SMiThaYe

3DMark11 Benchmark (Synthetic)

13 posts in this topic

Using Intel i7 2600K @ 4.8GHz + Nvidia GTX 680-SLI Ultra Charged

Proof with official 306.97 drivers + PhysX 9.12.0613 - http://www.3dmark.com/3dm11/4633304

Proof with official 306.97 drivers + PhysX 9.12.0613 - http://www.3dmark.com/3dm11/4633706

GPU-Z validation - http://valid.canardp....php?id=2541548

Overclocking with trusty MSI Afterburner 2.2.4 - these are my 24/7 settings apart from fan speed is set to Auto. My memory can top out to another 200MHz, the core maybe 8MHz but that's the limit. Temps would become a real issue and hit the thermal limitation of Kepler as vcore at 76c starts to drop from 1.175v. Third image displays the default Boost clock at 1215MHz.

0

Share this post


Link to post
Share on other sites

Maybe you could post a single GTX 680 results for a reference? I will put mine here on Sunday (away from pc).

0

Share this post


Link to post
Share on other sites

Great Benchmarks Dude ...Like Admin Said Can You Put Benchmarks With A Single 680...

0

Share this post


Link to post
Share on other sites

Sure, I'll sort that out :)

I also have a spare low-end GT440 that I wanted to try with GTX 680-SLI for separate PhysX/graphics gaming performance (it's actually good enough for Low/Medium PhysX), but checked that the 3rd PCIe 4x slot spacing is too tight with those big GPUs.

0

Share this post


Link to post
Share on other sites

best my hardware can do with a single GTX 480 and an OC'd Q9550:

0

Share this post


Link to post
Share on other sites

As requested here is my single GTX 680 run, also did another GTX 680 SLI run.

#1 - Using Intel i7 2600K @ 4.8GHz + Nvidia GTX 680 Ultra Charged PCIe 2.0 x16 (same settings as before)

Proof with official 306.97 drivers + PhysX 9.12.0613 - http://www.3dmark.com/3dm11/4649390

Proof with official 306.97 drivers + PhysX 9.12.0613 - http://www.3dmark.com/3dm11/4649360

Also re-ran the initial benchmark (I always run 3-times) whilst I had the time. Not sure why I had the jump in Xtreme mode but Performance stayed the within margin or error.

#2 - Using Intel i7 2600K @ 4.8GHz + Nvidia GTX 680-SLI Ultra Charged PCIe 2.0 x8/x8 (same settings as before)

Proof with official 306.97 drivers + PhysX 9.12.0613 - http://www.3dmark.com/3dm11/4649498

Proof with official 306.97 drivers + PhysX 9.12.0613 - http://www.3dmark.com/3dm11/4649552

I know I can get past X7,000 in Xtreme as the Ultra Charged memory can be pushed another 200MHz but this was simply to compare single GTX 680 with a moderate and easily achievable overclock to SLI.

The scaling performance - in a perfect world - from my quick calculation is 79.23% efficiency at PCIe 2.0 x16 Vs x8/x8 SLI.

Would have liked a bit more performance but would be interesting to see posts from others on their SLI efficiency.

1

Share this post


Link to post
Share on other sites

this makes me feel a bit better about my score :D , but it was ran on absolute max clocks at that time. When it gets a little colder outside (below 40F), I'm going to re-run my cold air induction test. Going to shoot for 4.2ghz on my Q9550 this time and maybe 930mhz or so on the GTX 480.

Nice run on your single GTX 680.

0

Share this post


Link to post
Share on other sites

Your toasty GTX 480 was only 43% of Graphics Score behind the GTX 680 ;) Still doing a fine job :) My room temp has been 22c of late (10c outside) and have a nice 12,000 BTU air conditioner standing by for the summer, come winter in the UK, the temps drop to -10 outside and that is ideal for further overclocks if I really want to push it like you. The 2600K hit 5GHz last winter on water at just 1.38vcore and can safely do v1.4v. Would go up to v1.45 but it all depends on the temps from the vcore as you gradually step it up.

Due to thermal limits being hit on the top-GPU, I still can't run those GTX 680s in SLI at maximum clocks (overclocked silly). Purely because of the GPUs being technically 3-slots and spaced next to each other.

I remembered only that the bandwidth was overclocked to a mighty 230GB/s when I first got them and spent hours doing various benches.

0

Share this post


Link to post
Share on other sites

You need to do like I was talking about last night and have your intake fans pulling air from the outside, especially if the temps are getting as low there as you say there (I'm going to have to wait another month or so probably to get the sub 40F temps I'm looking for). The only danger is condensation (very small chance, really depends more on humidity in the air and how hot the hottest piece of hardware gets, if my 480 didn't do it, I doubt anything would) so you can't run it too long setup like this, no more than 1-2 hours is what I go by. Basically you need to know the settings you are going to throw at it before you even open that window. Once the window is open do like I did and stuff pillows beside the case so you essentially seal off the computer, then watch the temps dive :D

I just realized something, if I get the case I want, Raven2, I won't be able to do this test unless I just set the entire rig outside due to that case pulling air in from under the case.

0

Share this post


Link to post
Share on other sites

I will try that ZoSo :) For everyday use I've used a couple approx 90CFM 120mm fans covering the GTX680s as a cheap solution as I can't currently afford to put them under water - I still want to do that because those extra fans spoil the the clean look. One 120mm fan is pushing the internal air directly between the GPUs and the 120mm side panel fan is pulling the hot air directly out the case into the room. I've spent ages testing the airflow/temps in my case but not gone too deep into it. The rear 140mm case fan on the Corsair 800D is pulling air in instead of exhausting it, that helped temps by a few degrees C. Condensation: My aircon is also a dehumidifier and extracts litres of water out of the air every hour during the summer and can further drive room temps down but the actual limit on the unit is 16c, will make a difference even in the winter. Ideally, I'd like to point the unit towards the front of the case with a few 5.25" panels taken off but my hose is too short to turn it around 180-degrees.

These aren't ideal solutions, I can only work with within my means and welcome suggestions such as yours. I won't be able to seal the case with insulation since my case pulls/exhausts air from the top, side, rear, and underneath for the PSU ventilation.

While I remember, I bought a black SLI connector to replace the orange Asus connector - this is atheistically better ....

0

Share this post


Link to post
Share on other sites

Isn't 3DMark11 like the previous version where we are supposed to turn PhysX off because it boost the score?

Anyway here is my last run with PhysX Off (set to CPU) as not to boost the score due to having a NVIDIA card.

very nice boost from my Q9550 run, 2049 pts higher with Ivy + Z77.

0

Share this post


Link to post
Share on other sites

I assume you are referring to Vantage where PPU was selectable, default was off as shown below:

All PhysX is run on the CPU as I suspected by Futuremark Community Manager: "It's also worth noting that the Physics Test runs entirely on the CPU using Bullet Physics. In the Combined Test, the CPU handles rigidbody physics while the GPU looks after softbody physics for the waving flags, again using Bullet Physics. In both cases PhysX is not used." [ source ]

ZoSo, any boost it was likely to come from your CPU where the PhysX is bound to, and updating your graphics drivers (your cheating if you have Lucid enabled so turn it off - press release).

When I can get a spare 20 minutes I will run 3DM11 with the latest betas, since they aren't WHQL (not allowed on leaderboards) they will be here purely out of curiosity like yours.

0

Share this post


Link to post
Share on other sites

Yea I've already read that press release about Lucid HyperFormance and how it skips frames and produces a false high score. I figured the same thing was true for PhysX as the last version of 3DMark, I had them both off. I did OC my GPU to 860/1920/4000 with the new CPU.

0

Share this post


Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now