An article on the “new” practice of low-tier manufacturers selling untested memory got attention on Slashdot this week.
This isn’t a new practice. I’ve known about it for about eight years.There’s a pretty good reason why all name-brand memory is priced pretty much the same. You can occasionally catch a break in pricing, but on average, a Kingston module is going to cost about the same as a Crucial module, and so will any other top-tier brand. Memory from a computer manufacturer like HP or Sun may cost a bit more still, ostensibly because the manufacturer tests for compatibility. They may or may not actually test the module you buy, but at least they’ll guarantee it not only works but works in the machine you put it in.
If you’re building your own PC, by all means buy Crucial or Kingston memory or go to a specialty high-performance memory like Mushkin. The same holds true for upgrading a name-brand PC. But pay the extra money for server memory from the company who made your server. An hour of downtime will obliterate the $100 you might save.
But there’s another tier of memory. I first became aware of it back in the days when a typical issue of Computer Shopper was as thick as the Greater St. Louis White Pages. Tucked away in the back, there was always someone who beat the typical memory prices and he usually beat it by a long shot–at least 30%. For several years, that was how I bought my memory, and for a long time I got away with it.
Then along came Slot 1 and Super 7. Once CPU rates broke the 233 MHz barrier, the systems became a whole lot harder on their memory. I don’t know what was special about 233 MHz, but that cheap commodity memory just didn’t cut it anymore. Suddenly, I started noticing that commodity memory often didn’t pass the rudimentary memory test that computers perform before they load the operating system. That’s akin to flunking grade-school recess, so I started looking into it.
What I found was that commodity memory generally isn’t tested, or it’s tested very loosely. What’s worse yet is that the chips on some commodity memory were tested, and failed. They were certified for use in things like pagers and other consumer devices, but not up to the higher demands of computers.
So, having known this for about 8 years, you can imagine what I thought when I read the headline “Why untested DRAMs are getting into more and more products.” I was thinking hey, an upgrade! Since it didn’t test bad, at least there’s a chance it’ll work!
Maybe this practice has evolved in the past few months, as the author of the article in question alleges. But it’s hardly a new trick. In the highly competitive no-name clone market, this has been going on since at least the days of the 486. What was going on in the days of the 386 is even scarier.
Will Dell and the boys follow suit, like the author fears? I doubt it. PCs are problematic enough as it is, and it only takes a few months to lose a reputation that was built over the course of a decade. Shipping commodity memory isn’t like outsourcing technical support to India–there’s a fair percentage of your customer base who will never use your tech support. All of your customers will use your memory.
I can’t imagine commodity memory ending up in any name-brand PC, unless it’s a name brand whose ship is sinking fast.
But I guess I shouldn’t be surprised that this old trick is showing back up again. The business is competitive, PC sales are down, the economy isn’t what it was 10 years ago, and profit margins are impossibly thin. If todays untested and/or defective memory is better than 1997’s, someone’s going to use it.
But part of the story never changes: Always buy your memory from a reputable manufacturer and distributor, so you know what you’re getting and whence it came. You’ll save a lot of frustration over the life of the PC that receives the upgrade.