Tenstorrent Blackhole p100a
Overview
The Tenstorrent Blackhole p100a is the entry card in the Blackhole line: 120 Tensix cores and 16 "big" RISC-V cores delivering 664 TFLOPS of Block FP8, with 180 MB of on-die SRAM and 28 GB of GDDR6 at 448 GB/s, on a 300 W actively cooled PCIe 5.0 x16 board. It is the one Blackhole card with no QSFP-DD networking, so it scales within a host but not between them. Tenstorrent publishes a single compute figure for this card and labels nothing dense or sparse, so only Block FP8 is recorded here and no sparse value is derived. Tenstorrent's architecture is unusual in that every Tensix core is a full RISC-V compute tile rather than a fixed-function matrix unit, and the whole stack down to the low-level kernel API is open source.
Performance Metrics
Peak theoretical throughput by precision type
| Precision | Bits | Peak TFLOPS | Efficiency |
|---|---|---|---|
| FP8 | 8 | 664.0 | 2.213 TFLOPS/W |
Power Specifications
TDP
300 W
Max Power
345 W
Power Connector
PCIe 8-pin
Cooling
Active
Memory Specifications
Capacity
28 GB
Type
GDDR6
Bandwidth
448 GB/s
Interface
--
Hardware & Design
Form Factor
PCIe
Architecture
Blackhole
Process Node
--
Launch Year
--
Variant
p100a
Market Segment
Data Center
Full Specifications
| Compute Engine | |
|---|---|
| Tensix | 120 |
| Memory | |
| VRAM | 28 GB |
| Memory Type | GDDR6 |
| Bandwidth | 448 GB/s |
| On-Die SRAM | 180 MB |
| Interconnect & I/O | |
| PCIe | 5.0 x16 |
| Power & Thermal | |
| TDP | 300 W |
| Cooling | Active |
| General | |
| Form Factor | PCIe |
| Architecture | Blackhole |
Documentation & Resources
Common Use Cases
The Tenstorrent Blackhole is optimized for high-performance computing tasks with Blackhole architecture delivering high TFLOPS of compute power.
Where to Rent
Compare cloud providers offering on-demand GPU instances for AI training, inference, and HPC workloads.
Browse GPU Cloud ProvidersSimilar GPUs
Stay Updated on GPU Releases
Get notified when new GPUs are added or specifications are updated.
No spam, unsubscribe anytime.