Tenstorrent Blackhole p150
Datacenter
VRAM
32 GB
Memory
GDDR6
Bandwidth
512 GB/s
TDP
300W
Suitable Workloads
Smaller Language Models
Inference for 7B-13B parameter models
Enterprise Deployment
Designed for 24/7 datacenter operations
Key Highlights
- Built on Tenstorrent Blackhole architecture
- 120 compute cores for parallel processing
- PCIe form factor
Memory
VRAM32 GB
Memory TypeGDDR6
Memory Bandwidth512 GB/s
Compute Performance
FP16 (Half Precision)332 TFLOPS
BF16332 TFLOPS
Architecture
ArchitectureTenstorrent Blackhole
Compute Cores120
Release Year2024
Power & Physical
TDP300 W
Form FactorPCIe
PCIe GenerationPCIe 5.0
Interconnect
Multi-GPU SupportYes
Notes
Open-source stack (TT-Metalium). 664 TFLOPS BLOCKFP8 / 332 BF16 per Tenstorrent; 120 Tensix cores (cards downgraded from 140 via firmware, Jan 2026). 4x QSFP-DD 800G ports pool memory across cards. $1,399 active or passive.
Models That May Fit on Tenstorrent Blackhole p150
Estimates based on INT8 quantization. Actual fit depends on framework and batch size.
Explore More
Added Jul 30, 2026
Last updated: Jul 30, 2026
Your AI stack shouldn't stand still.
Every month new models become cheaper, faster, and more capable. Inferbase ensures your application automatically benefits without changing a single API call.