Hi there, just copying a post I placed in another thread as I think it's also relevant here:
I must say I'm divided on the psu issue now that I've had a chance to test a little...
Bought my first miner with the red & black 12V connectors on the boards and ran it off an Antec Neo Eco 520C 80+ bronze which has a single 12V rail delivering 480W at 40A and got mad hardware errors at 400mhz so clocked it down thinking it was the power supply.
I then bought an Antec 750W High Current Pro 80+ Gold which has 4 12V rails delivering 744W at 40A per rail and the hardware errors were only slightly reduced but it was still not worth running at 400mhz.
So, I bought a second miner now with the yellow and black 12V connectors and connected it to the 520C and clocked it to 400mhz and it's rock solid with <1% errors:
IMAGE REMOVED...
In the end; same 520W PSU, 2 different generation miners and completely different results, one with >6% errors and one with 0.6% errors at 400mhz... go figure!
I also notice on the newer miner the fan runs faster as it seems to want to keep the boards around 40o whereas the older miner will keep them around 45o... I guess there must have been an update to the PWM management software?
to answer the question in the quoted post - it seems that either the fan drver or the fan model was changed, which is why the lower temperatures and frequently not spinning until hashing causes the fan to pass the minimum starting voltage
The fans look almost identical from the outside, but closer inspection reveals the first model is an NMB and the newer one is an FXDS. The fact that the newer ant keeps its boards at 40
o instead of the older 45
o makes me think it was a firmware change or something like that as the board temps should be regulated regardless of the fan model used... no?
EDIT: Even closer inspection reveals the older ant has half the RAM apparently:
OLD:
[ 0.000000] Determined physical RAM map:
[ 0.000000] memory: 02000000 @ 00000000 (usable)
NEW:
[ 0.000000] Determined physical RAM map:
[ 0.000000] memory: 04000000 @ 00000000 (usable)
I guess that would explain the hardware errors, but not necessarily the lower temps, unless the chips are doing more work and running hotter due to less RAM?
