Computer Component Standard


Components

Hardware to Avoid

Foxcon - http://boycottnovell.com/2008/07/25/foxconn-bios-acpi/ http://www.phoronix.com/scan.php?page=news_item&px=NjYyMA


Motherboard

Tyan S7002 or any other of the Tyan mother boards that support registered memory. A bit more money for more reliability is a deal. Tyan is the only mother-board manufacturer to supply sensors.conf files as of March of 2010 (some configs will need tweaking).

Sound card

AV200 Xonar DX for PCI-E slots.. lspci 05:04.0 Multimedia audio controller: C-Media Electronics Inc CMI8788 [Oxygen HD Audio]

CPU

I7 xeon E5504 - see http://en.wikipedia.org/wiki/Intel_Nehalem_(microarchitecture)#45_nm_processor_architecture for details.

Memory

Heat-sink

Thermaltake CL-P0501 CPU Fan For Intel LGA1366 (see Computer_Component_Standard)

Chassis

Rack mount is all I use any more - Wish they had a deal so you could rotate the CD ROM drives when sitting them on edge when used for a work station.. I use 4U only - 1U will either run too hot or have enough fan noise to drive one insane. I also think there should be a case that is sort of backward with all the connectors in front for running test set ups - I suppose one could build a 1U panel that extended all the connectors - I've not seen one available.

Power-supply

EPS12V  Used to recommend Antec - but they don't do SSI and the S7002 needs three MB connectors (one for each CPU besides the ATX)

Silverstone OP1200 Power Supply

Video card

dual DVI out

Keep trying to use ATI - keep having to go back to nvidia

VGA compatible controller: nVidia Corporation G92 [GeForce 9800 GT] (rev a2) At this time there isn't a 'single layer' nvidia card that supports dual head with a total of 1920x2 = 3840 resolution - it requires a double card. llpci VGA compatible controller: nVidia Corporation G92 [GeForce 9800 GT] (rev a2)

Key - go to nvidia site - look at the current Linux driver - find it's supported device list. Dual DVI connectors out.

Network switches

They apparently have decided to seriously support their ProSafe line of business class switches.

We are using several of their gigibit switches - the latest ones with some POE -

Hard-drive

SATA drives are all I will mess with - i am not sold that the SCSI are any more reliable than the good SATA

The Hitachi run quite a bit cooler than the Seagate drives - should result in better life and reliability. They feature the SMART monitoring system. I mostly get what ever size $100 or so will get me. Not sure when I will need all of the 60 160 500Gig 2TB size.

Thermal Pads Vs Thermal Grease - a critical failure point issue!

Thermal pads AKA the devils bubblegum is a serious failure point. Isopropyl usually not effective in dissolving the devils bubblegum.

After reading how thermal pads are as good as the grease I just used the heat-sink as it came. (I was starting to believe.) Even if the 'new' way works as well, after a year of cooking under a Athlon it clearly failed. The key issue seems to be poor lifespan. Yet, the film sellers claim that the grease works it way out over time - something I've never seen in my 35 years of working in electronics and they don't use pads in mil-qualified equipment that I've seen.

The dead computer had such a pad coating - but when I took it apart it no longer was at all flexible and over the CPU area it had clearly gotten VERY hard and brown. I cleaned it off with paint thinner (some recommend acetone) and had to scrape the top of the Athlon die with a screwdriver (it had turned very hard and brown like burned on food!). I put on some artic silver and now it works fine. (update: still working fine a year later). I have never seen thermal grease fail even at extreme temperatures that even burned the circuit boards!

There are two advantages the pads or film coatings offer:

  1. They are user friendly - and provide some cushion for those that can't seem to put the heat sink on gently (end users). With a little practice you can put heat sinks on processors with out stressing them. Find an old (dead) mother board and practice! uP manufacturers specify the pads so they don't have to tell customers they are clumsy.
  2. Pads are less messy. Working with grease takes a bit of practice. On the other hand, a smudge or two of grease isn't going to hurt a thing.


Bottom line - Grease works better

The - "works it way out over time" line is just advertising FUD that has no research to back it that I know of. I don't think there is much difference between the brands of the films, but my experience designing circuit boards that used heat sinks and thermal grease with power transistors over the years gives me a some expertise that the average computer user lacks. Grease will be a thinner film - all things equal, a thin layer will have less thermal resistance than a thick layer. Next, I have seen nothing in any engineering publication that shows any problem with thermal grease other than it is messy. If you have contrary information please post it here.

The few degrees of difference for the CPU is not a big deal, BUT once a pad starts to fail, it will fail completely. This run away failure is caused because as pad fails to transfer heat, its own temperature gets higher which causes the pad to fail even faster.

The temperature difference measured in the above link was when both the pad and grease were new. I will stick my neck out a bit here and guess that the 3-4 degrees C difference will increase with time and then very well could make a difference - random crashes etc.

Motorola once had a design note on how to measure transistor temperature - they had us drill a tiny hole in the metal part of the transistor tab so we could mount the thermistor as close as possible to the transistor junction. On a computer heat sink, it lays right on the chip top and there is no place to put a thermistor on the chip - but by drilling a tiny hole all the way through the heat sink, one can mount a tiny thermistor that could touch the top of the chip. Packing the hole with grease and keeping this hole VERY small (1/16") will allow you to measure the temperature of the grease interface between the heat sink and chip top. The extremely small thermistor for this are a specialty item, you probably would have to contact some thermistor manufacturers to find them. I also would not use such a drilled heat sink on a production system.

Reading chip temperature from the processor or motherboard won't have much accuracy to speak of, BUT as long as you don't change the mother board and processor the resolution of the on chip temperature sensor should be quite good and allow one to make valid comparisons between different heat-sinks and thermal interface products. Just don't expect to be able to make compare to your friends system with the same hardware.

Best heat-sink compound appears to be arctic-silver 5

BIOS settings

Testing

I've come up with a burn in procedure: You can set most BIOS to a slight over clock burn in setting. I put tissue paper over the air filters to raise the power supply inlet temp by 30 Deg F And ran I then boot up any tinny CD version of Linux and then I run cpuburn, https://packages.debian.org/search?keywords=cpuburn memtest, bonnie++, stress over the next nights.

Also see Breakin Memtest , vmt (Video Memory Test) ext. 


Discussion

The collection is based on price, keeping systems uniform and software hassle free when used on Linux. If you know of a reason that one of the parts is overly expensive or has incompatibility problems let me know. Better choices are always of interest. Post any information here. Be aware that I'm not looking for the Cheapest or Fastest. Reliability and low maintenance is the guide here.


Top Page wiki Index

Disclaimer

This information may have errors; It is not permissible to be read by anyone who has ever met a lawyer.
Use is confined to Engineers with more than 370 course hours of electronic engineering for theoretical studies.
ph +1(785) 841-3089

Email inform@xtronics.com

(C) Copyright 1994-2017, Transtronics, Inc. All rights reserved
TranstronicsĀ® is a registered trademark of Transtronics, Inc.
All trademarks are the property of their respective owners.