Mike wrote:
On Thursday 14 February 2008 15:54, Per Jessen wrote:
OK, yesterday morning I moved the box outside, hooked it up with network and power. Outdoor temperature was -1C .
Temperature rose to about 6 in the afternoon, but the machine actually ran pretty stable throughout the day. I had one lockup which needed a hard reset, but I haven't been able to reproduce that.
Whilst running 4 x mprime, the temperatures as seen by sensors stabilised at this:
work System CPU temp3 Outdoor 4 mprime 39 51 81 5 4 mprime 35 50 80 6
So it's all about cooling. The interesting thing is - the machine uses the AMD supplied CPU heatsink and fan. It's got additional fans and heatsinks to improve the air/heat flow. How do I improve on the cooling without looking at water-cooling and such?
The first thing I always do is get rid of the AMD supplied fans. I've used Thermaltake type heatsinks and fans. I know that Coolermaster and Thermaltake make pretty good equipment for CPU cooling.
How big is your case? Midtower, full tower?
Well - the system remains unstable and automatically reboots after a while, maybe 10-15mins, of running 4 x mprime. I've already got an RMA# so I think this board is going back after all.
Could be the problem.
Also remember.. Fans at the bottom of the case blow inwards. Fans at the top of the case blow outwards. In a quiescent state, CFM in = CFM out [CFM = Cubic Feet/Minute = (ft^3)/minute, essentially the same as m^3/minute except for a dimensionless conversion factor of about 30. or, to make it more clear CFM in = CFM through = CFM out. This is an application of Thevenin's Current Rule and is practically a mathematical identity. This is an identity. if CFM in > CFM out, the case would eventually explode; conversely, if CFM in < CFM out, then eventually the case would be completely evacuated). What does this mean? If you have 1 fan blowing in rated at 300 CFM, and 2 case fans and a PSU fan blowing out, each rated at 300 CFM, then initially, the out fans will blow out 600 CFM more than the in fans, until the air pressure inside the case has dropped enough that... the in-fan is bring in a little more than 300 CFM...and the out-fans are blowing out a bit more than 300 CFM total (a little over 100 CFM each, 1/3 of their rating!). Additional in-flow might be provided by leaks or ventilation holes in the case. In any event in the above scenario, the air pressure in the case is low, which means that any fans and heat sinks used inside the case for CPU cooling, etc are operating in an environment in which there is less air density than the thing is designed for, and will therefore not cool as well as if there was more air inside the case. If the situation is reversed (3 fans blowing in, 1 PSU fan blowing out), we still get to the same quiescent state of the flow-through being slightly higher than the flow rate provided by one fan. Initially, the in-fans blow 900 CFM in, while the out-fan blows out only 300 CFM...but as internal air pressure RISES, the input fans flow rapidly drops to about 100 CFM each (again), and the out-fan now runs slightly above rated capacity (a bit over 300 CFM). But now...the air WITHIN the case is under increased pressure. The CPU fan is blowing MORE air molecules over the heat sink...therefore IMPROVING the cooling ability of the heat-sink/fan apparatus. In this second example, using case perforations for more air flow is more acceptable than in the first example... In both examples, the air flow through performations only happens due to a difference in air pressure inside the case vs. outside the case, but in the second example, the perforation air-flow is due to higher air pressure (and thus, higher air density inside the case), while in the first example, the perforation air flow is due to a vacuum condition inside the case, which is detrimental to cooling the CPU heat sink -- vacuum conditions within the case should always be resolved by increasing the amount of air forced into the case. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org