eBay used NVIDIA V100 32GB PCIe card (ex-datacenter pull)
various server liquidators / eBay sellers · United States (eBay.com / picclick)
Fields
id
supply/ebay-used-tesla-v100-pcie-32gbKind
used-market
Vendor
various server liquidators / eBay sellers
Region
United States (eBay.com / picclick)
Channels
marketplace
Price (USD)
750
Price basis
USED, bare passive dual-slot PCIe card, no server, no carrier needed (plain x16). $750 = midpoint of the $695-900 ask band captured 2026-10-03..14: HPE P05913-001 $694.99, HP PG500-216 $809-899, Dell pulls ~$758. ASKING price on live listings, NOT a sold/completed price - realised typically below ask. PERISHABLE-FIGURE STAMP: OBSERVED 2026-10-14. RE-VERIFY BY: 2027-01-14.
Availability
in-stock
Lead time source
structural-na
Lead time basis
lead_time_weeks is null because this is a secondary market priced per listing; availability is a snapshot of live listings, not a schedulable delivery date.
Export controlled
no
Export control regime
EAR 3A001 or EAR99 (Volta generation below 3A090.a TPP/density thresholds) + EAR 744.23 / 744.6 party screening
Export control note
Volta-generation part falls below the 3A090.a TPP/performance-density thresholds; the advanced-computing licence regime does not attach. General EAR plus destination and party-list screening still apply.
notes
The LOW-DRAMA V100 route and the sweet spot of the used-datacenter slice. The PCIe premium (~$300-350 over the bare SXM2 module at $349-800) buys the SXM2 carrier/baseboard problem solved - this drops into any x16 slot. Passive cooling needs an airflow mod for desktop use. Real advantage over Pascal: Volta first-gen tensor cores give a genuine FP16 path (~112 TF) so vLLM FP16 works. 32 GB HBM2 at ~900 GB/s fits a 32B Q4 or an 8B at any quant. SXM2 variant (nvidia-v100-sxm2-32gb) is cheaper per card but needs a carrier+heatsink - bare module is useless alone.
Notes
The LOW-DRAMA V100 route and the sweet spot of the used-datacenter slice. The PCIe premium (~$300-350 over the bare SXM2 module at $349-800) buys the SXM2 carrier/baseboard problem solved - this drops into any x16 slot. Passive cooling needs an airflow mod for desktop use. Real advantage over Pascal: Volta first-gen tensor cores give a genuine FP16 path (~112 TF) so vLLM FP16 works. 32 GB HBM2 at ~900 GB/s fits a 32B Q4 or an 8B at any quant. SXM2 variant (nvidia-v100-sxm2-32gb) is cheaper per card but needs a carrier+heatsink - bare module is useless alone.