Verdict
All guidesResearchMethodologyAboutContact
Head-to-head · Best AI Workstations

NVIDIA DGX Spark vs Puget Systems Genesis II

Which is the better buy? Side-by-side on rating, price, strengths, and watch-outs — with the published ratings we averaged to get there.

Verdict is reader-supported. As an Amazon Associate we earn from qualifying purchases. Some links on this page are affiliate links — if you click through and buy, we may earn a small commission at no extra cost to you. Our ratings are sourced from independent publications, not sponsors.
The short answer

Puget Systems Genesis II comes out ahead by a narrow margin (4.4 vs 4.5). The gap is mostly about Best for enterprise — multi-GPU training and inference — read the strengths below before deciding.

NVIDIA DGX Spark
Ranked #3 in Best AI Workstations
NVIDIA DGX Spark
4 sources3 derived$4,649.99as of Aug 1

The NVIDIA DGX Spark is the productized version of Project DIGITS — a 150 mm cube housing the GB10 Grace Blackwell Superchip, 128 GB unified LPDDR5x, and the full CUDA AI stack out of the box. Tom's Hardware called it 'a well-rounded toolkit for local AI'; ServeTheHome called it 'must-have for AI developers'; LMSYS published the most thorough independent benchmarks. The 128 GB unified-memory ceiling is the headline feature: it loads models that would otherwise need a $30K+ multi-GPU rig. The catch is bandwidth-limited decode — LMSYS measured Llama-3.1 70B FP8 at 2.7 tokens/sec single-batch, while GPT-OSS 120B (MoE, ~17B active) hits ~14.5 tokens/sec per ServeTheHome. Best understood as a CUDA-native development box for buyers who need to iterate on big-model code without renting cloud GPUs.

Strengths
  • — 128 GB unified LPDDR5x memory — fits 70B FP8 / 120B Q4 / 405B with two clustered units
  • — Full CUDA + NVIDIA AI stack preinstalled; the most polished local-AI dev box on the market
  • — Compact 150 mm cube, 240 W max — fits any desk, runs cool and quiet
Watch-outs
  • — 273 GB/s LPDDR5x bandwidth caps decode tok/s on dense large models — 70B FP8 measures ~2.7 tok/s on a single unit
  • — Linux-only, no Windows or gaming use; specialist hardware for AI developers
  • — Price raised from $3,999 to $4,699 in February 2026 due to memory supply
$4,649.99 · Check Price on AmazonFull review →
Puget Systems Genesis II
Higher ratedRanked #2 in Best AI Workstations
Puget Systems Genesis II
3 sources$10,569as of Apr 25

Enterprise buyers who need configurations the major OEMs can't match are the audience for the Puget Systems Genesis II, a professionally built and deeply customizable workstation. Specs run to AMD Threadripper Pro, up to 4x RTX 4090 (or RTX Ada workstation cards), and 256 GB ECC DDR4, and Puget's hand-tuned assembly shows in both performance and build quality, with a quiet edition among the options. On local LLM work, a single RTX 4090 (24 GB VRAM at 1008 GB/s) runs 70B Q4 with system-RAM offload at roughly 10-20 tokens/sec; a 4x RTX 4090 configuration pools 96 GB of VRAM, holds 70B Q4 entirely in VRAM, and reaches 30-40 tokens/sec through tensor parallelism. The steep entry price limits it to buyers who can justify a $10K+ workstation. If local LLMs are the only goal, the DGX Spark below offers comparable model-size headroom at less than half the price.

Strengths
  • — Highly customizable with a wide assortment of mainstream and pro-channel components like Nvidia Ada workstation GPUs
  • — Serious professional build quality with careful component selection and assembly
  • — Exceptional customer experience including an easy-to-use configuration tool with detailed component comments
Watch-outs
  • — Extremely expensive, with review units costing over $10,000 and configurations reaching nearly $61,000
  • — Configurations with 4x RTX 4090 lose the NVLink that would have helped tensor-parallel LLM inference
  • — Price can be prohibitive compared to a DGX Spark or HP Z6 G5 A for buyers whose only need is local LLMs
$10,569 · Buy at pugetsystems.comFull review →

Derived means the reviewer published no score, so one was inferred from their written verdict. Everything else is a number the publisher printed. How we rank.

How they stack up

NVIDIA DGX Spark

The DGX Spark is the cheapest path to 128 GB of CUDA-addressable unified memory anywhere on the market. Versus the GMKtec EVO-X2 ($1,699) or Beelink GTR9 Pro ($2,000), it's roughly 2.5x the price but offers the full NVIDIA software stack the Strix Halo boxes can only approximate via ROCm or Vulkan. Versus the Puget Genesis II ($10K+), it's a single-purpose dev box — no multi-display creative workflow, no gaming, no general workstation duty. Pair two Sparks via the ConnectX-7 networking and you get 405B-class model coverage at roughly $9,400, the cheapest legal path to that ceiling.

Puget Systems Genesis II

The Puget Systems Genesis II is the enterprise pick. Versus the HP Z8 Fury G5, it offers comparable scale-up capability but in a quieter chassis with a more thoughtful configurator. Versus the HP Z6 G5 A, it's two tiers up in price and ceiling. Versus the NVIDIA DGX Spark, it's a different class of machine entirely — the DGX Spark is a 128 GB unified-memory dev box, the Genesis II is a multi-GPU training/inference workstation. For buyers whose only goal is running large local LLMs, the DGX Spark is the more cost-effective answer; the Genesis II earns its premium when training, fine-tuning, or multi-application workstation duty are part of the picture.

Specs side-by-side

SpecNVIDIA DGX SparkPuget Systems Genesis II
CPU20-core Arm (10x Cortex-X925 + 10x Cortex-A725)AMD Threadripper Pro 5975WX (32-core)
GPUNVIDIA GB10 Grace Blackwell SuperchipUp to 4x Nvidia RTX 4090 (96 GB pooled VRAM)
RAM128 GB unified LPDDR5x256 GB DDR4-3200 ECC
Storage4 TB NVMe M.2 (user-replaceable, self-encrypting)4 TB Sabrent Rocket 4 Plus NVMe
Memory Bandwidth273 GB/s~76 GB/s system; ~1008 GB/s per RTX 4090 VRAM
Form FactorCompact desktop (150 mm cube, 1.2 kg)Full tower (Fractal Define 7)
← See the full ranking of best ai workstations