We discussed a topic in the computer organization course. The professor told us that when the computer first turned on, all the transistors on the CPU are instantly electrified. So he said, keeping the computer on is better than constantly turning it off and on. Is this really true? Is electricity flowing through all transistors when first powering up?
6
-
\$\begingroup\$ In term of stress seen by the CPU during start-up it could make sense, but as overall system is the CPU the most sensitive part in a computer? I am sure the Power supply in the PC has a failure rate higher than the CPU, the electrolytic capacitor on the PSU are depending on the operating time and temperature,less it runs longer it works. \$\endgroup\$– DelpheskCommented Sep 29, 2020 at 10:11
-
2\$\begingroup\$ In a sense it is better to not turn things on and off constantly. As everything degrades slightly when turned on, and everything degrades slightly when just being on, perhaps a time period can be defined that shows when it is better to turn off a computer if not used than to keep it constantly on. But "all transistors getting electrified" sounds ridiculous. Are you sure that is what prof said, and what does prof mean by saying so, and what is the area of the professor? A computer scientist does not necessarily need to know about details of electronics you know. \$\endgroup\$– JustmeCommented Sep 29, 2020 at 10:12
-
2\$\begingroup\$ All of the transistors in CMOS logic are enhancement-mode, which means that no current flows at all until their threshold voltages are reached -- by which time, reset circuits make sure that the logic is in a "safe" state. \$\endgroup\$– Dave TweedCommented Sep 29, 2020 at 11:26
-
3\$\begingroup\$ ...instantly electrified I'm assuming that that was the literal phrase the Professor used. Then it does not sound like someone who knows very well how electronics work. Electrical engineers usually do not say that "circuits are electrified", they would say for example, that the supply voltage is applied. And not all circuits will have their supply voltage applied immediately. It is possible but not always the case. For example, the WiFi transceiver might remain switched off until it is actually needed. \$\endgroup\$– BimpelrekkieCommented Sep 29, 2020 at 11:28
-
\$\begingroup\$ CPU's have finely delineated power-on and power-off requirements. So much so, special "power supervisory" ICs were created to assist in power-on and power-off. Same functionality may be integrated today. If these were not in place, the CPU may indeed not power up correctly, which is what I think your prof is implying. PSIC's keep the CPU in an off/sleep/reset state until the power is verified to be good. And likewise when shutting down, reset/sleep/power off before the main supply sags, so the chip is not presented with undefined voltage levels. \$\endgroup\$– rdtscCommented Sep 29, 2020 at 12:10
|
Show 1 more comment
1 Answer
\$\begingroup\$
\$\endgroup\$
3
If he is of a certain age, he may be half remembering early 3-rail NMOS devices such as 16 kilobit (2 kilobyte) DRAMs. These required 5V for I/O, 12V for the storage cells (to get adequately large signals on the storage capacitors) and -5V back bias to turn the MOS transistors off by default.
On these devices it was important that the -5V supply was present before the other supply rails. All the FETs were NMOS, acting as low side switches, with either resistors between drain and VDD, or NMOS FETs configured as passive current sources.
With these chips, if the 12V supply came up first, over 16000 transistors all turned on at once, consuming enough power to destroy the device.
The next generation (64 kilobit) incorporated a charge pump on-chip, t supply the -5V supply automatically.
I never encountered any actual CPUs with this requirement but it's possible the Intel 8008 (no that's not a typo for 8080) did.
Nowadays it sounds like a quaint myth ... and with rare exceptions (possibly some high performance FPGAs) it is.
-
\$\begingroup\$ Thanks, this was the answer I was looking for. \$\endgroup\$ Commented Sep 29, 2020 at 12:56
-
1\$\begingroup\$ Brian, Those 16K DRAMS gave me grief (random errors every few days). They had to be treated not as digital circuits, but analog. Required very clean DC supplies, bypass caps everywhere. I recall some off-beat excuses from manufacturers for their fussiness. \$\endgroup\$ Commented Sep 29, 2020 at 13:03
-
\$\begingroup\$ My Richard Russell Board did the same ... until I added an R-C bodge to clean up the refresh signal timings. I had a coat over my head looking at scope traces to find that one! Rock solid afterwards as you would expect if you knew RTR. retrocomputing.stackexchange.com/questions/2402/… \$\endgroup\$– user16324Commented Sep 29, 2020 at 13:08