Questions we actually get

The Nexa/BCH Schnorr variant over secp256k1, as deployed by Nexa and Bitcoin Cash — not BIP-340 (Taproot). The exact accept/reject rules are pinned by the downloadable known-answer corpus on the K1 page, with verdicts derived from the reference implementation.

No. The card ships provisioned to run our encrypted production bitstreams, but it is never locked: you can load any bitstream you build — volatile over USB-JTAG, or written to flash — whenever you want.

No. Bitstreams you build can load over USB-JTAG into the FPGA’s volatile configuration memory — flash untouched, a power cycle returns the card to stock — or be written to the application flash slot like our own updates. Either way the frozen golden image and a JTAG port wired straight to the FPGA — independent of whatever design is loaded or flashed — mean even a completely broken bitstream can’t lock you out. Short of physical damage, there is always a way back.

Through the Blitz app, over PCIe: an updated production bitstream is written — encrypted end to end — to whichever of the flash’s two application slots is not currently running, and the card flips its current-application pointer only after the new image is confirmed to load. A failed update therefore just leaves the old bitstream in place. A small, frozen golden image sits in a separate slot as the last-resort fallback — it proves the board out and exposes the update port. USB-JTAG remains the development and last-resort path.

The first production run is a beta batch — near hardware cost, limited quantity, limited warranty (details on the Blitz page). There is no list to join: when the batch opens, pricing and a way to buy appear on this site.

Blitz is a standard M.2 2280 M-key card — the same size and socket as an NVMe SSD — and it needs one PCIe lane, which any NVMe-capable M.2 slot provides. Most desktop motherboards have a spare slot; many laptops do too. No spare M.2 slot? Inexpensive passive M.2-to-PCIe adapters put it in a regular PCIe slot. Rated maximum power is 8 W, within the normal envelope of the socket.

For light workloads you should — libsecp256k1 is excellent. Measured on the same scheme with the same library, one core of a desktop CPU (Intel Core i7-12700KF) does about 29,900 verifications per second; the four-core Blitz card sustains about 9.9 times that, in an M.2 slot, while your CPU stays free for everything else. Yes, multiplied across all twelve cores that CPU would edge out one card — but then the whole machine is doing nothing but verifying signatures, and nobody runs a node that way. And if the comparison is allowed more silicon, it cuts both ways: M.2 slots take more cards. That the cost is real is not our claim alone: signature verification is expensive enough that Bitcoin Core skips it for historical blocks by default (assumevalid) — the ecosystem’s most conservative software already treats checking every signature from genesis as too slow to insist on. The FPGA figure is the floor, not the ceiling: the same architecture in ASIC is the roadmap.

No. Blitz is a signature-verification accelerator and a general-purpose PCIe FPGA development platform. It verifies signatures; it does not ship with mining functionality.

The kernel driver exposes a device file: write() a 160-byte record, read() one verdict byte. Any number of processes at once — the driver multiplexes them onto the card. The runnable demos and driver source are on our GitLab.

Mostly — and the limits are deliberate: every metric on the K1 page carries the commit and date it came from; the known-answer corpus is downloadable in three formats; and an MIT-licensed open verification core runs the whole suite from a fresh clone with two commands. Throughput on our silicon you’ll be able to reproduce on your own card.