To: John Stichnoth who wrote (27339 ) 8/20/1999 3:04:00 PM From: Dave B Read Replies (1) | Respond to of 93625
John,First, I believe it has been agreed that Tom's Hardware Guide is an interesting, but unreliable, source. Tom's attitude is to question Intel; to be their sceptic. That's fine, but recognize his point of view and take his results with a grain of salt. Tom's value is to help raise questions, not to provide answers. I think it's a broader question with respect to Tom's hardware. As Ten has pointed out, Tom tends to run his tests with overclocked systems. Not only is that something that the average user can't do, but shouldn't do as well. While many systems may run fine, it can cause serious problems with some. How responsible is it to report results based on these types of systems? His tests should not be not taken with a grain of salt, they should not be considered at all, IMO. We should wait until an independent entity (e.g. one of the PC mags) tests production RDRAM in production systems before debating the performance (which means I'm willing to agree that the tests by the DRAM manufacturers and systems vendors who are trying to sell RDRAM may be skewed as well). After that, we can toss all the tests into the ocean, because I think the next two paragraphs in your post were dead-on, and are the key to the long-term success of RDRAM:Second, The question is not necessarily whether Rambus provides an edge on a current generation chip, whether pre-production or production. It is what box-makers will be putting into their machines in this and following generations. With the major box-makers all lining up to produce their newest and best machines incorporating rambus, we have our answer. Ultimately, the performance edge will be important, but that will likely occur as we move to faster and faster CPU's--from 600MHz to 800MHz and beyond. And it is in these faster models, appearing next year, that we will see the real performance edge. And while I'm not contesting either of your assertions regarding September chipset results, I haven't heard any contradiction to the assertion that RDRAM provides the better migration path. Dave