posted by Tensorwave 3 months ago
$2.95 / gpu / hour - 576 MI355Xs available from 04/01/2026 to 04/01/2027 (North America, multiple locations)
Unverified
Interconnect: RoCE v2
Cores per node: 8192
RAM per node: 2304 GB
NVME storage per node: 0
Location: North America, multiple locations
Cloud provider: Tensorwave
Cluster interface: K8, Slurm, Bare Metal
Minimium bookable GPUs: 32
Minimium bookable duration: 52 weeks
Additional details
1.5-2x greater inference throughput and 40-60% cost reduction on AMD's latest chips vs equivalent NVIDIA. Contact for benchmarking and free POC.
Can scale up into thousands of GPUs as needed. Private, dedicated, isolated, secure - No noisy neighbors!
contact seller
seller email domain is tensorwave.com