Steve, that's not my understanding (that Intel's plan has planned to move away from RDRAM for a long time).
My understanding has been that Intel's plan was to keep the P4 RDRAM only, and high-end only for a while; the P3 would be the mid-range system, and would support SDRAM (with the 815). Then as the price of both RDRAM and the P4 itself moved down, the P4 would move into the "mid-range", still with RDRAM. Until July, I never saw ANYTHING indicating publicly that Intel would offer any memory solution for the P4 with any memory other than RDRAM, and, indeed, continued to see Intel support for RDRAM and RDRAM exclusively.
Rambus has two ways to win: RDRAM becomes a major portion, if not the dominant portion of the memory market, or the patents give Rambus royalties on DDR and/or SDRAM in addition to RDRAM.
There were four major factors impacting the market position of RDRAM:
- Intel's undivided backing
- The performance advantages of RDRAM at high clock speeds
- The relative price of RDRAM
- The availability of RDRAM
Intel's position, at the moment, is EXTREMELY fuzzy. Their backing is not as undivided as it was prior to the July announcement. But the dimensions of this change are still VERY unclear.
The performance issue remains to be definitively evaluated. The pundits would tell you that it's been proven that RDRAM is actually inferior to SDRAM based on the tests on Intel's web site. Those test don't prove anything, they only show the relative performance of the 815E vs the 820 (820, NOT the 820E) chipset.
But, that said, they don't prove an RDRAM ADVANTAGE, either. Many of us have faith in a significant RDRAM advantage at high clock speeds, an advantage that increases with clock speed (and we know in which direction clock speeds are heading). However, this has yet to be proven. Indeed, one possibility is that the advantage is real and present, but doesn't matter in most applications because it is masked (to the benefit of slower memory architectures) by the on-chip memory cache.
Price, also, remains an unknown. It has been and is continuing to come down rapidly. Was $1,000 in January, now it's $279 or so and we are told by reliable sources (memory dealers) that they expect $249 by about October 1st. At $600 and up the price was absolutely prohibitive. Now we are in the range that is acceptable to the market IF there is a performance advantage. Consider that towards the high end of the range, people often will allow a CPU price to double, from the $400 range to the $700-$800 range for about a 20% jump in clock speeds (compare the price of a PIII 800 to a PIII 1 GHz). However, the performance benefit has to be there. If the price continues to fall, we may get to the point of price indifference. Since any given chipset supports either RDRAM or some other type of memory, but not both, the ultimate outcome here depends on price vs. the SYSTEM performance of systems with DIFFERENT CHIPSETS. You can't, exactly, isolate the memory performance in a meaningful manner.
Now we have the infineon suit. Remember what happened with Hitachi: Rambus filed the suit ONLY after negotiations broke down, then Hitachi settled (five months later). Apparently neogtiations with Infineon broke down and, bingo, Rambus filed suit. Then negotiations have apparently resumed, so the suit was kept low-key and no press release was made. But the suit itself has apparently not been withdrawn. That's fine, the suit won't go to trial for more than a year, in all likelihood. In the meantime, Infineon will PROBABLY settle, unless one of two things happens: Either Infineon deceides that a stand has to be made, all the way to trial and verdict, and they are going to do it (unlikely, but someone, at some point, may take that position), OR, there is a negotiated industry-wide settlement. I'm not expecting the industry-wide settlement, but it could well happen.
In the meantime, we know that RMBS is negotiating with "double digit" numbers of companies (my guess is 15 to 35). These negotiations all began at approximately the same time frame, and take, apparently, 4-8 months in most cases. So, in the absence of anything else, I'd begin to expect that at some point in the 4th quarter of the year, we'd begin to see a number of license agreements announced in relatively rapid succession. Or, alternatively, a number of lawsuits announced in relatively rapid succession, but if that happens, I'd expect the courts to combine them into a single action.
I have never thought, and still don't think, that there is ANY chance of FTC action against Rambus (an investigation, perhaps, but ending in the conclusion that there is no grounds for action). The Rambus pundits are trying to claim that the situation is similar to that with Dell and VESA, but it's not. First, Rambus never signed an agreement giving up patent rights, as Dell had done. But, more importantly, the patents were applied for 4-6 YEARS before the meetings in question occured. Thus, to claim that the subject of the patents is material that Rambus "stole" from the VESA meetings is, in my view, patently absurd (no pun intended).
However, the change in the Intel support (to whatever unclear extent it has changed), plus the Infineon situation, plus the issue with Cisco and Cypress (I think that's who it was), plus the remaining uncertainty of when the P4 will be announced, why it's REALLY delayed (if it even is delayed), and what the memory roadmap is with Timna, all create even MORE uncertainty than that which we have already been having for quite some time.
By the way, anyone who thinks that the Yahoo message board is working has my permission to cross-post this message on the Yahoo board (please identify me as the author). |
