Hi cordob; Like ptNewell, EvilDoctor is one of those easy to find people who knows less than he thinks about electronics. Oh well, let's take a look at his comments...
Re: "Intel does not want DDR to succeed. Intel has no vested interest in seeing DDR succeed." This is true, and it's true of Samsung as well. That's because those two companies were early into DRDRAM and consequently have a lead on the competition. But both Samsung and Intel have seen the writing on the wall and have defected to TeamDDR.
Re: "The people who want DDR to succeed are the ones who can't make RDRAM successfully (cost effectively), because they didn't want to pay for the learning curve." This is kind of silly. The cost of the learning curve is a part of every technology. If Rambus is so hard to "learn", then that is a defect. One of the features of DDR is that it is easy to "learn".
Re: "RDRAM has a performance advantage over DDR." This is simply not true. If RDRAM was the magic bullet that solved all engineering problems, everyone would be using it. Instead, there are far more design wins for DDR than there are for RDRAM.
Re: "Intel has spent billions of dollars to
develop technologies that best use RDRAM." True. But DDR is cheaper, and engineering is all about bang for the buck. Intel made a mistake, now they're getting over it.
Re: "AMD made a big bet on DDR. So did Micron and Via. It didn't work out." AMD Athlon DDR computers are being sold all over the place. AMD has a small share of the market. DDR became cheap, relative to SDRAM, (just like TeamDDR said it would) while RDRAM remains expensive, relative to SDRAM, (exactly the opposite of what Rambus said would happen). When DDR boards for the P4 are available, they'll likely be successful.
Re: "Why in the world would Intel step in and bail them out at this time? Why would Intel stake their future on a technology that has an inferior design?" Intel isn't stepping in to "bail them out". By going to DDR, Intel is rescuing their own failing chipset business. DDR is not an inferior design - it's been chosen over RDRAM by companies and universities all over the world. In fact, the companies most familiar with earlier versions of RDRAM (i.e. concurrent and base) dumped Rambus for later products.
Re: "Oh, yeah, DDR400 will not happen. Not in PC main memory. The technology does not scale to 400MHz, no matter how fast you make the silicon." Samsung is already sampling a 600MHz x32 DDR chip. Just one of these chips provides 2.4GBytes/sec, which is considerably more than any RDRAM chip being sampled. The writer is (1) assuming that PC main memory will always be user serviceable, and / or (2) assuming that he understands enough about electronics to predict what is and is not going to be possible. Note that PTNewell, a bonafide professor of Physics, said that PC2100 wouldn't work.
Re: "Intel knows what is going to happen to DRAM prices within the year. The margins will creep back up to 0%, and RDRAM will cost a little more than SDRAM." If Intel knew all this, and RDRAM really were such a cheap solution, Intel wouldn't be sending out a DDR solution for the P4. Quite simply, the actions of Intel do not indicate that Intel believes that RDRAM will cost "a little more than SDRAM." Predictions that RDRAM is soon to be at such a price have been continuously made by the Rambus camp since 1997, but RDRAM is still way more expensive than SDRAM. And since DDR is as cheap as SDRAM (from Micron), and since DDR provides higher performance than RDRAM (lower latency and higher bandwidth), and DDR is considerably cheaper to build controllers for (no royalties, more suppliers) RDRAM needs to get cheaper than DDR before it becomes attractive.
Re: "SDRAM will cost more than it does now - at some point these guys have to make money somewhere. At that time, DDR won't be any kind of bargain." It's not necessary to be able to predict whether SDRAM will be cheaper or more expensive a year from now. All that's necessary is to be able to predict that DDR will be at roughly the same price as SDRAM, and both of them are going to be cheaper than RDRAM. This is what the memory makers (even Samsung) have been telling us for several years, and so far, they've been spot on.
Re: "DDR requires a 6 layer board and more pins to achieve the same bandwidth, and it's still slower than RDRAM even then." This is simply not true. DDR was specifically designed to be implemented in cheap motherboards. By contrast, RDRAM requires much tighter tolerances. As far as comparing speeds, it is very difficult to compare different memory speeds because there are so many other things involved. But on the face of it, as far as performance goes, the DDR based Athlons are kicking the butts of the RDRAM based P4s.
Re: "It also has higher power consumption," This is not true. DDR is being designed into chipsets designed for mobile applications all over the place, while RDRAM is not.
Re: "... costs as much to make as RDRAM (yes, it's true - compared to 4-bank RDRAM)," This is not true. Samsung suggests that when they go to the 4i RDRAM version (which won't be out for a long time anyway), the parts will only be 5% more expensive to manufacture than DDR, but that is not equal in price. In addition, far more important for pricing is the fact that DDR is available from many more companies than RDRAM. When Samsung has 4i RDRAM available they'll be the only company in the world that has it. Are they going to sell it as cheap as DDR? Not a chance, LOL! Instead, Samsung will do what any company that has a unique technology does, they will fatten their profit margins accordingly.
Re: "and those signal-integrity problems are not going to magically clear themselves up as you go higher in speed." There have been no signal integrity problems with DDR. The overclockers take the stuff way beyond its rated limits, do you see anyone doing that to PC800 RDRAM? There were some signal integrity problems with the early AMD 761 chipset, but they were with the 266MHz FSB, not the 266MHz DDR bus.
More later...
-- Carl |
