View Full Version : Aguatec Blue Ice CP-101
dnar
20th July 2003, 04:46
LOL, bolt on H2O for the slackers... Interesting concept.
http://www.ocaddiction.com/reviews/cooling/aquatec_blue_ice/
Gets a very crap 2/10 rating BTW.
Side note: why do people beleive FAH to be the best thermal stress test for Intel processors? I recently upgraded my server to P3-600E@800Mhz, and have found FAH runs the CPU at 40c where as most other DC projects such as ECC2-109 run it at 48c..... AMD processors are heated equally across the DC projects such as FAH, GAH, ECC2-109 and Seti.
Bruce
20th July 2003, 14:04
Originally posted by dnar
Side note: why do people beleive FAH to be the best thermal stress test for Intel processors? I recently upgraded my server to P3-600E@800Mhz, and have found FAH runs the CPU at 40c where as most other DC projects such as ECC2-109 run it at 48c..... AMD processors are heated equally across the DC projects such as FAH, GAH, ECC2-109 and Seti.
I don't think anybody said FAH is the best thermal stress test. There are a variety of things that can be stressed, and, as I'm sure you know, any one of them can bring down a system. Neither Tinker or Genome ever claimed to be a stress test, but Gromacs does put a lot of stress on SSE, which, in turn, puts a lot of stress on floating point and memory. If that happens to crash your system at 40c it's just as much a crash as if some other test crashed it at 48c. You can't really call a system stable until you've stressed each component of your system - - which might, or might not, require several different tests. Gromacs should be one of them.
phil
20th July 2003, 14:40
I personally don't use DC clients to stress test my system because no error checking is involved. I use the Prime95 Torture test as it error checks all it's calculations. I have found that most systems can run happily at a much higher speed while still generating errors. For example, my P4's max Prime95 stable speed (2x clients as it is a HT enabled CPU) is 3.64GHz. It will easily boot into windows and run benchmarks at 3.9GHz but it fails Prime95 almost immediatley. Now, if you don't use a program that error checks it's results you would assume that 3.9GHz is totally stable all the while sending worthless results back to the DC project of choice....not good for the science.
I use Prime95 for CPU stability and memtest86 for memory subsystem checking. Only when both of those tests pass at least 12hours continuous running, am I 100% happy that the machine is stable enough to run a DC client.
dnar
20th July 2003, 18:11
I have found the BEST check, is compiling glibc. I can find instabilities compiling glibc that both memtest86 and Prime95 dont.... Make sure you have plenty of free disk space before compiling though!
verT
22nd July 2003, 04:09
Originally posted by dnar
I have found the BEST check, is compiling glibc. I can find instabilities compiling glibc that both memtest86 and Prime95 dont.... Make sure you have plenty of free disk space before compiling though!
pardon my stupidity but what is glibc? I have always used P95 and memtest as well and thought they were good enough....
veggyhed
23rd July 2003, 01:37
I think its a Linux thing. At least I cant load the latest lib version under RH8.
dnar
23rd July 2003, 09:11
glibc is the C library of GNU Linux, it's the next layer around the kernel so to speak, and without it you have NO OS!
ohms18k
23rd July 2003, 09:38
Gets a very crap 2/10 rating BTW.
It's to bad cause that baby would look good in my clear blue box. :rolleyes:
vBulletin® v3.7.4, Copyright ©2000-2012, Jelsoft Enterprises Ltd.