sorry for not being able to do some testing for my TX-DDR2, I was at my friend's house(ddww) the whole night testing his Hynix (OEM) C4-based 2x1GB modules (they aren't dual channel kits). The RAM was rated at DDR533 CL4-4-4-12.

As was expected... it ran all the way up to DDR1060 with CL5-5-5-12 without any issue (tested with SuperPi 16M/32M) on Intel P4-630 and Gigabyte DS3 with 2.2Vdimm. We used 1:2 divider to push the RAM to the max.

Here are some screenies (sorry....did not capture 32M)


SuperPi 16M @ DDR2-1060 5-5-4-12 (2.2Vdimm)


*Please note that these are OEM Hynix DDR533 1GB modules, not Kingston/Corsair/Apacer or whatsoever.

Unless they're old stocks, most OEM Micron nowadays are using the smaller rectangular BGA which are actually less overclockable compare to the classic 'fatbody' Micron or the newer square BGA Micron. The good square ones are supposed to be 'premium' chips....so they're the least expected on OEM DDR533/667 Micron. Those Micron DDR2 on Lelong.com.my are just the normal ones.

This post has been edited by soulfly: Aug 16 2006, 01:37 AM

my mobo cannot use ddr800 divider.... so this is what i can do for now....

mobo : Tforce550
ram : 2x512MB TeamXtreem DDR2-667 CL3
divider used : ddr667
vdimm : 2.41v (vdimm pencil mod, measured with dmm)

DDR2 benchmark challenge added to overclocking database.

Link : http://forum.lowyat.net/index.php?showtopic=313858

nah... the benchmark thread is strictly database, discussions will be done here

easier to manage

wrong....

unless it's LL or XBL series

not sure about the signature line.... but probably some 'mainstream' chips like normal Elpida, Nanya or something...

Try get good Kingston or Corsair la... at least you can see what chips they're using (not covered by heatspreader). Or if possible, look for Hynix C4 (oem, DDR533)... should be cheap enough. They clock great as well.

This post has been edited by soulfly: Aug 17 2006, 11:14 PM

Nothing better to do....

Just now I did an experiment on how Tras value affect memory bandwidth in AM2 K8 system. Surprisingly.... lower Tras doesn't mean better. The following values were taken from Memtest86+ bandwidth display.

AM2 memory divider : DDR667
CL-Trcd-Trp : 4-4-3
Tras : 5-18 (options accessible from Biostar TForce550 BIOS)



*Will do more updates and testing with other benchmark like SuperPi etc...

This post has been edited by soulfly: Aug 18 2006, 04:24 PM

Sorry but D9DQT can only be found in very old modules with RED heatspreaders. Newer ones had been replaced with either Elpida or the newer Micron (most of the time is Elpida).... or could be anything.

This post has been edited by soulfly: Aug 19 2006, 09:36 AM

newer Micron should be quite good. the advantages of the fatbody is that they can run on 2ns on Trcd and Trp... but at such tight timing they wont go very high as the new D9s do.

I think Hynix DDR533 should be C4. As far as I know.... many Hynix DDR2-533 now is C4.

I think not only Nanya. I would say the chips are random. Probably normal Elpida as well.

Patriot LL .... PC2-5300 is high probability that it's Elpida... little chance it might be Micron (maybe for 2x1GB sticks)

for PC6400LL.... i think it should be Micron, at least 95% chance for it.

Micron overclocks better.

I'm not sure whether 512MB modules using D9GKX or not.... but the 1GB modules does.... and it explains the 'premium price' eventhough it's only KVR (imagine Czone selling it at RM600 per 1GB stick)

team xtreem running ddr800 3-3-3-8 @ 2.4V (2.35v actual reading)

screenie coming later, now in class

1GB and 512MB ... that's a lot of difference.

it's a single piece 1GB module isn't it? it should show up as 1024MB

how about CPU-Z? got screenshot to show us?

LOL....

I guess maybe there's something wrong with the SPD programming....

but the interesting thing is that... the SPD timing is very similar to Hynix DDR2-533 CL4 modules

is there by any chance to test those to run at DDR1000 CL5-5-5-1 @ 2.2Vdimm?

This post has been edited by soulfly: Aug 24 2006, 11:16 PM

they manufacture memory modules... but sourcing the chips would probably from some random chips manufacturer

depends on the size and the precise alignment of the dot

XS forum has a lot of pics of Micron

btw.... the small rectangular Micron doesn't really do wonder.

It's the square ones which we want... D9GMH and D9GKX