NVIDIA DGX H100
A complete 8-GPU H100 server for organizations that need serious throughput today.
640 GB
GPU Memory vs. tier max
10200 W
Power vs. tier max
$390,000
Price vs. tier max
| The dedicated memory built into the graphics card. An AI model's entire set of "weights" (its learned parameters) must fit into this memory before the card can run it. If a model needs more VRAM than one card has, you need multiple cards working together, or a card with more memory. | 640 GB total across 8 GPUs (80 GB each) |
|---|---|
| The general-purpose "brain" of the server that manages the operating system, feeds data to the GPUs, and handles tasks the GPUs aren't built for. More cores let it keep many GPUs fed with data at once. | Dual Intel Xeon Platinum 8480C (112 cores total, 2.00GHz base / 3.80GHz boost) |
| Fast temporary memory used by the CPU (separate from GPU VRAM). Large AI servers need a lot of it to stage data, run the operating system, and manage many simultaneous requests. | 2048 GB (2.0 TB) |
| Thermal Design Power — roughly how many watts the card draws under full load. Higher-performance cards need more electricity and generate more heat, which affects your electricity bill and cooling needs. | 10200 W |
| Approximate cost to acquire the hardware. For consumer cards this is usually a fixed retail price. For datacenter-class hardware, list prices are rarely published — buyers negotiate directly with NVIDIA or its partners, so we show a typical range or say "contact for pricing" rather than inventing a number. |
$390,000
No official NVIDIA list price — quote-only. Reseller listings suggest roughly $300,000-$480,000. NVIDIA ended new orders Jan 2025 in favor of Blackwell-based systems.
|
Why this matters
A card or server needs at least as much
The dedicated memory built into the graphics card. An AI model's entire set of "weights" (its learned parameters) must fit into this memory before the card can run it. If a model needs more VRAM than one card has, you need multiple cards working together, or a card with more memory.
as an AI model requires,
or the model simply won't load. See Get a Recommendation
to see which models this hardware can run, or
add it to a side-by-side comparison.