NVIDIA 170HX Capacitor‑Modified AI Accelerator (64GB Mode Enabled, Base 8GB HBM2e)
NVIDIA 170HX Capacitor‑Modified AI Accelerator (64GB Mode Enabled, Base 8GB HBM2e)
NVIDIA 170HX Capacitor‑Modified AI Accelerator (64GB Mode Enabled, Base 8GB HBM2e)
NVIDIA 170HX Capacitor‑Modified AI Accelerator (64GB Mode Enabled, Base 8GB HBM2e)

NVIDIA 170HX Capacitor‑Modified AI Accelerator (64GB Mode Enabled, Base 8GB HBM2e)

copy Link
Facebook
X
LinkedIn
Line
Tumblr
NVIDIA 170HX compute‑only accelerator with custom capacitor modification, unlocking 64GB extended memory mode for AI video workflows. LTX 2.3 ready, blower cooling with dual 8‑pin power. Ideal for generative video, long‑sequence video generation and multimodal AI inference.
$3200.00
Quantity
1 items selected,Total $3200.00
Add to Cart
Buy Now
+8618819251189
Delivery to
Please add the address first.
Next: NVIDIA DGX Station™​ —— The Ultimate Deskside AI Supercomputer
Specifications
Detail
The NVIDIA 170HX is a data center compute accelerator originally equipped with 8GB HBM2e. This version undergoes a professional capacitor modification, which stabilizes power delivery under extreme burst compute load and enables the card to run in 64GB memory mode for AI video pipelines.
The upgraded capacitor circuit suppresses voltage sag and power spikes that happen during long video generation runs, which is the critical prerequisite to activate the extended 64GB memory addressing capability. It is fully compatible with the LTX 2.3 video generation framework, providing the memory headroom required for longer video clips and high‑resolution generative video tasks that cannot run on the stock unmodified 170HX.
It adopts blower‑style turbo cooling, exhausting hot air out the rear of the card. The dual 8‑pin power design fits standard rack‑mount servers and dense multi‑GPU AI workstation clusters. This is a compute‑only accelerator with no display outputs, dedicated solely to AI computing workloads.
This modified variant is widely sought after by AI video developers, small R&D labs and system integrators working on local generative video, long‑sequence video synthesis and multimodal agent AI workloads.
Important Disclosure: This is a custom hardware‑modified variant, not an original NVIDIA factory SKU. The 64GB memory mode is achieved via capacitor modification, the native factory HBM2e of this card is 8GB.

Key Specifications

  • Model: NVIDIA 170HX (Capacitor Modified)

  • Native Onboard Memory: 8GB HBM2e

  • Modified Feature: Extra capacitor upgrade, stabilized power delivery, supports 64GB memory addressing mode

  • Primary Use: AI Video Computing, Generative Video, Multimodal AI Inference

  • Framework Support: LTX 2.3 Ready

  • Cooling: Blower / Turbo cooling

  • Power Connector: Dual 8‑pin

  • Form Factor: Dual‑slot PCIe accelerator

  • Display Outputs: None (Compute‑only)

  • Recommended Scenarios: Long‑sequence AI video generation, LTX 2.3 video model testing, offline multimodal AI development

Notes:

  1. This is NOT an official NVIDIA factory configuration. The 64GB mode depends on the capacitor modification for power stability.

  2. Confirm compatibility with your server motherboard, power supply, BIOS and software environment before purchase.

  3. Stability of 64GB mode is subject to workload, PSU quality and thermal condition.


WhatsApp：+86-18819251189