In the past two weeks, I’ve built a new Windows Vista system and upgraded the CPU on my primary production machine. In each case, I’ve seen problems pop up afterward that caused the Windows Reliability Monitor to report errors and related problems on those machines, and have watched their reliability scores plummet accordingly.
On the CPU upgrade system, I had an application hang in Outlook that I still don’t understand, and I had daily crashes on Windows Media Player (WMP) until my research taught me that changing a significant hardware component, like the CPU, causes various DRM checksums to fail and thus also causes WMP to crash as well. Once I looked up the relevant Help and Support article “WMP DRM system may not work if your computer hardware changes” (KB 891664), I killed that problem. I still have no idea what caused my Outlook freeze, but it seems that closing the program when I quit for the day keeps it from crashing. That’s also before Carbonite comes along and backs things up nightly, thereby also resetting key values on Outlook.pst that force me to restart the app anyway. Go figure!
At this point, I think I’ve got the issues on the production machine taken care of, and the Reliabilty monitor is finally trending upward again. In the meantime, I dropped from a high of 9.34 (barely acceptable, to me) to a pretty horrific 6.56 as I dealt with problems in the wake of the upgrade. These included two WMP crashes before I did my homework on the DRM problem and rebuilt a new DRM environment, the aforementioned Outlook failure, and a trio of disruptive shutdowns when the system locked up on me three days in a row and I had to hit the reboot (warm restart) button to get the machine up again. At this point, I think I’ve got those problems licked and am back to a more or less stable and steady state. No crashes in the last 90 hours anyway.
Before the upgrade, I had been using a QX6800 65 nm 2.93 GHz quad core processor; after the switch, I was running the QX9640 45 nm 3.0 GHz quad core part instead. What the switch bought me, more than a slight bump in processing power, was a major drop in power consumption (from 155W at idle to just 102W, and from 224W at peak usage to around 160W) and operating temperatures (the QX6800 ran in a temp range from 58° to 80° C; the QX9650 runs in a temp range from 32° to 62° C). I had been getting kind of nervous that my Zalman CNPS 9500 cooler just wasn’t doing enough to keep the QX6800 cool, but it seems to have no problem with the 45 nm QX9650. With ordinary temps in a range from 35° to 45° C, I’m a lot happier now than I was with ordinary temps in the 58° to 68° range before. So much for the upgrade issues and things I learned or observed…
The new machine has been more of a trial-and-error adventure. I’ve experienced problems with older Gigabyte motherboards when trying to use 4 x 1 GB memory sticks, especially when trying to push memory speeds at the same time. The memory controller seemed to stumble when I tried to push past conservative memory speeds on P35 mobos, but I thought that had been fixed on the newer X38 boards.
Not so, as a little research quickly showed me when I started crashing like mad after upgrading my RAM from 2 PC2-10000 Corsair Dominator SIMMs to 4 of the same sticks. In fact, I found hundreds of user complaints about the same thing on this motherboard, with a great thread on Tom’s Hardware, one of the sites that I write for myself. Because my DIMMs are rated for up to 1250 MHz, I thought I was being conservative by running them at 1066. Again: not so! Further searching showed me that 1066 is particularly prone to problems when running 4 memory sticks of any kind. 800 MHz appears to be the only way to go.
I messed around with this system over the holiday weekend, trying various combinations of settings and speeds. But it wasn’t until I dropped back to 2 sticks and cut memory speed to 800 MHz that I got the system completely settled down. This after days of BIOS fiddling, reading, researching, and profound reflection (not to mention a few quotes from George Carlin’s “7 Words You Can’t Say on the Radio”). I’ve got 2 2GB DDR2-800 sticks that I will probably end up inserting in the machine instead, just so I can get memory up to about 3,581 MB (what’s left over after the graphics card, BIOS, and drivers grab their share of 4,096 MB of memory address space). So much for overclocking this rig: I just don’t have the time for it right now. But if anybody has gotten this configuration to work, please e-mail me your BIOS settings, and I’ll give it another go.
Right now, I’m running the QX9650 underclocked (at 2.66 GHz) with the memory at 800 MHz. This is an incredibly cool system, with case temps in the mid-to-high 20s, CPU temps between 38 and 43 (they stayed so low for so long I thought the sensors were stuck, but when I left HWMonitor running overnight I saw peaks in the low to high 50s instead), and power consumption mostly in a range from 100W to 120W. I think I’ve finally got most of the stability issues shaken out on this box, too, but it’s only been about 60 hours since my last crash, which makes it a little too early to tell. My current reliability rating hit a rock-bottom 3.86, but at least I have nowhere to go but up from here!