AMD Ryzen Threadripper PRO 7965WX (8-channel DDR5)
Compare AMD Ryzen Threadripper PRO 7965WX (8-channel DDR5) against another machine in the Arena
Reference: Threadripper on Wikipedia
- Class
- CPU
- Memory
- 256 GB DDR5 (8 channel, 8x32GB Kingston Fury KF560R32-32)
- Runs (Q4_K_M)
- —
- Bandwidth
- 358.4 GB/s
- TDP
- 350 W
- Released
- 2023-10-19
- Price (US)
- $9850 new as of 2026-08
Sourcing and disambiguation notes
24-core/48-thread. Memory speed inferred at DDR5-5600 from the Kingston Fury "KF560" part-number prefix (Kingston's naming encodes speed), not independently confirmed by the source; bandwidth computed as 8 channels x 5600 MT/s x 8 bytes. Threadripper PRO 7000WX supports up to DDR5-5200/8ch per AMD spec, so this may be an XMP/overclocked configuration - flagged as uncertain. 2026-08 build-total price (avadirect.com + retailer cross-checks): sTR5/WRX90, 8x DIMM. Breakdown: CPU Threadripper PRO 7965WX (24c) ~$2,875, ASUS Pro WS WRX90E-SAGE SE board ~$1,300, 8x32GB DDR5 ECC RDIMM ~$708.55 each = $5,668. Memory is again the dominant line (roughly 58% of this total) -- see the EPYC 9554 entry's note on the $520-$2,908/module DDR5 RDIMM spread; check it first. No used price: only one eBay listing found for a used 7965WX, an untested 'parts only' unit at $999.99 -- not a usable-condition data point, so this workstation part is treated as having no real secondary market yet.
Specification sources
- https://forum.level1techs.com/t/deepseek-deep-dive-r1-at-home/225826 — forum, 2025-02-01
Measurements
Decode is token generation — the speed you feel while an answer streams. Prefill is prompt processing — the wait before it starts. Why bandwidth predicts decode speed.
No records yet for this hardware. Know of a published benchmark on it? Submit the link.
What verified, single-source and estimated mean, and the same rows with every filter and sort in the benchmarks explorer.
Speed over time
What it can run
Fit is arithmetic, not a measurement — how it is computed. Against 256 GB; models that fit are listed largest first.
No modeled quant fits in 256 GB.