To: Bill Jackson who wrote (66695 ) 12/30/2001 11:10:02 PM From: Elmer Read Replies (2) | Respond to of 275872 Elmer, Ok, thanks for the course. You feel that AMD stays below the knee to hide a bad truth? This is possible. what other major company in CPUs engages in similar tactics.? I'm not going to be too hard on AMD on this one. I don't think it's necessary to publish this data because a processor won't be at 100Deg if it's not being clocked. This is more lab data than customer data and AMD isn't obligated to publish it. IMO. The only point is it prevents a direct comparison.Now since AMD routinely runs it's CPUs hotter than intel does this mean that the knee is process dependent? and Intel has a lower one? If the knee was at 65 and we see AMD operating above that is that not saying they have a higher knee? I must admit I haven't analysed this in depth. I know there is a knee but I don't know how it varies with channel lengths along the x axis(x being current and y being temp). All I know is the non-linearity.Usually both analog and digital transistors in the turned of state both have low leakage current. The analog may have a wider range over which the change their current from max to min. The digitals one would expect to have a far sharper curve? I would also expect this to make faster operating transistors? the faster they go the more they leak? or is that related to the increased number of transitions?I recall the P-3 ran into the problem and in fact was limited by it. This is getting complicated. Remember the transistors only need to propogate a logic state through the design, so if they can do that without completely shutting off, or turning on the current flow, then Bob's your Uncle, and they can do that faster by not completely turning on or off, relative to saturation and shutoff. A large amount of current is spent in switching, so the faster the switch the lower the effect of static leakage during the off state. So static current is a more useful measurment or the transistor characteristics than dynamic current, in regards to transistor leakage, which translates to channel lengths. Make sense? It is also true that dynamic currents correlate as well but you lose the resolution.Is a large standby current bad in that is increases the power duty cycle, thus making the item use more watts and thus need more cooling? In a desktop system the dynamic currents dominate. For mobile it becomes more significant. EP