(All-cores)
(Single-core)
Specifications
Server Metadata
Vendor ID | gcp |
Server ID | 1741005 |
Name | ct5p-hightpu-4t |
Description | 208 vCPUs, 448 GB RAM, 4 Google TPUs |
Family | ct5p |
Hw Virt | - |
Status | active |
Observed At | 2026-10-23T22:06:25.427285 |
Availability
| REGION / ID | SPOT | ONDEMAND |
|---|---|---|
Columbus (US) / us-east5 | 3.6131 USD/h | 16.8 USD/h |
Council Bluffs (US) / us-central1 | 3.7044 USD/h | 16.8 USD/h |
Moncks Corner (US) / us-east1 | 3.7044 USD/h | 16.8 USD/h |
Eemshaven (NL) / europe-west4 | 4.7166 USD/h | 17.01 USD/h |
St. Ghislain (BE) / europe-west1 | 6.7292 USD/h | 17.8378 USD/h |
Dallas (US) / us-south1 | 2.5227 USD/h | 19.116 USD/h |
Processor
vCPUs | 208 |
CPU Allocation | Dedicated |
CPU Architecture | x86_64 |
System Resources and Accelerators
| MEMORY | |
|---|---|
Memory Amount | 448 GiB |
| ACCELERATORS | |
|---|---|
Accelerator Count | 4 |
Accelerator Memory Min | 95 GiB |
Accelerator Memory Total | 380 GiB |
Accelerator Type | TPU |
Accelerator Manufacturer | Google |
Accelerator Family | TPU |
Accelerator Model | v5p |
Accelerators |
|
| STORAGE | |
|---|---|
Storage Size | 0 GB |
Storages |
| NETWORK | |
|---|---|
Inbound Traffic | 0 GB/month |
Outbound Traffic | 0 GB/month |
IPv4 | 0 |
Server Description
A high-density accelerator instance featuring dedicated compute cores and four specialized tensor processing units with substantial high-bandwidth memory.
Google Cloud Platform ct5p-hightpu-4t is a specialized accelerator instance featuring 208 dedicated x86_64 vCPUs and 448.0 GB of system memory, providing 2.15 GB of RAM per core. The instance is equipped with four Google TPU v5p accelerators that deliver a combined 380 GB of VRAM, designed specifically for tensor-intensive computations. It does not include local storage or complimentary public IPv4 addresses. This hardware profile is optimized for machine learning workloads, including deep learning model training, large language model fine-tuning, and high-throughput neural network inference. The consolidation of high-capacity TPU resources and dedicated compute cores offers a dense, single-node solution for complex AI workloads, eliminating the network overhead associated with multi-node clustering for compatible model sizes.
Economics
Alternatives
Servers of the Same Family
| INSTANCE | vCPUs | MEMORY | Accelerators |
|---|---|---|---|
| ct5p-hightpu-1t | 52 | 112 GiB | 1 |
| ct5p-hightpu-2t | 104 | 224 GiB | 2 |
Similar Servers
| INSTANCE | VENDOR | vCPUs | MEMORY | Accelerators |
|---|