Strange rendertime differences with remote/local RT GPU render

[PIXELBOX_SRO]

i am getting some really strange results when rendering over RTGPU
both computers have the same HW btw

1st test:
running animation with max opened on local comp but utilising remote graphics card and "use localhost unchecked" gives me time 3m38s

2nd test:
running animation with max opened on local comp and utilising only the local card gives me time 7m45s

3rd test:
running animation with ma opened on a local comp and utilising both cards gives me time 1m58s

these are some really weird results i have to say given the HW and systems are the same...even drivers
I was also surprised that the LC is being computed twice as seen on the remote server console ???

It looks like that the computer running MAX while computing is 50% slower!

any ideas here?

Cheers
M

[PIXELBOX_SRO]

so i guess this is normal since no one is not surprised about this???

[savage309]

Sorry, I didn't quite understood what is the setup.
How many machines with how many GPUs in each ..

Best,
Blago.

[PIXELBOX_SRO]

same machines Blago....each with one card GTX TITAN 6GB thats why its so weird.

i dont know if i explained it well but when i render the scene on my local computer and 1 local graphics only, it takes 7:45 minutes to render.
However, if i render it from my local computer BUT use for the computation the one card from the REMOTE computer, it renders at 3:38.

The difference in my opinion shows that the effect of having the 3DSMax open while rendering .....but over 50%??? htats a little too much.
Or it could be the infamous Win 10 problém reserving the VRAM....but then it shoudl be doing the same on the remote computer that has exactly the same specs...
Confused )

and then i tried rendergin locally but using both remote and local cards in DR mode and it rendered at 1:58

So its even more confusing...it doesnt add up or does it?

[savage309]

I doubt it is memory related and is something that hasn't been reported so far.
You can send the scene for us to check, but I am pretty sure it is something in the setup, not in the scene.

Are you testing with same noise level, rays per pixel, ray bundle size ? Mode (production / active shade) is the same during the tests ?

Best,
Blago.

[PIXELBOX_SRO]

I will do a video capture an send you it �� Have a nice weekend!