Ali, Re: " Reading comprehension problem, RAS-expert?
The guy said:
"We tried two 256M Hyundai DIMMs that it didn't like
as all where as two 256 M Hitachi DIMMs worked fine."
Yes, the YUK article said that, but it also said this:
If you use two RIMMs and they are from different manufacturers,
you can expect problems, according to very reliable sources.
I was keying off the mixing vendors statement. For anything that has to run as fast as the RIMM - chipset interface, I would put the no mixing of RIMM vendors "rule" into effect. Different vendors are bound to have subtle design and process timing differences, which added on top of lot to lot differences, even if in spec., could cause failures. Again, these beasties, as YUK calls them, are very fast and undoubtedly leave little room for timing variations. maybe they are pushing the envelope too much, I don't know. So, why can't someone assembling a mobo afford to just put the same Mfr's RIMMs together...in other words, NO MIX. Are they that hard to buy in pairs?
There is also this statement from YUK:
Intel engineers vehemently claim
that there is something wrong with most manufacturers' memory --
they told me 'most of them are more interested in making a buck
than in making PC-100 memory according to specs'".
So who do we believe, the YUK, or Intel? I certainly can't tell who's right.
From an investing point of view, though, looks to me like the Rambus 400 MHz (double pumped to get 800 Mbit/sec.) might be just too much over the edge right now. Too many possibilities for problems with all the too-tight tolerances among Intel-RIMM Mfrs. -Rambus. Sounds like every system might have to be crafted/tuned with a high speed scope and a logic analyzer. I'll certainly leave Rambus investing to those with more tolerance to roller coasters than I have. (RMBS +5 1/4, INTC +5/16), RTQs.
Tony |
