First, I acknowledge my system is very old and the processor is weak. I bought the A770 LE some months ago just because I really wanted it (it was being discontinued) and because I already planned a system upgrade by the end of the year.
My gaming rig comprise of a Xeon E5 1620v2 (slightly overclocked at 4.4 GHz via base clock), 16 GB of DDR3 ram running in quad channel at 1866 MHz and, now, an ARC A770 LE.
Before the Arc I had an RX570 4GB.
What I noticed is that the performance of the ARC gpu is much worse than the one of the RX570, specifically in Overwatch: while with the old RX570 I was able to lock on 144 fps without issues, with the A770 I see fps as low as 80 in some scenarios (usually when there are multiple characters on screen).
Also the usage of the CPU stays at around 50-60%, while for the GPU is 40-50%.
I really cannot think of an explanation for it: why would a much more powerful GPU perform worse on the same exact system and settings? Could it be a DX11 driver optimization issue? (in Cyberpunk, that is DX12, the ARC is performing as expected).
I was expecting to have at least the same performance while waiting to build the new system…
PS: thanks to the ReBarUefi project, I have Resizable BAR active.
Intel themselves said that running ARC on systems that don’t support resizable bar is not recommended and will not give optimal performance.
I have resizable BAR enabled tho…
Every PCIE 3.0 system has the hardware to support ReBAR/SAM, the only thing missing is the UEFI driver for it in the UEFI bios, luckly there’s a project that allows users of old platforms like me to add said driver to the bios and have ReBAR working.
I run ARC card before my new ryzen system on I7-9700k, i had Asus MB and they release BIOS that supports ReBar, it wasnt that great, it shows that is isnt supportet by Intel but it work at least.