
NVIDIA
NVIDIA A100 40GB PCIe
Ampere data-center card with 40 GB HBM2 at 250 W, a cost-controlled option for inference fleets and HPC clusters.
| Tình trạng hàng | Còn hàng |
|---|---|
| SKU | AXL-GPU-A100-40 |
| Số lượng đặt tối thiểu | 1 unit |
| Tình trạng | Tân trang |
| Giá | Theo báo giá |
Tổng quan
The A100 40GB PCIe belongs to the earlier Ampere generation and is now bought mainly on price per unit of throughput. It stays capable for inference serving, classical HPC, data analytics and research or teaching clusters, and Multi-Instance GPU lets one card be partitioned into as many as seven isolated instances for multi-tenant use. At 250 W it is easier to accommodate in older servers than current Hopper or Blackwell parts.
New production of this model has ended, so AXISLINK supplies verified refurbished units and says so plainly rather than advertising stock we cannot secure. Each card is inspected, tested under sustained load and tied to a documented serial number on the quotation. Cards ship loose or installed into a refreshed server platform. Export documents and freight are handled from Hong Kong, and end-user information may be required.
Thông số kỹ thuật
| Architecture | NVIDIA Ampere |
|---|---|
| Memory | 40 GB HBM2 |
| Memory bandwidth | 1555 GB/s |
| Multi-Instance GPU | Up to 7 instances of 5 GB |
| Form factor | PCIe 4.0 x16, dual-slot full-height full-length, passively cooled |
|---|---|
| NVLink | 600 GB/s bridge between two cards |
| Board power | 250 W |
|---|---|
| Cooling | Passive, requires server chassis airflow |
Yêu cầu báo giá
Báo giá cho NVIDIA A100 40GB PCIe. Chúng tôi sẽ phản hồi bằng hóa đơn chiếu lệ chính thức, thường trong vòng 24 giờ.
Sản phẩm khác trong GPU và bộ tăng tốc
Xem danh mục
NVIDIA
NVIDIA H200 141GB SXM
Hopper SXM module with 141 GB HBM3e for large-model training and memory-bound inference on HGX baseboards.

NVIDIA
NVIDIA L40S 48GB
Ada Lovelace data-center card with 48 GB GDDR6 ECC for inference, fine-tuning, rendering and virtual workstations.

NVIDIA
NVIDIA H100 80GB
Hopper-architecture data-center GPU for large-scale AI training and inference. PCIe and SXM variants.