Last updated on

ASUS Ascent GX10 vs DGX Spark: November price update

By RunAIHome Team · 12 min read
asus-ascent-gx10dgx-sparkgb10nvidialocal-llmmini-pchardwarecomparison

TL;DR: These are related GB10 systems, but ASUS is not automatically the cheaper choice. On November 2, NVIDIA Marketplace listed the 4TB GX10 at $7,999 and the 4TB DGX Spark at $6,950; both were out of stock. Compare matched configurations and actual availability before paying for either.

ASUS Ascent GX10 (1TB)ASUS Ascent GX10 (4TB)NVIDIA DGX Spark (4TB)
Best forBuyers with a verified lower quote and modest local storage needsBuyers who prefer ASUS support or find a better available quoteBuyers who want NVIDIA’s reference system
Memory / bandwidth128GB / 273 GB/s128GB / 273 GB/s128GB / 273 GB/s
Listing checked November 2, 2026Current like-for-like quote not verified$7,999; out of stock$6,950; out of stock
Buying constraintDo not substitute this trim’s price for 4TBHigher displayed price in this comparisonLower displayed price, but not an available checkout

Honest take: The old recommendation to buy a 4TB GX10 and save $550 no longer follows from the verified listings. The displayed ASUS price is $1,049 higher, approximately 15.1% above the Spark price. Since both listings show out of stock, request an available, delivered quote before choosing either.

The primary sources for this update are NVIDIA Marketplace’s DGX Spark listing and its ASUS GX10 4TB listing. These are dated displayed prices, not a claim about every retailer, a guaranteed checkout total, or stock elsewhere. Historical launch-price references remain in the source list for context; they are not today’s purchase quotes.

The decision is now about an available configuration, support, and workload fit. A lower historical price for a different storage trim is not evidence that the same saving is available today.

Same chip does not guarantee the same result in every workload

Both boxes are built around NVIDIA’s GB10 Grace Blackwell Superchip: a 20-core Arm CPU and a Blackwell GPU on one package, rated up to 1 petaFLOP of sparse FP4 compute, wired to 128GB of LPDDR5X unified memory at 273 GB/s. Same memory, same bus, same GPU, same DGX OS software stack. Matching headline specifications are a reason to compare the platforms, not proof that every software build and workload performs identically.

StorageReview’s GX10 review ran vLLM serving benchmarks across GPT-OSS, Qwen3-Coder, and Llama 3.1 and found the ASUS box “performed consistently with other Spark systems” — a useful comparison within that review’s test conditions. Do not transfer every result unchanged between machines or software versions. As a separate reference, the following published Spark result comes from llama.cpp’s official Spark thread:

$ llama-bench -m gpt-oss-120b-MXFP4.gguf -fa 1
| model               | test    |     t/s |
| ------------------- | ------- | ------: |
| gpt-oss 120B MXFP4  | pp2048  | 1723.07 |
| gpt-oss 120B MXFP4  | tg32    |   38.55 |

Call it 38–60 tok/s decode on the flagship 128GB-class MoE (the 60 is LMSYS’s tuned SGLang build), ~1,700–1,950 tok/s prefill, and the same ~2.6 tok/s slog on dense 70B models that every 273 GB/s box is condemned to — we walked through why in the Spark vs Mac Studio M5 Max comparison. If those numbers don’t fit your workload, no amount of choosing between these two boxes fixes it; skip to the last section.

With performance dependent on the tested configuration, compare these purchasing differences as well.

Difference 1: storage — and the non-negotiable SSD

The specific Spark listing compared here has 4TB. ASUS’s current US specifications list 1TB PCIe 4.0 and 4TB PCIe 5.0 options. The Marketplace GX10 description instead labels its 4TB SSD PCIe 4.0. Confirm the exact part number and storage specification with the seller before ordering; do not resolve that source discrepancy by guessing.

Storage tier matters because ASUS does not support customer hardware upgrades and excludes damage or malfunction caused by customer SSD replacement from warranty coverage. That is more specific than saying any SSD change automatically voids the entire warranty. Check the applicable vendor policy before planning an upgrade; buying sufficient supported capacity avoids relying on one.

Now do the math on 1TB. DGX OS plus the CUDA/NIM stack claims its share out of the box; gpt-oss 120B at MXFP4 is ~60GB; a 70B Q4 GGUF is ~42GB; add a couple of video-generation checkpoints, a fine-tuning dataset, and the Docker cache that NVIDIA’s playbooks assume, and a 1TB drive is doing inventory management inside six months. It’s workable with a NAS for cold weights, but “workable” is doing real lifting. The 4TB trims make the whole question disappear for years — which is exactly why NVIDIA only sells that one.

Difference 2: the update path — NVIDIA direct vs NVIDIA-via-ASUS

Both machines run DGX OS and only DGX OS — ASUS confirms the GX10 supports nothing else. The difference is the release pipeline: Spark updates come straight from NVIDIA, while GX10 builds are “adapted and validated by ASUS” before they reach the DGX Dashboard updater. DGX OS 7.5.0 (April 2026) reached the GX10, but GX10 owners have had to ask on NVIDIA’s forums where their firmware updates live — the answer being ASUS’s support site, on ASUS’s schedule.

For a box whose entire value is the NVIDIA software stack, one extra company in the update chain is a small but honest tax on the discount. It has already mattered once this year, which brings us to the fix section.

The 13 Gbps problem (and the fix that made it 196)

Both boxes carry two QSFP ports off an integrated ConnectX-7 NIC — 200 Gbps links whose whole purpose is pairing two GB10 units into one 256GB pool for models neither runs alone. ASUS sells the GX10 with a stacking cable as an official two-box bundle, and the wiring underneath is genuinely odd: the GB10 attaches the ConnectX-7 over two separate PCIe Gen5 x4 links in multi-host mode, so one physical port surfaces as two interfaces in the OS.

Here’s the problem that bit early adopters on both machines. You cable two units together, run iperf3, and get ~13 Gbps on a link that reports 200:

$ iperf3 -c 192.168.100.2
[  5]   0.00-10.00  sec  15.1 GBytes  13.0 Gbits/sec

Nothing is miscabled. The root cause, per the NVIDIA developer forum thread that ran it down, was ConnectX-7 firmware power-throttling itself off a bogus 27W power report on early builds (kernel 6.17.0-1008-nvidia, driver 580.126.09). The fix is anticlimactic:

$ sudo apt full-upgrade && sudo reboot

Post-update (kernel 6.17.0-1014-nvidia, driver 580.142), the same link runs ~196 Gbps, and independent two-box testing over RDMA/NCCL/vLLM confirms 185–190 Gbps sustained with RoCE and jumbo frames (MTU 9000). The takeaway for this comparison: the fix shipped through the DGX OS channel — Spark owners pulled it the day it landed, GX10 owners waited on the ASUS-validated build. Days, not months, but it’s the update-path difference made concrete.

Difference 3 and 4: the box itself, and the price of admission

The hardware around the chip is where ASUS actually did work. Per IT Pro’s review, the GX10 is a 150 × 150 × 51mm, 1.48kg metal box — lighter than the Spark, which ServeTheHome found packs Mac-Studio-like density — and it puts the power button on the front, which sounds trivial until you’ve reached around the back of a stacked pair for the tenth time. The Spark counters with the collectible factor: NVIDIA’s own champagne-gold chassis, which is either worth $0 or several hundred dollars depending entirely on how you feel about NVIDIA.

For a matched 128GB/4TB comparison, the displayed-price arithmetic is:

GX10 premium = $7,999 - $6,950 = $1,049
GX10 premium relative to Spark = $1,049 / $6,950 = 15.1% (rounded)
Two GX10 units = 2 x $7,999 = $15,998
Two Spark units = 2 x $6,950 = $13,900
Two-unit difference = $2,098

The two-unit totals are arithmetic on the listed single-unit prices. They are not verified bundle offers and do not include separately charged cables, shipping, or taxes. Both source listings were out of stock, so these totals are planning comparisons rather than executable purchase orders. A reseller quote may change the result; compare the same capacities and included accessories before calculating a new saving.

Should you buy either one?

A reality check before the decision table, because both machines fail the same buyers. At 273 GB/s, decode speed on big dense models is single-digit — these are the wrong boxes for chatting with a 70B, so compare the workload-specific evidence in our Mac Studio comparison and Strix Halo comparison before choosing a platform. And if everything you run fits in 24GB, a used RTX 3090’s 936 GB/s embarrasses this entire category. The GB10 case is specific: you want CUDA — fine-tuning (ASUS demonstrated LoRA on Llama 3 8B at 53K peak tokens/sec), prefill-heavy agentic work, NIM/TensorRT pipelines, day-one model support — in one small 240W box, with a petaFLOP of FP4 behind it. Run your model sizes through the VRAM calculator before buying a system to find out you needed bandwidth, not capacity.

What to actually buy

Use the current listing comparison above as a starting point, then obtain an available quote:

Your situationWhat to compareDecision rule
You need one 4TB GB10 systemExact GX10 and Spark part numbersPrefer the better delivered quote if support and workload requirements are satisfied
Both official listings remain unavailableAuthorized seller stock and return termsDo not treat an out-of-stock displayed price as an available deal
You need two systemsMatched nodes, supported network configuration, and cablesCalculate the complete two-node bill; do not assume a bundle discount
Your models fit existing hardwareYour current machine and required runtimeValidate the workload before buying more memory capacity
You are uncertain about GB10 compatibilityA representative GB10 test environmentA different GPU rental is not a direct benchmark of GB10 memory behavior

FAQ

Is the ASUS Ascent GX10 slower than the DGX Spark? They share core specifications, and the cited review found comparable results within its tests. Actual speed still depends on the model, quantization, runtime, context, and configuration. The cited Spark measurements are not a guarantee for either machine in your workload.

Is the DGX Spark still more expensive? Not in the matched 4TB listings checked for this update. The displayed GX10 price is higher. That reverses the old article’s price-based recommendation, while availability and delivered quotes remain decisive.

Can I upgrade the GX10’s 1TB SSD later? ASUS says customer hardware upgrades are unsupported and damage or malfunction caused by replacement is not covered. Plan around supported capacity or external storage; check the exact warranty terms rather than assuming a replacement is supported.

Can I mix a GX10 and a DGX Spark in a two-box cluster? They’re the same GB10 platform with the same ConnectX-7 QSFP ports, and the software stack is common — but the official pairing guidance and ASUS’s own bundles assume matched units. If you’re buying two anyway, buy two of whichever is cheaper that week and update both to current firmware first, or your 200 Gbps link may run at 13 (see the fix above).

When does the 1TB GX10 make sense? When an actual available quote is sufficiently lower and your measured storage needs fit it. No current 1TB quote was verified for this update, so the old $150 upgrade argument has been removed. Compare the current capacity premium and your recovery/storage plan instead.

Sources

Last updated November 2, 2026: official listing prices, availability, storage caveats, derived totals, and buying conclusions refreshed. Performance figures remain attributed examples, not new measurements. Verify final configuration and delivered price before purchasing.

Was this article helpful?

Get the numbers before you buy

New GPU and mini-PC benchmarks, VRAM thresholds, and price checks — sent when there's something worth acting on, not on a schedule. No spam, unsubscribe anytime.