Family | Subtypes | Purpose | Description |
A-series | A#_v2, A#m_v2 | General | Best suited for entry level workloads (development or test). |
B-series | B#s, B1ls, B#ms | General | Ideal for workloads that do not need continuous full CPU performance (burst). |
D-series | D#, DC#s, DS#, D#_v2, DC#s_v2, DS#_v2, D#_v3 D#s_v3, D#_v4,, D#a_v4, D#as_v4, D#d_v4, D#ds_v4, D#s_v4 | General / Memory optimized | Recommended for general purpose computing, balanced CPU-to-memory ratio. The D#_v2 and DS#_v2 are memory-optimized. Note: Ds_v4 machines do not have temporary storage, resizing to and from these machines is not possible (see warning under Resizing a VM) |
E-series | E#_v3, E#s_v3, E#_v4, E#a_v4, E#as_v4, E#d_v4, E#ds_v4, E#s_v4, E80ids_v4, E80is_v4 | Memory optimized | High memory-to-CPU ratio. Run on processors in multi- or hyper-threaded configurations. Note: E_v4 and Es_v4 machines do not have temporary storage, resizing to and from these machines is not possible (see warning under Resizing a VM) |
F-series | F#, F#s, F#s_v2, FX#mds | Compute optimized | Optimized for compute intensive workloads. Machines are hyper-threaded. Ideal for batch processing and analytics. |
G-series | G#, Gs# | Memory optimized | Provide high RAM and temporary storage, ideal for large database workloads. |
M-series | M#, M#ls, M#m, M#ms, M#s, M#ts, M#ds_v2, M#dms_v2, M#ids_v2, M#idms_2, M#is_v2, M#ims_v2, M#s_v2, M#ms_v2 | Memory optimized | Provides unparallelled computational performance to support large in-memory workloads. Note: M#s_v2 and M#ms_v2 machines do not have temporary storage, resizing to and from these machines is not possible (see warning under Resizing a VM. |
N-series | NC#, ND#s, ND#rs, NV#, NC#s_v2, NC#rs_v2, ND#rs_v2, NC#s_v3, NC#rs_v3, NC#as_T4_v3, NV#s_v3, ND#asr_A100_v4, NV#as_v4 | GPU | Ideal for compute and graphics-intensive workloads, such as high-end visualization, deep learning and predictive analytics. |
VM type | vCPU(s) | RAM (GiB) | Price per hour (estimated) |
| B1ms | 1 | 2 | € 0.0222 |
B2s | 2 | 4 | € 0.0445 |
B2ms | 2 | 8 | € 0.0889 |
B4ms | 4 | 16 | € 0.1778 |
B8ms | 8 | 32 | € 0.3556 |
D2_v4 | 2 | 8 | € 0.1065 (no temp) |
D2s_v3 | 2 | 8 | € 0.1111 |
D4s_v3 | 4 | 16 | € 0.2223 |
D8_v4 | 8 | 32 | € 0.4260 (no temp) |
D16s_v3 | 16 | 64 | € 0.8890 |
| D32_v4 | 32 | 128 | € 1.7039 (no temp) |
D48s_v4 | 46 | 192 | € 2.5460 (no temp) |
D64_v4 | 64 | 256 | € 3.4079 (no temp) |
F8 | 8 | 16 | € 0.4204 |
F16s_v2 | 16 | 32 | € 0.7186 |
F48s_v2 | 48 | 96 | € 2.1559 |
M128m | 128 | 3800 | € 34.4663 |
NC8as_T4_v3 | 8 | 56 | € 0.8705 (GPU: 1X T4) |
Important! It is not possible to resize machine types with temporary storage to machine types without temporary storage, and vice versa. Check the tables in this article or the Microsoft Azure website to see whether your desired VM type has temporary storage.