Sempron Overclocking - Starting with the bios
(Page 2 of 6 )
Starting in the DFI bios, there's more than a fair share of options to play with. For the purpose of this article however, I'll stick to just the commonly available ones. Blowing through the first couple of HTT steps was no problem. The board and CPU combo had no issues right up to 240MHz HTT, giving a total clock of 2160MHz.
At this point something I wasn't planning on occurred. On the DFI LanpartyUT nF3-250Gb, two of the SATA ports suffer from the same issues that many A64 boards do, in that the lock doesn't function properly at high HTT. After switching my RAID array to ports 3 and 4 however, it was back to the races. The memory divider was also augmented to CPU/11 instead of the stock CPU/9 since the known frequency limit for my KHX was approaching, and I felt the CPU still had more to give.
Adding a dash of voltage whenever instability reared its ugly head in Prime95, the HTT increase marched on to 288MHz. At this point, even pounding 1.85V into the chip would push it no further with absolute stability. Perhaps putting a freeze on the chip from a Vapochill or Prometia would make a difference, but this is as good as it was going to get with quality water cooling. This was also done with only one stick of 512MB KHX; adding the second caused errors within Windows. Backing the HTT frequency off to 285, and dropping the memory divider down one more step to CPU/12 cleared it up. With some extended burn-in, I'm hoping to force the CPU to accept both modules at CPU/11, but that's going to take time.

So what does an extra 800MHz do for this chip? Well, a hefty 45% clock increase bringing it up to FX-55 speed isn't going to hurt it, that's for sure. Before getting into numbers though, I'd like to look at what this projects for AMD's K8 design. The 2.6GHz number is starting to be awfully common for them. It's the highest speed they have a retail processor selling for, and more than a few times now it's the highest clock I've seen achievable for people overclocking without resorting to extreme methods. And that's on either process, the older 130nm that this Sempron 3100+ is built on, or the newer 90nm, from which there really isn't enough of a base yet on which to truly make any definitive statements.
Next: Testing: Adobe After Effects, STARS CFD Solver >>
More Hardware Guides Articles
More By DMOS