> Rambus uses less connections/chip, so it can be packaged smaller...Fewer pins means fewer traces and that makes boards cost less.
Almost, but not quite. The chip has fewer pins, but RamBust's advantage is that the bus width is smaller (16 bit) than the 64 bits used on an Intel processor. The _total system_ pin count is reduced. So one can use fewer chips. For example, if one has a 1Mx16 chip, one needs 4 of these to make a x64 bus. RamBust uses a 16 bit bus, so one can use one chip and still make the memory work.
OTOH, RamBust requires more die to accommodate the interface to what is basically a regular DRAM core. This will always make an RDRAM chip more expensive than the same sized SDRAM chip.
> It's not that it's bad technology, just misapplied to PC's when supply was not availible,and managed by a company with a overzealous legal department.
RamBust requires tightly controlled PC board impedances, routing is critical, and memory sockets mess up the signal quality. It's a very flaky design. Don't even think about mixing different chip vendors together. The best RamBust application is one that needs 2 memory chips, and the chips are soldered down directly on the board with extremely short traces to the memory controller. Anything else works poorly. If you don't believe me, check out the troubles introducing the Intel i820/i840 chipset.
Sure, it's easy to get a job now. Some of the best engineers I ever worked with didn't have degrees. Everything is great when jobs are easy to get. But times change. When engineers are being laid off, they know they are stuck in their jobs. They can't go anywhere else. They know no other company would even look at their resumes without the paper degree. You can see the gloom and despair in their eyes when the next round of layoffs come, knowing the future is dark and bleak if they lose their jobs.
If you want to retire in 5 years, don't go to college. If you plan on working the next 40 (ugh), get the degree NOW.
The "problem" is that Rambust owns patents on SDRAM. They were part of the JEDEC committee, while at the same time patenting the core technologies being finalized. If this isn't conflict of interest, I don't know how else to define it.
The biggest problem with Rambust is not that it is expensive, but that it doesn't work well. The technology is immature and flaky. Don't even think about mixing vendors on the same board. Also, it is not only the PC-800 that has problems. It is not like SDRAM, whereby faster chips run better in slower systems. There are problems across the board. It's just that Rambus isn't so popular that these problems have not come out.
If you buy a Rambus-based system, buy one from a first tier vendor, or you'll be sorry.
This Intel vs. AMD 1GHz race is really marketing positioning, or "mine is bigger than yours". Both will allegedly announce 1GHz chips this month, within days of each other (AMD first?), but they will not be available until 3Q. Intel has problems shipping 733 and 750 PIII's, and the latest available _engineering samples_ are "only" 866 MHz. And they were weeked to work. The 933's are not available even as engineering samples. NO ONE has a 1GHz PIII, not Dell, not anyone else. Oh, sure, they are coming, but don't have someone charge your credit card this month for one.;-)
Slashdot Mirror
> Rambus uses less connections/chip, so it can be packaged smaller...Fewer pins means fewer traces and that makes boards cost less.
Almost, but not quite. The chip has fewer pins, but RamBust's advantage is that the bus width is smaller (16 bit) than the 64 bits used on an Intel processor. The _total system_ pin count is reduced. So one can use fewer chips. For example, if one has a 1Mx16 chip, one needs 4 of these to make a x64 bus. RamBust uses a 16 bit bus, so one can use one chip and still make the memory work.
OTOH, RamBust requires more die to accommodate the interface to what is basically a regular DRAM core. This will always make an RDRAM chip more expensive than the same sized SDRAM chip.
> It's not that it's bad technology, just misapplied to PC's when supply was not availible,and managed by a company with a overzealous legal department.
RamBust requires tightly controlled PC board impedances, routing is critical, and memory sockets mess up the signal quality. It's a very flaky design. Don't even think about mixing different chip vendors together. The best RamBust application is one that needs 2 memory chips, and the chips are soldered down directly on the board with extremely short traces to the memory controller. Anything else works poorly. If you don't believe me, check out the troubles introducing the Intel i820/i840 chipset.
Sure, it's easy to get a job now. Some of the best engineers I ever worked with didn't have degrees. Everything is great when jobs are easy to get. But times change. When engineers are being laid off, they know they are stuck in their jobs. They can't go anywhere else. They know no other company would even look at their resumes without the paper degree. You can see the gloom and despair in their eyes when the next round of layoffs come, knowing the future is dark and bleak if they lose their jobs. If you want to retire in 5 years, don't go to college. If you plan on working the next 40 (ugh), get the degree NOW.
The "problem" is that Rambust owns patents on SDRAM. They were part of the JEDEC committee, while at the same time patenting the core technologies being finalized. If this isn't conflict of interest, I don't know how else to define it. The biggest problem with Rambust is not that it is expensive, but that it doesn't work well. The technology is immature and flaky. Don't even think about mixing vendors on the same board. Also, it is not only the PC-800 that has problems. It is not like SDRAM, whereby faster chips run better in slower systems. There are problems across the board. It's just that Rambus isn't so popular that these problems have not come out. If you buy a Rambus-based system, buy one from a first tier vendor, or you'll be sorry.
This Intel vs. AMD 1GHz race is really marketing positioning, or "mine is bigger than yours". Both will allegedly announce 1GHz chips this month, within days of each other (AMD first?), but they will not be available until 3Q. Intel has problems shipping 733 and 750 PIII's, and the latest available _engineering samples_ are "only" 866 MHz. And they were weeked to work. The 933's are not available even as engineering samples. NO ONE has a 1GHz PIII, not Dell, not anyone else. Oh, sure, they are coming, but don't have someone charge your credit card this month for one. ;-)