Graphcore Bow IPU vs Graphcore C600
The Bow IPU delivers 25% more FP16 throughput than the C600 (350 vs 280 TFLOPS dense).
Bow IPU: Colossus MK2, 2022 C600: Colossus MK2, 2022
FP16 dense lead
+25%
Bow IPU: 350 vs 280 TFLOPS
Graphcore Bow IPU
Full specs →Architecture
Colossus MK2
Year
2022
Graphcore C600
Full specs →Architecture
Colossus MK2
Year
2022
TDP
185 W
Specifications
Architecture
Colossus MK2Colossus MK2
Launch Year
20222022
Form Factor
—PCIe
Memory
——
Memory Bandwidth
——
TDP
—185 W
Process Node
——
Spec Bow IPU C600
Architecture Colossus MK2Colossus MK2
Launch Year 20222022
Form Factor —PCIe
Memory ——
Memory Bandwidth ——
TDP —185 W
Process Node ——
Performance (TFLOPS)
FP64
No verified data available No verified data available
FP32
87 TFLOPS No verified data available
TF32
No verified data available No verified data available
BF16
No verified data available No verified data available
FP16
350 TFLOPS 280 TFLOPS
FP8
No verified data available560 TFLOPS
FP6
No verified data available No verified data available
FP4
No verified data available No verified data available
INT8
No verified data available No verified data available
Precision Bow IPU C600
FP64 No verified data available No verified data available
FP32 87 TFLOPS No verified data available
TF32 No verified data available No verified data available
BF16 No verified data available No verified data available
FP16 350 TFLOPS 280 TFLOPS
FP8 No verified data available560 TFLOPS
FP6 No verified data available No verified data available
FP4 No verified data available No verified data available
INT8 No verified data available No verified data available
FLOPS by Precision
What actually differs
Both GPUs launched in 2022: the Bow IPU on Colossus MK2 and the C600 on Colossus MK2.
Dense throughput for the Bow IPU against the C600: FP16 350 vs 280 TFLOPS. Sparse figures, where the vendor publishes them, appear under each dense number in the performance table.
For LLM training and inference, weight the FP8 and FP16 rows and memory capacity most heavily. For scientific and HPC workloads, the FP64 row is the one to read.
Frequently asked questions
Is the Bow IPU faster than the C600?
At FP16 precision the Bow IPU reaches 350 TFLOPS dense against 280 TFLOPS for the C600. The performance table on this page lists every published precision for both GPUs.
Get Comparison Updates
New GPUs added weekly. Be the first to see how they compare.