What's included with the Lenovo ThinkStation PGX (NVIDIA DGX Spark) from EMARQUE?

The Lenovo ThinkStation PGX ships as a complete personal AI supercomputer: the GB10-based system, a 240W external USB-C power adapter and cable, and Lenovo's documentation, with NVIDIA DGX OS pre-installed. All I/O is USB Type-C, so plan for a USB-C-to-USB-A dongle for legacy peripherals. For teams that need deeper deployment help, solutions are available from the EMARQUE AI team on request (pre-validation testing, setup assistance, EMARQUE AI stack, workflow tuning).What sets Lenovo's variant apart in the DGX Spark family (Gigabyte, ASUS, Dell, HP, MSI and others each make their own): ThinkStation reliability and manageability brought to desk-side AI. Lenovo houses the GB10 platform in a palm-sized 150 x 150 x 50.5 mm chassis and differentiates on I/O and security — USB4 20Gb/s Type-C ports with DisplayPort 2.1, a 4TB PCIe Gen5 self-encrypting NVMe M.2 SSD, and the integrated NVIDIA ConnectX-7 with dual 200GbE QSFP ports for connecting two units. Two ThinkStation PGX units link to run models up to 405 billion parameters.

What are the exact specifications of this Lenovo ThinkStation PGX unit?

Published Lenovo ThinkStation PGX specifications: NVIDIA GB10 Grace Blackwell Superchip (20-core Arm CPU — 10x Cortex-X925 + 10x Cortex-A725 — plus a Blackwell GPU); 128GB LPDDR5X coherent unified memory at 273 GB/s; 4TB PCIe Gen5 x4 M.2 2242 NVMe SSD with Opal self-encryption; NVIDIA ConnectX-7 with dual 200GbE QSFP ports plus 10GbE RJ-45; 3x USB-C (USB4 20Gb/s, DisplayPort 2.1) and a separate USB-C power input (PD 3.1); 1x HDMI 2.1a; Wi-Fi 7 and Bluetooth 5.3; 150 x 150 x 50.5 mm chassis from ~1.2 kg. Software is NVIDIA DGX OS (Ubuntu Linux Pro base) with the NVIDIA AI software stack, PyTorch and Jupyter.Performance ceiling is approximately 1 petaFLOP (1,000 TOPS) of FP4 AI compute, supporting models up to 200 billion parameters on a single unit and up to 405 billion by linking two units over ConnectX-7. For exact stock-level configuration questions, email business@emarque.co.

How long is the warranty, and what support does EMARQUE provide?

The Lenovo ThinkStation PGX ships with Lenovo's 1-year Onsite base warranty, upgradable to 3-year Premier Support (24x7 access to senior technicians), covering hardware faults on the unit and its components. EMARQUE's built-in role is local handling of that warranty:Local point of contact — pre-sale, post-sale and warranty all routed through the same EMARQUE team. Real person, no ticket queue.RMA routing — if a hardware fault appears, we route the Lenovo warranty claim on your behalf so you don't have to navigate the manufacturer's process yourself.For deeper post-sale needs — pre-validation testing, OS reinstall, driver and network setup, Spark Stacking configuration, or EMARQUE AI stack deployment — solutions are available from the EMARQUE AI team on request. Scope and charges quoted on enquiry.

What real-world performance does the Lenovo ThinkStation PGX deliver?

The Lenovo ThinkStation PGX delivers the DGX Spark family's standard performance envelope — approximately 1,000 TOPS of AI compute (FP4), 273 GB/s LPDDR5X unified memory bandwidth, and support for local inference on models up to ~200 billion parameters (up to 405 billion by linking two units over their ConnectX-7 200GbE ports).Where Lenovo's variant differentiates:ThinkStation enterprise pedigree — Lenovo's workstation reliability, build quality and manageability brought to a desk-side GB10 AI node, aimed at corporate and regulated deployments.USB4 20Gb/s connectivity — three USB-C (USB4) ports with DisplayPort 2.1, faster than the USB 3.2 I/O on several other DGX Spark variants.4TB self-encrypting NVMe storage — 4TB PCIe Gen5 with Opal hardware encryption for secure-at-rest local storage of datasets, checkpoints and model weights.Independent reviews (ServeTheHome, XDA) confirmed the platform running 200-billion-parameter models locally. Variant-specific throughput and thermals across the Spark family depend on each partner's chassis and SSD configuration; for workload sizing relative to your specific use case, the EMARQUE AI team can review and quote.

1 6

Lenovo ThinkStation PGX NVIDIA DGX Spark Grace Blackwell Computer - 4TB

Name: Lenovo ThinkStation PGX NVIDIA DGX Spark Grace Blackwell Computer - 4TB
Brand: Lenovo
SKU: 30KL000GMY
Price: 24388.00 MYR
Availability: InStock

RM24,388.00

Sold out

Shipping calculated at checkout.

FREE PRE-SALES CONSULTATION with the EMARQUE AI team to understand the right setup and deployment approach for your organization

Corporate Desk-Side AI on the NVIDIA DGX™ Spark Platform

The Lenovo ThinkStation PGX is Lenovo's build of the NVIDIA DGX Spark platform — a palm-sized desktop AI supercomputer powered by the NVIDIA GB10 Grace Blackwell Superchip. Built for developers, AI researchers and data scientists, it brings ThinkStation reliability and manageability to local AI: prototype, fine-tune and run models up to 200 billion parameters on your desk, with your data staying on-premises.

Petaflop-scale AI performance: NVIDIA GB10 Grace Blackwell Superchip — a 20-core Arm CPU (10x Cortex-X925 + 10x Cortex-A725) with a Blackwell GPU, delivering 1 petaFLOP / 1000 TOPS of FP4 compute.
128GB LPDDR5X unified memory: coherent memory shared across CPU and GPU via NVLink-C2C at 273 GB/s, for running and fine-tuning models up to 200 billion parameters on a single unit.
Scale to 405B parameters: link two ThinkStation PGX units over their integrated NVIDIA ConnectX-7 200GbE (dual QSFP) ports to work with models up to 405 billion parameters.
4TB self-encrypting NVMe storage: 4TB PCIe Gen5 x4 M.2 2242 NVMe SSD with Opal hardware self-encryption for datasets, checkpoints and model weights.
Connectivity: 10GbE RJ-45, 3x USB-C (USB4 20Gb/s, DisplayPort 2.1), HDMI 2.1a, Wi-Fi 7 and Bluetooth 5.3.
Compact and efficient: a 150 x 150 x 50.5 mm chassis, ~1.2 kg, powered by a 240W supply.
Full-stack AI: NVIDIA DGX OS (Ubuntu Linux Pro base) with the complete NVIDIA AI software stack — PyTorch, Jupyter and framework images — pre-installed for local development, fine-tuning and inference.

Supplied in Malaysia by EMARQUE with local warranty and setup support. Talk to the EMARQUE AI team to scope the right setup for your workload.

WhatsApp Us: 012 627 2280 View full details

FAQ

Frequently asked questions

Everything you need to know about the Lenovo ThinkStation PGX — what's included, support, and Malaysian fulfilment.

What's included with the Lenovo ThinkStation PGX (NVIDIA DGX Spark) from EMARQUE?

The Lenovo ThinkStation PGX ships as a complete personal AI supercomputer: the GB10-based system, a 240W external USB-C power adapter and cable, and Lenovo's documentation, with NVIDIA DGX OS pre-installed. All I/O is USB Type-C, so plan for a USB-C-to-USB-A dongle for legacy peripherals. For teams that need deeper deployment help, solutions are available from the EMARQUE AI team on request (pre-validation testing, setup assistance, EMARQUE AI stack, workflow tuning).
What sets Lenovo's variant apart in the DGX Spark family (Gigabyte, ASUS, Dell, HP, MSI and others each make their own): ThinkStation reliability and manageability brought to desk-side AI. Lenovo houses the GB10 platform in a palm-sized 150 x 150 x 50.5 mm chassis and differentiates on I/O and security — USB4 20Gb/s Type-C ports with DisplayPort 2.1, a 4TB PCIe Gen5 self-encrypting NVMe M.2 SSD, and the integrated NVIDIA ConnectX-7 with dual 200GbE QSFP ports for connecting two units. Two ThinkStation PGX units link to run models up to 405 billion parameters.
What are the exact specifications of this Lenovo ThinkStation PGX unit?

Published Lenovo ThinkStation PGX specifications: NVIDIA GB10 Grace Blackwell Superchip (20-core Arm CPU — 10x Cortex-X925 + 10x Cortex-A725 — plus a Blackwell GPU); 128GB LPDDR5X coherent unified memory at 273 GB/s; 4TB PCIe Gen5 x4 M.2 2242 NVMe SSD with Opal self-encryption; NVIDIA ConnectX-7 with dual 200GbE QSFP ports plus 10GbE RJ-45; 3x USB-C (USB4 20Gb/s, DisplayPort 2.1) and a separate USB-C power input (PD 3.1); 1x HDMI 2.1a; Wi-Fi 7 and Bluetooth 5.3; 150 x 150 x 50.5 mm chassis from ~1.2 kg. Software is NVIDIA DGX OS (Ubuntu Linux Pro base) with the NVIDIA AI software stack, PyTorch and Jupyter.
Performance ceiling is approximately 1 petaFLOP (1,000 TOPS) of FP4 AI compute, supporting models up to 200 billion parameters on a single unit and up to 405 billion by linking two units over ConnectX-7. For exact stock-level configuration questions, email business@emarque.co.
How long is the warranty, and what support does EMARQUE provide?
The Lenovo ThinkStation PGX ships with Lenovo's 1-year Onsite base warranty, upgradable to 3-year Premier Support (24x7 access to senior technicians), covering hardware faults on the unit and its components. EMARQUE's built-in role is local handling of that warranty:
- Local point of contact — pre-sale, post-sale and warranty all routed through the same EMARQUE team. Real person, no ticket queue.
- RMA routing — if a hardware fault appears, we route the Lenovo warranty claim on your behalf so you don't have to navigate the manufacturer's process yourself.
For deeper post-sale needs — pre-validation testing, OS reinstall, driver and network setup, Spark Stacking configuration, or EMARQUE AI stack deployment — solutions are available from the EMARQUE AI team on request. Scope and charges quoted on enquiry.
What real-world performance does the Lenovo ThinkStation PGX deliver?
The Lenovo ThinkStation PGX delivers the DGX Spark family's standard performance envelope — approximately 1,000 TOPS of AI compute (FP4), 273 GB/s LPDDR5X unified memory bandwidth, and support for local inference on models up to ~200 billion parameters (up to 405 billion by linking two units over their ConnectX-7 200GbE ports).
Where Lenovo's variant differentiates:
- ThinkStation enterprise pedigree — Lenovo's workstation reliability, build quality and manageability brought to a desk-side GB10 AI node, aimed at corporate and regulated deployments.
- USB4 20Gb/s connectivity — three USB-C (USB4) ports with DisplayPort 2.1, faster than the USB 3.2 I/O on several other DGX Spark variants.
- 4TB self-encrypting NVMe storage — 4TB PCIe Gen5 with Opal hardware encryption for secure-at-rest local storage of datasets, checkpoints and model weights.
Independent reviews (ServeTheHome, XDA) confirmed the platform running 200-billion-parameter models locally. Variant-specific throughput and thermals across the Spark family depend on each partner's chassis and SSD configuration; for workload sizing relative to your specific use case, the EMARQUE AI team can review and quote.