Testing:
All testing was done on the following system:
• DFI NF4 Ultra-D motherboard
• A64 3200+ Venice core processor
• Thermalright XP120 heatsink with Coolermaster 120mm fan
• ATI Radeon X850XT video card
• Western Digital 36GB Raptor system drive
• 2 x 120GB Seagate drives in NV RAID 1 for data
The Raptor was formatted with a new installation of Windows XP PRO SP2 with all available updates. All drivers are the latest available manufacturer's revisions and the test games and software used the latest patches and versions. Only those programs used for testing were installed. Stability was tested with a single run of SuperPI 32M, and although this may not prove 100% long-term stability, it would at least point me in the right direction for running the benchmarks.
The Basics and Overclocking:
Having spent some overclocking time with my current Samsung TCCD fitted Corsair XMS TWINX1024-4400PT; I had a pretty good idea as to where to start to look for the limit to this new set of XMS TWINX2048.
 
My previous stable best with the TCCD chips was just a touch over 300MHz at a multiplier of 9 with 3,4,4,8 1T timings. I was pretty confident that with the additional GB of RAM, I was not going to be able to reach clear to DDR600 with the UCCC chipped TWINX2048 2GB set.
 
I began by clearing CMOS and loading defaults. The default SPD timings from the 2GB TWINX2048 came up as 3,4,4,10 at 2.6Vdimm. I changed the processor frequency from the stock 200MHz to 250 MHZ and dropped the multiplier from 10 to 9 for a speed of 2250MHz. SuperPi ran through the 32M iterations of prime without issue. Then it was time to jump up the frequency a bit and see if I could get more. Next step was 260MHz for a total of 2340MHz. Again SuperPI completed the 32M without issue. Figuring the luck has been good so far; I went another big step up to 270MHz to give 22430MHz. Windows booted and everything seemed stable enough but I got the dreaded "not convergent in prime" error from SuperPI. So although windows ran smoothly, I was at the beginning of having stability issues. I then dropped the multiplier to 7 to make sure I wasn't topping out the processor. No luck, SuperPI refused to finish again. I increased the Vdimm from 2.6V to 2.7V and found no relief from the "not convergent in prime" error. Further playing with Vdimm voltages, HTT, and processor Vcore produced little help to get stable at 270MHz; there was not much room to loosen the timings up much either. But please remember, many things will influence the overclock ability, my Venice core processor has never been able to impress me with outstanding clocks. There is also the big issue between the keyboard and the chair, which alone may be the limit of the overclocking.
To make sure I had a strong resemblance to stability for the following tests, I dropped the frequency back down to 265MHz and jumped the multiplier back to 9 for a final testing frequency of 2395MHz, 2.7Vdimm, an HTT of 4X, and timings of 3,4,4,8 2T.
|
|
