What do you make of this?

[rezerekted]

Just installed a GTX1070 and have been running some benches to make sure the card is stable. See the area I circled green, see the white line going through GPU Core clock and GPU memory clock. I have seen that twice now, once when using Unigine Valley and also when using Furmark. When that happens the graphics exhibit a very brief hiccup. Anyone know the cause of that?

 

[michalrz]

There's a power consumption dip and a perfcap dip coinciding with the clocks gap.
If you somehow caught it doing that you'd be able to see the perfcap reason. There are voltage related, thermal related and utilization related.
Keep the log file going and check it when it happens the next time, it should log the perfcap reason.

 

[rezerekted]

There's a power consumption dip and a perfcap dip coinciding with the clocks gap.
If you somehow caught it doing that you'd be able to see the perfcap reason. There are voltage related, thermal related and utilization related.
Keep the log file going and check it when it happens the next time, it should log the perfcap reason.

When I ran Furmark it said the perf cap reason was power and when I ran Unigene Valley (above screenshot) it said VRE. Would increasing voltage to video card maybe help? Or perhaps it is cpu maxing out the cores (i5 [email protected]).

I looked into these percap reasons and ppl say it is normal.

My PSU can use single rail or split rail so maybe if I switch it it will make a dif. ???

 

[FrgMstr]

Unless you are actually seeing that while in a real game, I would suggest there is not an issue. With both those non-game apps, you are basically trying to push the GPU to failure.

 

[michalrz]

When I ran Furmark it said the perf cap reason was power and when I ran Unigene Valley (above screenshot) it said VRE. Would increasing voltage to video card maybe help? Or perhaps it is cpu maxing out the cores (i5 [email protected]).

I looked into these percap reasons and ppl say it is normal.

VRE? Or Vrel?
If Power means "TDP limit reached" then increasing the card's volts should cause more dips related to 'PWR'.
I don't think it knows or cares about the CPU's TDP, so it's not your maxing out the i5.
If this happens at stock, I would care, but if it's only during furmark then I would not complain, because no game will reach Furmark load levels yet.
If this happens at stock during gaming, I would care because you're being denied some performance you paid for.

 

[rezerekted]

Yea, OK, thanks guys. I will test it during game play of BF4 and see what I get. I did see the boost clock go as high as 1950 when it is specced as 1866 but I saw it dip well below that too. This is a Gigabyte GTX1070 G1.

Maybe it was VREL, can't remember now.

 

[michalrz]

Yea, OK, thanks guys. I will test it during game play of BF4 and see what I get. I did see the boost clock go as high as 1950 when it is specced as 1866 but I saw it dip well below that too. This is a Gigabyte GTX1070 G1.

Maybe it was VREL, can't remember now.

MSI afterburner should be able to open up some headroom as far as the power envelope goes. Google says it allows for a +10 to +25% override. But the proper solution is apparently a hard mod (yay).

Theory says resistance goes up with heat, I wonder if it's possible to see a measurable difference in power consumption with aggressive cooling.

 

[rezerekted]

MSI afterburner should be able to open up some headroom as far as the power envelope goes. Google says it allows for a +10 to +25% override. But the proper solution is apparently a hard mod (yay).

Theory says resistance goes up with heat, I wonder if it's possible to see a measurable difference in power consumption with aggressive cooling.

I've done hard mod before so no problem but first you must do it via software to make sure the OC is stable. I will try MSI Afterburner and see what happens.

 

[xorbe]

That's the fun that started with gpu + boost. You try to push the gpu, you touch one of the various limits, and it recoils the mhz. Could also be a monitoring glitch.

Also, Litwicki already asked this question like 15 times, but I can't find the other threads atm.

 

[rezerekted]

That's the fun that started with gpu + boost. You try to push the gpu, you touch one of the various limits, and it recoils the mhz. Could also be a monitoring glitch.

Also, Litwicki already asked this question like 15 times, but I can't find the other threads atm.

Watch this video, it has the fix for this. That white line is because I hit either power limit or temp limit set on the card by the manufacturer, or both, and the fix is to raise the power limit, temp limit and fan curve. But I was using intensive benchmarking software and not a game so have to test over again with a game. Right now I have a slight fan noise issue that this card is not supposed to exhibit so have to try and fix that first.

 

[rezerekted]

OK, back from using Far Cry Primal benchamark numerous times and there is no pwr or vrel issue. In fact, my card is good and stays pretty cool, saw max of 59c deg and I get a boost clock of 1949. I found out the newer Nvidia cards boost over the factory boost if it determines your card can handle it, unlike AMD cards that never go over the boost clock setting.

But I will keep Furmark and Valley because I don't run them for long and use them to run in a window so I can monitor GPU-Z and CPU-Z at the same time.