we just open-sourced the entire akoya miner🦪
not a demo. the real thing, custom CUDA kernels RTX 2000 → B300, AMD/ROCm for MI300X, the full pool network stack.
one ./build.sh auto-detects your card and builds a ready-to-run binary. native windows too.
https://t.co/7mDbw6VyCW
we just open-sourced the entire akoya miner🦪
not a demo. the real thing, custom CUDA kernels RTX 2000 → B300, AMD/ROCm for MI300X, the full pool network stack.
one ./build.sh auto-detects your card and builds a ready-to-run binary. native windows too.
https://t.co/7mDbw6VyCW
we're also cutting fees to 2%
why give it away? a healthy network isn't one crew hoarding hashrate. it's independent miners and pools competing in the open. our kernels are public now, ask the other pools to do the same.
gl hf, go build something nice
New Akoya release is live.
H100s now hit 645TH (+5%)
5090s up to 300TH
4090s up to 235TH
30 series at full speed
L40S, A4000 and more now supported
Real speeds, tested on Vast rentals:
https://t.co/dlhTjojtpc
Akoya v2.0.1 is live.
– Fixed disconnects on high GPU count rigs
– Lower latency to pool
Docker, HiveOS, and Linux/WSL install instructions on the site.
https://t.co/7mDbw6VyCW
Akoya Pool V2 ⚒️
H100: 600-610 TH/s
5090: ~250 TH/s (CUDA 12.9, own kernel)
4090: ~210 TH/s
Load-balanced gateways, persistent sessions, one connection per miner. Faster and tougher than launch.
@rCrypto622 While running overnight tests for the V2 release, 2 and 1/3 blocks were mistakenly paid out to a single address. We caught it, fixed the bug, and added a guard so it can't happen again.
To make everyone whole, we covered the 2 blocks out of pool fees - payouts now match the data