Apparently, I’d been living under a rock for the past five years, but I didn’t even realize there were kids out there making hardware using GAs. I went to Korea House with this one last night and he set me straight. It’s kind of neat to think that you can make a cluster of computers produce a chip by evolving hardware to meet some conditions. While I think that doing an FPU this way would be effectively impossible using conventional hardware, what if quantum computers were used? I haven’t worked out the math yet, but it should be possible. I should note that I became kind of obsessed with GAs for a while when I was in college. I solved the famous N-Queens problem using one as an undergraduate, and was fairly amused by not understanding how they work, but seeing that on average they’re faster than conventional exhaustive solutions. It’s interesting to think of designing hardware this way, but I think it could scare some people to think of computers automatically designing hardware. Don’t be scared though, the process isn’t all that automatic.

Apparently, my Antec Neo Power 480 power supply was only pushing 11.17V to the 12V rail. I pulled it and put in an older Enerrmax 365 power supply (not as swanky), and now I’m reading 11.80V on my 12V rail. I hope this helps as my system has been fairly unstable since I built it in March…

I finally rebuilt my Windows XP box. I had previously been using a Via KT880-based Asus A7V board with a 3000+ Athlon XP. Unfortunately, it was never stable. I spent about 15 months tweaking things, adding copper coolers, ram sinks, different memory, new power supplies, and underclocking. After all of that, the machine _still_ had issues. I eventually put in an Athlon 1500+, and while that was more reliable, the machine would still occasionally flip out. Last night, I finally swapped out the Athlon motherboard and chip with an Asus P4P800 Deluxe and a 3.2GHz Pentium 4 (leaving all other components the same). Everything is working like the clappers now. The new board and chip are a lot quieter, and runs infinigons around the old one. Also, the P4P800 board actually supports all four DDR DIMMs that I bought for the KT880 board even though I had two different sets of matched pairs.

You may be wondering why I went Intel this time. I had been running an entirely AMD shop for a while now. Hudge, the machine that hosts involution.com is an Athlon XP 2400+/Via KT133A-based system, and Homestar, my ever reliable Linux firewall, is an Athlon 2600+/Via KT133A box. The problem was that I’m doing a fair amount of recording and video encoding now, and I have a ton of PCI gear that I don’t want to part with. So, I saw a good deal on the refurbished P4P800 board on Newegg for $85, and I decided to be processor agnostic again, and go with an Intel box. Before running the AMD desktop, I was using a 800MHz Pentium III on an Asus CUSL2 board. This was probably the most reliable machine that I have ever owned. So, all of this led to me going back over to the dark side and building another boring Intel rig.

The other big news is that I will not call my new machine Yellowdart as it was before. I have renamed it to Inifnigon after the terminology rediscovered by the metasnazz today. I thought about calling it Allupons too, but my LJ friends page coupled with my daft instinct prevented that from happening. This whole ordeal has allowed me to regenerate 23 happy cells in my brain.

Is it some kind of warning sign when it takes two days and 57 DVD+Rs to backup one’s home directory? I knew I should have gotten that DL burner…

Well, another piece of hardware failed here at HQ. The chipset fan on my main server seemed to have stopped, and it was causing the Via KT133A chip to run very, very hot. This was causing intermittent lock-ups and other odd behavior on my box which is typically very reilable. I made an afternoon trip to Fry’s to try to find a very esoteric chipset cooler that would attach to my motherboard. To my delight, they did manage to have a
Thermaltake Orb Chipset Cooler heedlessly tucked away on a clearance shelf. I think the Orb cooler is intended for Geforce cards, but it does fit precisely in the two pre-drilled holes on my A7V133 motherboard. So, after installing the new chipset cooler, all was still not well. Apparently, there was some problem with my PSU’s 5V rail. So, I replaced that with a spare that I had laying around. After that, everything seemed to be back in working order, and my chipset was running 7 degrees cooler.

I kind of want to build a new x86 computer, however, I’m not seeing the motherboard that I want. I kind of have a reputation for having some excessive, but relatively low cost boxes among my collection. The machine that this box runs on was built from a $50 motherboard that I bought on Ebay housing a 2600+ Athlon XP ($99) and 1536 MB of PC133 memory that I got from mwave.com for $150. So, it’s a pretty kickin’ machine for $300 worth of parts.

I’ve been thinking about getting some kind of Opteron system as a pure workstation-type machine. The only board that I can find that really interests me is the MSI Master3. It’s a swanky dual-Socket940 board with twelve DIMM slots. Unfortunately, it doesn’t AGP or a PCI-E x16 slot to run some kind of pimp video card. Word on the street is that Nvidia is going to make a killer chipset that supports dual PCI-E x16 slots for their new SLI cards. I’m looking for something similar to this MSI board with dual Socket940 support and lots o’ DIMMs with PCI-E or an AGP slot.

That’s right, some crazy people with access to LN2 and a 156-Watt TEC managed to get a P4 up to over 6 GHz. You can read about it here.
With a 6 GHz computer, Doom III is almost playable at 1280×1024.

Orion Multisystems is now flogging some seriously smoking workstations running Fedora Core. Essentially these machines put 12 Transmeta to 96 Efficeon processors into a single desktop or deskside box. I’d love to have one these things at work. I think the ideal desktop solution for me right now is a high-end Mac with two 30-inch cinema displays and a Orion DS-96 system. Technically, I do have access to way more powerful machines than these, but unfortunately, I don’t have root on them.

I added an article on what I did to get the Promise SX4000 to work on Redhat 9.