Tip on Nvidia Cards - power and temp target

[dburne]

Hey gang,

 

Most everyone in here is probably already familiar with this, but thought I would mention it in case it helps someone.

 

One thing I have noticed with my Rift, is oftentimes my GPU will max out at the 100% GPU useage. Some of this is attributable I guess to VR , due to it displaying two images imposed on each other. Some may be attributable to my system specs and how the CPU feeds the GPU.

 

Now as I have an EVGA video card, I use Precision XOC so I have control over all three fans. Used Afterburner for years prior to getting this card, and what I am talking about can be done with either program. Not talking about overclocking the GPU, but maxing out the power and temp targets.

 

When I load up Precision for a flying session, I have an aggressive fan profile I apply, and I slide the Power and Temp Target sliders all the way to the right which gives me Power Target of 117% and Temp Target of 90c. Default was like 100% and 79c.

 

I flew a session yesterday and forgot to move those sliders, leaving at default. So I thought I would sneak glances at Precision to see what it was doing. Well it was hitting that 100% Power target and clock speed was varying a lot, dropping down when it would hit the target. I compared this morning, whilst having the power  slider all the way over to 117%, and it never did this maintaining full GPU boost for the duration of the flight, dropping a small amount only occasionally depending on GPU temp.

 

So even though I do not overclock, I make sure I have my power and temp targets all the way , and it does help greatly in maintaining boost speed consistency.

[[CPT]CptJackSparrow]

Ignorant question time..

 

Could this lead to more fps or at the minimum stable fps across a broader range?

[dburne]

Ignorant question time..

 

Could this lead to more fps or at the minimum stable fps across a broader range?

 

It just helps keep the Nvidia GPU running at full boost ( max clock for power and temp)  without a bunch of dips up and down, at least in my case.

[SAS_Storebror]

I did this with MSI afterburner once on my GTX 970, max power target for that card is 111%, max temp I've set to 89°C (that's what I got is the max safe temp for the Maxwell Chips).

The result was a 10% performance increase in 3D Mark, at the cost of massive stuttering and warping in both Battle of Stalingrad and IL-2 1946.

 

Cheers!

Mike

[[CPT]CptJackSparrow]

Wow, I'll leave mine alone then.

[SCG_motoadve]

I did this with MSI afterburner once on my GTX 970, max power target for that card is 111%, max temp I've set to 89°C (that's what I got is the max safe temp for the Maxwell Chips).

The result was a 10% performance increase in 3D Mark, at the cost of massive stuttering and warping in both Battle of Stalingrad and IL-2 1946.

 

Cheers!

Mike

Same here, and actually lower FPS when overclocking, that was with my old PC i7 4770 and DDR3.

 

New PC i7 8700 , DDR 4, seems to like overcloking on the GPU, more FPS and smoother.

 

So its worth a try to see if your systems likes it or not.

[SAS_Storebror]

Apparently it depends on how much the CPU or any other component is the limiting factor already.

Yesterday I've been wondering why my framerates dropped down to 30 for instance, just to figure that I've hit the old ScaleForm API bug again. Turn off HUD and FPS are back to 90. Funny thing, especially since it only happens when there are AI planes on the map, without them even with HUD turned on the FPS was 80+. But that's just a sidenote to demonstrate how many things can influence the frame rate, stability etc.

 

Cheers!

Mike

[dburne]

To clarify gang,

 

I was not talking about overclocking the GPU or Mem, only increasing the headroom for the power and temp. Power and Temp Target is kind of misleading, it does not automatically throw additional power to the card, only allows more power to be used if it is already hitting the default threshold of 100%, rather than down-clocking because it hit this threshold.

 

With my power limit set to 117%, my card never hits that limit. But it was at default 100%, which gives me a much more constant GPU clock with Nvidia Boost. It will still down-clock slightly depending on GPU temp ( my temps though never really exceed 70c), so it stays pretty constant at full boost.

And again a lot also depends on how the cpu is feeding the gpu.

 

As with most things, your mileage may vary.

 

[SAS_Storebror]

Yes indeed, changing the power target doesn't necessarily change the clock rates, at least as long as the target is (usually) not being hit.

In my particular case it was definitely not hit, still the massive stutters occured.

One more thing I've noticed and which can affect frame rates and micro-stutters is hardware monitoring software.

For instance there's a couple of users reporting that Nvidia Drivers younger than 388.0 have a tendency to show micro stutters when a hardware monitoring tool like HWInfo, XOC or the like monitors the "Power" sensor.

I've made the test and disabled all sensors not plotted in Rainmeter (which means that I have disabled the GPU Power sensors of my 970 as well, two in total) and it feels a tad "fluffier" - hard to back up with numbers 

 

Cheers!

Mike

[dburne]

I am not really sure why, maybe luck of the draw, but mine has always been a very smooth experience.

I have been running Precision XOC since I got my 1080 Ti, and ran Afterburner for years prior to that. Always setting my power target to max.

[Guest deleted@134347]

i got a 1080ti that was aircooled out of the box. With my horrible PC case I had it running at 80C, which in its turn introduced a very nasty thermal throttling. In my subjective opinion the temperature of NVidia cards has a lot more to do with the performance than anything else such as power or overclocking. If power is maxed out and OC is maxed and the card is stable then you won't get any microstuttering if the temperature is below the thermal throttling gate. What that thermal throttling gate is different from card to card.

 

So I just bought a custom water cooling plate for it, hooked it up to an isolated loop and now the temp is 44C at 150% power limit and some crazy OC'ed stuff to 2050mhz gpu for hours on end. No microstuttering...

 

just my 2 cents.