# Instance Types
## About Instance Types
### Availability
This page provides a complete list of the AWS, Azure, and OCI instance types that are available for use on the DNAnexus Platform.
{% hint style="info" %}
Not all instance types are accessible to all customers. If you want to use an instance type listed here, but cannot access it when running an analysis, [contact DNAnexus Support](mailto:support@dnanexus.com).
{% endhint %}
A list of instance types to which you have access is available via the command-line interface (CLI), by entering the command `dx run --instance-type-help`. When using the user interface (UI) to configure a tool to run an analysis, you can see this list in the [Stage Settings pane of the Run Analysis screen](https://documentation.dnanexus.com/getting-started/key-concepts/apps-and-workflows#running-apps-and-workflows).
### GPU Instance Types
AWS, Azure, and OCI GPU instance types are available for use on the Platform. See the [full list below](#gpu-instance-types-1) and learn [how to specify NVIDIA driver version](https://documentation.dnanexus.com/developer/api/io-and-run-specifications#run-specification).
### FPGA Instance Types
AWS FPGA instance types are available for use on the Platform. See the [full list below](#fpga-instance-types-1) and [learn how to specify FPGA driver version](https://documentation.dnanexus.com/developer/api/io-and-run-specifications#run-specification).
### OS Support
Ubuntu Linux 24.04 and 20.04 are supported on all instance types in all regions.
### Resource Usage
When using any instance, the Platform uses some resources to run processes that support your job and provide API services. For example, around 5% of the available storage is reserved for Platform use. Your job's virtual file system can use the remaining space as local scratch. The Platform also allocates some available memory to its processes.
### New Versions of Existing Instance Types
DNAnexus regularly adds new versions of existing instance types, as cloud providers make them available. These new versions feature better hardware, such as a more powerful CPU, more memory, more local storage, or some combination of these elements.
On the Platform, each new instance type version has a version infix, such as `_v2`, in its name, to distinguish it from the original version of that instance type. See the [Instance Names section](#instance-type-names) for more information on version infixes.
Although the latest instance type versions deliver better performance, DNAnexus maintains both the original and updated versions of instance types. For example, you can access both `mem1_ssd1_x8` and `mem1_ssd1_v2_x8`. See the [full list of available instance types](#available-instance-types) for more information.
## Instance Type Names
Except for [GPU and FPGA instance types](#gpu-and-fpga-instance-type-names), instance type names are constructed according to the following scheme:
| Cloud provider prefix | | Memory infix | | Storage infix | | Version infix | | Core suffix |
| ------------------------------------------------------------------------------------------------------------------------------------------- | - | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | - | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | - | ---------------------------------------------------------------------------- | - | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| ∅ (AWS)
azure: (Microsoft Azure)
oci: (Oracle Cloud Infrastructure)
| + | mem1\_ (<=2GiB/core)
mem2\_ (\~4GiB/core)
mem3\_ (>=7GiB/core)
mem4\_ (\~14GiB/core)
mem5\_ (\~28GiB/core)
| + | ssd1\_ (<=20GB/core)
ssd2\_ (\~32GB-128GB/core)
ssd3\_ (>600GB/core)
hdd1\_ (>100GB/core)
hdd2\_ (>100GB/core)
| + | ∅ (version1)
v2\_ (version2)
| + | x1
x2
x4
x8
x16
x20
x32
x36
x40
x48
x64
x96
x128
|
Where:
* **Cloud provider prefix:** Denotes the cloud provider.
* **Memory infix**: Denotes the memory capacity (per core).
* **Storage infix**: Denotes the local storage technology and capacity (per core). `ssd` represents a solid-state drive, whereas `hdd` represents a regular hard disk drive.
* **Version infix** (optional): Denotes the version of the instance type.
* **Core suffix**: Denotes the number of cores.
For example, `mem1_ssd1_v2_x8` is the second version of an instance type available on AWS, featuring 8 cores, 16GiB of memory (2GiB/core), and 160GB of solid-state drive storage (20GB/core). Similarly, `azure:mem1_ssd1_x8` is an instance type available on Azure, featuring 8 cores, 15.7GiB of memory (\~1.9GiB/core), and 128GB of solid-state drive storage (16GB/core).
{% hint style="info" %}
Not all name combinations are available as instance types. See the [Available Instance Types](#available-instance-types) section for full detail on available instance types.
{% endhint %}
### GPU and FPGA Instance Type Names
GPU and FPGA instance type names following the schema described above, but with the inclusion of an additional infix. This infix is included before the version suffix, and provides information on the instance type (GPU or FPGA). Many such infixes also include a number, indicating the number of GPUs or FPGA included in the instance.
For example, the AWS FPGA instance type `mem3_ssd2_fpga1_x24` includes 1 FPGA. The AWS GPU instance type `mem2_ssd1_gpu4_x48`, meanwhile, includes 4 GPUs.
For more information, see the detailed lists of [GPU instances types](#gpu-instance-types-1) and [FPGA instances types](#fpga-instance-types-1).
## Available Instance Types
### Standard AWS Instance Types
| Instance Type | Cores | Memory (GiB) | Storage (GB) |
| ------------------------ | ----- | ------------ | ------------ |
| mem1\_hdd1\_x2 | 2 | 3.75 | 200 |
| mem1\_hdd1\_x4 | 4 | 7.5 | 400 |
| mem1\_hdd1\_x8 | 8 | 15 | 800 |
| mem1\_hdd1\_x16 | 16 | 30 | 1,600 |
| mem1\_hdd1\_x36 | 36 | 60 | 3,200 |
| mem1\_hdd1\_v2\_x2 | 2 | 4 | 200 |
| mem1\_hdd1\_v2\_x4 | 4 | 8 | 400 |
| mem1\_hdd1\_v2\_x8 | 8 | 16 | 800 |
| mem1\_hdd1\_v2\_x16 | 16 | 32 | 1,600 |
| mem1\_hdd1\_v2\_x36 | 36 | 72 | 3,600 |
| mem1\_hdd1\_v2\_x72 | 72 | 144 | 7,200 |
| mem1\_hdd1\_v2\_x96 | 96 | 192 | 9,600 |
| mem1\_ssd1\_x2 | 2 | 3.8 | 40 |
| mem1\_ssd1\_x4 | 4 | 7.5 | 80 |
| mem1\_ssd1\_x8 | 8 | 15 | 160 |
| mem1\_ssd1\_x16 | 16 | 30 | 320 |
| mem1\_ssd1\_x32 | 32 | 60 | 640 |
| mem1\_ssd1\_x36 | 36 | 72 | 900 |
| mem1\_ssd1\_v2\_x2 | 2 | 4 | 50 |
| mem1\_ssd1\_v2\_x4 | 4 | 8 | 100 |
| mem1\_ssd1\_v2\_x8 | 8 | 16 | 200 |
| mem1\_ssd1\_v2\_x16 | 16 | 32 | 400 |
| mem1\_ssd1\_v2\_x36 | 36 | 72 | 900 |
| mem1\_ssd1\_v2\_x72 | 72 | 144 | 1,800 |
| mem1\_ssd2\_x2 | 2 | 3.8 | 160 |
| mem1\_ssd2\_x4 | 4 | 7.5 | 320 |
| mem1\_ssd2\_x8 | 8 | 15 | 640 |
| mem1\_ssd2\_x16 | 16 | 30 | 1,280 |
| mem1\_ssd2\_x36 | 36 | 60 | 2,880 |
| mem1\_ssd2\_v2\_x2 | 2 | 4 | 160 |
| mem1\_ssd2\_v2\_x4 | 4 | 8 | 320 |
| mem1\_ssd2\_v2\_x8 | 8 | 16 | 640 |
| mem1\_ssd2\_v2\_x16 | 16 | 32 | 1,280 |
| mem1\_ssd2\_v2\_x36 | 36 | 72 | 2,880 |
| mem1\_ssd2\_v2\_x72 | 72 | 144 | 5,760 |
| mem1\_ssd2\_v3\_x2 ² ³ | 2 | 4 | 118 |
| mem1\_ssd2\_v3\_x4 ² ³ | 4 | 8 | 237 |
| mem1\_ssd2\_v3\_x8 ² ³ | 8 | 16 | 474 |
| mem1\_ssd2\_v3\_x16 ² ³ | 16 | 32 | 950 |
| mem1\_ssd2\_v3\_x32 ² ³ | 32 | 64 | 1,900 |
| mem1\_ssd2\_v3\_x48 ² ³ | 48 | 96 | 2,850 |
| mem1\_ssd2\_v3\_x64 ² ³ | 64 | 128 | 3,800 |
| mem1\_ssd2\_v3\_x96 ² ³ | 96 | 192 | 5,700 |
| mem1\_ssd2\_v3\_x128 ² ³ | 128 | 256 | 7,600 |
| mem1\_hdd2\_x1 | 1 | 1.7 | 160 |
| mem1\_hdd2\_x8 | 8 | 7 | 1,680 |
| mem1\_hdd2\_x32 | 32 | 60.5 | 3,360 |
| mem2\_ssd1\_x2 | 2 | 7.5 | 40 |
| mem2\_ssd1\_x4 | 4 | 15 | 80 |
| mem2\_ssd1\_x8 | 8 | 30 | 160 |
| mem2\_ssd1\_v2\_x2 | 2 | 8 | 75 |
| mem2\_ssd1\_v2\_x4 | 4 | 16 | 150 |
| mem2\_ssd1\_v2\_x8 | 8 | 32 | 300 |
| mem2\_ssd1\_v2\_x16 | 16 | 64 | 600 |
| mem2\_ssd1\_v2\_x32 | 32 | 128 | 1,200 |
| mem2\_ssd1\_v2\_x48 | 48 | 192 | 1,800 |
| mem2\_ssd1\_v2\_x64 | 64 | 256 | 2,400 |
| mem2\_ssd1\_v2\_x96 | 96 | 384 | 3,600 |
| mem2\_ssd2\_x2 | 2 | 8 | 160 |
| mem2\_ssd2\_x4 | 4 | 16 | 320 |
| mem2\_ssd2\_x8 | 8 | 32 | 1,280 |
| mem2\_ssd2\_x16 | 16 | 64 | 2,560 |
| mem2\_ssd2\_x40 | 40 | 160 | 3,200 |
| mem2\_ssd2\_x64 | 64 | 256 | 5,120 |
| mem2\_ssd2\_v2\_x2 | 2 | 8 | 160 |
| mem2\_ssd2\_v2\_x4 | 4 | 16 | 320 |
| mem2\_ssd2\_v2\_x8 | 8 | 32 | 640 |
| mem2\_ssd2\_v2\_x16 | 16 | 64 | 1,280 |
| mem2\_ssd2\_v2\_x32 | 32 | 128 | 2,560 |
| mem2\_ssd2\_v2\_x48 | 48 | 192 | 3,840 |
| mem2\_ssd2\_v2\_x64 | 64 | 256 | 5,120 |
| mem2\_ssd2\_v2\_x96 | 96 | 384 | 7,480 |
| mem2\_ssd2\_v3\_x2 ² ³ | 2 | 8 | 118 |
| mem2\_ssd2\_v3\_x4 ² ³ | 4 | 16 | 237 |
| mem2\_ssd2\_v3\_x8 ² ³ | 8 | 32 | 474 |
| mem2\_ssd2\_v3\_x16 ² ³ | 16 | 64 | 950 |
| mem2\_ssd2\_v3\_x32 ² ³ | 32 | 128 | 1,900 |
| mem2\_ssd2\_v3\_x48 ² ³ | 48 | 192 | 2,850 |
| mem2\_ssd2\_v3\_x64 ² ³ | 64 | 256 | 3,800 |
| mem2\_ssd2\_v3\_x96 ² ³ | 96 | 384 | 5,700 |
| mem2\_ssd2\_v3\_x128 ² ³ | 128 | 512 | 7,600 |
| mem2\_hdd2\_x1 | 1 | 3.8 | 410 |
| mem2\_hdd2\_x2 | 2 | 7.5 | 840 |
| mem2\_hdd2\_x4 | 4 | 15 | 1,680 |
| mem2\_hdd2\_v2\_x2 | 2 | 8 | 1,000 |
| mem2\_hdd2\_v2\_x4 | 4 | 16 | 2,000 |
| mem3\_ssd1\_x2 | 2 | 15 | 40 |
| mem3\_ssd1\_x4 | 4 | 30.5 | 80 |
| mem3\_ssd1\_x8 | 8 | 61 | 160 |
| mem3\_ssd1\_x16 | 16 | 122 | 320 |
| mem3\_ssd1\_x32 | 32 | 244 | 640 |
| mem3\_ssd1\_v2\_x2 | 2 | 16 | 75 |
| mem3\_ssd1\_v2\_x4 | 4 | 32 | 150 |
| mem3\_ssd1\_v2\_x8 | 8 | 64 | 300 |
| mem3\_ssd1\_v2\_x16 | 16 | 128 | 600 |
| mem3\_ssd1\_v2\_x32 | 32 | 256 | 1,200 |
| mem3\_ssd1\_v2\_x48 | 48 | 384 | 1,800 |
| mem3\_ssd1\_v2\_x64 | 64 | 512 | 3,200 |
| mem3\_ssd1\_v2\_x96 | 96 | 768 | 3,600 |
| mem3\_ssd2\_x4 | 4 | 30.5 | 800 |
| mem3\_ssd2\_x8 | 8 | 61 | 1,600 |
| mem3\_ssd2\_x16 | 16 | 122 | 3,200 |
| mem3\_ssd2\_x32 | 32 | 244 | 6,400 |
| mem3\_ssd2\_v2\_x2 | 2 | 15.25 | 475 |
| mem3\_ssd2\_v2\_x4 | 4 | 30.5 | 950 |
| mem3\_ssd2\_v2\_x8 | 8 | 61 | 1,900 |
| mem3\_ssd2\_v2\_x16 | 16 | 122 | 3,800 |
| mem3\_ssd2\_v2\_x32 | 32 | 244 | 7,600 |
| mem3\_ssd2\_v2\_x64 | 64 | 488 | 15,200 |
| mem3\_ssd2\_v3\_x2 ² ³ | 2 | 16 | 118 |
| mem3\_ssd2\_v3\_x4 ² ³ | 4 | 32 | 237 |
| mem3\_ssd2\_v3\_x8 ² ³ | 8 | 64 | 474 |
| mem3\_ssd2\_v3\_x16 ² ³ | 16 | 128 | 950 |
| mem3\_ssd2\_v3\_x32 ² ³ | 32 | 256 | 1,900 |
| mem3\_ssd2\_v3\_x48 ² ³ | 48 | 384 | 2,850 |
| mem3\_ssd2\_v3\_x64 ² ³ | 64 | 512 | 3,800 |
| mem3\_ssd2\_v3\_x96 ² ³ | 96 | 768 | 5,700 |
| mem3\_ssd2\_v3\_x128 ² ³ | 128 | 1024 | 7,600 |
| mem3\_ssd3\_x2 | 2 | 16 | 1,250 |
| mem3\_ssd3\_x4 | 4 | 32 | 2,500 |
| mem3\_ssd3\_x8 | 8 | 64 | 5,000 |
| mem3\_ssd3\_x12 | 12 | 96 | 7,500 |
| mem3\_ssd3\_x24 | 24 | 192 | 15,000 |
| mem3\_ssd3\_x48 | 48 | 384 | 30,000 |
| mem3\_ssd3\_x96 | 96 | 768 | 60,000 |
| mem3\_hdd2\_x2 | 2 | 17.1 | 420 |
| mem3\_hdd2\_x4 | 4 | 34.2 | 850 |
| mem3\_hdd2\_x8 | 8 | 68.4 | 1,680 |
| mem3\_hdd2\_v2\_x2 | 2 | 16 | 500 |
| mem3\_hdd2\_v2\_x4 | 4 | 32 | 1,000 |
| mem3\_hdd2\_v2\_x8 | 8 | 64 | 2,000 |
| mem4\_ssd1\_x128 ¹ | 128 | 1,952 | 3,840 |
Notes:
* ¹ High memory: Expensive. Use with caution.
* ² Limited access: Available only to special licensed users.
* ³ Not available in the `aws:me-south-1` [region](https://documentation.dnanexus.com/developer/api/regions).
### Standard Azure Instance Types
| Instance Type | Cores | Memory (GiB) | Storage (GB) |
| ------------------------ | ----- | ------------ | ------------ |
| azure:mem1\_ssd1\_x2 | 2 | 3.9 | 32 |
| azure:mem1\_ssd1\_x4 | 4 | 7.8 | 64 |
| azure:mem1\_ssd1\_x8 | 8 | 15.7 | 128 |
| azure:mem1\_ssd1\_x16 | 16 | 31.4 | 254 |
| azure:mem2\_ssd1\_x1 | 1 | 3.5 | 128 |
| azure:mem2\_ssd1\_x2 | 2 | 7 | 128 |
| azure:mem2\_ssd1\_x4 | 4 | 14 | 128 |
| azure:mem2\_ssd1\_x8 | 8 | 28 | 256 |
| azure:mem2\_ssd1\_x16 | 16 | 56 | 512 |
| azure:mem3\_ssd1\_x2 | 2 | 14 | 128 |
| azure:mem3\_ssd1\_x4 | 4 | 28 | 128 |
| azure:mem3\_ssd1\_x8 | 8 | 56 | 256 |
| azure:mem3\_ssd1\_x16 | 16 | 112 | 512 |
| azure:mem3\_ssd1\_x20 | 20 | 140 | 640 |
| azure:mem4\_ssd1\_x2 | 2 | 28 | 128 |
| azure:mem4\_ssd1\_x4 | 4 | 56 | 128 |
| azure:mem4\_ssd1\_x8 | 8 | 112 | 256 |
| azure:mem4\_ssd1\_x16 | 16 | 224 | 512 |
| azure:mem4\_ssd1\_x32 | 32 | 448 | 2,048 |
| azure:mem5\_ssd2\_x64 ¹ | 64 | 1,792 | 8,192 |
| azure:mem5\_ssd2\_x128 ¹ | 128 | 3,892 | 16,384 |
Notes:
* ¹ High memory: Expensive. Use with caution.
### Standard OCI Instance Types
| Instance Type | Cores | Memory (GiB) | Storage (GB) |
| ------------------------ | ----- | ------------ | ------------ |
| oci:mem1\_ssd1\_v3i\_x2 | 2 | 3.9 | 50 |
| oci:mem1\_ssd1\_v3i\_x4 | 4 | 7.8 | 100 |
| oci:mem1\_ssd1\_v3i\_x8 | 8 | 15.6 | 200 |
| oci:mem1\_ssd1\_v3i\_x16 | 16 | 31.3 | 400 |
| oci:mem1\_ssd1\_v3i\_x36 | 36 | 70.3 | 900 |
| oci:mem1\_ssd2\_v3i\_x2 | 2 | 3.9 | 160 |
| oci:mem1\_ssd2\_v3i\_x4 | 4 | 7.8 | 320 |
| oci:mem1\_ssd2\_v3i\_x8 | 8 | 15.6 | 640 |
| oci:mem1\_ssd2\_v3i\_x16 | 16 | 31.3 | 1,280 |
| oci:mem1\_ssd2\_v3i\_x36 | 36 | 70.3 | 2,880 |
| oci:mem2\_ssd1\_v3i\_x2 | 2 | 7.8 | 75 |
| oci:mem2\_ssd1\_v3i\_x4 | 4 | 15.6 | 150 |
| oci:mem2\_ssd1\_v3i\_x8 | 8 | 31.3 | 300 |
| oci:mem2\_ssd1\_v3i\_x16 | 16 | 62.5 | 600 |
| oci:mem2\_ssd1\_v3i\_x32 | 32 | 125 | 1,200 |
| oci:mem3\_ssd1\_v3i\_x2 | 2 | 15.6 | 75 |
| oci:mem3\_ssd1\_v3i\_x4 | 4 | 31.3 | 150 |
| oci:mem3\_ssd1\_v3i\_x8 | 8 | 62.5 | 300 |
| oci:mem3\_ssd1\_v3i\_x16 | 16 | 125 | 600 |
| oci:mem3\_ssd1\_v3i\_x32 | 32 | 250 | 1,200 |
| oci:mem3\_ssd2\_v3i\_x2 | 2 | 15.6 | 475 |
| oci:mem3\_ssd2\_v3i\_x4 | 4 | 31.3 | 950 |
| oci:mem3\_ssd2\_v3i\_x8 | 8 | 62.5 | 1,900 |
| oci:mem3\_ssd2\_v3i\_x16 | 16 | 125 | 3,800 |
| oci:mem3\_ssd2\_v3i\_x32 | 32 | 250 | 7,600 |
| oci:mem3\_ssd3\_v3i\_x2 | 2 | 15.6 | 1,250 |
| oci:mem3\_ssd3\_v3i\_x4 | 4 | 31.3 | 2,500 |
| oci:mem3\_ssd3\_v3i\_x8 | 8 | 62.5 | 5,000 |
| oci:mem3\_ssd3\_v3i\_x12 | 12 | 93.8 | 7,500 |
| oci:mem3\_ssd3\_v3i\_x24 | 24 | 187.5 | 15,000 |
| oci:mem3\_ssd3\_v3i\_x32 | 32 | 250 | 20,000 |
### GPU Instance Types
The following table shows all the GPU instance types available on AWS, Azure, and OCI.
{% hint style="info" %}
**CUDA support**
The CUDA version available on your GPU instance depends on which NVIDIA driver version you specify in your [`nvidiaDriver` configuration](https://documentation.dnanexus.com/developer/api/io-and-run-specifications#run-specification). The default driver version `R470` supports CUDA 11.4, while the newer version `R535` supports CUDA 12.2.
If you need a newer CUDA version than what these drivers provide, you can use [NVIDIA Forward Compatibility](https://documentation.dnanexus.com/faqs/developing-apps-and-applets#how-do-i-run-newer-nvidia-gpu-accelerated-software-in-my-app) to run more recent GPU-accelerated software.
{% endhint %}
| Instance Type | Cores | Memory (GiB) | Storage (GB) | GPU | GPU Memory (GiB) |
| -------------------------------- | ----- | ------------ | ------------ | ------------- | ---------------- |
| mem2\_ssd1\_gpu\_x16 | 16 | 64 | 225 | 1 NVIDIA T4 | 16 |
| mem2\_ssd1\_gpu\_x32 | 32 | 128 | 900 | 1 NVIDIA T4 | 16 |
| mem2\_ssd1\_gpu\_x48 | 48 | 192 | 900 | 4 NVIDIA T4 | 64 |
| mem2\_ssd1\_gpu\_x64 | 64 | 256 | 900 | 1 NVIDIA T4 | 16 |
| mem2\_ssd1\_gpu1\_x32 | 32 | 128 | 900 | 1 NVIDIA T4 | 16 |
| mem2\_ssd1\_gpu1\_x64 | 64 | 256 | 900 | 1 NVIDIA T4 | 16 |
| mem2\_ssd1\_gpu4\_x48 | 48 | 192 | 900 | 4 NVIDIA T4 | 64 |
| mem2\_ssd2\_gpu1\_x4 ² | 4 | 16 | 250 | 1 NVIDIA A10G | 24 |
| mem2\_ssd2\_gpu1\_x8 ² | 8 | 32 | 450 | 1 NVIDIA A10G | 24 |
| mem2\_ssd2\_gpu1\_x16 ² | 16 | 64 | 600 | 1 NVIDIA A10G | 24 |
| mem2\_ssd2\_gpu1\_x32 ² | 32 | 128 | 900 | 1 NVIDIA A10G | 24 |
| mem2\_ssd2\_gpu1\_x64 ² | 64 | 256 | 1,900 | 1 NVIDIA A10G | 24 |
| mem2\_ssd2\_gpu4\_x48 ² | 48 | 192 | 3,800 | 4 NVIDIA A10G | 96 |
| mem2\_ssd2\_gpu4\_x96 ² | 96 | 384 | 3,800 | 4 NVIDIA A10G | 96 |
| mem2\_ssd2\_gpu8\_x192 ¹ ² | 192 | 768 | 7,600 | 8 NVIDIA A10G | 192 |
| mem2\_ssd2\_gpu1\_v2\_x4 ² ³ | 4 | 16 | 250 | 1 NVIDIA L4 | 24 |
| mem2\_ssd2\_gpu1\_v2\_x8 ² ³ | 8 | 32 | 450 | 1 NVIDIA L4 | 24 |
| mem2\_ssd2\_gpu1\_v2\_x16 ² ³ | 16 | 64 | 600 | 1 NVIDIA L4 | 24 |
| mem2\_ssd2\_gpu1\_v2\_x32 ² ³ | 32 | 128 | 900 | 1 NVIDIA L4 | 24 |
| mem2\_ssd2\_gpu1\_v2\_x64 ² ³ | 64 | 256 | 1,880 | 1 NVIDIA L4 | 24 |
| mem2\_ssd2\_gpu4\_v2\_x48 ² ³ | 48 | 192 | 3,760 | 4 NVIDIA L4 | 96 |
| mem2\_ssd2\_gpu4\_v2\_x96 ² ³ | 96 | 384 | 3,760 | 4 NVIDIA L4 | 96 |
| mem2\_ssd2\_gpu8\_v2\_x192 ¹ ² ³ | 192 | 768 | 7,520 | 8 NVIDIA L4 | 192 |
| mem3\_ssd1\_gpu1\_x16 ² ³ | 16 | 128 | 600 | 1 NVIDIA L4 | 24 |
| mem3\_ssd1\_gpu1\_x32 ² ³ | 32 | 256 | 900 | 1 NVIDIA L4 | 24 |
| azure:mem3\_ssd2\_gpu4\_x64 ¹ | 64 | 488 | 2,048 | 4 NVIDIA V100 | 64 |
| oci:mem2\_ssd1\_GPU\_v2\_x24 | 24 | 70.3 | 480 | 1 NVIDIA P100 | 16 |
| oci:mem3\_ssd1\_GPU\_v3\_x12 | 12 | 87.9 | 240 | 1 NVIDIA V100 | 16 |
| oci:mem3\_ssd1\_GPU\_v3\_x24 | 24 | 175.8 | 480 | 2 NVIDIA V100 | 32 |
| oci:mem3\_ssd1\_GPU\_v3\_x48 | 48 | 351.6 | 960 | 4 NVIDIA V100 | 64 |
| oci:mem3\_ssd2\_GPU\_vA10\_x30 | 30 | 234.4 | 960 | 1 NVIDIA A10 | 24 |
| oci:mem3\_ssd2\_GPU\_vA10\_x60 | 60 | 468.8 | 1,920 | 2 NVIDIA A10 | 48 |
Notes:
* ¹ High memory: Expensive. Use with caution.
* ² Limited access: Available only to special licensed users.
* ³ Not available in the `aws:me-south-1` [region](https://documentation.dnanexus.com/developer/api/regions).
### FPGA Instance Types
The following table shows all the FPGA instance types available on AWS.
| Instance Type | Cores | Memory (GiB) | Storage (GB) | FPGA |
| ------------------------------------- | ----- | ------------ | ------------ | ---- |
| mem3\_ssd2\_fpga1\_x8 (deprecated) ³ | 8 | 122 | 470 | 1 |
| mem3\_ssd2\_fpga1\_x16 (deprecated) ³ | 16 | 244 | 940 | 2 |
| mem3\_ssd2\_fpga1\_x24 ² | 24 | 256 | 940 | 1 |
| mem3\_ssd2\_fpga1\_x64 (deprecated) ³ | 64 | 976 | 3,760 | 8 |
| mem3\_ssd2\_fpga2\_x48 ² | 48 | 512 | 1,880 | 2 |
| mem3\_ssd2\_fpga8\_x192 ¹ ² | 192 | 2,048 | 7,520 | 8 |
Notes:
* ¹ High memory: Expensive. Use with caution.
* ² Limited access: Available only to special licensed users.
* ³ Deprecated: Consider using `mem3_ssd2_fpga1_x24`, `mem3_ssd2_fpga2_x48`, or `mem3_ssd2_fpga8_x192` instead.