Dead Elpida chips, what next?

More musings on the Elpida failure saga.

Being a joint venture between big companies can be a plus in terms of finances and strength, but it can also be a minus when the mixed-up corporate bureaucracy has to solve problems. Elpida, as such a result of Japanese "keiretsu" corporate combinations to save national face in the DRAM market, is having its first major technical quality test with the high-end DDR3 memory failures that have been seen now across many module brands using its dies.

I have had my own experience with this, as you can see here on the dead Kingston HyperX module during the last weekend's test:

This problem is not a joke. Since it first showed off these dies at the last Intel Fall IDF in San Francisco last August, Elpida has openly claimed the best DDR3 performance, like DDR3-2500 speed potential with acceptable latencies and Core i7 memory controller voltage levels, compared to what was at that time the only other major Core i7 compatible memory from Qimonda, which did DDR3-1900 CL9 nicely. However, it took Elpida quite a long time to get those DIMMs out, and its first manufacturing partner was an odd memory company called, of all names, Buffalo, not exactly a familiar name to top-end enthusiasts. More prominent names like Kingston and Corsair seemingly got the goodies later.

And yes, the DIMMs really performed! Even DDR3-2000 CL7 was possible with some modules, quite a feat. At least, as long as the DIMMs lasted, which seemingly wasn't all that long for many of them. Geek forums all over have been abuzz on the topic for awhile now.

Aside from waiting for the next process or at least a die redesign, maybe now, together with its Taiwan Memory Company partner, Elpida will have to undertake a serious damage control and repair campaign as its next immediate step, in my mind. Then, that IDF promise of DDR3-2500 at near stock voltage will have to be delivered to safeguard its repaired image. Otherwise, Samsung and Micron, and even Hynix for that matter, will definitely benefit from this episode, and these behemoths have new die revisions, both at higher speeds and lower voltages, all the way down to 1.2 volts.

I have contacted my good friends at Elpida's Tokyo headquarters who are in charge of DDR3 marketing about the problem, too. These are the same fine chaps who so highly promoted these ultrafast memory dies at the aforementioned San Francisco IDF, after all. While up until now we have heard no response from Elpida, we sincerely hope that the company will decide to address this issue soon.

After all, the famous Japanese image for producing high quality products is at stake here.