BLOG · GUIDE · AUGUST 2026

Nutanix and GPUs: building an AI cluster

IN BRIEF
  • Six NX platforms take GPUs today, across the G9 and G10 generations. The per-node ceiling: four L40S or H100 NVL in the NX-9151-G9
  • The model code does not tell you: the NX-8150G-G10 takes three cards, the NX-8155-G9 in the same 2U1N chassis is not on Nutanix’s GPU list at all
  • The L40S does not split with MIG at all; the RTX PRO 6000 Server Edition splits into four 24 GB instances
  • MIG-backed vGPU is listed for vSphere, Red Hat KVM and Ubuntu KVM; AHV is not named
  • MIG instances cannot talk peer-to-peer across GPUs and NCCL does not work with MIG, so distributed training will not assemble on slices

Which nodes take cards, and which do not

A hyperconverged node is a server where disks, compute and networking are fused into one block under AOS. Room for a full-size card is not always there: drive bays and fans eat it. So some NX models take up to four accelerators and others take none.

NODEFORM FACTORGPUs
NX-9151-G92U1Nup to 4× L40S or H100 NVL 94 GB
NX-3155-G92U1Nup to 2× L40S, or up to 4× L4
NX-8150G-G102U1Nup to 3× L4, A16 or L40S
NX-8155AS-G102U1Nup to 3× L4 or A16
NX-1175S-G101U1Nup to 2× L4
NX-1150S-G92U1NGPU option for the edge (L4)
NX-8170, NX-3060, NX-3035, NX-8155, NX-8155A, NX-8173AS, NX-1120AS1U1N / 2U1N / 2U2N / 2U4Nnot on the GPU list

Per Nutanix’s platform FAQ and the G10 launch material (G10 nodes ship since October 2025); confirm the spec sheet of the exact generation before ordering

Reading GPU capability from the model code does not work; the table shows it directly. There are two reliable checks: the GPU line in the model’s spec sheet, and the card list validated for that chassis. The NX-9151-G9 accepts the H100 NVL; the NX-3155-G9 with the same 2U1N form factor stops at two L40S; the H100 80 GB PCIe is qualified but not sold by Nutanix. Power and airflow set the limit, not case size.

What blocks a fifth card

The power spread across the line is large: an edge node draws a few hundred watts, while an NX-9151-G9 with four 350 W cards needs a power budget of several kVA and a matching cooling budget, a separate conversation with your data centre about the phase and row cooling.

The second reason: server cards are passive. The L40S and RTX PRO 6000 Server Edition have no fans of their own and live off front-to-back chassis airflow. NVIDIA’s forum has a telling case: an L40S in a Dell Precision workstation idling at 100 °C. Outside a server, this cannot be fixed.

Three ways to hand a card to a VM

METHODHOW IT DIVIDESISOLATION & LICENCE
Passthroughwhole card to one VMfull; no vGPU licence needed
Time-sliced vGPUtime division, processes take turnsdedicated frame buffer per VM, shared compute, no fault isolation; licence mandatory
MIG-backed vGPUhardware slices working in parallelhardware isolation; NVIDIA’s sizing guide measured 20% higher performance than time-sliced at the 48 GB profile size

Then comes the limit that breaks half of all multi-card node plans. The MIG documentation states it plainly: with MIG enabled there is no peer-to-peer between instances on different GPUs, over PCIe or NVLink, and NCCL is unsupported. NVIDIA’s forum position: distributed training over MIG can theoretically be pushed through GPUDirect RDMA over the network, but it will be slower than taking a bigger slice or disabling MIG. Partitioning a card and fast card-to-card exchange are mutually exclusive modes.

Which card splits how far

CARDMAX MIG INSTANCESSMALLEST SLICE
RTX PRO 6000 Blackwell SE, 96 GB41g.24gb: 24 GB, 46 SM
RTX PRO 4500 Blackwell, 32 GB21g.16gb: 16 GB
H100 NVL, 94 GB (the H100 Nutanix sells)71g.12gb: 12 GB
H200 NVL, 141 GB71g.18gb: 18 GB
L40S, L4, A2, A16no MIGpassthrough or time-sliced vGPU only

The last row deserves attention when picking a node: the L40S is the workhorse card for the NX-9151-G9 and NX-3155-G9, and it is exactly the one that does not slice. “How do we partition the L40S then?” is a question asked on NVIDIA’s forum by people who chose the card by TFLOPS and looked at the partitioning model afterwards.

What works on AHV

NVIDIA maintains a dedicated Nutanix AHV support page and classes the hypervisor as generic Linux with KVM. Ordinary time-sliced vGPU works. But in the consolidated feature table, MIG-backed vGPU is ticked only for vSphere, Red Hat KVM and Ubuntu KVM from release 19.0. Verify compatibility for your AOS version before ordering.

Two more things surface after deployment. Live migration of a vGPU VM requires the identical physical card on the receiving host and a driver from the same release branch that is the same or newer, and never works across different MIG profiles. And unlicensed behaviour: 20 minutes at full speed, then a 15 fps cap with degraded CUDA; after 24 hours 3 fps and no CUDA. C-series compute profiles come only with NVIDIA AI Enterprise (licensed per GPU), graphics vWS per concurrent user: two separate budget lines.

How many VMs per card

vGPU TYPEMEMORY PER VMVMs PER CARD
DC-96C96 GB1
DC-48C48 GB2
DC-24C24 GB4
DC-12C12 GB8
DC-8C8 GB12

Compute profiles for the RTX PRO 6000 Blackwell Server Edition

For virtual desktops NVIDIA’s sizing guide gives two to four heavy users per RTX PRO 6000 (DC-12Q to DC-24Q profiles) and six to eight light users per RTX PRO 4500 Server Edition (DC-4Q). And the budget-saving caveat: the best-effort scheduler often allows 2–3× higher user density. The monitoring rule is simple: VM video-memory fill should rarely cross 90% and average below 70%.

WHAT WAS MEASUREDNO PARTITIONINGWITH MIG
flan-t5-base on A100 40 GB (Red Hat)~4,200 tok/s~9,800 tok/s on 7 slices
P95 latency next to noisy neighbours (vLLM, community test)~2,499 ms~1,319 ms
Single small task on a 1/7 slice (NVIDIA forum microbenchmark)0.00533 s0.02640 s

The last row is the price of partitioning: an individual slice is roughly five times slower than the whole card. Aggregate throughput still grows, because a small model does not saturate an A100, and scaling is sublinear.

Where to start and how to grow

The working starting configuration is three nodes: RF2 replication runs on three, and the cluster survives losing one node without stopping VMs. Cards are not needed in every node; the pattern that most often works: three nodes for general workloads, one or two of them with accelerators for inference.

Keep the GPU pool homogeneous, and the reason is live migration. A second node with the same card gives headroom and a place for VMs to move during maintenance. A mixed cluster with an L40S in one node and an RTX PRO 6000 in another works too, but every planned maintenance turns into an evening of service stops.

What we supply

Eurokommerz supplies the Nutanix NX line together with the GPUs Nutanix lists for it: L40S, L4, A16 and H100 NVL, and the RTX PRO 6000 Server Edition on the newest nodes, with EU invoicing and manufacturer warranty. Deployment and virtualisation engineering comes from our partner Vixen.UNO under the same contract.

FAQ

Which Nutanix NX nodes support GPUs?
Six platforms across the G9 and G10 generations: NX-9151-G9 (up to four L40S or H100 NVL), NX-3155-G9 (up to two L40S or four L4), NX-8150G-G10 (up to three L4, A16 or L40S), NX-8155AS-G10 (up to three L4 or A16), NX-1175S-G10 (up to two L4) and the NX-1150S-G9 edge node.
How many GPUs does an NX-3155-G9 hold?
Up to two L40S or up to four L4, in a 2U single-node chassis; the H100 80 GB is qualified but not sold by Nutanix. Power and airflow, not the case, set the ceiling.
Does NX-8155-G9 support H100?
No. The NX-8155-G9 is not on Nutanix’s GPU list at all; its GPU sibling is the NX-8150G-G10 with up to three L4, A16 or L40S. The H100 NVL is limited to the NX-9151-G9.
The L40S does not support MIG. How do we divide it?
Time-sliced vGPU. The L40S-8Q profile gives six instances in a homogeneous layout, four in a mixed one. There is no hardware memory or fault isolation; the card switches between VMs in time.
Do we need AI Enterprise to hand a MIG instance to a VM?
For C-series compute profiles, yes: they ship only with AI Enterprise, licensed per GPU in the server. MIG itself on bare Linux needs no licence; the money starts where a slice meets a hypervisor.
The card is capped at 450 W instead of 600. Is it defective?
Most often it is the cable: the 12V-2×6 sense configuration sets the power limit, and some bundled server cables are configured for 450 W. The 600 W part is a separate SKU; put it in the BOM.
What do two cards in a node give over one?
Not one memory pool. Two 48 GB cards are two independent 48s. The value: several models resident side by side (reranker, embeddings, main LLM), or twice the users.
Can we mix L40S and RTX PRO 6000 nodes in one cluster?
It works, but live migration requires identical cards, so each planned maintenance means stopping services on the odd node. Keep the GPU pool homogeneous where you can.

Tell us which models you run and how many people work with them. An engineer will pick the node and card count, price the licences, and say where the configuration hits the rack’s power limit. We reply within one business day.

Talk to an expert
Talk to an expert

We reply within one business day

By sending this form you agree that we process your details to answer your enquiry – see our privacy policy.

request@eurokommerz.at  ·  +43 1 585 1405 50  ·  Jordangasse 7, 1010 Vienna