Well... I'm not an Intel engineer but I work in this field so I understand their position.
Basically, marketing and business pple like to say to get the job right the first time. Most senior engineers will also agree with this point, but it is easier said than done!
Every new technology has its own quirks and perks. Many engineers will hope that the perks will cover the quirks but that's usually not the case. Coupled with the fact that the business side wants the product YESTERDAY, engineers are faced with the dilemma of getting it right or getting it out of the door. Unfortunately, most choose the latter or the company can go bust.
I'm a research engineer doing low-level firmware protocol stack stuff, not unlike those Intel comrades doing micro-code (but their's are 2 order of magnitude lower than mine) and it's very hard to cover everything. That's why we have revisions upon revisions, it's a necessary evil but it ensures that end-users get to see results, else, the company can go down even before the end-users get to even sniff the thing.
The first version is always one that has some known issues that will be solved when given time, but THEY do solved. The trick is balancing the number of known issues that are shipped to the amount of time you can get the issues fixed.
So bear with them. AMD's engineers TOO took a long time to get the Athlon out and the first issue had tons of problems as well, but just look at the TBird now!
So I say that the P4 is definitely NOT crap and RDRAM does have its place. Actually part of the latency problem is caused by Intel's mobo. If there's a controller per RIMM instead of 2 RIMM to 1 controller now, the latency will be much lower. This will also solve the problem of the latency increasing as you add more RIMMs.
------------------
Dual Celeron 300A oc 450, Gigabyte GA-6BXD, ATI Radeon 64MB VIVO 166 oc 180, 256MB SDRAM, Abit HotRod66, IBM DS 34GXP 20.1GB, Pioneer 10x Slot-in DVD, Yamaha 8424 CD-RW, Yamaha YMF744 XG sound, W2K SP1, Mitsubishi Diamond Pro 900U.