Specifications
PNY NVIDIA A100 40GB Graphics Card (TCSA100M-PB) – CUDA Cores: 6912, Streaming Multiprocessors: 108, Tensor Cores | Gen 3: 432, Multi-Instance GPU (MIG) Support: Yes, up to seven instances per GPU, NVLink: 2-Way Low Profile, 2-Slot, NVLink Interconnect: 600 GB/s Bidirectional, GPU Memory: 40GB HBM2e, Memory Interface: 5120-bit, Memory Bandwidth: 1555 GB/s, System Interface: PCIe 4.0 x16, Thermal Solution: Passive, vGPU Support: NVIDIA Virtual Compute Server with MIG support, NEBS Ready: Level 3, Power Connector: 8-pin CPU, Maximum Power Consumption: 250 W
What's in the Box
PNY NVIDIA A100 40GB Graphics Card (TCSA100M-PB), Documentation
Warranty
3 Year Warranty
ArchitectureĀ Ā Ā Ā Ā
Ampere
Process SizeĀ Ā Ā Ā Ā Ā
7nm | TSMC
TransistorsĀ Ā Ā Ā Ā Ā Ā Ā Ā
54 Billion
Die Size
826 mm2
CUDA CoresĀ Ā Ā Ā Ā Ā
6912
Streaming Multiprocessors
108
Tensor Cores | Gen 3Ā Ā Ā Ā
432
Multi-Instance GPU (MIG) SupportĀ Ā Ā Ā Ā Ā Ā Ā Ā
Yes, up to seven instances per GPU
FP64Ā Ā Ā Ā Ā
9.7 TFLOPS
FP64 Tensor CoreĀ Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā
19.5 TFLOPS
FP32Ā Ā Ā Ā Ā
19.5 TFLOPS
TF32 Tensor CoreĀ Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā
156 TFLOPS | 312 TFLOPS*
BFLOAT16 Tensor CoreĀ
312 TFLOPS | 624 TFLOPS*
FP16 Tensor CoreĀ Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā
312 TFLOPS | 624 TFLOPS*
INT8 Tensor CoreĀ Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā
624 TOPS | 1248 TOPS*
INT4 Tensor CoreĀ Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā
1248 TOPS | 2496 TOPS*
NVLink 2-Way Low Profile,
2-Slot
NVLink InterconnectĀ Ā Ā Ā Ā
600 GB/s Bidirectional
GPU MemoryĀ Ā Ā
40 GB HBM2e
Memory InterfaceĀ Ā Ā Ā Ā Ā Ā Ā Ā Ā
5120-bit
Memory BandwidthĀ Ā Ā Ā Ā Ā
1555 GB/s
System InterfaceĀ Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā
PCIe 4.0 x16
Thermal SolutionĀ Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā
Passive
vGPU SupportĀ Ā
NVIDIA Virtual Compute Server with MIG support
Secure and Measured Boot Hardware Root of TrustĀ Ā Ā Ā Ā Ā Ā
CEC 1712
NEBS ReadyĀ Ā Ā Ā Ā Ā Ā
Level 3
Power ConnectorĀ Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā
8-pin CPU
Maximum Power Consumption
250 W
There are no reviews yet.