A stress testing method, system and equipment for hl-100 reasoning card
A stress test and stress technology, applied in the field of stress testing of HL-100 reasoning card, can solve problems such as time-consuming and labor-intensive, and achieve the effect of ensuring product quality, improving customer satisfaction, and highlighting substantive features
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0041] like figure 1 A stress test method for the HL-100 reasoning card shown, including the following steps:
[0042] S1: Check that the HL-100 reasoning card can be recognized normally under the current system. Specifically, it includes: counting the number of HL-100 inference cards identified in the current system and saving it to HL-100_num.txt; checking the working mode of the PCIe interface of the HL-100 inference card and saving it to the PCIeSpeed.txt file.
[0043] S2: Use the command #lspci-d 1da3: to obtain the bus IDs of all HL-100 inference cards under the system; use a for loop to traverse the bus IDs of all HL-100 inference cards, and perform pressure cooling on the HL-100 inference cards one by one for 1 hour. Test, 1 hour BERT stability test and 1 hour data upstream and downstream transmission bandwidth test of server memory and memory on the card, and save the test results, showing pass or fail.
[0044] S3: Perform a 24-hour stress cooling test on all HL-1...
Embodiment 2
[0047] like figure 2 The stress test method of a HL-100 inference card shown, the specific implementation steps and the corresponding script content are as follows:
[0048] 1. Check that the HL-100 reasoning card can be recognized normally under the system:
[0049]#lspci–d lda3:|wc–l|tee–a HL-100_num.txt
[0050] Count the number of boards identified under the system and save them to HL-100_num.txt;
[0051] #hl-smi–L|grep–E “Bus Id|Link Speed|Link Width|Max|Current”|tee–aPCIeSpeed.txt
[0052] View the working mode of the PCIe interface of the board and save it to the PCIeSpeed.txt file.
[0053] 2. Use the command #lspci-d 1da3: to obtain the bus IDs of all HL-100 inference cards under the system, use a for loop to traverse the bus IDs of all inference cards, and continuously perform 1h pressure heat dissipation test and 1h BERT stability test on the boards one by one Interact with the data upstream and downstream transmission bandwidth test of the server memory and t...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More