Supported cards and measured rates

NVIDIA from Turing to Blackwell, AMD RDNA2 to RDNA4. All GPU code is built into the program: you need only the normal graphics driver.

armadig-miner --list shows what would run on each of your cards. A card that is not supported is refused with a line that names the reason; the miner never silently runs a wrong kernel.

Pearl

FamilyCardsStatus
NVIDIA Turing (tensor cores)RTX 20xx, RTX 2060 12 GB, RTX 2060 SUPERsupported
NVIDIA Turing (no tensor cores)GTX 1660 SUPER, 1650, 1660, 1650 SUPERsupported, opt-in (--algo pearl)
NVIDIA Turing mining cardCMP 30HXsupported, low rate
NVIDIA AmpereRTX 30xx, RTX A4000supported
NVIDIA Ampere mining cardCMP 90HXonly with an unlocked driver (probed at start)
NVIDIA Ada / BlackwellRTX 40xx / RTX 50xxsupported
AMD RDNA4RX 9070 / 9070 XTsupported (Linux, Windows)
AMD RDNA3RX 7800 XTsupported (Linux; Windows not yet run)
AMD RDNA2RX 6700 XTsupported (Windows; Linux not yet run)

Quantus (QTC)

FamilyCardsStatus
NVIDIA PascalP104-100supported
NVIDIA TuringGTX 1660 SUPER, RTX 2060 SUPER, CMP 30HX, CMP 90HXsupported
NVIDIA AmpereRTX 3060, RTX A4000 and other sm_86 cardssupported
NVIDIA Ada / BlackwellRTX 40xx / RTX 50xxsupported
AMD RDNA4RX 9070 / 9070 XTsupported (Linux, Windows)
AMD RDNA3RX 7800 XTsupported (Linux; Windows not yet run)
AMD RDNA2RX 6700 XTsupported (Linux, Windows)

Cards of a supported family that we have not measured start from a generic point, and auto-tune then finds the card's own best settings. For the models we measured the program carries a tested starting point.

CMP 90HX: at start the miner measures the card's int8 tensor rate. An unlocked card mines Pearl; a locked card is refused for Pearl with a line that names the measured rate, and can still mine QTC.

Measured rates

Our own measurements on our own cards, with auto-tune on. These are compute rates (work finished divided by wall time), before the dev fee. Rates move with silicon, cooling, power limit and pool conditions; your card can be faster or slower.

CardPearlQTC
RTX 2060 SUPER50.36 TH/s236.37 MH/s
RTX 306053.33 TH/s232.45 MH/s
RTX 4060 Ti98.11 TH/s418.35 MH/s
RTX 5060 Ti107.47 TH/s460.84 MH/s
RTX A400066.53 TH/s309.7 MH/s
CMP 90HX82.46 TH/s374.35 MH/s
CMP 30HX-173.31 MH/s
GTX 1660 SUPER8.50 TH/s158.16 MH/s
RX 6700 XT-193.05 MH/s
RX 7800 XT33.94 TH/s270.13 MH/s
RX 9070 / 9070 XT95.63 TH/s247.50 MH/s

The pool credits a few percent less than the compute rate: see why the pool shows less.

← Config fileNext: FAQ →