vs
EN
NVIDIA H20 141GB HBM3e vs Enflame L600
Side-by-side specifications, performance, and rental pricing for datacenter AI workloads.
H20: Hopper, 2025 L600: GCU-CARA 4, 2025
Memory
141 vs 144 GB
3 GB more for the L600
TDP
400 W vs 700 W
H20 draws 300 W less
NVIDIA H20 141GB HBM3e
Full specs →Architecture
Hopper
Year
2025
Memory
141 GB
TDP
400 W
EF
Enflame L600
Full specs →Architecture
GCU-CARA 4
Year
2025
Memory
144 GB
TDP
700 W
Specifications
H20 L600
EF
Architecture
HopperGCU-CARA 4
Launch Year
20252025
Form Factor
SXMOAM
Memory
141 GB144 GB
Memory Bandwidth
4.0 TB/s—
TDP
400 W700 W
Process Node
4nm—
Spec H20 L600
EF
Architecture HopperGCU-CARA 4
Launch Year 20252025
Form Factor SXMOAM
Memory 141 GB144 GB
Memory Bandwidth 4.0 TB/s—
TDP 400 W700 W
Process Node 4nm—
Performance (TFLOPS)
H20 L600
EF
FP64
1 TFLOPS No verified data available
FP32
44 TFLOPS No verified data available
TF32
74 TFLOPS
sparse not published
No verified data availableBF16
148 TFLOPS
sparse not published
No verified data availableFP16
148 TFLOPS
sparse not published
No verified data availableFP8
296 TFLOPS
592 TFLOPS sparse
No verified data availableFP6
No verified data available No verified data available
FP4
No verified data available No verified data available
INT8
296 TOPS
sparse not published
No verified data availablePrecision H20 L600
EF
FP64 1 TFLOPS No verified data available
FP32 44 TFLOPS No verified data available
TF32 74 TFLOPS
sparse not published
No verified data availableBF16 148 TFLOPS
sparse not published
No verified data availableFP16 148 TFLOPS
sparse not published
No verified data availableFP8 296 TFLOPS
592 TFLOPS sparse
No verified data availableFP6 No verified data available No verified data available
FP4 No verified data available No verified data available
INT8 296 TOPS
sparse not published
No verified data availableFLOPS by Precision
What actually differs
Both GPUs launched in 2025: the H20 on Hopper and the L600 on GCU-CARA 4.
Memory is 141 GB against 144 GB. More memory per GPU means larger models fit before you have to shard across cards, which often matters more than raw TFLOPS for inference.
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
Which has more memory, the H20 or the L600?
The L600 carries 144 GB of memory versus 141 GB for the H20.
How much power do the H20 and the L600 draw?
The H20 is rated at 400 W TDP and the L600 at 700 W. On FP32 throughput per watt, the H20 is the more efficient part.
Get Comparison Updates
New GPUs added weekly. Be the first to see how they compare.