Function Computing Power Support
Last updated: 2025-04-30 19:16:29
SCF currently supports CPU and GPU computing power. This document introduces the detailed parameter information of different computing powers.
CPU Computing Power
Memory
Specify the available memory size for function runtime. The minimum is 64 MB, the maximum is 120 GB, the default is 128 MB, and the incremental step length is 128 MB. When the memory is greater than or equal to 6 GB, a large specification resource application is required.
CPU
The CPU processing capability of SCF is directly proportional to the function configuration memory. You can increase the memory configuration to get larger CPU computing power. 1280 MB corresponds to 1 accounting power, and 3072 MB corresponds to 2 accounting powers. When the memory is equal to or greater than 6 GB (6144 MB), the CPU computing power is further released, as shown in the table below:
Execution Memory (MB) | CPU Computing Power (Cores) |
64 | 0.1 |
128 | 0.1 |
256 | 0.2 |
384 | 0.3 |
512 | 0.4 |
640 | 0.5 |
768 | 0.6 |
896 | 0.7 |
1024 | 0.8 |
1152 | 0.9 |
1280 | 1 |
1408 | 1.1 |
1536 | 1.2 |
1664 | 1.3 |
1792 | 1.4 |
1920 | 1.5 |
2048 | 1.6 |
2176 | 1.7 |
2304 | 1.8 |
2432 | 1.9 |
2560 | 2 |
2688 | 2 |
2816 | 2 |
2944 | 2 |
3072 | 2 |
6144 | 4 |
14336 | 8 |
30720 | 16 |
61440 | 32 |
122880 | 64 |
GPU Computing Power
SCF supports multiple computing powers such as compute-oriented and rendering-oriented for T4 cards, as well as compute-oriented for A10 cards. The detailed parameter information of supported specifications is shown in the table below.
GPU Instance Type | Specification | Number of Cards | Card Type | CPU (Cores) | Memory (GB) | Disk (GB) |
Compute-Optimized GN7 | GN7.LARGE20 (discontinued) | 0.25 | T4 | 4 | 20 | 10 |
| | GN7.2XLARGE40 (discontinued) | 0.5 | 10 | 40 | 10 |
| GN7.2XLARGE32 (default specification for Stable Diffusion AI painting application scenario, not yet open for other scenarios) | 1 | | 8 | 32 | 10 |
| | GN7.5XLARGE80 | 1 | 20 | 80 | 10 |
Rendering GN7vw | GN7vw.LARGE16 | 0.25 | T4 | 4 | 16 | 10 |
| | GN7vw.2XLARGE32 | 0.5 | 8 | 32 | 10 |
| | GN7vw.4XLARGE64 | 1 | 16 | 64 | 10 |
Compute-Optimized PNV4 | PNV4.7XLARGE116 | 1 | A10 | 28 | 116 | 10 |
Must-Knows
Due to the scarcity of resources for large-spec instances and GPU instances, their elasticity capabilities are insufficient. Recommend you use preset concurrency for resource handling to guarantee resource supply stability.
Recommended Usage of GPU
Preset Concurrency: Preset concurrency supports concurrent instances to start in advance according to the configuration. At the same time, the SCF will not proactively reclaim them and will guarantee as much as possible that there are a corresponding number of concurrent instances that can process requests. You can use this feature to set the preset concurrency quota for a specified version of the function. By configuring preset concurrency, you can prepare computing resources in advance and reduce the time taken caused by cold startup, initialization of runtime environment, and business code initialization.
GPU Computing Power Billing Mode
GPU function computing power and CPU computing power share the same billing mode, which charges based on memory usage within a certain time. For details, see Billing Overview.
GPU Computing Power Hybrid Card Type Scheduling
SCF GPU hybrid card type scheduling feature supports businesses to select and assign different GPU models, addressing insufficient resources of a single card type and enhancing cluster computing power resource usage rate and supply capacity.
Operation Steps
1. For functions of container image type, specification selection is supported during creation and updating by clicking.
2. Check corresponding GPU resource specifications in the pop-up: On existing GPU types, users can check multiple card types and sort the checked ones to determine scheduling priority.
Was this page helpful?
You can also Contact Sales or Submit a Ticket for help.
Yes
No
Feedback