consume high % GPU "3D"

hhspiny

Member
not sure if this is unique again to my machine or if this is a concern.

intel iGPU+Nvidia dGPU on server. when connecting via remotly, task manager iGPU "3D" stays above 40% (40-60%). Encoding usage only a few %. this is the case regardless which GPU is assigned for encoding.
 

mirillis

Administrator
Staff member
Hi hhspiny,

Sorry for the late response. We are working on v.1.17.3 that will fix all recently reported issues and bring more improvements.

"3D" usage by iGPU is probably caused by RGB to NV12 conversion (or HDR RGB to NV12/P010 for HDR displays). This number can vary depending on current iGPU clock frequency (you might see higher number as the clock is sometimes reduced).

I have seen this on Nvidia too. But after checking Gpu-Z I found out that the GPU clock is very low.
 

hhspiny

Member
the high "3D" usage might be from dwm.exe and only happens within remotly session. (i.e. not with other software). with any mouse movement, it shoots up to 50-70%. idle about 20%. this is with 1.17.3. not sure if this is the cause of the sluggishness I feel. as if the computer is slower than I am used to and most GUI interactions take a tiny fraction of second longer than it should.
 

mirillis

Administrator
Staff member
the high "3D" usage might be from dwm.exe and only happens within remotly session. (i.e. not with other software). with any mouse movement, it shoots up to 50-70%. idle about 20%. this is with 1.17.3. not sure if this is the cause of the sluggishness I feel. as if the computer is slower than I am used to and most GUI interactions take a tiny fraction of second longer than it should.
Can you make the same test running Remotly.exe from the installation folder (starting without admin rights).

I think I know where this "sluggishness" feel might be coming from. As Remotly is very light on GPU the clock is running lower than it should.
I've seen this on Nvidia (not on Intel). After starting something that stresses GPU more the clocks are increased and it "feels better".

If you try the non-admin test Remotly will use the "standard" capture engine and not the low-level one. After starting the standard capture engine you should see less dwm.exe stress. GPU clocks are increased automatically in this case.

Please let me know how this looks on your side in non-admin mode.
 
Last edited:

mirillis

Administrator
Staff member
Another funny thing. On Window 8.1 with and old Nvidia GTX 745 and an old Core i5 CPU performance is better than on newer machines.
Everything runs smoothly (no sluggishness) and CPU runs at 0.8Ghz during connection in 1080p at 60fps 10Mbps.
No idea what is going on on newer PCs and new Windows but it is not as "green" as they are marketing it.
 

mirillis

Administrator
Staff member
Made some quick tests on this old PC.

1. Low-level capture mode 1080p@60Hz 10Mbps GPU clock runs at 320Mhz
1. Standard capture mode 1080p@60Hz 10Mbps GPU clock runs at 1050Mhz

I can feel that when the GPU clock runs at 320Mhz there is 1 frame more lag. After starting something that stresses the GPU and clock rises to 1050Mhz the lag is gone.

I guess being "green" is not good in all situations.
 

hhspiny

Member
was running remotly in user account. cpu-z does not show iGPU info. HWinfo shows but only intermittently, not sure if the # is right.
when connecting to admin instance -- "3D" is little, ~5%, ~230MHz up down
when connecting to user instance -- "3D" is ~50%, also ~230MHz up down
 

hhspiny

Member
still experiencing video play stuttering, incl. online streaming, every minute or so on the server machine (Intel iGPU & Nvidia) -- this is just with remotly running in user account, but not in remote session. issue disappears once exiting remotly. could this be related?
 

mirillis

Administrator
Staff member
still experiencing video play stuttering, incl. online streaming, every minute or so on the server machine (Intel iGPU & Nvidia) -- this is just with remotly running in user account, but not in remote session. issue disappears once exiting remotly. could this be related?
Is this happening only on this one specific machine? It might be related with this service. In the next update it will have lower priority and will be optimized by a factor of 4-5x in terms of CPU usage.
 
Top