16:30:11
DataHoarder:
https://shop.bitmain.com/product/detail?pid=00020251226185131923DUjfvPIO06BC
16:30:55
DataHoarder:
> Shipping in July 2026
16:32:20
DataHoarder:
I wonder how many boards are stacked up here, 17kg
16:33:13
DataHoarder:
oh, same weight as their previous board stack, but 200 -> 1000 KH/s they advertise
16:47:45
nioc:
can't pay in monero lool
16:50:11
sech1:
So they claim 400 h/s/watt basically
16:50:14
sech1:
Interesting
16:50:31
sech1:
Best tuned EPYC systems get 200-250 h/s/watt
16:50:48
kico:
let's fork!
16:51:08
sech1:
My plan is to finish RandomX V2 in Q1 2026 :P
16:51:14
kico:
<3
16:51:25
DataHoarder:
FCMP++ already has the commitment stuff in the pipeline
16:51:26
sech1:
RandomX v2 is missing the RISC-V code, but this is exactly what I've been working on in December
16:51:28
kico:
thanks for doing the needful sir _/\_
16:51:29
DataHoarder:
V2 probably if ready
16:51:43
DataHoarder:
V2 is "minor" overall, the AES part + rounding changes
16:51:54
sech1:
There might be some other changes ;)
16:51:58
DataHoarder:
;)
16:52:00
sech1:
It's not set in stone yet
16:52:02
kico:
noice
16:52:18
kico:
would be fun to brick these inb4 release
16:52:19
sech1:
I think it should be tuned for Zen4/Zen5/Zen6 CPUs
16:52:22
sech1:
*tuned more
16:52:51
sech1:
As it's the main source of RandomX hashrate now, and ZenX CPUs will dominate the CPU market in the next 5 years or so
16:52:56
DataHoarder:
AVX512 128-bit specific masked operands :)
16:53:10
DataHoarder:
so 128-bit conditional lanes, which also exist in other instruction sets
16:53:30
DataHoarder:
kico: if they are some RISC-V tbh they are "cpu"s so unless they cheap out on what they include it can probably adapt
16:54:09
kico:
DataHoarder, yeah I mean a fork could at least make them less profitable
16:54:23
kico:
"brick" was not the best word
16:54:46
kico:
they could need a FW update to adapt no?
16:54:46
DataHoarder:
I mean, if they are after all just plain old CPUs just more power efficient
16:55:07
kico:
sure
16:55:17
DataHoarder:
purpose of randomx about "someone implementing this just needs to make a more efficient CPU than existing"
16:55:38
DataHoarder:
no ssh needed in a performant way
16:55:46
DataHoarder:
they can cheap out there
16:56:06
DataHoarder:
they can kill all their branch prediction to reduce consumption and silicon
16:56:48
kico:
ic
17:00:07
kico:
Operating altitude(2-5), m ≤2000
17:00:13
kico:
no mining in tibet!
17:00:16
kico:
ah!
17:00:31
kico:
based jihan
21:12:36
sech1:
This PR completes XMRig's RISC-V support https://github.com/xmrig/xmrig/pull/3749
21:12:41
sech1:
Nothing more to do there
21:12:49
sech1:
Now I can focus on RandomX v2 :)
21:13:13
sech1:
Oh, but first I need to move all new RISC-V code to the upstream repo
21:13:24
sech1:
In January probably
21:18:56
sech1:
https://github.com/tevador/RandomX/pull/274#issuecomment-3697565152
21:54:35
sech1:
https://www.reddit.com/r/Monero/comments/1pyzdpc/randomx_v2_update/