EstimatesHardware prices change daily, so prices here are estimates. Add what you need to your basket and request your final price: we reply within one working day.
AMD · SKU AMD-MI350P

AMD Instinct MI350P 144 GB PCIe

PCIe card with 144 GB HBM3E for running generative AI inside the servers you already have. Priced on quote.

Pricing
On quote
Built to order · quote

AMD Instinct MI350P 144 GB PCIe specifications

AMD Instinct MI350P 144 GB PCIe specifications
Board power600 W (450 W configurable)
Compute units128 (8,192 stream processors)
Peak FP16 / BF16 matrix1.15 PFLOPS
Peak FP8 (OCP)2.3 PFLOPS (4.6 with sparsity)
Memory bandwidthUp to 4 TB/s
Power connector12V-2x6
Peak MXFP4 / MXFP64.6 PFLOPS
Peak FP6436 TFLOPS
SoftwareAMD ROCm (Linux x86-64)
CoolingPassive (server airflow)
GPU memory144 GB HBM3E, 4096-bit
Form factorPCIe 5.0 x16, full height, double slot, 267 mm
ArchitectureAMD CDNA 4

View the AMD product page (opens in a new tab)

Request a quote

You'll get a quote link and PDF immediately; our team follows up within one working day.

AMD Instinct MI350P 144 GB PCIe is built to order and priced per project (configuration, delivery, installation and support).

Video from AMD

Introducing the AMD Instinct MI350 Series GPUs: Ultimate AI & HPC Acceleration. Played from the official AMD YouTube channel. Watch on YouTubeProduct names shown are trademarks or registered trademarks of AMD. Shown to describe the products we supply; no endorsement is implied.
Will it run?

Check a model against our machines before you buy.

Pick the model and quantisation you plan to run. We'll show whether it fits in memory, how much room is left for context, and the throughput to expect. Estimates come from memory bandwidth and are replaced by our own bench figures where we have them.

Machine

Quantisation

Tokens of context you'll actually use16k
Fits
Llama 3.3 70B at Q4_K_M runs on the DGX Spark
16k context · 128 GB available · bandwidth estimate
Needs
58 GB
Headroom
+70 GB
Est. speed
3.8 tok/s · est.

Comfortable. 70 GB spare — room for 128k context, a second small model, or a RAG index in memory. Expect roughly 3.8 tok/s single-stream.

Machines that run this configuration

Not sure? Book a free 20-minute sizing call

Weights 42.8 GB at 4.85 bits/param · KV cache 5.0 GB at 16k (fp16, approximate) · 10 GB reserved for OS and runtime. Speed is single-stream decode, memory-bound; prompt processing and batching behave differently. Or WhatsApp an engineer on .

The MI350P puts AMD's 4th-generation CDNA architecture on a standard double-slot PCIe card. With 144 GB of HBM3E on a 4096-bit interface and up to 4 TB/s of bandwidth, a single card holds models that would need two smaller GPUs, and the new MXFP4 and MXFP6 data types speed up inference further. Board power is 600 W, or 450 W where your rack's power or cooling budget is tighter.

It runs on AMD ROCm, the open software stack, with PyTorch, TensorFlow, JAX, Triton, SGLang and ONNX Runtime support and no per-GPU licence fees.

Fitted in a Workstation or Dell server, or supplied standalone where the form factor allows. Before it ships we check it against your chassis (power, airflow, slot layout and platform support), burn-in test it and install the AMD ROCm drivers and software stack. Warranty claims and advance replacements are handled by us, and support comes from our own UK engineers.

Priced on quote: tell us what you need and we will confirm price, availability and lead time.

Questions

What warranty do I get?

Systems we build carry the warranty stated on your quote: three years return-to-base as standard. Parts carry their manufacturer's warranty, and we help you claim it.

Are prices shown with VAT?

Prices are shown ex VAT. VAT (20%) is added in your basket and on your quote. You can put your own purchase order number on a quote.

How much is delivery?

UK delivery is £35.00 ex VAT, shown before you pay. For high-value systems we may book a delivery slot or ask for a signature.

Can I return it?

Business customers can return unopened stock items within 30 days of delivery (ask us for a returns number first). If you buy as a consumer, you can cancel within 14 days of receiving the goods. Faulty goods are always covered.

Full details in our Terms of Sale

Related guides

Similar machines

NVIDIA
NVIDIA H200 (SXM)
  • Form factor: SXM (HGX)
  • GPU memory: 141 GB HBM3e
  • Architecture: NVIDIA Hopper
Pricing
On quote
Built to order · quote
NVIDIA
NVIDIA H100 PCIe 80 GB
  • Interface: PCIe
  • GPU memory: 80 GB HBM2e
  • Form factor: Dual-slot, full height, full length
Pricing
On quote
Built to order · quote
NVIDIA
NVIDIA L40S 48 GB
  • Interface: PCIe
  • GPU memory: 48 GB GDDR6 with ECC
  • Form factor: Dual-slot, 4.4" H × 10.5" L
Pricing
On quote
Built to order · quote