Moving up a class
Dell Pro Max with GB300 vs Dell Pro Max with GB10
Memory, bandwidth, software stack, price and the models each one can hold. Generated from the two sourced machine records; estimates are labelled.

NVIDIA GB300
Dell Pro Max with GB300
496 GB CPU + 252 GB GPU
Open the record
NVIDIA GB10
Dell Pro Max with GB10
128 GB unified
Open the recordWhat actually differs
- Different software stacks: the Dell Pro Max with GB300 runs CUDA, the Dell Pro Max with GB10 runs CUDA. Check that your runtime and model format are supported before comparing numbers.
- The Dell Pro Max with GB300 has 124 GB more model memory (252 GB vs 128 GB). That decides which models fit at all.
- The Dell Pro Max with GB300 reads memory 26.0× faster (7100 vs 273 GB/s). For a model that fits on both, that is the ceiling on generation speed.
- The Dell Pro Max with GB10 lists about $167,276 lower (≈ $175k vs ≈ $8.2k). Offers are dated snapshots in different regions and tax conditions.
- Form factor: Deskside workstation vs Desktop AI system.
Which models fit
Estimate · 4-bit · 8K context · 1 request| Model | Dell Pro Max with GB300 | Dell Pro Max with GB10 |
|---|---|---|
| Llama 3.1 8B | Fits 12 GB≤ 1251 tok/s | Fits 16 GB≤ 48 tok/s |
| Qwen3 14B | Fits 16 GB≤ 721 tok/s | Fits 20 GB≤ 28 tok/s |
| Qwen3 32B | Fits 28 GB≤ 338 tok/s | Fits 32 GB≤ 13 tok/s |
| Llama 3.3 70B | Fits 50 GB≤ 164 tok/s | Fits 54 GB≤ 6.3 tok/s |
| gpt-oss-20b (MoE) | Fits 19 GB≤ 2871 tok/s | Fits 23 GB≤ 110 tok/s |
| gpt-oss-120b (MoE) | Fits 76 GB≤ 2008 tok/s | Fits 80 GB≤ 77 tok/s |
| Qwen3 235B-A22B (MoE) | Fits 147 GB≤ 499 tok/s | Too large 151 GB |
| DeepSeek V3 / R1 671B (MoE, MLA) | Too large 404 GB | Too large 408 GB |
Specifications
Sourced and dated| Dell Pro Max with GB300 | Dell Pro Max with GB10 | |
|---|---|---|
| Chip / accelerator | GB300 Grace Blackwell Ultra | GB10 Grace Blackwell Superchip |
| Platform | NVIDIA GB300 | NVIDIA GB10 |
| Model memory | 496 GB CPU + 252 GB GPU | 128 GB unified |
| Memory bandwidth | GPU 7.1 TB/s · CPU 396 GB/s | 273 GB/s · GB10 platform |
| Software stack | CUDA | CUDA |
| Operating system | Ubuntu 24.04 Pro | NVIDIA DGX OS 7 |
| Storage | 4 × 4 TB NVMe · listed configuration | 4 TB PCIe Gen4 |
| Networking | Confirm selected configuration | Confirm selected configuration |
| Form factor | Deskside workstation | Desktop AI system |
| Dimensions | Confirm selected configuration | Confirm selected configuration |
| Weight | Not verified | Not verified |
| Power rating (not consumption) | 1,600 W PSU · output depends on input voltage · Dell specifies a 20 A circuit | 280 W USB-C adapter |
| Supplier list price | ≈ $175k · Supplier list price · United States · Sept 2026 | ≈ $8.2k · Supplier list price · United States · Sept 2026 |
| Availability | Listed for order · recheck with supplier | Listed for order · recheck with supplier |
| Model configurations in the Lab | 4 | 6 |
| Setups in the Lab | 3 | 5 |
| Record reviewed | 2026-09-13 | 2026-09-07 |
Before you choose
Dell Pro Max with GB300
- 748 GB coherent memory comprises 252 GB HBM3e GPU plus 496 GB LPDDR5X CPU memory. These tiers have different bandwidths.
- Regional snapshots cover specific configurations with RTX PRO 2000, four 4 TB drives and 12-month ProSupport. They are not entry prices or directly comparable landed costs.
- UT’s founder reports a deployment of this system; comparable reproducible measurements are not yet published in this lab.
Dell Pro Max with GB10
- Confirm region, storage, warranty and exact shipping configuration with Dell.
- A shared chip does not imply identical thermals, service terms or measured performance.
Dell Pro Max with GB300 compared with
NVIDIA DGX StationDell Pro Max with GB10 compared with
NVIDIA DGX SparkOwn one of these? A measured run (model revision, runtime, context, tokens per second) is worth more than this whole page. Publish it.