P54869-001: HPE A100 80GB HBM2e PCIe 4.0 GPU Accelerator (SPS-PCA, Ampere architecture, dual-slot, for HPE ProLiant/Apollo servers)
MPN: P54869-001|By: HPE
Starting from
$60,370
Indicative. Final pricing is confirmed with your quote.
In Stockships worldwide
- Condition:
- New (authorized channel)
- Warranty:
- Full manufacturer warranty
Prefer your own inbox? Email us about this part
Bulk order? Request volume pricing
Authorized & warrantied
Sourced through authorized distribution, with the full manufacturer warranty.
Compatibility checked
Validated against your server platform before dispatch.
Delivery & lead time
Confirmed with your quote, including expedited freight where available.
Business & public sector
POs, net terms and volume pricing for qualified organizations, government and edu.
Product overview
Part number P54869-001 identifies this HPE HPE/NVIDIA A100 80GB HBM2e PCIe 4.0 GPU accelerator (SPS-PCA, Ampere architecture, dual-slot, for HPE ProLiant/Apollo servers). Passively cooled accelerators need chassis airflow to survive and will throttle or shut down in a workstation. Tell us where it is going and we will confirm the cooling before you order.
Quotes are itemised down to the accessories, so nothing that is needed to make the part work is left off and discovered later.
Specifications
General
- Manufacturer part number
- P54869-001
- Product Type
- GPUs & Accelerators
- Manufacturer
- HPE
Technical specifications
- Memory
- 80GB
These are the 3 facts we can state about P54869-001 without inferring anything. Compatibility beyond this is confirmed per platform, on request, before you order.
Where the HPE P54869-001 is deployed
- Like-for-like replacement of P54869-001 in an existing system
- Building a training or inference node to a fixed budget
- Standardising a cluster on one accelerator model
Who this part is for
- Procurement buying accelerators on a purchase order at lead time
- ML teams adding accelerators to servers that are already racked
- Research groups sizing a training or inference node to a budget
What you get ordering this from us
Ordered by part number
Quoted and supplied against P54869-001 specifically, so what arrives is what your build sheet asks for.
Compatibility checked
We confirm the part is supported on your platform before it ships, rather than leaving the check to you.
Power chain checked end to end
From the card connectors to the supply to the rack circuit, we check the whole chain rather than the card alone.
Questions buyers ask about this part
Is P54869-001 genuine and is the warranty intact?
Yes. Supplied new and unopened, so the manufacturer warranty and firmware update path stay supported. No grey-market stock and no refurbished units sold as new.
Will P54869-001 work in my server?
Send the chassis model. Slot width, auxiliary power connectors and airflow decide this before the silicon does, and a passively cooled card in the wrong chassis will throttle or shut down.
Do you accept purchase orders and quote at volume?
Yes. We quote against a purchase order and hold the price for the quote period, with tiered rates at quantity. Net terms are available once the account is approved.
Compare other GPUs & Accelerators we stock
| Model | Brand | Key spec | Availability | From |
|---|---|---|---|---|
| NVIDIA L4 24GB PCIe Data Center GPU Accelerator | NVIDIA | Memory: 24GB GDDR6 | In Stock | $3,120 |
| AMD Instinct MI210 64GB HBM2e Data Center GPU Accelerator | AMD | Memory: 64GB HBM2e | In Stock | $10,920 |
| Intel Data Center GPU Flex 140 12GB GDDR6 PCIe Accelerator | Intel | Memory: 12GB GDDR6 | In Stock | $935 |
| Dell NVIDIA L4 24GB GPU Kit with Riser and Bracket for PowerEdge Servers | Dell Technologies | Memory: 24GB GDDR6 | In Stock | $4,420 |
| Xilinx Alveo U50 Data Center Accelerator Card | Xilinx | N/A | In Stock | $7,930 |
| ASRock Intel Arc Pro B70 Creator 32GB graphics card | ASRock | Memory: 32GB | In Stock | Request quote |
Browse all GPUs & Accelerators or everything we stock from HPE.
Specifying a system, not just the part?
Accelerators are usually bought as part of a node rather than on their own, because the chassis, power and cooling have to match the card.