AWS EC2 Instances›GPU Instance›p2›p2.8xlarge

Compute

vCPUs32
Memory488 GiB
Mem / vCPU15.25 GiB
ProcessorXeon E5-2686 v4
Clock speed2.3 GHz
Architecturex86_64
GPU8
GPU WattageN/A
GPU ArchitectureN/A
Video Memory96 GiB
FPGA0
ffmpeg FPSN/A
CoreMark / secN/A

Storage

TypeEBS only
BandwidthN/A
ThroughputN/A
IOPS0
Base BandwidthN/A
Base ThroughputN/A
Base IOPS0
Devices0
NVMe DriveNo
SSDNo
Swap PartitionNo

Networking

Performance10 Gigabit
Max ENIsN/A
IPv6Yes
GPUNVIDIA Tesla K80
Placement GroupSupported
TrunkingNo
Branch Interface0
IPs / ENIN/A
Baseline BandwidthN/A

NUMA Architecture

Uses NUMANo
Node CountN/A
Max DistanceN/A
Cores / NodeN/A
Threads / NodeN/A
Memory / NodeN/A
L3 Cache / NodeN/A
L3 Cache SharedN/A

AWS Metadata

Instance typep2.8xlarge
Full nameP2 General Purpose GPU Eight Extra Large
FamilyGPU instance
Variantp2
GenerationCurrent
Bare MetalNo
EMR supportYes
DedicatedSupported
Introduced2016-09-29
Family sizes

p2.8xlarge

AWS's p2.8xlarge is a GPU instance with 32 vCPU and 488 GiB. Rackspace Spot has no equivalent to offer: 8 × NVIDIA Tesla K80 wired together behind a high-speed interconnect is one tightly coupled machine, not 8 independent nodes, and that interconnect is usually what the instance was bought for.

Rackspace Spot's largest VM tops out at 16 vCPU and 128 GiB, and even that is not the missing piece here. The table below prices what AWS charges; the honest answer on the Rackspace Spot side is that this workload stays where it is.

Server comparison caveats
No equivalent

The AWS p2.8xlarge is 8 GPUs (96 GiB VRAM) wired together inside one chassis. The interconnect is what you bought (that is what makes it one accelerator with pooled memory rather than several machines), and Rackspace Spot has nothing that reproduces it. The Rackspace Spot GPU class is built on the NVIDIA A30 (Ampere, 24 GB VRAM), and separate nodes of it on ordinary datacentre networking are not a substitute for a single interconnect-joined node.

Split the job across independent single-GPU nodes only if it is data-parallel with no per-step synchronisation and the model fits one card. Model-parallel work, or anything synchronising gradients every step, does not survive the move. Rackspace Spot sells x86_64 servers up to 16 vCPU, plus one NVIDIA A30 GPU class.

Pricing by region

1 hr

Regions are Rackspace Spot locations.

RegionAWS on-demandAWS Spot1-yr Reserved3-yr ReservedRackspace on-demandRackspace SpotOn-demand price savingsSpot price savings
US East, N. Virginia$7.200not offerednot offerednot offerednot offerednot offeredn/an/a
US Central, Chicagono AWS regionn/an/an/anot offerednot offeredn/an/a
US Central, Dallasno AWS regionn/an/an/anot offerednot offeredn/an/a
Hong Kongnot sold heren/an/an/anot offerednot offeredn/an/a
Sydney, Australia$12.336not offerednot offerednot offerednot offerednot offeredn/an/a
London, UKnot sold heren/an/an/anot offerednot offeredn/an/a
US West, San Josenot sold heren/an/an/anot offerednot offeredn/an/a
US Central, Dallas 2no AWS regionn/an/an/anot offerednot offeredn/an/a
US East, N. Virginia 2$7.200not offerednot offerednot offerednot offerednot offeredn/an/a

Prices updated 17 Nov 2026.

When to use it

Good for
  • ●Model inference and embedding generation
  • ●Video transcoding and image processing
  • ●Machine learning experimentation and fine-tuning
  • ●Workloads that fit inside 488 GiB of RAM
  • ●Parallel work that can occupy 32 vCPU
  • ●Network-heavy services (AWS rates this size at 10 Gigabit)
Avoid if
  • ●Starting a new deployment. p3.8xlarge is the newer generation at this size.
  • ●The working set needs more than 488 GiB of RAM
  • ●More than 15.25 GiB of RAM per vCPU is needed, since that ratio is fixed for the whole family
  • ●The work is single-threaded. Most of these 32 vCPU would sit idle.
  • ●Local NVMe or instance-store disk is a hard requirement
  • ●Anything the CPU can do fast enough. GPU hours are the most expensive on the price list.
Rackspace Spot sells spare capacity at market prices. You pay what the market decides, not a list price.
  • Rackspace Spot sells spare capacity at market prices. It is the cheapest compute on the internet.
  • Prices are set by supply and demand. You pay what the market decides.
  • Compute is delivered as fully managed Kubernetes. All infrastructure is taken care of.
  • Egress costs $0.01/GB after the free allowance. AWS charges $0.09/GB.
Build your Cloudspace

Frequently asked questions

How much does p2.8xlarge cost?

The AWS p2.8xlarge costs $7.20/hr on-demand in US East (N. Virginia), about $5,256/mo running 24/7. Rackspace Spot also sells in Chicago and Dallas, where AWS operates no region at all, so those rows have no AWS price to compare against.

How much is p2.8xlarge per month?

At 730 hours a month, p2.8xlarge in US East (N. Virginia) is $5,256/mo on AWS on-demand.

Does Rackspace Spot have an equivalent to p2.8xlarge?

No, and it is worth saying so plainly: 8 × NVIDIA Tesla K80 wired together behind a high-speed interconnect is one tightly coupled machine, not 8 independent nodes, and that interconnect is usually what the instance was bought for. Rackspace Spot sells market-priced CPU and GPU capacity, which is a good fit for the batch, web and inference work around this instance, but not for p2.8xlarge itself.

Is p2.8xlarge still worth using?

For workloads already running on it, yes: it works and it is priced accordingly. For anything new, p3.8xlarge is the current answer. 3 newer generations of the p family have shipped since (p3, p4d and p5), which is the signal that matters for infrastructure you expect to keep running.

What is the difference between p2 and p3?

p3 is the newer generation. It runs on Intel Xeon E5-2686 v4 (Broadwell) against p2's Intel Xeon E5-2686 v4 (Broadwell). On-demand in US East (N. Virginia) the newer size is 70% more expensive. Neither generation changes the Rackspace Spot side: both map to the same Rackspace Spot servers.

What is a cheaper alternative to p2.8xlarge?

Not on Rackspace Spot, for the reason above. Within AWS the usual levers are a smaller size in the same family, a reserved or savings-plan commitment, or AWS Spot capacity if the work can tolerate interruption.