Afterburner logs a "1" in the Power Limit column when the card hits it.
My guess is that the manufacturers play games with the power circuits to
differentiate the cards for marketing purposes.
56RAF_phoenix
On 10/04/2021 20:41, Joe Stronach (Redacted sender jstronach for DMARC)
wrote:
Power limit huh...Well that's interesting, and not something I'd previously thought about. Are you going from Afterburner's reading of the same Phoenix?
My 3080 registers 100% usage with several other applications, so why would power limit be hit in Il2 before 100% utilisation but not in other games or benchmarks I wonder? Those other applications are not VR, which is the most obviate difference I guess.
Roke
Get Outlook for Android <https://aka.ms/ghei36>
------------------------------------------------------------------------
*From:* 56raf_firebirds-bounce@xxxxxxxxxxxxx <56raf_firebirds-bounce@xxxxxxxxxxxxx> on behalf of 56RAF_phoenix <phoenix@xxxxxxxxxxxxxxxxxxxx>
*Sent:* Saturday, April 10, 2021 8:10:41 PM
*To:* 56raf_firebirds@xxxxxxxxxxxxx <56raf_firebirds@xxxxxxxxxxxxx>
*Subject:* [56raf_firebirds] Re: Benchmarking, was Drivers for VR
I meant "triangles" rather than "diamonds".
I had hoped that the GPU was being under used and I could rack up something like FXAA or "sharpen" to get better spotting. But it looks like I'm stuck at my current settings.
However, SYN_Vander's track is on the Veliky Luki map with all those trees, so 65 /s Framerate there is 90 elsewhere.
56RAF_phoenix
On 10/04/2021 19:50, 56RAF_phoenix wrote:
So Roke, using the SYN_Vander benchmark mission, it seems that even when the GPU is at 60-80% load, it's actually being limited by the power draw. In the attached graph, I've plotted Frame Rate, GPU Usage with the Power Limit being indicated by a diamond, and the CPU Usage (though I'm using Hyperthreading, so 50% means 100% of the hardware).
So contrary to my earlier thoughts, this track is GPU limited with my current settings of no shadows, medium clouds, FXAA=2 and lower than 100% "resolution" on my G2.
The SYN_Vander track is a pretty hard test, close to the ground.
56RAF_phoenix
On 06/04/2021 10:06, Joe Stronach (Redacted sender jstronach for DMARC) wrote:Yup, I was talking to Bo about this on Sunday. My 3080 never gets above 85% usage according to Afterburner in Il2 VR, and it doesn't always run at 85% either. That's true even when settings are cranked right up.
Roke
Get Outlook for Android <https://aka.ms/ghei36>
------------------------------------------------------------------------
*From:* 56raf_firebirds-bounce@xxxxxxxxxxxxx <mailto:56raf_firebirds-bounce@xxxxxxxxxxxxx> <56raf_firebirds-bounce@xxxxxxxxxxxxx> <mailto:56raf_firebirds-bounce@xxxxxxxxxxxxx> on behalf of 56RAF_phoenix <phoenix@xxxxxxxxxxxxxxxxxxxx> <mailto:phoenix@xxxxxxxxxxxxxxxxxxxx>
*Sent:* Monday, April 5, 2021 10:17:51 PM
*To:* 56raf_firebirds@xxxxxxxxxxxxx <mailto:56raf_firebirds@xxxxxxxxxxxxx> <56raf_firebirds@xxxxxxxxxxxxx> <mailto:56raf_firebirds@xxxxxxxxxxxxx>
*Subject:* [56raf_firebirds] Benchmarking, was Drivers for VR
I've finally just begun collecting data to try and optimise my rig for il2.
So far, I've been dialing down the "resolution" below the 100% that I believe gives native resolution /at the centre of the field of view,/ and dialing up FXAA to get better apparent sharpness further out from the centre and increased framerates with my G2. But it's been a seat-of-the-pants operation, rather than systematic.
I've flashed my BIOS ready for BAR, currently running the 457.51 NVidia GPU driver and I'll try the 457.30 and the latest with BAR.
I've just started collecting numbers using SYN_Vander's benchmark. I'm not at all sure it's representative of a multiplayer workload.
The really, really, shocking thing is how little the 3080 GPU is being used. I've yet to figure out how to log CPU usage of il2 at the same time, there are temperatures galore, but not much else. I'm beginning to strongly suspect that the real dips in framerate are when new objects come into the field of view and the CPU has to shovel data at the GPU. So BAR may help.
Watch this space.
56RAF_phoenix
On 01/04/2021 09:20, Colin (Redacted sender bart_56 for DMARC) wrote:
Defiantly seems smoother using this driver :).
BAR has been enabled in the latest nvidia driver, might need a bios update on your motherboard/gpu.
https://www.nvidia.com/en-gb/geforce/news/geforce-rtx-30-series-resizable-bar-support/ <https://www.nvidia.com/en-gb/geforce/news/geforce-rtx-30-series-resizable-bar-support/>
Interesting to see if that improves speeds.
Bart
