Side-by-side, with the numbers that matter.

Two rigs, one table. Higher numbers win in cobalt; TDP and price flip.

Rig A

SambaNova · accelerator

Change

Rig B

NVIDIA · workstation

Change

0 of 31 fields comparable

Spec	SambaNova SN40L Reconfigurable Dataflow Unit	NVIDIA RTX PRO 6000 Blackwell
AI Tops	—	4,000
Bandwidth GB S	—	1,792
Bf16 Tflops PER Socket	638	—
Cuda Cores	—	24,064
DDR Bandwidth GB S	200	—
Ddr5 Capacity TB	1.5	—
Displayport Count	—	4
Displayport Version	—	2.1
ECC	—	true
Form factor	—	5.4 in H x 12.0 in L, dual slot
Hbm3 Capacity GB	64	—
Memory BUS Bits	—	512
Memory Speed Gbps	—	28
Nvdec Count	—	4
Nvdec GEN	—	6
Nvenc Count	—	4
Nvenc GEN	—	9
ON Chip Sram MB	520	—
Packaging	3x CoWoS-S, 2 accelerator dies	—
Pattern Compute Units	1,040	—
Pcie GEN	—	5
Process Node	TSMC 5nm	—
RT Cores	—	188
RT Cores GEN	—	4
Slot Width	—	2
Streaming Multiprocessors	—	188
TDP (W)↓ better	—	600
Tensor Cores	—	752
Tensor Cores GEN	—	5
VRAM (GB)	—	96
Vram Type	—	GDDR7 ECC