click to enlarge
The first step was to check out how this TwinMOS PC 2700 is programmed. The SPD reading on the TwinMOS PC2700 memory as shown by Sandra Pro declares this memory indeed as PC2700, running up to 133MHz in CL 2.0 and up to 167 MHz in CL 2.5. Very well!
The next step was to quantify the bandwidth the TwinMOS PC2700 is delivering at various steps starting at 133 MHz and hopefully up to 166 MHz. . We can see that PC 2700 running at the 8KHA+ is delivering excellent bandwidth. But we can also see that TwinMOS PC2700 is more or less delivering the same data throughput as PC 2100. Disappointing was that we could take the TwinMOS PC2700, that is supposed to run at 166 MHz, just a few MHz higher than the TwinMOS Nanya PC2100. Instead of 156MHz with PC2100, we reached 158MHz with PC2700. It might have been possible to squeeze out 1 or 2 more MHz by reducing the memory timings to less aggressive settings or increasing the DIMM voltage. We did not bother do so, because it was clear, that this TwinMOS PC2700 memory we tested is not the breakthrough for bandwidth hungry AMD 760 or KT266A chipset based main-board owners. We have to stress again that we did not test this memory with a mainboard/chipset that is specified to run at 166MHz. It might be even possible to run this PC2700 at 166 MHz at other KT266A boards, but at least we can say, that this is far from guaranteed and that the bandwidth output is almost identical with PC2100.
The following 2 Sandra screenshots show the bandwidth with TwinMOS PC 2700 running at 134 MHz (actually 133.5 MHz) and 148 MHz.
 |
 |
|
|
|
The TwinMOS PC2700 memory with the Hynix module HY5DU28822AT we tested looks rather mysterious to us: (1) First of all: although rumored, discussed, and expected, a JEDEC PC333 / PC2700 standard seems not to exist yet, at least it is not published. The JEDEC website / data base does not mentioning anything about PC 2700/PC333. JEDEC is a private consortium, and it's of course their decision whether and when to publish their standards. That a PC2700 standard is not published at this time, does not necessarily mean it doesn't exist. (2) Secondly, and more confusing: the specification sheet for the Hynix module HY5DU28822AT, which Hynix and TwinMos are marketing as PC2700/PC333 does not contain any reference about 166 MHz! The specification sheet for this memory module describes the HY5DU28822AT module as running at MHz 100, 125, and 133; but does not mention anything about 166 MHz. (3) Now, if we combine our bandwidth and overclocking results as shown above with (1) and (2) we could draw a possible conclusion: that the TwinMOS PC2700 DDR with the Hynix HY5DU28822AT module is PC2100, which might have been handpicked for its capability to run at 166 MHz (which it did not do in our case) or whatever reason. On the other hand: the tested TwinMOS PC 2700 is not validated for the 8KHA+ (or any other KT266A board) and speculations like the above ones are therefore more or less pointless.
The more important point and at the same time our conclusion here is the following: It looks like that the TwinMOS PC2700 memory we tested behaves and yields like PC2100 memory when running at a KT266A board like the 8KHA+. It also does not overclock better than PC2100. The PC2700 bandwidth output was excellent, but not better than with PC 2100. Given a price twice as high as CL 2.0 rated PC 2100, we can not recommend PC2700 for the purpose of overclocking mainboards specified to run up to 266 MHz.
Thank you very much to all OCAU members who contributed to the "fact finding" about this memory in this thread.
 Gainward GF3 Golden Sample |
 Hardcore Storage! |
 Rheobus Kit |