Nvidia Pascal cards still exhibit high-refresh-rate "power bug"

A while back, we reported on a weird issue with GeForce graphics cards and multiple-display setups with refresh rates higher than 60 Hz. The short version is that connecting multiple high-refresh-rate displays to the DisplayPorts on a GeForce card causes it to run at much higher clock speeds than the Windows desktop would seem to demand, increasing power draw and heat production. As Pascal cards start to trickle into the TR labs, we figured it was as good a time as any to revisit this issue and see whether it's been ironed out. Sadly, the problem appears to persist with these next-gen cards.

First, a brief word about our test setup. The particular graphics card we're using is Gigabyte's GeForce GTX 1080 Xtreme Gaming. My test bench has two monitors on it: my personal Eizo Foris FS2735, a 144-Hz FreeSync display, and Asus' ROG Swift PG279Q, a 165-Hz G-Sync display. Both of these screens have DisplayPort inputs, and I hooked them up to the GTX 1080's corresponding DisplayPort outputs. I monitored clock speeds using Gigabyte's Xtreme Gaming Engine software.

Our GTX 1080 running at idle, as it should be

To reproduce this bug, I first tested the card with the PG279Q connected by itself. Whether the display was set to run at 60Hz or 165Hz at the Windows desktop, the GTX 1080's core clock speeds fell to the expected 291MHz at idle. Going by our trusty Watts Up power meter, the system consumed about 65W to 75W on average. Setting the PG279Q to run at 165Hz and plugging in the Eizo at 60 Hz caused the card to clock up to 1304 MHz. At those speeds, our system draws about 107W on average. Setting the Eizo display to run at 144Hz didn't cause clock speeds or power consumption to increase any further.

Once again, though, things can get weird with this setup. Running the Eizo display at 144 Hz with the PG279Q set at 60 Hz causes the GTX 1080 to clock down to 291 MHz at idle, as expected. As we saw above, though, the opposite isn't true. Disabling G-Sync on the PG279Q doesn't cause the GTX 1080 to clock down, either. I'm at a loss to explain why any of those behaviors are as they are.

Our GTX 1080, reproducing this "power bug"

Thanks to Pascal's apparent efficiency gains, the average power draw of our system with the GTX 1080 clocked up is a pretty significant reduction from the 130W to 150W our test system pulled from the wall with a similarly-afflicted GeForce GTX 980 Ti installed. Still, it's clear that Pascal can run even more efficiently yet when its clock speeds aren't being pushed above what's apparently needed to drive the Windows desktop. Users with multiple high-refresh-rate displays are still going to pay more for power over the life of the card than they otherwise might if the card clocked down appropriately at idle, and it appears this problem resides somewhere in the driver.

For now, I'm resorting to setting both of my displays to run at 60 Hz at the Windows desktop so that I'm not inflating my power bill and unduly heating up my office. I'm still hoping Nvidia finds it in its heart to investigate and fix this issue, because fiddling with Windows display settings every time I want to take full advantage of the PG279Q's 165-Hz refresh rate is far from ideal. We'll keep you updated if we hear anything more about this issue from the green team's HQ.

Tip: You can use the A/Z keys to walk threads.
View options

This discussion is now closed.