NVIDIA Rubin CPX vs NVIDIA Rubin SXM
Side-by-side specifications, performance, and rental pricing for datacenter AI workloads.
Rubin CPX: Rubin, 2026 Rubin: Rubin, 2026
Memory
128 vs 288 GB
160 GB more for the Rubin
NVIDIA Rubin CPX
Full specs →Architecture
Rubin
Year
2026
Memory
128 GB
NVIDIA Rubin SXM
Full specs →Architecture
Rubin
Year
2026
Memory
288 GB
Specifications
Rubin CPX Rubin
Architecture
RubinRubin
Launch Year
20262026
Form Factor
—SXM
Memory
128 GB288 GB
Memory Bandwidth
—22.0 TB/s
TDP
——
Process Node
——
Spec Rubin CPX Rubin
Architecture RubinRubin
Launch Year 20262026
Form Factor —SXM
Memory 128 GB288 GB
Memory Bandwidth —22.0 TB/s
TDP ——
Process Node ——
Performance (TFLOPS)
Rubin CPX Rubin
FP64
No verified data available33 TFLOPS
FP32
No verified data available130 TFLOPS
TF32
No verified data available2,000 TFLOPS
sparse not published
BF16
No verified data available4,000 TFLOPS
sparse not published
FP16
No verified data available4,000 TFLOPS
sparse not published
FP8
No verified data available17,500 TFLOPS
sparse not published
FP6
No verified data available17,500 TFLOPS
sparse not published
FP4
No verified data available No verified data available
INT8
No verified data available250 TOPS
sparse not published
Tensor Core
FP64 (TC)
No verified data available200 TFLOPS
Precision Rubin CPX Rubin
FP64 No verified data available33 TFLOPS
FP32 No verified data available130 TFLOPS
TF32 No verified data available2,000 TFLOPS
sparse not published
BF16 No verified data available4,000 TFLOPS
sparse not published
FP16 No verified data available4,000 TFLOPS
sparse not published
FP8 No verified data available17,500 TFLOPS
sparse not published
FP6 No verified data available17,500 TFLOPS
sparse not published
FP4 No verified data available No verified data available
INT8 No verified data available250 TOPS
sparse not published
Tensor Core
FP64 (TC) No verified data available200 TFLOPS
FLOPS by Precision
What actually differs
Both GPUs launched in 2026: the Rubin CPX on Rubin and the Rubin on Rubin.
Memory is 128 GB against 288 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 Rubin CPX or the Rubin?
The Rubin carries 288 GB of memory versus 128 GB for the Rubin CPX.
Get Comparison Updates
New GPUs added weekly. Be the first to see how they compare.