I recently tested out upgrading several jobs in my team’s workflows to the m4pro.medium resource class. While I noticed a significant boost to build and test times, anything dependent on network was significantly slower. For a 3.52 GB project, the checkout time went from under 1 minute on M1 or M2 to 4-6 minutes on M4. I’ve been able to workaround this with a custom shallow checkout step, but just wanted to raise this here for further investigation.
We are currently working on improving network performance for M4 pro.
Has there been any update on this? I tested a standard checkout on our M4 jobs again, and it’s still much slower than it was on M2. I’m a little concerned about this with the upcoming deprecation of the older resources.
From my recent test:
- M2: 43s
- M4: 3m 33s
Any update here a year later? We see jobs on M4 Pro Medium fall on either of these opposing sides when checkout code (using the same checkout method):
Receiving objects: 100% (20436/20436), 465.06 MiB | 24.39 MiB/s, done.
…which takes less than a minute to complete.
Receiving objects: 100% (20436/20436), 465.13 MiB | 966.00 KiB/s, done.
…which takes 8+ minutes to complete.
The latter seems too slow, especially at the current pricing for M4 Pro Medium at 200 credits/minute.