-
gingeropolousso motherboard manufactures could get into the race as well. I'm sure there are plenty of things on a mobo that could be cut out to make a randomx-mining-mobo
-
gingeropolousalso realized we can get into some fun semantics if "asics" ever do get created for randomx, and there exists a crowd that thinks that because asics exist, we should fork to more asic-friendly pow....
-
gingeropolousand we get into discussions.... it may be impossible to prove that a randomx asics exists, because it will probably technically just be a general purpose computer
-
gingeropolousso that could devolve into "well, its specialized mining hardware"
-
gingeropoloussomehow ended up here: venraytechnology.com/Papers/economics_of_cpu_in_DRAM2.pdf
-
gingeropolousooooh, i wonder if this is still active: venraytechnology.com/Test_Drive.htm
-
hycweb page copyright 2017, might still be active
-
hyconly a 32bit core though, not too interesting
-
hycthe overall approach is interesting
-
hycthese guys are similar upmem.com
-
hycthe venray twitter profile reads more like they're a patent troll house, not a serious electronics vendor
-
sech1
-
hycmust be about at the max memory bandwidth by now
-
cohcho600 * 32, beautiful
-
sech129 threads there
-
cohchoooops, I didn't notice
-
Inge-I've been hearing about maxing memory bandwidth for a long time now :)
-
sech119.2 kh/s is far from maximum bandwidth
-
sech1It's 19.2 GB/s
-
Inge-well. bandwidth, latency.. maxing out the memory subsystem with RX ..
-
Inge-20.9KH/s (unstable) on 3950x