LenovoThinkSystem SR665 V3

A 2U server with one or two AMD EPYC processors of up to 160 cores, up to 6 TB of memory and room for eight single-slot GPUs or three double-width ones. Lenovo uses it as the base of Hybrid AI 241, its inference configuration for starting small.

Ask for price and lead time See what it supports

Rack 2U

What it supports

The SR665 V3 is built to order: these are its limits according to Lenovo's guide. Lenovo Press product guide

Processor1 or 2 AMD EPYC9005 and 9004 series · up to 160 cores and 400 W

GPUUp to 8 single-slotor 3 double-width · RTX PRO 4500, L4, L40S

MemoryUp to 6 TB24 DDR5 slots · up to 6400 MHz with EPYC 9005

StorageUp to 40 × 2.5"or 20 × 3.5" · NVMe, SAS or SATA

Network1 OCP 3.0 and up to 12 PCIe10 rear slots and 2 front

Form factorRack 2UAir-cooled

Power: Two 750 to 2,600 W power supplies, Platinum or Titanium, or 1,100 W at -48 V DC
Cooling: Air, with optional Neptune Core for the processors
Management: XClarity Controller 2
Operating systems: RHEL, SLES, Ubuntu, Windows Server and VMware ESXi
Warranty: 1 or 3 years depending on the model, with an on-site engineer next business day

Inference, and more, in 2U

It suits a mid-sized language model for one department, a RAG setup over company documents, or a server that runs a few GPUs alongside virtual machines and a database. With up to 40 × 2.5" drives, it also works as a storage node.

If the model doesn't fit in 32 GB per GPU, the SR665 V3 takes three 96 GB RTX PRO 6000 Max-Q cards. For four server-edition RTX PRO 6000 GPUs, you need the SR650a V4.

What we do

We deliver it with RHEL, Ubuntu, SLES, Windows Server or VMware ESXi, with the NVIDIA drivers and whatever runs on top: vLLM serving the model, Kubernetes with the GPU Operator or RAG over your documents. More on this in RAG with Docling.

Supported GPUs

GPU	Memory	How many
NVIDIA RTX PRO 4500 Blackwell	32 GB	Up to 4
NVIDIA L4	24 GB	Up to 7, single-slot
NVIDIA L40S	48 GB	Up to 3, double-width
NVIDIA RTX PRO 6000 Max-Q	96 GB	Up to 3, double-width

All are air-cooled. For four server-edition RTX PRO 6000 GPUs you need the SR650a V4, which carries the GPUs at the front.

What your data centre needs

It takes 2U and is air-cooled with up to six fans, with Neptune Core as an option for the processors. It runs on two power supplies from 750 to 2,600 W, or -48 V DC; Hybrid AI 241 ships with two 1,800 W units.

Hybrid AI 241: inference on a single server

This is the configuration Lenovo offers as a low-cost starting point for serving models on one node or a few. The name comes from what's inside: two processors, four GPUs and one network adapter.

Processor: 2 AMD EPYC 933532 cores each
Memory: 256 GB8 × 32 GB DDR5 modules at 6400 MHz
GPU: 4 NVIDIA RTX PRO 4500 Blackwell32 GB each
Storage: 2 × 3.84 TB NVMePlus 2 × 960 GB M.2 for the operating system
Network: 2 × 25 GbE portsNVIDIA ConnectX-6 Lx
Power: 2 × 1,800 W power suppliesPlatinum

Compared with the rest of the family

Model	What it has	When to choose it
ThinkSystem SR650a V4	Up to 4 RTX PRO 6000 · Intel Xeon 6	The entry point: one to four GPUs in 2U, with Intel.
ThinkSystem SR675 V3	Up to 8 PCIe GPUs · AMD EPYC	Up to eight GPUs with AMD, for several models or teams at once.
ThinkSystem SR680a V4	NVIDIA HGX B300 · 8 GPUs	Eight B300s linked by NVLink, for training and serving large models.
ThinkSystem SR680a V3 with B200	NVIDIA HGX B200 · 8 GPUs	The previous HGX, also air-cooled, when lead time or budget comes first.
ThinkSystem SR780a V3	NVIDIA HGX B200 · liquid-cooled	The same eight B200s in 5U, if your data centre has a water loop.

See all Lenovo AI and GPU servers

What the delivery includes

From the factory

Server built to your configuration by Lenovo
1- or 3-year warranty depending on the model, with an on-site engineer next business day

From us

GPU, memory and network sizing based on your model
Rack mounting, XClarity and up-to-date firmware
Operating system, NVIDIA drivers and the AI stack
Testing with your workload before it goes into production

If you need it

24×7 on-call support with SLA
Lenovo warranty upgrades: 4- or 2-hour response, or one or two extra years

How we build it

What runs on top of these servers, explained in detail:

RAG over your own documents with Docling AI inference with Ceph, OpenStack and Kubernetes Rocky Linux for AI, with NVIDIA drivers and CUDA ready All our AI services

Support and training for your team

Independent support from the people who installed it Local LLM and GPU infrastructure course Deep learning course with PyTorch and TensorFlow Red Hat OpenShift training

Questions about the ThinkSystem SR665 V3

SR665 V3 or SR650a V4?

The SR665 V3 is a general-purpose 2U AMD server that takes GPUs of up to 350 W, and in return offers far more drive bays and slots. The SR650a V4 is built for GPUs: it carries up to four RTX PRO 6000 cards at the front. For a large model with many users, the SR650a V4; for modest inference alongside other workloads, the SR665 V3.

What is Hybrid AI 241 for?

Lenovo positions it as the starting point for serving models on one server or a few, on a modest budget. Four RTX PRO 4500 GPUs add up to 128 GB split across four cards, so the model has to fit within 32 GB per card.

What warranty does it come with?

One or three years depending on the model, with an on-site engineer next business day. Hybrid AI 241 is ordered on the three-year-warranty model. Beyond that, we can provide our own 24×7 on-call support with an SLA.

Is the SR665 V3 right for what you're building?

Tell us which model you want to serve, for how many users and what else will run on the server, and we'll come back with the configuration, the price and a confirmed lead time.

Ask for price and lead time WhatsApp +34 91 198 02 43