Confused about GTX 690 overclocking

[AdamK47]

So uhhh, yeah...

I've been using MSI Afterburner. It shows voltage control by offset. Does this even work? All the settings I've tried don't reflect in Afterburner when the GPUs are fully stressed. It seems to go to 1.175V or 1.150V at load no matter what.

It's fine if it doesn't work. The GPUs seems to overclock well as is.

 

[lavaheadache]

duuuude, you are crazy! =) I'm crazy too, crazy jealous.

 

[chimaxi83]

Those don't have unlocked voltage control, like almost all other Kepler cards. 1.175V is the max.

Also, love the GPU insanity in those pics

 

[AdamK47]

OK, thanks for the confirmation. I've only played with them for a few hours. 1150MHz is what I have now and it's been stable with a 1 hour loop of Unigine. On to more testing!

BTW - I'm not sure what to think of the green glowing letters. It does have a somewhat nice contrast againt all of the blue.

 

[lavaheadache]

it does look pretty nice. The logo that was on the side of my reference 680 looked so cheesy it was almost embarrasing. It could have been nice if done differently.

 

[3DVagabond]

I can't imagine that you "need" to O/C those things.

I'm curious to see how much of an improvement they are over your old trifire setup.

 

[notty22]

Awesomeness. Brought a smile with all those gpu's. 5 of them with the gtx 650ti at the bottom.
I'd like to see a screenshot of your NV control panel where it shows what you can do with those gpu's. configuration options/explained.

Not mine just for example.

 

[AdamK47]

The GTX 650Ti has been removed since taking the shots. Did some PhysX testing with the 650Ti as dedicated compared to PhysX being performed on the Quad SLI setup. I didn't see much of a difference. The control panel did show all five GPUs. The two 690s in SLI and the 650Ti as separate. The MSI Afterburner graphs were a mess with all 5 GPUs in it.

 

[AdamK47]

Just did a PCI-E stress test with GPU-Z and it shows all GPUs running at PCI-E 3.0. I thought nVidia disabled 3.0 on X79 and defaults to 2.0. Perhaps it detects the dual PLX chips on the ASRock Extreme11 and not X79?

 

[Lepton87]

Why not another 7970? Guess money is of absolutely no concern to you.

 

[AdamK47]

Why not another 7970? Guess money is of absolutely no concern to you.

Thought about it. I would need another power supply. The AX1200 is not enough. It doesn't have enough power or enough connections for 4 cards. The EVGA 1500W I was looking at is $500. Add another $400 for the card and it's $900. Sure, the GTX690s are very expensive, but I can sell the three 7970s and recoup a lot of the money.

 

[AdamK47]

Been playing around with the voltage control and I'm convinced it doesn't do anything, even if there is a 1.175V limit.

This was tested using Unigine Tropics. A good power test since it renders high framerates in DX11.

The first bump in the graph is with a +25 offset. The second bump is with a +37 offset. The third and last bump is with a +0 offset.

It's exactly the same with all three. It starts at 1.175V for a few seconds and then goes to 1.162V.

 

[chimaxi83]

On the cards with locked voltage, Kepler dynamically decides what voltage it needs. Well even on my volt modded 680, voltage would change depending on GPU load. Doesn't really matter what the offset is set to, if it doesn't need that 1.175V, it won't go that high.

 

[3DVagabond]

Thought about it. I would need another power supply. The AX1200 is not enough. It doesn't have enough power or enough connections for 4 cards. The EVGA 1500W I was looking at is $500. Add another $400 for the card and it's $900. Sure, the GTX690s are very expensive, but I can sell the three 7970s and recoup a lot of the money.

If you ever feel you need a 1500W PSU, don't buy an NEX1500. Not only is it ridiculously priced, it's just not a good unit.

 

[Tempered81]

Amazing man. Should you use Precision instead of Afterburner? Volts locked. Hows framerate compare to 7970 trifire?

 

[AdamK47]

3DMark 11 results comparison:

http://www.3dmark.com/compare/3dm11/5252426/3dm11/5171069

Could use more tweaking I suppose.

 

[3DVagabond]

~20% improvement is about what I'd expect. As far as the gross scores go, yeah, they seem a bit low, but they should be comparable against each other.

 

[Grooveriding]

Amazing!

There is no voltage control on these cards... but

http://www.xtremesystems.org/forums/showthread.php?284014-KGB-Kepler-BIOS-Editor-Unlocker

This does work, while at the same time voiding your warranty. Most get about 50-100Mhz more out of doing this. Your temperatures usually go up a bit as well.

I got my cards to 1280 when I tried it and temps went up about 5C.

 

[AdamK47]

~20% improvement is about what I'd expect. As far as the gross scores go, yeah, they seem a bit low, but they should be comparable against each other.

The GTX 690 SLI score was with only a +85 core offset for an even 1000MHz base clock.

 

[blackened23]

Been playing around with the voltage control and I'm convinced it doesn't do anything, even if there is a 1.175V limit.


It's exactly the same with all three. It starts at 1.175V for a few seconds and then goes to 1.162V.

You can't control voltage with any software with the kepler, Adam, messing with the voltage slider only changes the "offset". It doesn't change the max voltage which is always determined by GPU load, maxing out at 1175. There is really absolutely no point in messing with any voltage controls, i've fiddled with it and concluded it helps nothing. As Max voltage is what you want to increase for max overclocks, manipulating the vDroop offset doesn't do much (in my experience)

Another thing to be aware of is kepler throttle: anytime your GPU temp goes above 70C, your GPU boost speed will throttle. So keep in mind _no matter what_ when you overclock you need to keep your GPUs under 70C always. That may be a challenge with the 690, when I tinkered with them I would hit 80-85C during heavy gaming loads. YMMV though, I live in a humid area - and those temps only occured in games like crysis 2 / metro 2033. Normal games with vsync didn't have such a high temp. I've seen others mention no such problems on their 690s, so you may be fine in that respect.

For that kind of cash, have you thought of getting MSI lightning 680s? I really enjoy my pair which OC to 1402 - You can still flash the BIOS to get unlocked voltage with afterburner. Either way, sick rig, man. :thumbsup: Really like your case layout!

 

[wand3r3r]

You can't control voltage with any software with the kepler, Adam, messing with the voltage slider only changes the "offset". It doesn't change the max voltage which is always determined by GPU load, maxing out at 1175. There is really absolutely no point in messing with any voltage controls, i've fiddled with it and concluded it helps nothing. As Max voltage is what you want to increase for max overclocks, manipulating the vDroop offset doesn't do much (in my experience)

Another thing to be aware of is kepler throttle: anytime your GPU temp goes above 70C, your GPU boost speed will throttle. So keep in mind _no matter what_ when you overclock you need to keep your GPUs under 70C always. That may be a challenge with the 690, when I tinkered with them I would hit 80-85C during heavy gaming loads. YMMV though, I live in a humid area - and those temps only occured in games like crysis 2 / metro 2033. Normal games with vsync didn't have such a high temp. I've seen others mention no such problems on their 690s, so you may be fine in that respect.

For that kind of cash, have you thought of getting MSI lightning 680s? I really enjoy my pair which OC to 1402 - You can still flash the BIOS to get unlocked voltage with afterburner. Either way, sick rig, man. :thumbsup: Really like your case layout!

I had the same experiences, same cards.

The 690 cannot be kept under 70 though, I haven't heard of it yet anyway. With a little overclock it's around 80.

The lightning hit ~1390ish with voltage, voltage contributed about 50MHz. I wouldn't do it with the 690, they are thermally constrained as it is and voiding the warranty's not too great.

What I'm waiting for are a few benches showing some games FPS with trifire vs. quad sli.

 

[AdamK47]

blackened23, thanks for the tips. I've decided to keep the voltage offset at +0 since it doesn't seem to do anything.

I did torture test the 690s for a couple hours by looping Unigine Tropics in DX11. I've found this test to be much more GPU intensive than looping Unigine Heaven. Temps reached the upper 80s on all 4 GPUs. Didn't monitor GPU clock speeds at the time, I'll have to test again and see what kind of throttling is done. Played Borderlands 2 for a while yesterday with MSI Afterburner running in the background. All settings were maxed in game and PhysX was on high. Vsync was also enabled. The GPU temps were in the low 60s.

Thought about getting three 4GB GTX 680s, but decided on getting the two GTX 690s. I get 2GB less memory per GPU, but I run one 30" display, so it should be enough. The 4th GPU makes up for the less memory. This quad SLI setup is also much more power efficient. Looking at the power consumption tests done at TechPowerup, the GTX 690 uses 28W less on average load compared to a single GTX 480 and 94W less than 680 SLI. That's pretty impressive. Plus, having only two cards allows for better airflow.

 

[OVerLoRDI]

Madness man.

What are you doing with those 7970s?

 

[Elfear]

If you don't mind me asking, what made you decide to switch?

 

[AdamK47]

Microstutter! No, just kidding. I've always been able to prevent it. It hasn't really been an issue for me with the 7970s.

Just wanted to try something different. Never owned a dual GPU card before. I like the idea of Quad SLI in a two card config that's around the same power consumption of Tri-Fire 7970s while at the same time being a bit faster.