Grex Staff Notes: Assembly of Grex 2003 Computer

In the spring of 2003, Cyberspace Communications started the job of moving the Grex from it's old Sun computers to an x86 system running OpenBSD. This will be the first time Grex has ever run on a computer that was purchased new. A fund raiser was begun, and Grex users quickly donated enough money to purchase the needed components. Many thanks to our generous users.

After much discussion with other staff, STeve André selected the following components for the new system:

description vendor price
Antec PLUS1080AMG fileserver case with 430 watt power supply. CompUSA $159.99
2 extra Antec case fans. CompUSA 2 x $15.99
2 Asus A7VBX/GBL/13 motherboards (one as a spare). Strategic Computer Solutions  2 x $115.00
2800HMz AMD Athlon XP2 CPU chip, with heatsink/fan. Strategic Computer Solutions $250.00
Adaptec 29160 SCSI controller card. Strategic Computer Solutions $190.00
Seagate ST318453LC 18.4 GB had Disk. NewEgg $212.00
Western Digital WD800JB 80 GB IDE Disk. NewEgg $96.00
3 Crucial 512MB 184-pin DIMM memory cards. Crucial 3 x $78.99
Sony CRX220A1 CD-R/RW drive. CompUSA $69.99
Sony MPF920-Z 3.5 inch floppy disk drive. Strategic Computer Solutions $20.00
OpenBSD 3.3 distribution CD. openbsd.org $40.00
TOTAL: $1504.95

The two disk drives and the OS hadn't arrived in time for the initial assembly. The high performance SCSI drive will be the main drive, and a lower perfomance IDE drive will be for rarely used data. Dan Gryniewicz donated a PCI video card. Special thanks to Leeron Kopelman of SCS for good deals and extreme helpfulness in getting the parts we wanted.

Assembly Fest One: May 4, 2003

Present for the initial assembly on May 4, 2003, were Mark Conger (aruba), Dan Gryniewicz (dang), Jan Wolter (janc) and Dan Romanchik (danr). Dan Romanchik served as supervising engineer and Jan served as recording journalist and assistant supervising engineer, so this left Mark Conger and Dan Gryniewicz to do actual useful work. Mark had prepared by studiously reading all the manuals and writing up some notes on what to plug in where. Dan Gryniewicz had built similar computers in similar chassis before.

CPU board with processor installed Assembly started with the insertion of the CPU into the motherboard, as shown at right. The back of the board is at the top of this picture. Six PCI slots are at the back left (the blue one has special support for an Asus wireless interface but can be used as a normal PCI card slot). An AGP slot is to the right of that. CPU is already in the socket at the right, just in back of the camera flash reflection. The memory slots are just in front of it.

All motherboard DIP switches were left in their default factory settings. They were all pretty obviously not relevant to Grex.

We departed from standard procedure slightly in the installation of the heat sink. The heat sink rests directly on top of the processor chip, stablized by the four round rubber pads visible in the corners, and contacting the shiny rectangle in middle, which is directly over the working part of the processor chip. For the heat sink to be effective, it must have a good thermal connection to this little rectangle. Our heat sink, like most, came with an adhesive pad on the bottom. Several of the web pages that we read suggested that these pads tended to reduce the effectiveness of the heat sink and using a good thermal compound instead was a better approach. So we took this route, guided primarily by the directions we found at the DUX Computer Digest website.

Application of thermal compound to CPU We first removed the existing pad from the bottom of the heat sink. The heat sink can be seen (upsidedown) at the bottom of the picture at right, with the faint rectangle on the bottom marking where the pad was. This was scraped off carefully to avoid scratching the metal surface, since scratches interfere with the contact between heatsink and CPU. It was then cleaned with isopropyl alcohol. Mark squirted some silver thermal compound onto the fingertip of his glove (worn to prevent skin oils from messing with the thermal compound), and dabbed a thin layer onto the chip. The best thermal contact is direct contact between the heat sink and the CPU, so only a thin layer of thermal compound is wanted to fill tiny scratches and cracks. Care was taken not to smear it around anywhere except where it was needed since it is an electrical conductor.

The heat sink is held firmly in place using spring clips at the back and front of the socket. It is first attached at the back and then carefully tipped down to clip in front. Mark had tried this out beforehand without the CPU chip in the socket.

This departure from the official procedure probably provides better cooling, but we learned later that it also probably voids the warrantee on the CPU chip.

Next we started installing various components in the chassis. The Antec case is rather nice. It comes with two fans in back, and one in the removable side panel. We had purchased two more to go in the front. All these fans just snap into place. For the CD drive, we just screw a mounting strip onto each side of the drive, then it snaps into place. (Extra mounting strips are attached to the bottom of the chassis). The bays for 3.25 inch drives also just snap in.

Installation of motherboard After shuffling loose cables aside and inserting the CPU's back plate, the motherboard was inserted into the case. It is held in place by nine screws. We then inserted the three memory cards.

We spent the next half hour plugging cables in. All the fans had both power and control cables. We didn't connect any of the control cables. If the fan control cables aren't connected, the fans just run at full speed all the time. This is just what we want, and OpenBSD doesn't support fan speed control anyway.

The Antec chassis has USB and Firewire ports on the front of the chassis. We don't expect to use either one, but who knows? The USB cable connected to the motherboard reasonably well, but the firewire connector was declared to be not worth the effort. Each of the six conductors in the cable had a separate little plug that has to be attached. These were hard to get in because of cramped quarters, and then fell out at a touch. So we hooked up the firewire connectors on the back instead. These have a real plug and was easily installed. So the USB plug in front works, but the Firewire plug in front doesn't.

The CDRW drive was set up as master on the second IDE channel. Our IDE disk, when we get it, will be master on the first channel.

People opening the case in the future should be aware that there is a power cable connected to the fan in the door.

Working computer The video card and SCSI controller card were inserted into PCI slots, and, two hours after we started, we were ready to power it up, watching closely that all seven fans were running (five in the case, one each in power supply and CPU heatsink - a bit of a step down from the nine fans in our previous computer). This is important, since the CPU can fry itself very quickly if its fan is not running.

The BIOS came right up (dimly visible on the monitor at right), and asked for the CPU speed. The actual CPU speed for this processor is 2083 MHz. The only other change to the BIOS was to disable the full screen logo. The BIOS found all drives, memory and controllers. The BIOS memory checks were fine. We booted a Linux CD and everything looked good from there too.

We then got a memtest86 CD and started that doing continuous memory tests. We'll leave the machine running that for a while. So far no errors have been found and no smoke has emerged from the machine.

Assembly Fest Two: May 11, 2003

A second assembly fest was schedule for a week later. In the mean time, the board decided to buy two additional 18.4 GB SCSI drives. Though the drives arrived in time for the assembly fest, they all turned out to have the wrong kind of SCSI connectors - the drives had 80-pin connectors, we want 68-pin connectors. This left us with nothing to install but the IDE disk.

Nobody except Mark showed up for the second assembly fest, so he installed the IDE drive without assistance. He then installed Windows 98 on the drive, just to use to test the system more thoroughly. Windows 98 was badly confused by the huge amount of memory in the system, and Mark had to edit the system.ini file to make it work. But he was them able to get the system on his LAN and confirm that the ethernet interface worked.

Mark returned the wrong-connector 18.4 GB SCSI drives and ordered three new ones from Strategic Computer Solutions.

Jan Wolter - Sun May 4 23:33:16 EDT 2003 - first draft.
Jan Wolter - Tue May 13 09:48:58 EDT 2003 - update.