I like to tell another thing, my 2 PCI-E Power cables to S3 were a little hot.
My S3 hashrate was not stable at the rated 441Ghs but was averaging at 420. I ordered thermal paste.
Before thermal plates were arriving, I tried to plug two additional PCI-E power cables to S3.
The Antminer S3 is running at stock clocks 218.75 and PCI-E power cables are cool now.
The main reason for this post is to report that the miners stabilised at 441. Earlier it was 420.
Two thing I'm pointing out,
1. The miners plugged in continuously for 4 or 5 days and seems like it was running-in and stabilised at-last at 441.
2. Plugging the all four connectors even if you are not overclocking will some way help you.
Yes, but did they stabilize because of a 4 days run-in or because of 4 vs 2?
I am also in the middle of one experiment. A unit was fluctuating between 420 and 428 on 212.5 (~400 only on 218.75) on four connectors, now I switched to 2 connectors in front (first pair close to the ethernet port; connectors are from the same double cable from EVGA1300, incoming cable is warm (42C with a laser temp probe), temp of actual connectors is 31 and 29C-unit is at 429gh (after 18hr), so possible slight improvement.
Conclusion: a possibility that 2 connectors are better than 4 (at least for some machines).
Will report on latter stats and will try to go to 218.75, which it was NEVER able to properly do before (on 4 connectors).
Another new unit I am "burning in"-checking it with just a single double cable from CX500M, connected to the "back" pair of connectors-incoming cable is at 42C, first connector is at 32C, second at 30C with ambient at 25-26C, speed 441.55GH at 218.75 for 20hr.
Conclusion: CX500M is OK at regular speed, temp not that high, speed good
Can be just coincidence too.
Higher gauge or multiple strands will always lower cable heat even though it doesn't matter much here.