Hi cordob; More re the TMF poster child:
Re: "Intel is not stupid. They are supporting SDRAM now because of the price point, and they will support any kind of DDR that is stable and if it makes sense, but DDR seems to have completely dropped off their "official" memory roadmap." It's hard to understand this statement in the face of Intel's support for PC2100 in 3Q02.
Re: "So far, DDR has proven to be unstable and unreliable. Intel is not, however, going to spend a lot of money improving the DDR standard so that it works." DDR is doing fine. Heck, even Samsung says they're putting 10~15% of their production currently into DDR and that's in early July long after Micron forced the price down.
Re: "Let me remind everyone, DDR is not a better technology than RDRAM. DDR cannot be made into a better technology than RDRAM. DDR is an attempt to push an old technology one step further before letting it die a natural death. It's old, and decrepid, and the marketers are pushing it like it's some young olympian who just won his fifth gold medal." DDR put Nvidia on the map. VIA stole most of the market share in chipsets that Intel lost (with PC133 support) and now VIA is fully supporting DDR. AMD is ignoring RDRAM. If RDRAM is so great, how come the industry (other than Intel and Sony) is ignoring it? Where are the ALi RDRAM chipsets, they signed with Rambus? Where are the ServerWorks, SiS, Sun, Cisco, IBM chipsets? Is it really the case that only Intel can see the value of RDRAM? And if Intel does see it, then why is Intel supporting DDR?
Re: "Take a look at the number of systems that have DDR. Take a look at the number of systems that have RDRAM." There are far more systems containing DDR than contain RDRAM. The primary reason that the DDR systems aren't obvious is that there isn't much advertising for them. When Cisco used DDR in their systems they don't tell anyone. Rambus advertised the heck out of every design win they got, and for good reason. They needed to sell stock to victims like EvilDoctor. TeamDDR isn't selling stock, so their advertising (actually the advertising of the DDR supporting memory makers) went to publications read by design engineers. TeamDDR's advertising was far more effective at promoting DDR use, while Rambus' advertising was far more effective at selling shares to punters.
Re: "Take a look at the number of systems that contain RDRAM that have been recalled (answer: none)." Actually, RDRAM was supposed to have 4 RIMMs per RSL channel, but it was reduced to 3 RIMMs when 4 proved impossible. Then, after Timna had already shipped samples all over they found a signal integrity issue that forced Intel to reduce the maximum number of RIMMs to two. The RIMM limit is still two, and this goes a certain way to explaining why it is that Intel is working so hard on DDR.
Re: "No, Intel is not working on a 400MHz version of DDR so they screw Rambus and get back in the good graces of AMD. Intel is not looking to make a technology that really helps out Micron because they like them so much, and to hell with all this stuff about "performance". Intel is not looking to give a helping hand to AMD or Via, or any of its other competitors, who happen to still be stuck in DDR land without knowing how to get out." Actually, what Intel is trying to do is win back their share of the chipset market. Their customers asked for DDR, and Intel is giving it to them. DDR2 is to start with 400MHz (800MHz data rate) and that is a natural for Intel's P4, so naturally Intel wants it. That's why Samsung, Toshiba, IBM, Infineon and God knows who else already have working samples of DDR2. Big companies like Intel have to determine what their memory usage (in terms of what type) long before the products come out the door. That's why it's taken so long for Intel to get a DDR chipset, and that's why Intel is now working on the chipsets they will need in 2003 when DDR2 shows up.
Re: "When you hear this kind of thing, first consider the source (none mentioned)." It hardly takes a genius to figure out that Intel is working on 400MHz DDR. Everybody else in the industry is, at least for long term designs that are just now starting out. VIA has already publicly pledged their support. Intel is in a different situation, they have to keep selling RDRAM, so they characteristically keep quiet about DDR.
Re: "Second, use common sense - is this something that Intel would consider doing, and what benefit would they gain?" Use common sense: why would Intel design DDR products if they didn't want to ship them?
Re: "What might it cost them to go this route? Does it get them any competitive advantage? Is this something that would help or hinder their competitors?" All their competitors already went the DDR route. What Intel is doing now is trying to catch up.
Re: "i845, SDRAM, makes sense right now. It allows Intel to target a price-point that is just a bit lower than they could otherwise, allowing them to phase out P3 a little earlier." Yeah, and i845 DDR SDRAM makes sense in early 2002, allowing Intel to target a price-point a bit lower than RDRAM.
Re: "Keeping P3 around aids AMD, since their processors run faster on P3 optimized code than a P3 does. A year from now (if not sooner), it will be phased out because it will suck." Intel's transition to the P4 isn't being done to help RDRAM, it was delayed until now because of the lack of availability of memory options other than RDRAM. Wall Street's been saying this a lot recently, and it makes sense.
Re: "DDR will also continue to suck, as it has always sucked and as it sucks now, and as it will always continue to suck." If DDR "sucked", then why is Intel supporting it? And why is the whole industry ignoring RDRAM? And why are the AMD DDR Athlons so famous for high performance? That DDR "sucks" is just a broken Rambus long dream.
Re: "Even if someone figures out a way to make DDR not suck, what would
Intel gain by doing this? Would it help AMD? Would it hurt AMD to make them think that Intel was going to support DDR, when Intel has no intention of doing that?" Most of the gain that AMD and VIA have made against Intel over the last year have been due to Intel's missteps. The primary misstep was supporting Rambus. Now that Intel is getting the Rambus out from around their neck, maybe they'll quit losing market share to VIA and AMD.
-- Carl |
