Hi John Fowler; Re RDRAM and "mobile growth is huge, according to Intel."
Chart showing power consumption for DDR, RDRAM, and SDRAM as a function of bandwidth in a typical notebook:
dramreview.com
DramReview is well known to be friendly with Rambus, (the website, while claiming to be unbiased, is run by an ex Rambus VP: #reply-13405397 and is not yet a member of CMP), so it is particularly significant that they should praise the source of this data as follows:
The data for this analysis comes from a presentation at Platform 2000 by Micron Technology. Micron designs and manufactures SDRAM, DDR, and RDRAM, as well as PCs, so they are in an excellent position to understand the complex analysis required to arrive at an accurate estimate of power consumption.
dramreview.com
These charts make it pretty clear why Intel cancelled its RDRAM notebook program. DDR is lower in power than RDRAM at any given bandwidth. The most important power consumption figure for mobile computing is that at very low bandwidth, as that is how most computers spend most of their execution time. At that bandwidth (far below peak bandwidths), even SDRAM is a lot less power hungry than RDRAM. But DDR is the winner at all bandwidths.
Knowing this, if you had to pump RDRAM, how would one make it look best for mobile? First, leave DDR out of the picture - only compare yourself to SDRAM. Second, only do the comparison at the maximum bandwidth, not typical ones. In short, simplify the issue to the calculation where you look the best, ignore the global calculation.
Naturally, this is exactly what Rambus does in their white paper on the subject. Note the Rambus calculations include only bandwidth per watt, not giving any hint as to what the power consumption is at the typical low bandwidths required by the typical application:
rambus.com
The above analysis should make obvious the reasons behind the news on RDRAM and notebooks:
Direct RDRAM a no-show in mobile PCs?
Sources last week said Intel Corp. has pulled the plug on Greendale, the company's first mobile chipset to support Direct Rambus memory, and through 2000 will instead look to devices that interface with PC100 and PC133 SDRAM.
techweb.com Dec 20, 1999
This is not good news for Intel, as AMD plans on delivering DDR notebooks this year:
Ron Huff, chipset marketing manager at Advanced Micro Devices Inc., Sunnyvale, Calif., said the company will deliver DDR-enabled desktop Athlon processors in the third quarter, and will have DDR-equipped notebook PCs shipping later this year.
techweb.com May 22, 2000
That DDR is great for mobile is well known in the industry:
"DDR will make big inroads in portable products because it operates at 2.5 V, a big power saver over the 3.5-V single-data-rate SDRAMs,"
techweb.com Jan 17, 2000
When Intel followed Rambus into the valley of the shadow of dead dead dead technologies, Intel had a certain amount of a backup plan in desktops, as it still had serviceable SDRAM solutions. But DDR provides a more significant advantage in power consumption to notebooks than it does a performance advantage to desktops. Consequently, a good bit of the damage to Intel from Rambus won't be apparent until DDR notebook solutions hit the market.
-- Carl |
