Team,
Just finished the first part of a detailed deep dive on RX Vega 64 on testing various combinations of GPU/MEM/Power Limit settings on a range of coins. BBT Carter performed this on a livestream for everyone to see the process of trying multiple perturbations. If you have a RX Vega 64 or are thinking of getting one hopefully this information helps your decision making for your particular situation (card availability/cost/power situation). Thanks for participating in this discussion with your own results! We plan taking another few rounds with this card after we get a more reliable release of Afterburner and/or a update to the existing beta driver.
Would be interested in feedback if these figures match other RX Vega 64's, specifically the "limited edition" versions (silver shrouds). We will do the same comprehensive test with the RX Vega 56 when we can get our hands on one.
You say there was no way to know anything in advance, while ignoring the fact that the Vega Frontier Edition has the SAME GPU and SAME ram (but twice as much) and could give a pretty good indication of what the consumer version was going to be able to do.
You list "Digibyte" but you don't list WHICH ALGO you were testing it on - DGB has *5* different algos it can be mined with, 2 of which are GPU viable (Skein and Groetsl) the other 3 are ASIC dominated (SHA256, Scrypt, Qubit).
Given the listed hashrate, it *appears* you were testing the Groetsl option, but that's NOT in the article anywhere.
Confused by your post? The Vega Frontier Edition while leveraging the same GPU and Memory does not predicate how the Vega RX 64 is going to function/hash. Take for instance any Polaris RX predecessor. An ASUS ROG STRIX vs a MSI Reference Edition both would have the same physical memory and GPU ASIC, say for instance a RX480 8GB, but potentially asic quality, thermals, vrms included, memory 'type' and timing all are different. This results in noticeable differences in hash/performance. That is the point of taking a large sample of output then asking the community, in their experience and testing, do they see differences in 'like' cards within a series.
Lastly on your question of Digibyte, updating the article to include the option used, yes, it was Groetsl. The video shows the miner and output if people were really curious.