PC Hardware: Dell Precision 220

I recently changed servers from a custom-built PC I had to my old Dell Precision 220. Mostly I made the choice because the processor in the custom PC is a tad bit faster than the 220's. It's running an Athlon 1GHz (of the Thunderbird variety) where as the 220 is running a Pentium III. True, the 220 can accommodate two P3s ... but not many of the apps I use take advantage of that fact. The Dell is also a little bit better engineered so it works a little better as a server.

I purchased a number of upgrades to turn the 220 into a worthy server. In no particular order:

• A drive cage from a newer Dell model (off eBay), capable of holding four drives versus the 220s measly two
  
• Two p3s running at 1GHz (off eBay)
• A new Kingston 512MB RDRAM module from memorysuppliers.com (the 220 doesn't need the modules to be paired)
  
• A new Dell-compatible power supply from PC Power & Cooling, the Silencer 360 Dell
• An APC Back-UPS RS 800VA
• A Promise FastTrak TX4200
• Two SATA II 320GB Seagate 7200.10 drives (model # ST3320620AS).
  

I bought the memory prior to really getting into eBay. Once I did I wished I had gone ahead and bought the memory there as well. I could have gotten 1GB for the same price I got 512MB for a new module. I looked for a while but nobody was selling single RDRAM modules, only pairs which were going at prices higher than I wanted to spend. So I'm stuck at 768MB ... which seems to be fine.

Most of the the other components I bought new because I didn't want used parts. This machine will be pretty well abused with all the services running on it: Beyond TV, music, web, mail, etc.

A note on the SATA controller/hard-drives I bought. I was specifically looking for parts capable of a specific SATA feature, Native Command Queue. From what little I've read this feature doesn't provide much use on a desktop (at least not yet). But since this machine is being used more as a server I think it'll be important.

Most of the upgrades were painless, but I did run into some minor problems with the power supply:

First, Dell uses a custom pin configuration on the motherboard's power connector. During installation I noted that the PCPower part had a different pin-out from the Dell part. After some conferring with the PCPower techs, however, I was assured that the Silencer would work without issue ... and it did.

Second, while installing the power supply I realized that Dell customized their part further (beyond the pin out customization) by moding the shell. Two notches are cut in the top of the power supply that help hold it in place, along with a lone screw. I didn't want to risk any problems with the Silencer falling out of place so I got around the issue by using two zip ties.

Third, some time after installing the Silencer I noticed that my APC was switching to battery mode enough to drain the battery. That was a big problem. At first I thought the unit was having problems due to continuous high voltage (hovering around 128 quite often). But after looking through the APC knowledge base I found a few articles that talked about Total Harmonic Distortion (THD) and how it can cause the UPS to switch to battery mode. Switching the sensitivity of the UPS to "medium" seems to have addressed that problem.

Windows Weirdness: SB Live!

"Holy crap, you're still using a Live!?" Yes. Yes I am. This is my work pc so it doesn't need the latest an greatest. Hell, even my play pc isn't that up to date anymore (though it's good enough for Final Fantasy VII).

Anyway ... this post isn't (necessarily) about antiquated technology. Ever since getting this pc repurposed as a work machine I've had a problem with the sound stuttering. I let it go for a while, but decided I couldn't take it any more and so decided to try to do something about it.

My first clue that something was amiss was in looking at the XP device manager. When viewing by resource type just about all the PCI devices on my system were assigned IRQ 9. Weird. Not totally unheard of thanks to IRQ sharing in modern PC's ... but weird non-the-less. It gave me an idea for what could be wrong so I did a few searches on Google.

Suffice it to say there seems to be plenty of blame to go around. Nobody can agree on who's at fault, but it looks to me like a confluence of factors causing issues.

1. VIA's Apollo KT133 chipset is implicated in a few places for the IRQ madness. Perhaps true, but my Asus A7V seems more than capable otherwise so I'm willing to give VIA some slack on this one.
2. Creative's SB Live! has gotten it's share of dress-downs. No doubt the drivers are a little suspect for XP since the card is fairly old. And I don't doubt it's a bit of a resource hog (on the PCI bus that is) since it's the only component that was having problems. I saw at least one post noting that the card has a poor ACPI interface header. I don't know much about that last one, so I won't comment on that particular ... um ... comment.
   
3. Windows XP once again shows it's immaturity. "Wha?" you say. I feel Microsoft can take some (or a lot) of the blame for this one. ACPI is supposed to alleviate problems with resource conflicts, but somehow it's just made them worse in this instance. Why doesn't the system see a preponderance of resource usage piling up on a single IRQ and address the problem? I know it probably has to partially due with the ACPI standard, but Microsoft had a hand in defining and implementing that standard. I think they could have done better.
   

Normally I like to cite resources, but unless anyone really wants to see the information I read I'm not going to bother on this one. There's plenty out there on Google web and Google groups.

I would like to point out two resources, however. The first one I'm pointing out not because it's necessarily the best one out there. I'm pointing it out because it got me where I needed to go by leading me to the second reference once I started down the path of this work-around.

So first we have Irq weirdness on usenet. Since the problem appears to be related to overuse of a single IRQ resource the obvious work-around would be to modify the IRQ used by the various devices. Unfortunately, if your computer type is specified as ACPI (see the "Computer" group in Device Manager) you're out of luck. Since the goal of ACPI is to automatically manage resource allocation to avoid conflicts, manual allocation is disabled.

To regain this functionality you have to modify the computer type to be "Standard PC" ... which I did by "updating" the driver for the ACPI device to the "Standard PC" driver. Making this change can have its own pitfalls (not the least of which is resource conflicts). Luckily my motherboard is more than capable of managing resource usage so once I rebooted everything worked fine. I still had a shared IRQ between the SB Live! and the built-in ATA66 contoller. I still couldn't change the settings from within Windows so I rebooted, entered the BIOS, and specified the IRQ I wanted to use for the PCI slot containing the sound card.

Doing this caused Windows to lose track of ... well ... everything. All system drivers (and I do mean all) had to be rediscovered. This meant redoing the display settings, redefining my wireless connections, etc. Luckily I have yet to run into any major problems otherwise.

The one issue I did encounter that was going to be extremely annoying was that my computer would no longer shut down by itself. I was getting the classic "It is now safe to turn off your computer" message. I found the answer to this one, and my second reference, from Microsoft. See the aptly named KB "It is Now Safe to Turn Off Your Computer" error message when you try to shut down your computer. The procedure basically has to do with disabling the ACPI driver and installing the driver for APM support.

After all is said and done I now seem to be stutter free.