← All hardware
Power-efficient inference chip: 512 TFLOPS FP8 at 150-200W
Pros
- Outstanding performance per watt
- Air-coolable inference option
- Strong LLM efficiency
- Independent NVIDIA alternative
Cons
- Small software ecosystem
- Inference-focused (not training)
- Limited availability
- Newer vendor risk
✓ Where it shines / best for
- Efficient LLM inference
- Power-constrained datacenters
- Enterprises seeking NVIDIA alternatives
- Korean AI infrastructure
✕ Not the best fit for
- Large-scale training
- CUDA workflows
- Edge
- Bleeding-edge frontier scale
Features
- ✓ Tensor Contraction Processor design
- ✓ 512 TFLOPS FP8 / 256 TFLOPS BF16
- ✓ 512 TOPS INT8
- ✓ 150-200W power envelope
- ✓ TSMC 5nm
- ✓ RNGD Server form factor
Pricing
| Plan | Price | Billing | Notes |
|---|---|---|---|
| RNGD card / server | Enterprise (contact) | one-time | Pricing on request |
Pricing verified from the official source. Prices change often — confirm on the vendor's site before buying.
Specifications
| use | Data center inference |
| power | 150-200W |
| memory | HBM3 |
| performance | 512 TFLOPS FP8 / 512 TOPS INT8 |
| architecture | Tensor Contraction Processor (TSMC 5nm) |
Sponsored
A full review is being generated for this product and will appear here shortly.