I started this compare a while back when the Arc SKUs were new and gave up because the performance of the Intel Discreet Graphics cards were garbage. I then had a lot of friends telling me that they “fixed” a lot of the issues with the Arc drivers and that I should give it another go. So I did! I got a list of the average frames on a bunch of games that I had installed on the framework using the 12GB 3060. I then dusted off the A770 box, removed all the nVidia drivers using DDU and did a fresh install of the newest intel drivers to test the 16GB A770. Well… bit of a spoiler alert here: there is no real noticeable improvement… you will still NEED to have a ReBAR enabled system or you will lose a LOT of frames. I have attached a screen shot of the spreadsheet I was using to collect the data but gave up part way through since it was clear that I was not going to get any good results (ignore the red, a friend was going to loan me an equivalent AMD card to test but did not). Some people have also mentioned that the issue could be that’s it’s an eGPU case and not in a desktop… well some of that “could” be since I am now testing a 3070 and barley getting any more frames than the 3060 was getting and not nearly as many as the 3060 or 3070 get in a much older generation intel desktop… but still… the Intel should on paper do as good as the 3060 right?
… considering selling the 3060 and the A770 now and buying a 4070ti instead… JK JK JK… $1,300 is what the average 4070TI will be going for in Canada and that is a total rip off!
I wonder what needs doing to enable reBAR on the FW or if it is even possible. Curious that mobile didn’t get reBAR but desktop did and that it isn’t part of the Thunderbolt spec. Intel seriously goofed since all 3 relevant parts are Intel designed.
Given that it’s a PCI feature I’m hoping it just doesn’t need to be explicitly part of the TB spec.
If someone feels like bricking a laptop, maybe they can try this?
I am not sure if Framework mainboard would accept unsigned BIOSes.
At least on Linux by using the same commands whether ReBar is enabled, I see it appearing.
I previously modified my Z97-Pro mainboard to enable ReBar (pinned as one of the working mainboards) and patched to shift the memory space to the 64-bit space so that I can use a much larger ReBar size. I am using it with the A770 now though I have a RX 6600M too.
But then, I had a dual BIOS mainboard which Framework doesn’t. ^^;
Now that I’ve got mine working, I’m eager to see how it stacks up compared to your earlier results, especially if it was on Windows, pre driver update that dropped last week.
I’m guessing that for an eGPU setup the vRAM is very important, as it helps to reduce the need for data transfers through the cable. That could explain your surprising results with the 3060 and the 3070.
@DannyT I can do Apex and CS:GO and potentially Far Cry depending on what game it was. I only own Far Cry 1-3. Was there any particular methodology used? I’d like my results to be as comparable as possible.
@Mapleleaf, I forgot to mention, I reinstalled the nVidia drivers again when testing with my buddies 8gb 3070 and it ended up preforming way better than my 12gb 3060. About 15-20 mire frames in almost anything.
@GhostLegion, My tests were rather simple, I did an uninstall of all other graphics drivers using DDU, then a fresh install of whatever driver I needed. Ran whatever game at the highest possible settings while still being playable to get a base line. Then repeated the unintall / install of other drivers and running the same games with the same settings.
Even with updated drivers / improvements, I don’t think the games will preform any different without ReBAR support.
My buddy who lent me the 3070 tried my A770 in an older desktop of his that wasn’t supposed to have ReBAR support but ended up supporting it anyways and he claimed similar frames to my 3060 results. So he did better than me… and that was older drivers now.
/j