Different platforms, same budget
NVIDIA DGX Spark vs AMD Ryzen AI Halo
Memory, bandwidth, software stack, price and the models each one can hold. Generated from the two sourced machine records; estimates are labelled.
NVIDIA GB10
NVIDIA DGX Spark
128 GB unified
Open the record
AMD Ryzen AI Max
AMD Ryzen AI Halo
128 GB unified
Open the recordWhat actually differs
- Different software stacks: the NVIDIA DGX Spark runs CUDA, the AMD Ryzen AI Halo runs ROCm and Vulkan. Check that your runtime and model format are supported before comparing numbers.
- Both have 128 GB unified, so the same models fit.
- Memory bandwidth is not in the record for the AMD Ryzen AI Halo, so the Lab cannot compare generation speed ceilings.
- No public list price in the record for the AMD Ryzen AI Halo. Ask the supplier for a configured quote.
Which models fit
Estimate · 4-bit · 8K context · 1 request| Model | NVIDIA DGX Spark | AMD Ryzen AI Halo |
|---|---|---|
| Llama 3.1 8B | Fits 16 GB≤ 48 tok/s | Fits 16 GBBandwidth not in record |
| Qwen3 14B | Fits 20 GB≤ 28 tok/s | Fits 20 GBBandwidth not in record |
| Qwen3 32B | Fits 32 GB≤ 13 tok/s | Fits 32 GBBandwidth not in record |
| Llama 3.3 70B | Fits 54 GB≤ 6.3 tok/s | Fits 54 GBBandwidth not in record |
| gpt-oss-20b (MoE) | Fits 23 GB≤ 110 tok/s | Fits 23 GBBandwidth not in record |
| gpt-oss-120b (MoE) | Fits 80 GB≤ 77 tok/s | Fits 80 GBBandwidth not in record |
| Qwen3 235B-A22B (MoE) | Too large 151 GB | Too large 151 GB |
| DeepSeek V3 / R1 671B (MoE, MLA) | Too large 408 GB | Too large 408 GB |
Specifications
Sourced and dated| NVIDIA DGX Spark | AMD Ryzen AI Halo | |
|---|---|---|
| Chip / accelerator | GB10 Grace Blackwell Superchip | Ryzen AI Max+ 395 · Radeon integrated graphics |
| Platform | NVIDIA GB10 | AMD Ryzen AI Max |
| Model memory | 128 GB unified | 128 GB unified |
| Memory bandwidth | 273 GB/s | See source; no lab measurement |
| Software stack | CUDA | ROCm and Vulkan |
| Operating system | NVIDIA DGX OS | Windows / Linux · ROCm |
| Storage | 4 TB NVMe | Confirm with manufacturer |
| Networking | 10 GbE · ConnectX-7 · Wi-Fi 7 | See manufacturer configuration |
| Form factor | Desktop AI system | Desktop AI system |
| Dimensions | 150 × 150 × 50.5 mm | Confirm with manufacturer |
| Weight | 1.2 kg | Not verified |
| Power rating (not consumption) | 240 W adapter · 140 W SoC TDP | No comparable wall-power measurement |
| Supplier list price | ≈ $4.7k · Supplier list price · United States · Sept 2026 | Price from the manufacturer · No public list price · manufacturer page |
| Availability | Listed for order · recheck with supplier | Manufacturer lists US purchase option; confirm stock |
| Model configurations in the Lab | 6 | 2 |
| Setups in the Lab | 5 | 1 |
| Record reviewed | 2026-09-13 | 2026-09-09 |
Before you choose
NVIDIA DGX Spark
- Memory capacity does not establish interactive speed or simultaneous request capacity.
- UT has documented a Spark deployment. Confirm exact software versions and support scope.
AMD Ryzen AI Halo
- The reference configuration is Ryzen AI Max+ 395 with 128 GB unified memory.
- AMD lists a PRO 495 version supporting 192 GB as coming soon; it is a separate configuration.
- The Lab has not tested this system or validated the UT stack.
NVIDIA DGX Spark compared with
ASUS Ascent GX10 Dell Pro Max with GB10 HP ZGX Nano G1n Lenovo ThinkStation PGX Acer Veriton GN100 GIGABYTE AI TOP ATOM MSI EdgeXpert Apple Mac Studio · M3 Ultra Apple Mac Studio · M5 Ultra Framework Desktop · Ryzen AI Max+ MINISFORUM MS-S1 MAX HP Z2 Mini G1a NVIDIA DGX StationOwn one of these? A measured run (model revision, runtime, context, tokens per second) is worth more than this whole page. Publish it.