The p3.2xlarge is a GPU AWS EC2 instance: 8 vCPU and 61 GiB of memory, 7.63 GiB per vCPU. In the name, "p" is accelerated computing, "3" is the generation and "2xlarge" is the size within the p3 family. It runs on Intel Xeon E5-2686 v4 (Broadwell) at 2.3 GHz, and AWS introduced it in October 2017. AWS has since shipped p4d and p5. Rackspace Spot runs one GPU class, gpu.vs2.xlargeplusplus — a single NVIDIA A30 with 24 vCPU and 64 GiB, in San Jose only. Against this instance's 1 × NVIDIA Tesla V100 that is a downgrade per card, honest for inference and not for the training this instance was bought for.
Match qualityWeaker GPU
p3.2xlarge carries 1 GPU (16 GiB VRAM). The Spot GPU node is built on the NVIDIA A30 (Ampere, 24 GB VRAM) with 24 vCPU and 64 GiB of host memory. Per card that is a downgrade, and it should be read as one: it is honest for inference, batch work and small-batch fine-tuning, and it is not the card p3.2xlarge was bought for if the job is a full training run. The catch is location rather than hardware: US West, San Jose is the only Spot region that offers it, so latency to your users and egress to your data are the things to check.
Pricing by region
1 hr
Region
AWS on-demand
AWS Spot
1-yr Reserved
3-yr Reserved
Rackspace Spot
You save
US West, San Jose
$0.191
When to use it
Good for
●Machine learning experimentation and fine-tuning
●Remote graphics workstations and rendering
●CUDA-accelerated batch jobs
●Workloads that fit inside 61 GiB of RAM
●Parallel work that can occupy 8 vCPU
●Network-heavy services — AWS rates this size at Up to 10 Gigabit
Avoid if
●The working set needs more than 61 GiB of RAM
●Anything the CPU can do fast enough. GPU hours are the most expensive on the price list.
●More than 7.63 GiB of RAM per vCPU is needed — that ratio is fixed for the whole family
●Local NVMe or instance-store disk is a hard requirement
Rackspace Spot sells spare capacity at market prices. You pay what the market decides, not a list price.
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.
