1/ New post: Jolt now proves RISC-V programs with 64-bit registers (RV64IMAC), at speeds exceeding those we previously reported for 32-bit. 1.5M cycles/sec on a 32-core CPU, 500k cycles/sec on a MacBook. Here’s why this matters 🧵

2/ Why care about 64-bit proving? Two reasons: 🔹 Performance: many apps need fewer cycles with 64-bit regs. Example: a 256-bit multiply takes 6× fewer cycles. 🔹 Language support: some languages (like Go) don’t compile to RV32, so RV64 is needed to prove programs like Geth

3/ So how did Jolt hit RV64 speeds that match our RV32 results? Two reasons: 1️⃣ Jolt is built for large registers. Other zkVMs split 64-bit values into bytes (8 field elements per value), a huge overhead. Jolt’s large-characteristic fields handle them directly.

4/ As a result, when other zkVMs double register size, they’re often 2× slower per cycle. Jolt is only 1.5× slower.

5/ 2️⃣ We keep optimizing. Alongside known techniques, we’ve been adding new, research-level advances. These gains already more than offset the 1.5× slowdown, so RV64 proving now runs faster than RV32 did two months ago.

6/ Memory usage keeps dropping too, now at ~700 MB per million cycles. Today, in <2 GB, you can prove knowledge of many digital signatures or that dozens of KB of data were correctly SHA-hashed.

7/ With Twist & Shout integrated, we’re on a steady milestone cadence. Next up: – Recursion – The streaming prover (keep prover space under 2 GB for arbitrarily large cycle counts, w/o recursion) – Zero-knowledge