Hi, I just wanted to share some info from my latest experiments with my old GPU GTX 980 TI. I have had a lot of issues from CTD’s to bad performance. The GPU is working at 100% all time even when navigating in the menus which is very strange.
Anyway, I installed MSI Afterburner and lowered the max temp for the GPU from 83 (!) to 75. That causes less power to be used by the GPU and should reduce performance but it seems to be the other way round. When lowering the temperature the thermal throttling is reduced and therefore the GPU is able to cope much better with the big load the sim is putting on it. This has caused the sim to be more stable with a much smoother experience.
Still I have had a few CTD’s but they seem to be related to heavy stress. Not sure what is happening but is it typically happening during bad weather. The sim starts to chop, freezes for a couple of seconds and then just dies. Maybe bad memory handling or bad multitasking logic, maybe some thread is doing something it shouldn’t do at certain times. I really have no idea, just guessing but there must be bugs in the sim popping up especially under heavy load. One typical scenario is when on short final maybe because the sim is loading in a lot of airport details.
I have been trying to get a better GPU since September! This delay makes me to postpone a lot of other purchases and from trying out VR. Anyway, until new hardware can be acquired I’ll continue my struggle with the old 980…
Just another observation… I lowered the max temp even further to 72 degrees Celsius using Afterburner and now the GPU utilization is about 75% when flying around and the fps went from about 30 to 22. Strange, with a higher max temp it is used at 100% all the time, lower temp decreases it to 75%??? At ground it goes up to 100% though. Seems like a mystery, you would need the source code to be able to explain that.
That seems spot on to me. When you lower the temps, the GPU cannot sustain full utilization as long until it has to throttle to lower the temps back down.
In most cases, CPU and GPUs will ramp up to 100% bc the temps are still low. As the temps increase at 100%, then the GPU has to scale down to keep the temps at what you set.
Think about it like a car engine. If it overheats, it needs time to cool down before ramping up again.
Ok thanks, will keep it at 75 then, it seems to be the sweet spot. Strange that you have to kind of underclock this thing to get it to work properly! It is also strange that the GPU is running at 100% when navigating the menus, I cannot understand why that should cause such a load on the GPU.
2 things on your last point. In the menus, you’re only driving the GPU. When playing, the CPU is in play, which means you still have the system competing bw CPU and GPU, so there’s a lot more overhead and context switching. Second, I keep reading that the Windows GPU utilization graphs are not always accurate. Take that with a grain of salt.
Did they fix something in the latest update? Not a singe CTD since then but I’m running the sim underclocked and right now I have about 14 fps (!) in the FlyByWire Neo. Better a bit flickering than CTD’s in the middle of everything. We will se if this has stabilized or not, time will tell.
but I’m running the sim underclocked and right now I have about 14 fps (!) in the FlyByWire Neo. Better a bit flickering than CTD’s in the middle of everything. We will se if this has stabilized or not, time will tell.