PDA

View Full Version : KHX3200K2/512 under heatspreader [pics]



eva2000
03-23-2004, 01:28 PM
Well i have a few KHX3200K2/512 kits, and i decided to take the heatspreader off one stick to see what's under it - i chose the poorest performing stick I tested so far to take the heatspreader off just in case i damage the ram i won't be ruining an awesome oc'er. :)

I only managed to get 1/2 the heatspreader off, the backside of the ram where there is no ram chips (single sided) is too hard to get off :( Any tips ?

Anyway underneath the heatspreader is indeed BH-5 but 4 chips are from week 40 and 4 chips are from week 38 ??? Aida32 3.93 reports the manufacture date in SPD as week 49, 2003 IIRC

Would chips from different weeks mess up oc'ing ?

http://www.fileshosts.com/memory/kingston/khx3200/256/BH5/1/BH5_005_thumb.jpg (http://www.fileshosts.com/memory/kingston/khx3200/256/BH5/1/BH5_005.html)

i noticed that the 4x week 38 chips are alot hotter than the 4x week 40 chips on the stick with no heatspreader

week 40 chips = 35C
week 38 chips = 40C

using infra-read thermometer

AndyT
03-23-2004, 11:32 PM
I suspect that the week has less to do with the issue than the fact that they were just using whtever was on hand for the modules. If they were trying to make some good overclocking modules, then certainly a module made with chips from a "good week" and proper testing of the module for the higher speeds would have better results. I would venture a guess and say that Kingston does less chip selection and more testing and module binning. They have quite the setup for module testing there and make truck loads of the stuff.

eva2000
03-24-2004, 03:16 AM
seems this particular stick doesn't respond as well to higher voltages i.e. 3.18v real vdimm. But lower voltages seem fine.

so far memtest86+ v1.11 test #5 looping:

217mhz 2-2-2-6, NB Strap = 800FSB - GAT = TAADD with 2.86v real vdimm = 2035 passes error free 11hrs 05mins 50s

225mhz 2-2-2-6, NB Strap = 800FSB - GAT = TAADD with 2.98v real vdimm = 125 passes before seeing 10 errors


.

eva2000
03-25-2004, 05:16 PM
well seems this stick doesn't like vdimm over 2.9v real but at lower vdimm is stable on AI7

Now it's rock stable on P4C800-E at 2.85v vdimm for testing

217mhz 2-2-2-6
burst length = 8
64usec
PAM = Enabled
PM = Standard

been memtest86+ v1.1 test#5 loop error free for 3,142 passes for past 17hrs 18mins