can get some solid bench numbers in hand.

Sapphire 5970 NIB, fresh from Amazon.
103_6336.jpg


2xHemlock, side by side.
103_6341.jpg


Each 5970 gets its own 12v 38A Rail thanks to the Antec TPQ-1200's modular design.
103_6343.jpg


Crossfire Bridge connected. We're in business!
103_6344.jpg
 
Last edited:
crossfire x with 5970's has been a bit sketchy for me. i'm looking forward to seeing how it works out for op.
 
Say good bye to overclocking. Running two 5970 side by side just seems like temperature management horror issue to me.
 
Say good bye to overclocking. Running two 5970 side by side just seems like temperature management horror issue to me.

Kind of like 3 280 then, but nothing water can fix and this card just beggs for it.
 
Yeah you need water for OCing with quadfire, unless you have some serious airflow through the case and well spaced out cards.
 
warn your roommates ahead of time that a helicopter is NOT landing in your room when the fans kick in.
i dont know if you plan on keeping both, or what kinda resolution you're running...but that's a steep price to pay for good benchmark numbers.

bring on the benchmark numbers!
 
So I just played Borderlands for about an hour with FRAPS on. The framerates did not vary at all really. And there are no video options to tweak. I think that Nvidia's TWIMTBP vendor lock must be in effect on this game. Can't find settings for AA or even Anistropic Filtering. Looks like Nvidia has 'patented' those video abilities as well!

Then again Min 59 Max 63 Avg 61.967 is nothing to sneeze at. Looks like Vsync has also been forced on which limits the framerates top end. Conclusion: There's not much benefit to high end video cards on Borderlands except for rock steady 60FPS. I need to find a way to tweak settings through CCC I think.

Oh, and my second video card got REALLY HOT! Afterburner reported temps of over 100C on cores 3 and 4! I think I need to find an extra long Crossfire Bridge so that I can space the video cards better. Does anybody know where I can find one? Cos the standard 3 inch CF bridge included with the video cards won't reach if I move the 2nd 5970 down more.
 
So I just played Borderlands for about an hour with FRAPS on. The framerates did not vary at all really. And there are no video options to tweak. I think that Nvidia's TWIMTBP vendor lock must be in effect on this game. Can't find settings for AA or even Anistropic Filtering. Looks like Nvidia has 'patented' those video abilities as well!

Then again Min 59 Max 63 Avg 61.967 is nothing to sneeze at. Looks like Vsync has also been forced on which limits the framerates top end. Conclusion: There's not much benefit to high end video cards on Borderlands except for rock steady 60FPS. I need to find a way to tweak settings through CCC I think.

Oh, and my second video card got REALLY HOT! Afterburner reported temps of over 100C on cores 3 and 4! I think I need to find an extra long Crossfire Bridge so that I can space the video cards better. Does anybody know where I can find one? Cos the standard 3 inch CF bridge included with the video cards won't reach if I move the 2nd 5970 down more.

The 62-63 max in borderlands isn't vsync its the result of a framerate smoothing options in the game config.

AA doesn't work because it's an Unreal game, and the devs didn't bother putting it in.

There are threads all around about how to tweak that game, but regardless it's horrible as a benchmark tool. I get the same framerates as you with my rig, even with crossfire disabled. Run something that puts that beast through its paces!
 
Impressive rig thumbs up. I can tell you right now you're biggest concern you should watch out for are your VRM temps overclocking or not. These 2 side by side with all the benchmarking and gaming during a hot summer will thank you and last a long time with at least a full coverage WC block.
 
plenty of unreal based games don't have in-game options for aa. force it through catalyst. as for extended cf bridges, try newegg.
 
you definitely need to space those cards out to slot 1 and 3...could you post a GPU-Z screen shot how they are now please.

When I had my cards in slots 1 and 2 GPU-Z was showing x16 x 8.

I have an extra long XFire bridge....just PM me and pay for shipping and its yours.
 
you definitely need to space those cards out to slot 1 and 3...could you post a GPU-Z screen shot how they are now please.

When I had my cards in slots 1 and 2 GPU-Z was showing x16 x 8.

I have an extra long XFire bridge....just PM me and pay for shipping and its yours.

As requested, a GPU-Z screenshot.

GPUZ5970QF.jpg


As you can see, GPU3 and GPU4 run about 25C hotter than GPU1 and GPU2. And this is with a side fan actively blowing air onto the video cards. So it looks like one of the cards is choked for fresh air due to proximity.
I think I need a 12cm Crossfire cable for the bridge to reach both video cards to when spaced further apart.
 
As requested, a GPU-Z screenshot.

GPUZ5970QF.jpg


As you can see, GPU3 and GPU4 run about 25C hotter than GPU1 and GPU2. And this is with a side fan actively blowing air onto the video cards. So it looks like one of the cards is choked for fresh air due to proximity.

I think I need a 12cm Crossfire cable for the bridge to reach both video cards when spaced further apart.
 
UPDATE: So I took a quick run through Dragon Age: Origins with the Quadfire setup in the Mage Tower's Basement area. Seems that this game scales better than Borderlands at least.

Fraps benchmark report:
Min FPS 55
Max FPS 354
Avg FPS 118.3

These looks better than the last set of numbers but perhaps someone with 5870CF would like to compare. These benches were done at stock clocks since the 3-4 gpus are running so hot right now.
 
I have 5870s in CrossfireX in my EVGA Classified.

I have the cards in PCI-e slots #1 and #3.
To do so I bought an extended Crossfire bridge(s) from the ASUS store. It's the only place I could find them for a reasonable price. They are 5 inches long. That is plenty to reach the slots I indicated above.
Very nice set up.
I found in Borderlands my Crossfire got the same framerate results as yours. The game is capped somehow.:D

Link:http://estore.asus.com/shop/item.asp?itemid=4662
 
Vjcsmoke can you post a screenshot of the GPU-Z sensors tab. Interested to see what temps you are getting in there.
 
amazing rig

yes some gpu-z sensor tabs. You might want to open 4 gpu-z's, and put each on sensor tab. (select each 5900 series in the drop down box on main tab)

Then a 5th gpu-z on the main tab.
 
Damn.. cooking at idle.. what kind of load temps you getting? Spacing will really help them.
 
just a little trick i remember from Kyle on one of his reviews.. i think it was the 9800GX2 in quad SLI.. he took a small piece of paper.. folded it a bunch of times and stuck it between the cards just to push them apart enough for the top card to get some cool air.. ya might want to try that and see if it helps with the current setup..
 
Thanks for the tips guys. I will be out of town for the rest of the week so I won't be able to do more benches until then.

Here's a bench I took this morning of the 5970 Quad system on Need for Speed Shift with FRAPS at stock speeds.

Min FPS 95
Max FPS 179
Avg FPS 132.125

I'll be back Sunday. Until then, play [H]ard!
 
I'd get a longer CF bridge as suggested so they can be spaced for cooling.
 
Yeah, see if you can space them for cooling with a longer bridge as suggested or start zip-tieing fans on to see what helps the most.
 
just a little trick i remember from Kyle on one of his reviews.. i think it was the 9800GX2 in quad SLI.. he took a small piece of paper.. folded it a bunch of times and stuck it between the cards just to push them apart enough for the top card to get some cool air.. ya might want to try that and see if it helps with the current setup..

This helped a lot with my GTX280s.

I could actually have a second 5970 right now but instead I'm ebaying it and bought a 5870 here for precisely the issues OP will run into. I think the 5870 will run fine even with its fan covered by the 5970, and I'll still be able to OC the 5970.
 
Back
Top