Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Rambus (RMBS) - Eagle or Penguin -- Ignore unavailable to you. Want to Upgrade?

To: gnuman who wrote (72263)	5/8/2001 4:49:36 PM
From: Bilow	Read Replies (2) \| Respond to of 93625

Hi Gene Parrott; I thought I'd illustrate some facts (not good or bad, just facts) about the RSL bus by picking apart the latest drivel from ptnewell. (Rambus LTBH types are now getting back to the technology, patents are too painful to discuss. I like that for several reasons.) First, the post: Power dissipation ptnewell, Yahoo RMBS thread #272518, May 8, 2001 The classic simple rule for power dissipation in the high frequency limit is P ~ C(Vmax-Vmin)2 x f2 where C=capacitance, Vmax-Vmin is the voltage swing, and f is the frequency. Since 800 MHz RDRAM has 5 time the frequency of PC133, it should use 55=25 times the power right? Not so. The voltage swing in SDRAM is 3.5V. The voltage swing in RDRAM is 0.8 volts, taken in two steps of 0.4 V. The the delta V term is only about 1/9 the size of PC133. After squaring, it is about 1/81 the size. This more than compensates for the 25-fold frequency term (25/81 <1). Thus on this naive basis, PC800 RDRAM should use less power than PC133. Of course other factors are at work. For one thing, in PC133, multiple units are active at once, while in RDRAM it might be just one unit active. Also, RDRAM comes with multiple power saving modes, which, if used, (at some cost to initial latency) can greatly reduce power whenever the RDRAM is not accessed. Intel seems to believe that an RDRAM based mobile PC would save considerable power over the PC133 version. messages.yahoo.com First of all, his formula for power dissipation is broken. A corrected formula is as follows: P = C * (Vmax - Vmin) * Vsupply * Freq There is no square in the frequency (at least for any particular transistor), and the squared voltage term is actually split into two parts. The first part, (Vmax - Vmin) accounts for the change in voltage across the capacitor. The second part, Vsupply, is larger than (Vmax - Vmin), and the difference is that my equation gives total power dissipation, including the dissipation in the current sources &c. With CMOS, (Vmax - Vmin) = Vsupply, so his formula is okay there, but this isn't CMOS, so the analogous formula needs to split the voltage into the two parts. Secondly, since the RSL bus is not a CMOS bus, using a formula for CMOS power dissipation is inappropriate. In particular, the RSL bus has pull up resistors that dissipate continually whenever a "1" is applied. This means that there is no frequency dependence at all in that part of the power consumption, and the CMOS power equation is entirely unsuitable. From there, his comparison flies off into the usual never never world of the typical mom and pop Rambus fantasy candy-land, but he makes a comment that gets back to our output compliance discussion: "The voltage swing in RDRAM is 0.8 volts, taken in two steps of 0.4 V." If you work through the example I gave you, that of a near (to the controller) RDRAM chip transmitting a "1" followed by a far RDRAM chip also driving a "1", you will find that there are actually 3 steps to the voltage swing. Another basic problem with his post is that he is ignoring DDR, which provides a much lower power consumption than SDRAM. Here's a good link for spreadsheets allowing one to calculate power consumption for SDRAM and DDR, as well as .pdf files explaining how all this is calculated: micronsemi.com -- Carl