The effect of temperature on your processor! *SCIENCE WARNING* - Overclock.net - An Overclocking Community
Forum Jump: 

The effect of temperature on your processor! *SCIENCE WARNING*

Reply
 
Thread Tools
post #1 of 39 (permalink) Old 08-05-2010, 05:42 PM - Thread Starter
New to Overclock.net
 
wcdolphin's Avatar
 
Join Date: Dec 2009
Location: 192.168.1.1
Posts: 2,577
Rep: 225 (Unique: 200)
Cooling? Why do I want to cool my CPU?

Contents:
  1. Processor Architecture
  2. Overclocking, the effects of temperature.
  3. Cooling System Design Principles

Processor Architecture
Understanding what a processor really is:
In short, your processor is an on/off switch...Well…A bunch of on/off switches, a billion or so. These switches use the most basic logic known to man, known as Boolean logic: everything is either on, or off, either a one (on) or a zero (off). Meanwhile the explanation of how computers use Boolean logic (which leads to Binary), is not necessary for the purpose of this post. These switches are of course tiny, around 32nm, depending upon the transistor in question, and have been described as containing "merely a couple of electrons" (citation needed)
The Switch
To understand the fundamental issues at hand here, we must understand this elusive switch; how it works and what it is made of. This switch is a switch whose state as "on" or "off" is governed by voltage; unlike your lighting switch at home, where up perhaps is "on", in the case of the switches we are talking about (High K Metal Gate Transistors), a high voltage is "on", and a low voltage is "off" N.B. 1. As some things are easier to visualize, I have included an image below.

As described, the state of the switch is determined by the voltage between the source and the drain (labeled S and D). If the voltage is HIGH, known as Vcc, the switch is said to be "on", if it is low, known as Vss, it is said to be "off".
Clock Cycles-Frequency, Speed
The frequency, or speed of a processor is, of course, related to the underlying structure of the switch. All of these terms are roundabout ways of describing the rate at which your processor's switches can turn on and off, or switch between the high voltage that designates "on" and the low voltage that designates "off" (Vcc and Vss). The rate at which this switching occurs is really, very rapid, as I am sure you know by this stage!
However, what is interesting is that the clock speed can be thought of as a pulse like signal, switching back and forth from on to off very rapidly.
This can be shown as follows:

Where Vcc is the high, "on", and Vss is the low "off" voltage. The Y axis here is Voltage, the X axis here is time.
However, this is not an accurate depiction of what is actually occuring, rather this is the ideal binary situation; the voltage is either exactly "on" or exactly "off", and it switches from the high voltage to the lower voltage instantaneously, which from physics we know is impossible.
Rather, our switches work on tolerances, accepting anything over some value Vcc as "on," and anything less as "off". It also takes some time to transition from one state to another, as you can see below:

Overclocking, the effects of temperature.
What it means to overclock:
What does it mean to overclock? Sure, you know what it means...Go faster, press some buttons, amp up your SuperPi times, but what does it really mean, or what does it really do? Remembering from above, the clock speed is actually the clock frequency, or the number of times that the transistors inside of your processor switch from on to off per second. When we overclock a processor, we force it to switch between the two states more rapidly, and some problems arise. Remembering the image above, there is some time required for the state of a given switch to change from on to off-there is some time required for the voltage to drain from the gate; increasing the number of times the state must switch per second, decreases the amount of time your processor has to transition between the two states. You all have experienced the issue associated with your processor being unable to make the transition in the time allotted: instability. Your processor cannot tell the difference between on and off, as the voltage never reaches the recognized "on", Vcc:

The issue at hand is simple: the voltage cannot switch fast enough between the two states, however there are two solutions, on one hand you can somehow change the value that is recognized as on, or else increase the speed at which that transition occurs. Whether or not you know it, you have been doing the former every time you have increased the voltage of your processor in order to maintain stability; while you actually raised the peak voltage, you effectively moved Vcc down in relation to the maximum voltage, providing for the same effect. While it is not quite that simple, as the rate at which electricity flows is also affected by voltage (V=IR, for a given resistance, an increase in voltage incurs an increase in current, or flow of electricity per unit of time), let us for simplicity's sake assume that is all that you are changing in the first case, and allow the rate at which electricity flows to be altered separately.

Decreasing the time it takes to switch states
The following contains some basic physics, I will try and provide links to wikipedia for as many concepts as I can, as some people have less experience than others, and further reading may be valuable to them!
Before discussing how to alter the time it takes to switch states, we must first discus what it is that governs the rate at which this switching can occur. While current is not technically the rate at which voltage dissipates, it can be related for any given situation, and we will use it as a proxy to describe the rate at which voltage dissipates.
First, the equation for current that we will be using is: I = V/R. Looking at this mathematically, we can see that there are two things we can do to increase the value of I (current):
  1. Increase the voltage
  2. Decrease the resistance
As we already covered point 1. in the previous paragraph (rather sneakily, I must admit!), we will merely concern ourselves with decreasing the resistance. What is resistance? Simply speaking, it is the ability of a given material to resist the flow of electricity through it. Resistance is affected by temperature: R = kT (where R is resistance, k is some constant of proportionality, and T is temperature). Thus we can see (both intuitevly perhaps, if you think about the random movement of electrons that inhibits the general path of flow, and perhaps mathematically) that resistance increases as temperature increases. On the other hand, as temperature decreases, so does resistance, effectively decreasing the time it takes to switch states, from high to low; solving our problem and providing a simple solution.
A Little more on Temperature'
While the previous paragraphs describe the increased stability at lower temperatures, because of the lesser amount of time to switch between ON and OFF (effectively allowing you to clock higher), it serves to mention that there are far more intricate processes at work here. As a measly engineering undergraduate with at best a 'decent' understanding of physics, I cannot explain all of these processes, however there is one that is commonly discussed and is relatively simple: electron-migration refers to the movement of particles within a conductor induced by the momentum of electroncs. This concept can be seen as a sort of bowling ball effect:
imagine bowling with tiny marbles. Once in a while, or over time, you will eventually move, hit, or change the pins, because while tiny in comparison to the pins, your marbles have momentum, and some force behind them!



wcdolphin is offline  
Sponsored Links
Advertisement
 
post #2 of 39 (permalink) Old 08-05-2010, 05:42 PM - Thread Starter
New to Overclock.net
 
wcdolphin's Avatar
 
Join Date: Dec 2009
Location: 192.168.1.1
Posts: 2,577
Rep: 225 (Unique: 200)
N.B1: Voltage is measured across, or between two locations, or objects. In simulation terms, we refer to this as an 'across variable'



wcdolphin is offline  
post #3 of 39 (permalink) Old 08-05-2010, 05:53 PM
 
Join Date: Apr 2010
Posts: 5,697
I like it.


Segovax is offline  
Sponsored Links
Advertisement
 
post #4 of 39 (permalink) Old 08-05-2010, 06:06 PM - Thread Starter
New to Overclock.net
 
wcdolphin's Avatar
 
Join Date: Dec 2009
Location: 192.168.1.1
Posts: 2,577
Rep: 225 (Unique: 200)
Making cookies as we speak, time to translate this all from my presentation last year.



wcdolphin is offline  
post #5 of 39 (permalink) Old 08-05-2010, 06:54 PM
New to Overclock.net
 
hybrid-kernel's Avatar
 
Join Date: Jun 2008
Posts: 407
Rep: 60 (Unique: 56)
Quote:
Originally Posted by cdolphin View Post
As described, the state of the switch is determined by the voltage between the source and the drain. If the voltage is HIGH, known as Vcc, the switch is said to be "on", if it is low, known as Vcc, it is said to be "off".
yeah.
hybrid-kernel is offline  
post #6 of 39 (permalink) Old 08-06-2010, 07:02 PM - Thread Starter
New to Overclock.net
 
wcdolphin's Avatar
 
Join Date: Dec 2009
Location: 192.168.1.1
Posts: 2,577
Rep: 225 (Unique: 200)
About halfway done, a more realistic thermodynamic approach to understanding cooling systems to come.



wcdolphin is offline  
post #7 of 39 (permalink) Old 08-06-2010, 07:05 PM - Thread Starter
New to Overclock.net
 
wcdolphin's Avatar
 
Join Date: Dec 2009
Location: 192.168.1.1
Posts: 2,577
Rep: 225 (Unique: 200)
If anyone has any input thus far, it would be appreciated, also mods, where do you think the best location is for this? Don't judge yet, as the real cooling section is to come

EDIT: Dang, hit new post instead of edit *may the gods merge me upwards*



wcdolphin is offline  
post #8 of 39 (permalink) Old 08-06-2010, 08:13 PM
 
Join Date: Mar 2009
Location: San Diego, California
Posts: 8,014
tl:dr

cliff notes please
88EVGAFTW is offline  
post #9 of 39 (permalink) Old 08-06-2010, 08:18 PM
 
Join Date: Jun 2009
Location: Texas
Posts: 1,991
Quote:
Originally Posted by 88EVGAFTW View Post
tl:dr

cliff notes please
to increase speed we must decrease temperature.


michaeljr1186 is offline  
post #10 of 39 (permalink) Old 08-06-2010, 08:20 PM
New to Overclock.net
 
Join Date: May 2010
Location: Montreal, Quebec, Canada
Posts: 3,372
Rep: 138 (Unique: 125)
Blostorm is offline  
Reply

Quick Reply
Message:
Options

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in



Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off