ASUS ET900N G3 Price: US Resellers List DGX Station Build From $111,999.99
Two US resellers list the ASUS ET900N G3 at $111,999.99 and $113,998.86 as of Nov. 4; ASUS has no list price. A creator's July test of a 120B model in NVFP4 read about 2,600 aggregate tok/s at 128 requests, engine undisclosed.
Local AI DeskEdited by Ege Erten
5 min readAI-assisted
Follows our story: NVIDIA DGX Station Review: 748 GB, 1.6 kW and a $160,195 Dell List Price

The ASUS ExpertCenter Pro ET900N G3, the company's build of NVIDIA's DGX Station.
Image: ASUS
Source · Press image from the ASUS product page, used for news reporting.
The ASUS ExpertCenter Pro ET900N G3, the company's build of NVIDIA's DGX Station, is listed at $111,999.99 by Central Computers and at $113,998.86 by SabrePC, as read on Nov. 4, 2026. ASUS publishes no list price.
These are the first US prices for the ASUS tower in Voltgrid's coverage. Our Nov. 1 DGX Station review found a list price from Dell but none from ASUS, and our Oct. 24 report set out the platform: one GB300 superchip, 252 GB of HBM3e and 1,600 W of system power. Central Computers marks its ET900N G3-2SSD-V2 available but quote-only, and SabrePC shows its PRO ET900N G3-2SSD out of stock.
Where the ASUS Price Sits Against MSI, Exxact and Dell
Rival GB300 towers, read the same day, bracket it. Exxact's Valence DGX Station shows a starting price of $98,450.00 on its own site; the page does not state the storage or card in that base price. Per the Nov. 4 update to our DGX Station review, Newegg lists MSI's XpertStation WS300 at $99,999.00 as a request-a-quote listing, out of stock. Dell's Pro Precision with GB300 still shows $160,195.00 for its default build, which adds an RTX PRO 2000 display card and 16 TB of SSDs. ASUS's spec sheet says its tower ships with two 2 TB drives, so the builds differ.

Configurations differ across these listings, and ASUS publishes no list price; the two ASUS figures are reseller prices.
Illustration (AI-generated)
Notebookcheck had reported $99,999 at Central Computers in a story updated Oct. 5, without saying when that price was read or for which model, so the two figures do not show a price change.
StorageReview found the ASUS within 2% of MSI's WS300 on 24 of 28 vLLM tests and said this second build of NVIDIA's reference platform confirms the first. StorageReview's results show near-identical throughput on most tests; neither the listings nor the review say what the price difference pays for.
Nemotron 120B Throughput Measured on the ASUS
The most detailed public run of a 120B model on the ASUS comes from YouTube creator Alex Ziskind, in a video published July 30. The machine was a loaner; in an Oct. 7 video with Level1Techs he said it had to go back to ASUS. A 70-second Short cut from the first video followed on Aug. 15.
He ran NVIDIA's Nemotron 3 Super 120B in NVFP4, NVIDIA's own 4-bit format, and kept adding simultaneous requests. One request ran at about 180 tokens per second. Combined output then read 250 at two requests, 400 at eight, over 800 at 16, 1,000 at 32, 1,500 at 64 and 2,600 at 128.

Totals across all requests in one creator's July 30 test, not per-user speed; the serving engine and prompt lengths were not disclosed.
Illustration (AI-generated)
Those are totals across all requests, not what each user sees. Above one request, Ziskind gave no per-request speed and said only that each request takes a bit longer. He credited continuous batching: the GPU groups waiting requests into one pass, so the total climbs while each request slows. He did not disclose the serving engine, its version, or the prompt and output lengths. No lab has published a run of the same model, quantization and tool on the ASUS, so his figures cannot be set beside lab results.
His power readings came from monitoring software on the machine, not a wall meter. With one coding agent, nvtop showed up to about 715 W and about 550 W on average. In a separate agent test, his dashboard showed about 750 W with 16 agents, 900 W with 64 and about 1,000 W at most with 128. In the October video, the same unit idled at 319 W, read on the battery station powering it.
StorageReview's lab run, published Sept. 23, states its conditions: vLLM, 512-token prompts, 512-token answers and aggregate output. On the ASUS, MiniMax M2.7 in NVFP4 went from 214 tokens per second on one stream to 4,793 on 128, and DeepSeek V4 Flash in FP8 from 152 to 1,766 at 32 streams. Nemotron 3 Ultra 550B in NVFP4 peaked at 159 against the MSI's 168, 5.2% lower; StorageReview calls every figure a single run on each tower and does not pin the gap on the chassis. The lab removed the RTX PRO 2000 card ASUS had fitted and did not measure power.
ASUS ET900N G3 Specs: Storage, Power Supply and Memory
ASUS's spec sheet settles, for its model, the storage conflict our review left open. It lists two M.2 slots at PCIe 5.0 x4, set up as a RAID 1 pair for the operating system, plus two PCIe 6.0 x4 slots for training data. That matches ServeTheHome's table for the MSI WS300, not NVIDIA's listing of four Gen5 slots.
The power supply is a 1,600 W Titanium-rated ATX unit; ServeTheHome lists a Platinum-rated 1,600 W supply in the MSI. ASUS's North American listing is 125 V AC with an IEC C19 to NEMA 5-20P cord, which it lists as optional, and says maximum power output depends on the input voltage and the cord; NVIDIA requires a 20 A circuit. StorageReview notes two carry handles and a fan aimed at the ConnectX-8 optics. ASUS claims about 50 dB in Auto Mode at full load without a graphics card, a figure no lab has checked.
The memory figure has changed too. ASUS's CES release on Jan. 7 gave 775 GB of coherent memory. Its June 15 launch release and spec page give 748 GB, and NVIDIA's own figure fell the same way.

The ASUS tower pairs a GB300 superchip with 748 GB of memory in two pools; Windows editions are expected from partners in December.
Illustration (AI-generated)
How the 748 GB Memory Pool Works
The 748 GB is two memory pools, not one. ASUS lists 252 GB of HBM3e, memory stacked beside the GPU, at 7.1 TB/s, and 496 GB of LPDDR5X on the Grace CPU at 396 GB/s. NVLink-C2C, the chip-to-chip link, joins them, and NVIDIA lists it at 900 GB/s (StorageReview calls that bidirectional). The link lets the GPU read the CPU's pool directly.
A model that fits in the 252 GB stays in the fast pool. A larger one spills into the slower pool. StorageReview ran GLM-5.2 in NVFP4 with about 215 GB of expert weights in Grace memory and measured 35 output tokens per second for one stream on vLLM. The 748 GB sets which models load, and the 252 GB at 7.1 TB/s sets which can stay in the fast pool.
What to watch
ASUS still ships Ubuntu with NVIDIA's developer tools and says Windows support is planned, with no Windows model listed. NVIDIA said on Oct. 21 that DGX Station for Windows is expected from partners in December, and that systems are expected to begin shipping later this year from ASUS, Dell Technologies, GIGABYTE, HP, MSI and Supermicro. ASUS's support page lists one BIOS, version 0402 from June 10, and BMC firmware 1.6.55 from Sept. 29 as the newest, without change notes. Still missing are an ASUS list price, a full-load wall-power measurement from any lab, and the engine and settings behind Ziskind's numbers.
How it compares
Full table| Product | Memory | Bandwidth | AI compute | Measured speed | Price |
|---|---|---|---|---|---|
| ASUS ExpertCenter Pro ET900N G3ASUS (NVIDIA DGX Station GB300 platform) | 748 GB coherent: 252 GB HBM3e (GPU) + 496 GB LPDDR5X (CPU)[1] | 7.1 TB/s (GPU HBM3e); 396 GB/s (CPU LPDDR5X)[1] | up to 20 PFLOPS AI performance (ASUS gives no precision or sparsity label) (sparse)[2] | GPT-OSS-120B MXFP4, vLLM, 512/512, 1-128 streams: 9,280 aggregate output tok/s at 128 streams; 5,023 on 8,192/1,024 (StorageReview, ASUS-supplied unit, no add-in GPU)[3] | $111,999.99 at Central Computers, US (model ET900N G3-2SSD-V2, request a quote, unavailable online, read Nov 4, 2026); no ASUS list price[4] |
| HP ZGX Fury AI StationHP | 252 GB HBM3e GPU + 496 GB LPDDR5x CPU (4x 128 GB SOCAMM); 748 GB unified[5] | 7.1 TB/s (GPU HBM3e); 396 GB/s (CPU LPDDR5x)[5] | — | Two-unit cluster with an ASUS ET900N G3 over two 400G DACs: GLM-5.3 NVFP4 5,018 output tok/s at 256 streams, 512/512, vLLM 0.27.1 TP2 with MTP (StorageReview); no single-unit HP result[6] | Japan list price from 28,490,000 yen incl. tax (sales from Oct 27, 2026); no US list price published[7] |
| NVIDIA DGX StationNVIDIA (systems built by ASUS, Dell, GIGABYTE, HP, MSI, Supermicro and Exxact) | 748 GB coherent: 252 GB HBM3e (GPU) + 496 GB LPDDR5X (CPU)[8] | GPU HBM3e 7.1 TB/s; CPU LPDDR5X 396 GB/s; NVLink-C2C 900 GB/s between them[8] | 15 PFLOPS FP4 (without sparsity)[8] | gpt-oss-120b MXFP4, stock vLLM, 512/512 chat workload: 9,280 output tok/s at 128 streams (5,023 on 8,192/1,024 prompts), single GB300 (StorageReview, ASUS ExpertCenter Pro ET900N G3)[9] | No NVIDIA list price; lowest US partner listing read Nov 4, 2026: Exxact Valence VWS-158270643 from $98,450 (Tom's Hardware found $94,930 in Aug); Dell Pro Precision with GB300 lists $160,195[10] |
| NVIDIA RTX PRO 6000 Blackwell Workstation EditionNVIDIA (sold through PNY, TD SYNNEX and workstation makers) | 96 GB[11] | 1,792 GB/s[11] | 4,000 AI TOPS (FP4, with sparsity) (sparse)[11] | gpt-oss-120b MXFP4, stock vLLM, 512/512 chat workload: 3,064 output tok/s at 128 streams (1,169 on 8,192/1,024 prompts); Server Edition at 600W (StorageReview)[9] | $16,000 on NVIDIA's US marketplace (Workstation Edition), listed out of stock[12] |
| Apple Mac Studio (M5 Ultra)Apple | up to 512 GB unified (96 GB on 30-core M5 Ultra; 256 GB or 512 GB on 36-core/80-core M5 Ultra)[13] | 1.2 TB/s[13] | — | Qwen 3.8 27B 4-bit MLX in LM Studio: about 55 tok/s on M5 Ultra (36-core, 256 GB) vs about 40 on M3 Ultra (BGR)[14] | $22,299 minimum with 512 GB ($6,799 36-core CPU/80-core GPU M5 Ultra config plus $15,500 memory upgrade; Oct 28, 2026)[15] |
| NVIDIA RTX Spark (N1X) PCsNVIDIA (MediaTek collaborated on the custom CPU design) | Up to 128 GB LPDDR5X (20-core CPU / 6,144-core GPU version); up to 64 GB (18-core / 5,120-core laptop version)[16] | Up to 300 GB/s: LPDDR5X-9400 on a 256-bit bus (NVIDIA's product page lists no bandwidth)[17] | — | gpt-oss:20b via Ollama in a Codex CLI coding agent, 64K context: 58.5 tok/s generation, 3 of 4 tasks passed (Dell XPS 16, 20-core N1X, 64 GB; StorageReview)[18] | Laptops from $2,599; pricing and configurations vary by region[19] |
| ASUS Ascent GX10ASUS | 64 GB or 128 GB LPDDR5X-8533 unified[20] | 273 GB/s (8533 MT/s), as listed by StorageReview; ASUS's spec page states no bandwidth[21] | Up to 1 PFLOP FP4 (GB10; NVIDIA's footnote: theoretical, using the sparsity feature). ASUS's pages say 1 petaFLOP / FP4 with no sparsity label (sparse)[22] | gpt-oss-120b, vLLM online serving, equal ISL/OSL: 70 tok/s at batch 1 rising to 680 tok/s aggregate at batch 64 (StorageReview)[21] | $5,999 (128 GB, 1 TB, GX10-GG0010BN, ASUS eShop US, Oct 30, 2026)[23] |
| Dell Pro Precision with GB10 (model FCM1253; launched as Dell Pro Max with GB10)Dell | 128 GB LPDDR5x up to 8533 MT/s onboard (64 GB option also listed)[24] | 273 GB/s (128 GB LPDDR5x, 256-bit)[25] | — | gpt-oss-120b, vLLM online serving, equal ISL/OSL: 68.57 tok/s at batch 1 rising to 670.83 tok/s aggregate at batch 64 (StorageReview)[26] | $7,967 (128 GB, 2 TB Gen4 QLC SSD default config, Dell.com US, read Nov 2, 2026 16:40 UTC)[24] |
| Framework Desktop (Ryzen AI Max+ PRO 495, 192 GB)Framework | 192 GB LPDDR5X-8533 soldered[27] | 273 GB/s (256-bit)[27] | — | — | $6,799 (Max+ PRO 495 192 GB DIY Edition pre-order, Nov 3, 2026; storage not included)[28] |
| GIGABYTE AI TOP ATOMGIGABYTE | 128 GB unified system memory[29] | 273 GB/s (LPDDR5x-8533, 256-bit)[30] | — | gpt-oss-120b, vLLM online serving, equal ISL/OSL: 60.71 tok/s at batch 1 rising to 649.62 tok/s aggregate at batch 64 (StorageReview)[30] | $7,312.00 advertised at CDW US against a $7,999.99 list (128 GB, 4 TB, ATAGB10-9000; read Nov 2, 2026)[31] |
| GMKtec EVO-X5 Pro AI Mini PCGMKtec | 192 GB LPDDR5X-8533 onboard (8 x 24 GB, non-upgradeable); up to 160 GB allocatable as VRAM[32] | Up to 273 GB/s[33] | — | GMKtec Laboratory table: GLM-5.3-Flash UD-IQ4_XS (146.05 GB) decode 17.79 tok/s, prefill 70.18, TTFT 14.71 s; Qwen3.8-Flash-Next ROCmFP4-FAST (93.48 GB) decode 45.08 tok/s[32] | $6,799 (192 GB/2 TB) and $7,099 (192 GB/4 TB), GMKtec US store, Nov 2, 2026[32] |
| HP ZGX Nano AI Station (G1n)HP | 128 GB coherent unified system memory[34] | 273 GB/s (128 GB LPDDR5x)[35] | GB10: up to 1 PFLOP FP4, theoretical, using the sparsity feature; HP and reviewers list 1,000 TOPS FP4 (sparse)[22] | gpt-oss-120b, vLLM online serving, equal ISL/OSL: 69.6 tok/s at batch 1 rising to 722.9 tok/s aggregate at batch 64 (StorageReview)[36] | $7,399.00 at the HP Store for the 4 TB model; $4,759 on Amazon (Notebookcheck, review updated Oct 5, 2026)[37] |
| Lenovo ThinkStation PGXLenovo | 128 GB LPDDR5X unified (256-bit)[38] | 273 GB/s[38] | — | — | about $4,100 for the 1 TB configuration (Mar 2026)[39] |
| MSI EdgeXpert (MS-C931)MSI | 64 GB or 128 GB LPDDR5x unified (256-bit)[40] | 273 GB/s[40] | 1 PFLOPS FP4 (sparse)[40] | — | No list price; MSI sells via quote (Get a quote)[40] |
| NVIDIA DGX SparkNVIDIA (partner GB10 systems from Acer, ASUS, Dell, GIGABYTE, HP, Lenovo and MSI) | 128 GB; a 64 GB configuration is sold only through OEM partners[22] | 273 GB/s[22] | Up to 1 PFLOP FP4 (theoretical, using the sparsity feature) (sparse)[22] | Qwen 3.8 27B (dense) with multi-token prediction: decode of about 20 tokens/s, with time-to-first-token staying interactive out to the full 262K context.[41] | $6,950.00 (128 GB, 4 TB) on NVIDIA's enterprise marketplace with an Add To Cart button in a copy cached 07:59 UTC on 2026-10-31; stock unverified (listed out of stock on 2026-10-21); 64 GB partner systems from $4,999[42] |
| NVIDIA DGX Spark 64 GBNVIDIA (sold by Acer, ASUS, Dell, Gigabyte, HP and MSI) | 64 GB LPDDR5X unified (two units pool 128 GB over ConnectX-7)[43] | 273 GB/s (256-bit; NVIDIA spec table covers both 64 GB and 128 GB)[22] | 1 PFLOPS FP4 (sparse; NVIDIA spec table covers both memory configurations)[22] | — | from $4,999 (partner systems, Oct 2026; no Founders Edition)[43] |
| AMD Ryzen AI Halo Developer PlatformAMD | 128 GB LPDDR5x-8000 unified[44] | 256 GB/s[44] | — | gpt-oss-120b, vLLM online serving, equal ISL/OSL 256/256: 222 tok/s aggregate at batch 64 vs 701 on DGX Spark (StorageReview)[45] | $3,999 with 2 TB SSD (Jul 2026)[45] |
| Framework Desktop (Ryzen AI Max+ 395)Framework | 128 GB LPDDR5x-8000 soldered (32 GB and 64 GB options)[46] | 256 GB/s[47] | — | gpt-oss-120b MXFP4 in LM Studio (out-of-the-box setup): roughly 45 tok/s (ServeTheHome; Ryzen AI Max+ 395, 128 GB, Windows 11 Pro)[48] | $3,449 (DIY Edition, Max+ 395 with 128 GB, system only; out of stock, Nov 3, 2026)[49] |
| Acer Veriton GN100 AI Mini WorkstationAcer | 128 GB LPDDR5x coherent unified[50] | — | — | gpt-oss-120b, vLLM online serving, equal ISL/OSL: 69.65 tok/s at batch 1 rising to 713.18 tok/s aggregate at batch 64 (StorageReview)[51] | from $3,999 in North America (announced Sep 3, 2025; prices vary by region)[50] |
| ASRock Industrial AI BOX-A395ASRock Industrial | 128 GB LPDDR5X-8000 soldered[52] | — | — | AgentSTH V7, 32 threads: 15% ahead of the next-closest Ryzen AI Max+ 395 box (single-core within 1%); estimated SPEC CPU2026 int rate: top of a 44% spread across four 395 systems; measured by ServeTheHome, Ubuntu 26.04[53] | — |
| Dell Pro Precision with GB300 (model FCT6263; announced as Dell Pro Max with GB300)Dell | 252 GB HBM3e GPU + 496 GB LPDDR5X CPU[54] | — | — | — | $160,195.00 (Dell.com US default config FCT6263 with RTX PRO 2000 display GPU, read Nov 4, 2026); RTX PRO 6000 Workstation Edition option adds $13,560[54] |
| GMKtec EVO-X2 AI Mini PCGMKtec | 64 GB or 128 GB onboard LPDDR5X-8000 (non-upgradeable)[55] | — | — | — | $3,499.99 (128 GB/1 TB, GMKtec US store, Oct 30, 2026; listed against $3,999.99)[55] |
| HP Z2 Mini G1a WorkstationHP | up to 128 GB unified, up to 96 GB assignable to the GPU, up to 8000 MT/s[56] | — | up to 50 NPU TOPS[56] | gpt-oss-120b GGUF (63.39 GB, full GPU offload) in LM Studio: close to 40 tok/s (StorageReview)[57] | $4,781 list for StorageReview's review configuration; about $3,300 street (B&H top spec with 2 TB SSD at $3,342.65, Aug 2025)[57] |
| Microsoft Surface RTX Spark Dev BoxMicrosoft | 128 GB unified memory[58] | — | up to 1 PFLOP of AI compute (Microsoft; precision and sparsity not stated on the page) (sparse)[58] | — | From $5,999.99 (Microsoft Store US, pre-order, read Nov 1, 2026)[58] |
| MSI XpertStation WS300MSI (NVIDIA DGX Station platform) | 252 GB HBM3e GPU + 496 GB LPDDR5X CPU (748 GB coherent)[59] | — | — | StorageReview vLLM: gpt-oss-120b MXFP4 9,294 output tok/s at 128 streams (512/512), 5,038 on 8,192/1,024; MiniMax M3 NVFP4 1,041 at 32 streams with EAGLE3-GQA at a forced acceptance length of 3 (best-case)[59] | $99,999.00 'Start From' at Newegg US (model WS300T60L, 2x 2 TB NVMe RAID1), sold by Newegg as a request-a-quote listing (read Nov 4, 2026)[60] |
| Tenstorrent TT-QuietBox 2 (Blackhole)Tenstorrent | 128 GB GDDR6 on the accelerators + 256 GB DDR5-5600 system memory[61] | — | — | — | $9,999 (Oct 30, 2026)[61] |
| WiCi One (RTX 5060 Ti edition)WiCi | 16 GB GDDR7 VRAM; 1 TB NVMe 4 SSD for TurboStream weight paging (WiCi's claim: NVMe used as working VRAM). System RAM not stated[62] | — | — | — | $2,599 list (WiCi shop, Oct 2026); a $600-off code (WICIWELCOME) is applied at checkout. Early-access price was $1,999 (Notebookcheck, Engadget, Sep 2026)[62] |
| WiCi One (RTX 5090 edition)WiCi | 32 GB GDDR7 VRAM; 4 TB NVMe 5 SSD for TurboStream weight paging (WiCi's claim: NVMe used as working VRAM). System RAM not stated[62] | — | — | — | $9,999 list (WiCi shop, Oct 2026); a $600-off code (WICIWELCOME) is applied at checkout[62] |
Italic values are reported by an outlet or measured by an independent lab; the rest come from the maker. Measured speeds use different models, quantizations and tools, so read each with its source rather than against the others. Voltgrid News does not test hardware itself.
Sources (62)
- [1] ASUS
- [2] ASUS
- [3] StorageReview
- [4] Central Computers
- [5] HP
- [6] StorageReview
- [7] HP Japan
- [8] NVIDIA
- [9] StorageReview
- [10] Exxact
- [11] NVIDIA
- [12] NVIDIA
- [13] Apple
- [14] BGR
- [15] MacRumors
- [16] NVIDIA
- [17] ServeTheHome
- [18] StorageReview
- [19] NVIDIA
- [20] ASUS
- [21] StorageReview
- [22] NVIDIA
- [23] ASUS eShop
- [24] Dell
- [25] HotHardware
- [26] StorageReview
- [27] Framework Blog
- [28] Framework
- [29] GIGABYTE (PR Newswire)
- [30] StorageReview
- [31] CDW
- [32] GMKtec
- [33] GMKtec
- [34] HP
- [35] HotHardware
- [36] StorageReview
- [37] Notebookcheck
- [38] Lenovo PSREF
- [39] ServeTheHome
- [40] MSI
- [41] Tom's Hardware
- [42] NVIDIA Marketplace
- [43] NVIDIA Blog
- [44] AMD
- [45] StorageReview
- [46] Framework
- [47] Framework Blog
- [48] ServeTheHome
- [49] Framework
- [50] Acer
- [51] StorageReview
- [52] ServeTheHome
- [53] ServeTheHome
- [54] Dell
- [55] GMKtec
- [56] HP
- [57] StorageReview
- [58] Microsoft
- [59] StorageReview
- [60] Newegg
- [61] Tenstorrent
- [62] WiCi
Sources
- 01This was a data center a year ago… Now it's on my deskAlex Ziskind (YouTube) · youtube.com
- 02ASUS ExpertCenter Pro ET900N G3-2SSD-V2 AI Workstation (Central Computers listing)Central Computers · centralcomputer.com
- 03ASUS PRO ET900N G3-2SSD NVIDIA DGX Station (SabrePC listing)SabrePC · sabrepc.com
- 04ASUS Launches ExpertCenter Pro ET900N G3 Built on NVIDIA DGX Station ArchitectureASUS · press.asus.com
- 05ASUS ET900N G3 | NVIDIA GB300 | DGX Station Architecture - Tech SpecsASUS · asus.com
- 06ASUS ExpertCenter Pro ET900N G3 Review: The GB300 DGX Station Gets Handles, Titanium Power, and Cooled OpticsStorageReview · storagereview.com







