Running Nemotron 3 Ultra on RTX PRO 6000 server
Nemotron 3 Ultra has 550 billion parameters. RTX PRO 6000 server offers 768 GB of GDDR7 across 8 GPUs memory. This page puts the two side by side.
How much memory does Nemotron 3 Ultra need?
Weights take the parameter count multiplied by the format size, plus a 15% runtime margin for activations and buffers. In FP8 the calculation reads: 550 billion parameters × 1 byte (FP8) = 550.0 GB of weights; × 1.15 runtime margin = 632.5 GB. The attention cache sits on top: it grows with context length and with the number of concurrent requests, so it is sized case by case.
| Format | Weights in memory | On this machine | What the format costs |
|---|---|---|---|
| FP16 | 1,265 GB | does not fit | full precision, the quality reference |
| FP8 | 632 GB | tight | negligible loss on most tasks |
| NVFP4 | 316 GB | fits | Blackwell format, native FP4 compute |
| Q4 | 379 GB | fits | block quantisation, broadly supported |
Model sheet: 550 billion parameters, 55 billion active per token, 262,144-token context, OpenMDW 1.1 licence. 550 billion announced; the BF16 repository holds 560 billion elements (1,121 GB). The card claims “up to 1M” context, config.json declares 262,144.
Nemotron 3 Ultra activates 55 billion parameters per token out of 550 billion. Active parameters govern speed, not footprint: every expert stays resident in memory. Confusing the two under-sizes the machine by an order of magnitude.
What is left to serve requests?
Nemotron 3 Ultra fits comfortably on RTX PRO 6000 server. In FP8, its weights take about 632 GB of the 768 GB available, leaving 136 GB for the attention cache and concurrency. The remaining margin decides how many concurrent requests and how much context the machine sustains. By QDNA convention a machine “fits” when the weights take at most 75% of its memory; beyond that it “barely fits” and forces a limit on context or on concurrency. The attention cache is not calculated on this page. These models’ architectures (compressed latent, linear or sparse attention, Mamba layers) share no per-token formula; it is measured on the machine, with the target context and concurrency [TO BE MEASURED].
RTX PRO 6000 server: price on quotation. No manufacturer publishes a price for an 8-card server; the card alone is listed at $16,000 on the NVIDIA marketplace as of 1 September 2026 (as reported by thundercompute.com). The machine targets the PME, GE segment.
Where else can Nemotron 3 Ultra run?
- Nemotron 3 Ultra on Mac Studio Ultra
- Nemotron 3 Ultra on DGX Station
- Nemotron 3 Ultra on H200 SXM server
- Nemotron 3 Ultra on B200 SXM
- Nemotron 3 Ultra on B300 SXM
See the Nemotron 3 Ultra and RTX PRO 6000 server fact sheets.
How much memory does Nemotron 3 Ultra need?
In FP8, weights take about 632 GB including the runtime margin. The attention cache sits on top and depends on context.
Does Nemotron 3 Ultra fit on RTX PRO 6000 server?
Nemotron 3 Ultra fits comfortably on RTX PRO 6000 server. In FP8, its weights take about 632 GB of the 768 GB available, leaving 136 GB for the attention cache and concurrency.
Method and sources
Memory footprints are calculated, not measured: a reading on real hardware may differ depending on the engine and the exact weight format. Every figure below carries its source and the date of its last check.
- Nemotron 3 Ultra: 550 billion parameters, 55 billion active, 262,144-token context, OpenMDW 1.1 licence. Hugging Face repository (config.json, API, model card), checked on 2 September 2026. 550 billion announced; the BF16 repository holds 560 billion elements (1,121 GB). The card claims “up to 1M” context, config.json declares 262,144.
- RTX PRO 6000 server: 768 GB of memory; 96 GB GDDR7 and 1,597 GB/s per card, 600 W, hence 768 GB for 8 cards. Manufacturer sheet checked on 2 September 2026.
- RTX PRO 6000 server price: on quotation. No manufacturer publishes a price for an 8-card server; the card alone is listed at $16,000 on the NVIDIA marketplace as of 1 September 2026 (as reported by thundercompute.com).
- Bytes per parameter: FP16 2 (16 bits per parameter, hence 2 bytes); FP8 1 (8 bits per parameter (E4M3), hence 1 byte, block scales not counted); NVFP4 0.5 (4 bits per parameter, hence 0.5 byte; 4-bit published weights weigh 0.54 to 0.56 byte per parameter with their scales (DeepSeek V4 Pro 865 GB for 1,599 billion, Kimi K3 1,561 GB for 2,780 billion)); Q4 0.6 (taken here as llama.cpp Q4_K_M, 0.6 byte per parameter including scales; Q4_K_S weighs 0.56, AWQ and GPTQ 0.55). Source: connaissance/faits.yaml, quantifications family, checked on 2026-08-31.
- Runtime margin × 1.15: QDNA operating assumption, not measured: 15% above the weights for activations, buffers and fragmentation.
- “Fits” threshold at 75% of memory: QDNA assumption, not measured, which keeps the remaining quarter for the attention cache and concurrency.