Central Registry

The Compute Index

A clinical index of the world's most powerful AI accelerators. Search, filter, and inspect variants for technical and market compatibility.

Index Depth15 Variants

Architects1 Brands

New '24-250 Nodes

Accelerator Identity	Architecture	Cores	FP32 TFLOPS	VRAM	Bandwidth	Wattage	Launched
SOTA GB200NVL4	-	-	-	-	-GB/s	-W	???
SOTA GB200NVL72	-	-	-	-	-GB/s	-W	???
SOTA GB300NVL72	-	-	-	-	-GB/s	-W	???
SOTA H100NVL	NVIDIA Hopper™	16,896 (Per GPU)	-	94	3.9GB/s	-W	2022-01-01
SOTA H100NVL1	-	-	-	-	-GB/s	-W	???
SOTA H100PCIe	-	14,592	51 teraFLOPS	-	-GB/s	-W	???
SOTA H100SXM	NVIDIA Hopper™	16,896	-	80	3.35GB/s	-W	2022-01-01
A10080GB PCIe	-	6,912	19.5 TFLOPS	-	-GB/s	-W	???
A10080GB SXM	-	6,912	19.5 TFLOPS	-	-GB/s	-W	???
HGXB300	-	-	-	-	-GB/s	-W	???
HGXRubin NVL8	-	-	-	-	-GB/s	-W	???
L40SL40S	-	-	-	-	-GB/s	-W	???
L40SNVL	-	18,176	-	-	-GB/s	-W	2023
L40SPCIe Gen4 x16	-	18,176	-	-	-GB/s	-W	2023
RTX PRO 6000 Blackwell Server EditionPCIe Gen 5	-	-	-	-	-GB/s	-W	???

Registry Depth: 15 UnitsEnd-of-Stream