I found out that my processor no longer runs efficiently at oc'd speeds, or even stock speed. Seems that in 3DMark2001, the vertex shader score drops drastically as the CPU score increases. At 1.3GHz, the score is 15 FPS, 1.26GHz is 26 FPS, 1.2GHz (stock) is ~50 FPS, and at 1.13GHz the score is a robust 80 FPS. I used to have the CPU up at 1.4GHz, but after a while, it would no longer go 12 hours in Prime95 (or 12 minutes), so I dropped it to 1.3GHz, where it was fine. Then my Ti4400 showed up, giving me dismal vertex scores, though everything else was fine. Thought it was my motherboard, but instead, on a whim, I brought the CPU to stock speeds, and the vertex score shot up. So, I think that running at or more than 60C for the last 7 months has damaged it. Ah well, it was time to upgrade anyway. I just wish that my motherboard actually supported more than the pre-AXPs.
Speaking of which, I need suggestions for a new motherboard, preferably a non-VIA-based one. I liked the IWill XP333 I put into a friend's computer, but if possible, I want something... better. Too many VIA 'boards out there, and since I've had bad experiences with multiple VIA boards, I'll stick to the ALis' and SiSs' of the world.
One more thing that I just thought about. Given the fact that the vertex score is so greatly affected by the CPU (or at least the ability of it to process things correctly), is it possible that the NV30 and R900 are incorporating some kind of T&L-like hardware onto their boards to process DX8 (and maybe DX9) on-board instead of waiting for the CPU to handle it and send it over on the AGP bus? This comes up as I've heard reports of the NV30 having a seperate processor on board to handle T&L, and being able to process DX8 commands on the board itself could account for the supposed 50-100% speed increase Nvidia documents are reporting. I ask, cause both my experience and CPU scaling by Anandtech seems to show that DX8 titles on DX8 cards scale extraordinarily well with faster CPUs (like Jedi Knight 2 and Comanche 4, and maybe even UT 2).
-SammyBoy
Speaking of which, I need suggestions for a new motherboard, preferably a non-VIA-based one. I liked the IWill XP333 I put into a friend's computer, but if possible, I want something... better. Too many VIA 'boards out there, and since I've had bad experiences with multiple VIA boards, I'll stick to the ALis' and SiSs' of the world.
One more thing that I just thought about. Given the fact that the vertex score is so greatly affected by the CPU (or at least the ability of it to process things correctly), is it possible that the NV30 and R900 are incorporating some kind of T&L-like hardware onto their boards to process DX8 (and maybe DX9) on-board instead of waiting for the CPU to handle it and send it over on the AGP bus? This comes up as I've heard reports of the NV30 having a seperate processor on board to handle T&L, and being able to process DX8 commands on the board itself could account for the supposed 50-100% speed increase Nvidia documents are reporting. I ask, cause both my experience and CPU scaling by Anandtech seems to show that DX8 titles on DX8 cards scale extraordinarily well with faster CPUs (like Jedi Knight 2 and Comanche 4, and maybe even UT 2).
-SammyBoy