Dual Tesla V100 32 GB on EPYC 7402 (bnk engine)
| GPU(s) | 2x Tesla V100 32 GB (PG500-216, sm_70) Amazon · eBay (used) · Newegg |
|---|---|
| VRAM / unified memory (GB) | 64 |
| CPU | AMD EPYC 7402 (32 vCPUs) Amazon · eBay (used) · Newegg |
| Motherboard | |
| RAM | 128 GB DDR4 (192 GB for DeepSeek) Amazon · eBay (used) · Newegg |
| PSU | |
| Other | PCIe 3.0 x16; Ubuntu 24.04, driver 580, CUDA 12.8 |
| Reported total cost | |
| Power |
Measured speed (generation tokens/sec)
📏 How measured: reported by the original source; RigBase re-checked the numbers at that source
| Model | Quant | Engine | tok/s | Details |
|---|---|---|---|---|
| Qwen3.8 27B | Q8_0 | bnk dense engine, tensor parallel, no drafting | 42.4 | 📏 Maintainer decode tok/s from the bnk README |
| Qwen3.8 27B | Q8_0 | bnk dense engine, tensor parallel, MTP drafting | 104 | ⚡ spec-decode (MTP)📏 Maintainer decode tok/s from the bnk README |
| Qwen3.8 Flash-Next | FP4-Local | bnk v0.3.1, decode at 4K context | 114 | context: 4K📏 Maintainer decode tok/s from the bnk README |
| Qwen3.8 Flash-Next | IQ3_S | bnk v0.3.1, decode at 4K context | 121 | context: 4K📏 Maintainer decode tok/s from the bnk README |
| DeepSeek V4.1 Flash | GSQ-RCO 2-bit | bnk v0.5.0, experts split VRAM/RAM (CPU offload) | 13 | 📏 Maintainer decode tok/s from the bnk README |
All figures from the bnk README (maintainer's own machine). The DeepSeek V4.1 Flash run needs 192 GB system RAM and reads the Engram tables from NVMe; '~13 tok/s' in the README. Found via localmaxxing (discovery only); numbers re-checked at the original source on 2026-10-24.
Source: https://github.com/christ-pher/bnk
Blank fields = not stated in the source. We never estimate build data; GPU street prices above are separately sourced and dated.
Rent before you buy: try this model on a cloud GPU for a few dollars first: Runpod (referral link; sign up with Google to get the new-user bonus)
Quick answers: Can 2x Tesla V100 32GB run DeepSeek V4.1 Flash? · Can 2x Tesla V100 32GB run Qwen3.8 27B? · Can 2x Tesla V100 32GB run Qwen3.8 Flash-Next?