Tyan 2460 / Duron OC success!

I've constructed a few systems for myself and my friends over the past four or so years (though only three were overclocked), including a couple of rather spiffy Athlon rigs recently. My own hardware tended to lag quite a bit behind, though.

Indeed, my trusty K6-2 500 (@ 500 ) was getting rather long in the tooth after two and a half years. Thus the decision to upgrade was a no-brainer-- but I still needed to make my new machine somehow cooler than those of my friends. I'd always wanted to do SMP anyway, so a nefarious plan was born.

I acquired three 1.0 GHz Morgan-cored Durons, with the notion that I'd use the two that liked each other the best in SMP and hopefully OC'd. The third was for a cheap SDRAM board to upgrade someone else's system. After analysis of the fragmentary data on dual Durons on specific boards, I settled on the old standard Tyan Tiger MP. Not an overclockers' board, but my plan was to simply jumper it up to 133 MHz FSB and hope it worked. I could have gotten 1.3 GHz Durons about as cheaply, but as you know a dually needs all the RAM bandwidth it can get. I figured the 1 GHz units wouldn't have a problem with hitting 1.33 GHz in the absence of voltage adjustment. In spite of bad stories about Crucial RAM on certain boards, I picked up a cheap 256M reg ECC stick from them. Technically I could return it if it didn't work, since I'd jumped through their hoops when selecting it... The coolers were Foxconn PK04535 copper-base units that I selected because I was an idiot and thought they were entirely copper. They only engage one socket lug, but otherwise don't seem bad. At least they fit with something resembling clearance.

Well, my low-cost plan paid off. The CPUs turned out to be AHLCA (week 12, iirc) units with consecutive serial numbers. The first two I tried POST'd just fine together. I jumpered up to 133 with no issues, and indeed the system has served me for a week now with no crashes that I don't blame on software (certain games get mad when they expect EAX support and don't find it).

I recycled my year-old Radeon 7000 and rather old Turtle Beach Montego II (Aureal Vortex 2 based) sound card, bought new drives and a bunch of fans I'll figure out where to mount eventually, and I was good to go.

Oh yes-- I didn't skimp on the power supply. It's an Antec True 430, as I've heard too many horror stories about crappy power being the immediate or eventual downfall of otherwise good systems. And in my opinion, I now have a good system. =)


Would I do anything differently? Well, I'd probably get better-known CPU coolers-- not that my Foxconns haven't handled the situation. I had pondered getting a pair of Athlon MP 1.2 GHz units for 80$ each instead of the 40$ each I dropped on the Durons. Sure, that would've lowered my clockspeed, but the larger cache is useful for duallies... and at the time I didn't know for certain that My Little Morgans would function at all-- let alone overclocked.

I now feel that saving money at 'startup' was the right decision. One thing my new system has is plenty of room for additional RAM and zippier processors in the future. I realize that I won't get any significant overclock out of future processors with this motherboard, but hey-- no situation is perfect. "Cerberus," the low-budget, two-headed dog of Doom, will serve me well for a long time to come.

I can't even get that out of my Morgan on a single board, to get two of them at 1.33GHz on a Tyan, no less, is amazing! Excellent work.

Welcome to the forums as well!

Funny you should mention overclockability, as the third CPU did indeed go on a Gigabyte GA-7ZXE (KT133A board) during the same upgrade-fest when I assembled Cerberus.

That system had formerly been powered by a socket-370 Celeron 366 @ 550 on an Abit BM6 when I built it three years ago, but one of the two CPU fans burned out (long story). Not able to acquire a 50mm replacement until recently, he'd managed to hobble his system along at 83 x 5.5 = 458 for the last couple years. Everything in his system was happy with the not insignificant PCI and AGP overclock. including the Adaptec 2930U2 Ultra2 SCSI card and his new Radeon 7200. The 640M of RAM was effectively PC133 (another long story), and therefore was run at cas2.

The motherboard swap was fairly simple. The Gigabyte board had a jumper/BIOS FSB adjustment combo, such that we could select 100 to 120 MHz in 1 MHz increments, or step up to 133 and then proceed all the way to 200 (!) in the BIOS.

After making sure everything worked, we jumpered up to 133 and everything was good-- until his machine faced the ultimate stress test: a few hours sitting in "The Bazaar" in Everquest. This would eventually cause a crash to desktop, so we retreated to 120MHz x 10. The board has a BIOS option to run the RAM asyncronous, though we haven't yet figured out (haven't benchmarked it at all) whether that means a 4:3 vs. the FSB, or if it simply goes async to 133 MHz RAM clock. I'd guess it's 4:3, but anyway...

The proud owner isn't complaining in the least, since he's now cooking at 1.2 GHz, and I did the upgrade for a song and his old board and chip. Still funny, though, that it failed to quite hack it at 1.33 GHz while my CPUs (with the two previous serial numbers) succeed. I don't feel like endless chip-swapping experiments, so we'll never know whether the chip is less capable than its brothers or if something else is at fault.

Incidentally, my machine passes the "Bazaar Test." =)