Přijímáme kryptoměny — takže to, co držíte, můžete utratit za skutečný hardware.Plaťte kryptem, utrácejte ho za skutečný hardware. Pojištěná doprava zdarma po celém světě nad 399 $. Neutrální přepravní krabice bez loga, odesílané z Německa.Pojištěná doprava zdarma nad 399 $. Jakmile se otevře pokladna, je váš kurz zablokovaný na 30 minut.Kurz zablokovaný na 30 minut. Každý výrobek je zapečetěný, jeho sériové číslo je zkontrolované a vztahuje se na něj 24měsíční záruka.Zapečetěno, záruka 24 měsíců.

Sledované položky Přihlásit se
Full-height dual-slot graphics card with silver shroud, four DisplayPort outputs and NVLink connector

NVIDIA

NVIDIA RTX A6000 48G workstation graphics card 48GB

SKU CH-EW7K-PZ2DJ

Professional workstation GPU with 10,752 CUDA cores, 48GB GDDR6 ECC memory and 84 RT cores for rendering and AI workloads

  • Chipset ManufacturerNVIDIA
  • 10,752 CUDA cores deliver 38.7 TFLOPS single-precision compute for professional workloads
  • 48 GB GDDR6 ECC memory on a 384-bit bus provides 768 GB/s bandwidth with error correction
  • 84 RT cores and 336 Tensor cores combine ray-tracing and AI acceleration in one GPU
  • Four DisplayPort 1.4 outputs drive multi-8K displays and 2-way NVLink expands memory capacity
  • 300 W TDP on 8 nm process with active cooling sustains heavy rendering and scientific tasks

Professional workstation GPU with 48GB GDDR6 ECC

The NVIDIA RTX A6000 48G is a workstation graphics card built for professionals who need large memory capacity and error-correcting memory. It targets users in modeling, rendering and film post-production where data integrity and frame-buffer size are critical. The Ampere architecture drives 10,752 CUDA cores for 38.7 TFLOPS of single-precision compute. Buyers who need multiple GPUs in a single chassis without NVLink support should look elsewhere.

48GB GDDR6 ECC memory on a 384-bit bus

The card provides 48GB of GDDR6 memory with ECC across a 384-bit interface delivering 768 GB/s bandwidth. This capacity handles large-scale scenes, high-resolution textures and complex simulation datasets that would overflow smaller buffers. ECC correction protects against silent data corruption during long compute runs. Two-way NVLink allows memory pooling with a second identical card for workloads that exceed the single-board limit. Workflows requiring more than 48GB per GPU without a second slot cannot be accommodated.

300W TDP with active cooling on 8nm process

The board carries a 300W TDP managed by an active cooler on the 8nm process node. Under sustained rendering or scientific compute loads the cooler maintains stable clocks without throttling. Acoustic output rises with fan speed during peak utilisation but remains within typical workstation expectations. The four DisplayPort 1.4 connectors drive multi-display 8K setups without additional power draw. Systems without adequate case airflow or a sufficient power budget for 300W plus the rest of the platform will hit thermal or electrical limits.

Přednosti

  • 10,752 CUDA cores deliver 38.7 TFLOPS single-precision compute for professional workloads
  • 48 GB GDDR6 ECC memory on a 384-bit bus provides 768 GB/s bandwidth with error correction
  • 84 RT cores and 336 Tensor cores combine ray-tracing and AI acceleration in one GPU
  • Four DisplayPort 1.4 outputs drive multi-8K displays and 2-way NVLink expands memory capacity
  • 300 W TDP on 8 nm process with active cooling sustains heavy rendering and scientific tasks

Specifikace

BrandNVIDIA
SeriesNVIDIA RTX A6000 48G
ModelNVIDIA RTX A6000 48G
Chipset ManufacturerNVIDIA

Otázky k tomuto produktu

Will the 48GB GDDR6 ECC memory and 10,752 CUDA cores work with my existing PCIe 3.0 workstation motherboard?

The card installs in a standard PCIe slot and will operate at the motherboard's supported lane width and generation; the 48GB GDDR6 ECC memory and 10,752 CUDA cores function normally once the system POSTs and the NVIDIA driver loads.

How can I verify the 84 RT cores and 336 Tensor cores are active after driver installation?

Open the NVIDIA Control Panel or run nvidia-smi; the GPU details page lists the RT core and Tensor core counts, and a ray-tracing or AI benchmark will show the 75.6 TFLOPS ray-tracing and 309.7 TFLOPS tensor throughput.

Does the box include the NVLink bridge for the 2-way memory expansion, or must I buy it separately?

The card ships with four DisplayPort 1.4 connectors on the bracket; the NVLink bridge for 2-way memory expansion is not included and must be purchased separately.

Také v této kategorii

Zákazníci také porovnávali

Stejná kategorie, stejná pokladna — osm měn a kurz zamčený na 30 minut.