We take cryptocurrency — so the coins you hold can buy real hardware.Pay in crypto, spend it on real hardware. Free insured shipping worldwide over $399. Plain, unbranded boxes, sent from Germany.Free insured shipping over $399. Your rate is locked for 30 minutes once checkout opens.Rate locked 30 minutes. Every item sealed, serial-checked and covered by a 24-month warranty.Sealed, 24-month warranty.

Saved items Sign in
Full-height dual-slot graphics card with matte black shroud and silver bracket

GIGABYTE

GIGABYTE 25FC9-RTX004-P2R 96GB ai accelerator card

SKU CH-WMJW-8AKGS MPN 25FC9-RTX004-P2R

AI accelerator card with 96 GB GDDR7 ECC memory, 24,064 CUDA cores, 752 Tensor cores and 188 RT cores in a 4.4" H x 10.5" L dual-slot form factor

  • 96 GB GDDR7 with ECC powers massive AI models
  • 24,064 CUDA cores accelerate parallel compute workloads
  • 752 fifth-gen Tensor cores speed up deep learning training
  • 188 fourth-gen RT cores enable real-time rendering
  • 4" by 10.5" dual-slot form factor fits standard server chassis

96 GB AI accelerator card for server workloads

The GIGABYTE 25FC9-RTX004-P2R delivers 96 GB of GDDR7 memory with ECC, 24,064 CUDA parallel processing cores, 752 fifth-generation Tensor cores and 188 fourth-generation RT cores built on the NVIDIA Blackwell architecture. It targets data-centre teams running large-model inference, training or high-performance compute jobs that need maximum VRAM capacity in a single slot.

Check power delivery and chassis clearance before

The card uses a full-height full-length dual-slot form factor measuring 4.4 inches high by 10.5 inches long, so the server chassis must accept that length and height. Verify that the power supply provides the required auxiliary connectors and that the motherboard offers a compatible PCIe slot with sufficient lane width for the intended workload.

FHFL dual-slot build with ECC memory for reliability

The dual-slot bracket and full-length PCB are designed for dense server racks where adjacent cards sit close together. GDDR7 with ECC helps prevent silent data corruption during long training runs, while the sturdy metal bracket resists flex when the card is seated in a vertical backplane. Connectors are positioned along the rear edge for clean cable routing in rackmount enclosures.

How it compares

Measured against everything we stock in GPUs for AI & Data Centers, at today’s prices. Recalculated every night.

Price $13,475.99
Pricier than about 81% of the 36 comparable items (median $5,369.99).

New to this? Buying server hardware with crypto — what to check, what it costs to pay, and what is never asked for.

Specifications

Size & build
Shipping weight1.57 kg
Package size356 × 216 × 76 mm

Questions about this item

Does the 96 GB GDDR7 card ship with any cables, brackets or software licences, or must they be purchased separately?

The card is supplied as a standalone FHFL dual-slot unit; no cables, brackets or licences are included in the box.

The model is listed as a 6000-class 96 GB Blackwell accelerator — is there a lower-capacity version in the same lineup?

This specific MPN is the 96 GB configuration; other memory capacities would carry different part numbers.

With 24,064 CUDA cores and 752 fifth-generation Tensor cores, what workload size fits comfortably in the 96 GB frame buffer?

The 96 GB GDDR7 memory with ECC allows large language models and high-resolution datasets to reside entirely on the GPU without host-memory paging.

More in GPUs for AI & Data Centers

Similar items