Quadro FX 1600M (8700M GT) undervolting?

Does anybody know if this card can be undervolted?

Spoiler

Tomorrow, I'm getting a new one in the mail and I'd like to undervolt it as much as possible so it survives longer than the last few did.

I'll dump vBIOS with GPU-Z, edit voltages in NiBiTor, and flash with NVFlash.

Hopefully, I can get P0 down to 1.15V and P10/P12 down to 1V while keeping clocks the same.

 

I guess another option is to overclock P10 to P0 and use a batch file to force P10 all the time, but I'd prefer making it permanent in the vBIOS if possible.

 

Bump. Delivery got delayed so I still don't have my card yet. Would be interested in hearing from other current and former mobile Tesla GPU owners and their experiences with undervolting.

 

I recently bought a semi-defective M570RU with a broken power jack (was definitely not used for gaming) with an 8700M GT installed.
Today after having repaired it, I put the machine together again and installed Windows 7. Then did some temperature tests with Furmark and decided to try undervolting the card. At the moment, the card is successfully running stable under Furmark @ 1.15V (flashed into the V-bios for all performance levels) at 60°C. Next go will be to lower the voltage even further on the Throttle3D and 2D profiles to 1V, which it surely will be able handle (tomorrow, it is kinda late now in Europe ). So, yes, there should be plenty of room for undervolting.
It probably won't make it down all the way to 1V under Max. profile, but that's as far as I've got now.

 

Great, thanks for the info. My card was stolen or lost by the shipping company so I'll need to buy another one after I get my refund. I'll update this thread after I've had a chance to play around with it.

BTW, what undervolting method did you use? Did you flash it into vBIOS or adjust thru software?

 

Flashed it. Would have already tried the next step, but Windows keeps updating all the time, and I got to go to bed now

 

Which "next step" are you referring to?

 

Trying to get down to 1V for all performance levels. As I said, I don't really expect the card to run stable at max clocks with 1V, but the proof is in the pudding.
Also, I haven't lowered the voltage for the Throttled3D and 2D profiles, yet, but that should work out just fine.

 

Yeah, 1V for full 3D clocks is a massive undervolt. I really don't think it's stable. The 8700M and FX 1600M need a lot of volts as they're 80nm GPU's and clocked quite high for G84M. I'd be ecstatic if I can just get 1.15V on P0 and 1V on the others.

 

That's what I am running at the moment Dunno if I ever try 1V for full clocks, cos it won't lower the temps down any further,
as it may only make the fan spin one step slower, I guess. It may be beneficial to overall lifespan though.
We'll see...

 

Hard to believe, but it actually does run stable at 1V full clocks , tested with Furmark for 15 minutes, no lockups or driver crashes whatsoever :thumbsup:

 

<del>What about Unigine Heaven overnight?</del>

Edit: I meant Unigine Valley maxed out overnight. Keep forgetting this card isn't DX11 . Heaven without extreme tessellation isn't as demanding as Valley.

 

Ok, seems as if I have to backpedal a little here. I've encountered an odd phenomenon. Lowering the voltage down to 1V on max clocks actually raised the temps back to 1.35V levels (67°C).

Hadn't really paid much attention to this yesterday, cause I ran Furmark in a well heated room, but I've counterchecked today, and it holds true in a much cooler environment. Also has nothing to do with a different fan level, as the fan speed is temperature controlled anyway,

Applying 1.15V for max clocks definitely runs cooler than using 1V!

It seems as if the V-Bios doesn't allow the card to use 1V, and instead applies max voltage of 1.35V instead.

This could very well be true for all performance levels, so I guess leaving everything at 1.15V and calling it good is the way to go, as it yields the lowest temps with the card topping out at 60°C, as opposed to 67° on "apparently-not" 1V.

Your thoughts, octiceps?

P.S. Will test stability with Unigine Valley later, got to take a nap now

 

Interesting, but not unheard of. It's probably hard locked to a certain limit, so if you try to go below that limit, it just jumps back up to stock. Can you verify whether the lower p states can take 1V or if it's the same deal? I'd like to get them down to 1V to save battery, but if not, oh well, I'll live with 1.15V across the board.

 

How about trying to aim for 1.1V instead?
If 1.15V is stable... try lowering it to 1.1 to see if that makes a difference.

 

There are only four voltages available to be selected: 1, 1.15, 1.25 and 1.32V in the V-Bios, and I've just verified my findings.

Idle temperatures on "supposedly" 1V are significantly higher for low 2D, as opposed to choosing 1.15V.

I didn't even take the time to make any temperature comparisons under load on the lower two performace states, but flashed right back to 1.15V for all three p-states.

For me it's "case closed", I'd say.

 

Ok...
out of curiosity... what were your temperatures before undervolting, and now after undervolting?

 

Under max clocks and max load with Furmark 67°C before, 60° after and I guess two fan levels slower.
The M570RU has rather narrow-stepped fan tripping points.

 

Wow, that's pretty damn low. Does it have dual fans or something? Even after a fresh repaste, my HP 8710w would idle close to 50C and get over 80C, sometimes close to 90C, under full load, and that's with max fans running. The cooling system is pretty weak. Only a single fan and one thin heatpipe each for both CPU and GPU.

 

Yes, the M570Ru has a dedicated heatsink and fan for the GPU. Though, I am defintely going to replace those loud- CPU and GPU fans against some quieter ones. I've done it before on a different machine, with only negligible temperature increase.

Oh yeah, and I also tested Unigine Valley for 30 minutes without any problems, temps maxed out at only 58, so i guess it isn't as demanding as Furmark for stability testing.