Low power limit(GTX 1080)

I hope someone, can help me figure it out.
For some reason, my MSI 1080 SeaHawk is limiting itself to 0.90-0.92V.
This limits me to ~1700MHz, far from the 2XXXMHz other card can do.
This is with a Core voltage at 100% and Power limit at 105%(Max limit on stock firmware).
I’ve observed it over 1V but only for a split second.

The temperature seems to be in check GPU temp under 60 at full load(Furmark).
I don’t know about the VRM, but they do have there own dedicated cooler.

I tried flashing a Founders Edition firmware, and while the default clock speed is lower, the Power limit can be set at 120%.
This allowed 0.95V, but thats still far from what I need for higher clocks.

And now I’m out of ideas…

Edit: Forgot that I also tried putting it in OC mode, using MSI’s utility.
And it made no to little difference.

Edit 2: It seems the driver is detecting Furmark, and limiting the performance, to keep me “safe” I guess…
It’s not the first time, that furmark has been targeted like that. CoughingAMDdriversCoughing…

Short your shunts with Liquid Metal… That’s how to get around it at the cost of insane power usage. This fools the driver into thinking it’s using less power than the card is actually pulling.

So at the risk of damaging the card…

1 Like

It’s an overclock card… He said the VRMs are sufficiently cooled and it’s all under water. Just use nail polish on SMD components and just be careful to remove the liquid metal on the shunts once benchmarks are done.

If you disagree with the power limits, get around them.

You know these exist right?
image

1 Like

This is too vague to draw a decent conclusion.

Um… 0 Ohm will trigger the driver to put the card into safe mode.

More like 33 or 25 ohm. Just enough to trick the driver it’s using less power than it actually is.

1 Like

So how is using liquid metal any different from a shunt resistor? Gallium has a resistivity of 1.4×10-7 m Ω. As an electrical engineer it seems safer to use a shunt resistor opposed liquid metal

It is. And you’re not wrong. People just use it on top of pre-existing 50 ohm shunts to get their overclocks in then remove it once the benchmarks are done.

I don’t think the OP is trying to suicide his GPU for a dick measuring contest.

It looks like the goal is to get a decent overclock OVERALL.

So why would you suggest a potential GPU killing OC just to dick measure when the OP implies he wants a long term stable OC?

Adding more liquid metal will be more effective, but you begin greatly increasing risk of (1) LM dripping and shorting other components, thereby killing the card

I think its reasonable to assume OP wants a stable OC not records

he’s only looking for a few hundred extra megahertz

Then don’t use Furmark. Run a Unity game with Vsync off that uses AVX instructions. Any modern Unity game is extremely stressful on the GPU with Vsync off.

Heck, Unigine Superposition looped by using the Game mode -> Cinematic mode trick should do too. The driver would think Superposition is a game.

um… AVX is a CPU instruction set doesn’t even get sent to the gpu

I use it for CPU stability too. Don’t question my methods. I’ve had Prime95 AVX pass, but a AVX instruction Unity game BSOD after confirming Prime stability.

Really now?

This doesn’t sound surprising as Furmark is basically created to push the GPU to ridiculous levels

There is no reason why drivers shouldn’t limit card clocks when Furmark is running. Both vendors do it to protect users from themselves.

MY opinion on it is if the OP genuinely cared about getting past that limit, I’m giving him an answer to the solution.

Yes.

It isn’t.

Assuming this is the Sea Hawk with the AIO and not the one that is just has an EK block on it, then they are not under water and have meh cooling. The Sea Hawk is just a Corsair 120 AIO slapped on the GPU. The memory and VRMs all just use the standard blower cooler. While this should be sufficient, you don’t wanna go hog wild.

Also, a 1080 will boost to 1900Mhz even sub 1V. Overall power limit is your biggest limiting factor.

Also GPU Boost is extremely sensitive to temperature. Every 5 degrees can cut your boost by ~50Mhz or more.

I don’t see too big of an issue with it though. Furmark is pretty pointless IMO. Games are what matters and you can manually OC. You should easily be able to hit 2000Mhz at stock power limits.

My waifu’s 1080 will boost to over 2000Mhz by itself if given 105%. With basic tweaking 2100 isn’t much issue

OP had no issue with temperature, but if you’re concerned about the rest of the board, then yes, the AIO Sea Hawk is not 100% full coverage if you were technical about the true details.

Edit: Yes, Games and Unigine Synthetics sound more reasonable for OC targets.

MSI Afterburner should have an option to unlock the Vcore.
voltage
I don’t know if you have tried this yet, but its the only thing I can thing of doing without physically modifying and potentially destroying your card