Microprocessor Performance Explained
Posted on Nov 16, 2007 by
Paul WhiteI remember back in 1995, when windows 95 was just starting to hit shelves. These were the days that the first pentium computers started to be noticed. For over a decade microprocessors were pretty simple. The 286 by Intel was the first in a series of 32 bit processors. From the release of the 286 back in the mid 80s to the pentium 3s around year 2000. It was very easy to tell what was better, or faster. Every chip had a clock speed. Since most chips were all intel and the architecture was pretty much the same during this 15 year span. All you had to do was compare clock speed. This was great for consumers, and great for manufactures. The easier it is for a consumer to know what they are getting, the more units the
manufacturers will sell. But as Intel started to reach high clock speeds, problems started to rise. All these years Intel and other companies had focused on Clock speed. If you could just make a chip faster than the competition then you would maintain you status as the market leader, and therefore sell more units. But while the clock speed was getting faster the rest of computer was not.
If intel made
cars they had originally started with a lawn mower chasis with a lawn mower engine. within 15 years they had essentially advanced the engines to that of super
cars. Meaning your push lawn mower looked the same, but now had a ferrari V12 under the hood. This may sound cool and all, but hard rubber 6 inch wheels are not going to perform correctly for an engine like this. Once Intel started to make the Pentium 4 they realized they really needed to rethink the way the Chip would interact with the rest of the board.
Computers are only limited by the bottlenecks within their design. Just as a V12 on your lawnmower is not very efficient, neither is most computer chips on their native
motherboards. Whenever you run a program or find yourself waiting for your computer to do something, you are not waiting on the Chip, but rather everything else. Your chip is constantly sending data back and forth between the RAM,
hard drives, and other devices. Most of this time your Chip has requested information from one of your storage devices. When your computer makes this request. The device doesn't immediately send the information. The device has to find the data you requested, and then send it back to the CPU. In the meantime your CPU is sitting there idle doing nothing. Your CPU spends most of its time waiting on other devices. A good analogy would be to think of you being the CPU, and your hard drive being a shipping warehouse on the other side of the globe. What you need right now is some part for your car. You call in the order, and the shipping warehouse then ships you the part. Once you get the part you only spend 30 minutes to install it. Though you had to wait for 5
business days for the part to get here. That is similar to how your computer works. the CPU spend very little time working with the data, and most of its time waiting for parts to arrive.
Making your clock speed faster isn't going to speed up the warehouse. So for the last 10 years the focus has been more on how we can speed up the warehouse, and shipping rather than the CPU. At the same time computer architecture has changed where instead of having a single list of things todo, of which you must computer item number 1 before moving on to item # 2. We now have two or more lists. If a computer is held up waiting for a part to arrive, it can focus its attention on a second list. In the computer world we would call this hyperthreading. If you have two people filling orders and each person has two lists, this would be called duel core hyperthreading. If you have 4 people this would be a quadcore.
In modern Chips we have moved away from hiring one person to do the job, and instead hire multiple people of lesser power, but together they are more powerful, and get the job done quicker. It would be like hiring Mr Universe to go pick fruit from a field. The cost to hire him would be high and the even though he is powerful, and he could pick fruit very fast, you would be better to hire a dozen illegal aliens for less money, and they will get the job done faster. The fact is most computers don't need one big fast CPU, but intead need many smaller CPUs to take on tasks. This is where we are today. The focus is on how we can make current chips more efficient, rather than how to just make the Clock speed faster.
So what makes a chip efficient? For one how much energy it takes for the chip to run. The first computers took up an entire room, and created lots of heat. Rooms of airconditioners had to be used just to keep the chips from over heating. Today a small 80mm fan is all a chip needs to stay cool, but that still doesn't mean its efficient. With computers moving to a mobile platform, the focus is on how to make it run for days at a time on a small lithium ion battery. To do this they run the chips at a lower voltage, and use different conductors. The other problem with chips has been the limited amount of Cache on the Chip. For years 256K of Cache was pretty much standard. Now we have chips with up to 12 Megs of Cache. Over the last few years we have seen the amount of cache increasing while clock speed has stayed the same. A good analogy would be to compare your Chip to shopping at the grociery store. If you could only buy what you could carry in your hands that would be similar to what the old chips with 256K of cache were like. Then Chips got 1/2 Meg and 1 Full Meg of Cache ( you now have a basket to shop with ). Now we have 12 Megs of cache on a chip, which is like driving your pickup truck down the isles of your grociery store. Cache is very important.
So if you are picking what CPU to use on your computer, look at the number of cores, clock speed and cache. Also even though you will only have two options for Manufactureres ( Intel and AMD ). I highly recommend Intel. I have always used Intel and have never had any problems with them. I have friends that have ran AMD chips, but they seem to burn out and have issues.
Discussion
No Comments have been submitted
WhiteSites Blog