11 Likes
Love the enthusiasm in the video, and Battlemage looks super promising!
Would love to see a transcode comparison with Arc and/or NVENC and/or VA-API for AMD; could be simultaneous transcode capabilities, speed per single transcode at various resolutions, post-transcode file size comparison… Especially Alchemist v Battlemage, I feel like a big selling point for Alchemist for early adopter was the transcode engine, so seeing how it compares would be awesome.
3 Likes
I wonder what the old system specs are since I was thinking of putting this gpu with my Ryzen 7 2700X which has 32gb of ddr4 3200 and a B450 motherboard. Watching this video putting this gpu with this older cpu might not be the best but still better than the RX 6600XT that is in that system now?
Also the B770 might or might not ever be released I keep hearing?
2 Likes
Good point lol I should include that. He said it in the video I think but:
10700K, Z590 Aorus Elite, 16GB DDR4-3600
2 Likes
I enjoyed the coverage of the card – i am incredibly optimistic, though i wouldnt mind a writeup/video on how it compares specifically to the 1080ti since the 1080/1080ti still has a fair number of holdouts.
additionally, i did see that the card supports double precision compute, so in theory the card can be used for folding at home. I would appreciate if someone could get some performance numbers when/if it becomes supported.
1 Like
Hi all, great review. I suspect the lower performance on the 10700k system was caused by the new B580 card being limited to a x8 PCIe connection.
Can you test the B580 on the 9800x3d limited to PCIe3.0 bandwidth?
I am looking at this card for an older system limited to PCIe 3.0
Thanks in advance!
1 Like
Are we going to get a level1linux video on this? Also do we have a watt/frame graph?
2 Likes
I think he plans on a linux video, and I don’t have a watt/frame graph. I’ve never done one of those… I should probably figure that out sometime lol
2 Likes
Obligatory question: SR-IOV? Yes/no? Maybe in the future? With a hacky workaround perhaps?
6 Likes
It would be a nice graph to see to measure efficiency if all else is equal on a system. I’m running a two gpu system doing vfio and looking glass so increasing efficiency would be nice.
I’m also looking forward for a linux video. Not only with gaming numbers, but also home lab use with LLMs, transcode and idle power draw on a headless system
1 Like
The direction is impressive and i hope it stays the course with celestial/ druid maybe with some higher offerings.
With the talk of how the intel driver works from the YT video it got me thinking.
What way does the driver lean?
now what do i mean by that? recall that hardware unboxed did some testing for AMD vs Nvidia that you touched on where in CPU constrained scenarios the AMD driver has much less overhead. With similar thinking i was seeing if anyone would think it would be interesting to see how the intel driver is (probably relatively to AMD/Nvidia).
i also think that the devils advocate of it is also interesting :: does the overhead even matter?
With the current positioning assuming intel and AMD towards the lower end and Nvidia towards the top of GPU perf pole, hypothetically speaking then only intel/AMD would have to worry about that as they would be more likely to be CPU bottle-necked (paired with a lower ranked CPU).
What are yalls thoughts on this? My ramblings might be mad, but at least im in good company 
(i am paraphrasing heavily and glossing over many details, just wanted to put something out there to see public response)
Did @wendell do any testing on older systems? For instance, I’ve got an i5-8600k that still runs BG3 or whatever GPU-limited and was considering this to upgrade that old rx580. But, all I see are vague warnings that for “optimum” performance we need resizable bar - we talking a 5% or more like a 50% hit for using an older motherboard??
And I guess getting into Linux drivers… I’m slightly skeptical. Phoronix put out an article today that makes me think they’ll be way better than the Alchemist launch, but still… an rx 6800 can be had for the same price - which’ll perform better given the state of open source drivers?
“Also the B770 might or might not ever be released I keep hearing” is not a question.
Ltt specified 1660’s and 10 series that have been holding out finally have something worth upgrading to that’s still on a budget.
Also want to express my gratitude for doing the benchmarks with an older system.
And want to also ask if you all would be willing to do some tests* with pcie 3.0. I have a system that I would potentially want to put a single slot battlemage care in, if one is ever released, and would want to see how it performs.
AND would like to ask about the pcie passthrough stuff. My above pcie 3.0 usecase is to have it in a dedicated vm for remote gaming at 1080p60.
Again, thanks for all the work on this. Really thought the points about using fsr with the b580 was something that none of the other reviews that I have watched mentioned at all.
- EDIT: more tests, sorry! For context, I am running a Xeon W Skylake system.
1 Like
I’ve seen some requests similar to this, I will try to make time one day soon 
2 Likes
I appreciate the acknowledgement and will hold nothing against you all if you don’t!
iirc they didnt actually compare the performance of the 1080ti. which is far above the 1060/1070. but ill have to give the video another look.
Update to clarify in the older system is a 10700k, which is limited to PCIe 3.0. The motherboard it was in, the Z590, supports PCIe 4 but the CPU does not.
So, these benchmarks are at PCIe 3.0 speeds.
4 Likes