RAM rundown

Before I begin, let me apologize. I will be using a lot of technical terms that most of you might not know. I will make sure to make a page on my website with a list of definitions so you know what the heck I am talking about. You will be able to find it on my main page’s navigation bar under “Pages.” I’ll label it, “Computer Definitions.” I’ll have it posted tomorrow. Until then, I guess you’ll be really confused!

Computers have been moving so rapidly lately. It’s been hard to keep up. DDR3 hit the market, ATI has their new x2k series out, nVidea has their new 8000 series out, and AMD has their quad core design near completion. I’ve just going to briefly give you a run down based on the numbers.

DDR3 is not really the impressive in terms of numbers. Their timings are just horrible. What’s the point in having a high bandwidth if you can’t do anything inside the RAM? I know that that DDR2 is starting to become more powerful than DDR, but I think it will be a while before DDR3 becomes the new standard. It would be a good idea for you to get a motherboard that supports DDR3, but don’t buy DDR3 RAM until it becomes a more practical solution.

I should explain something about micro processors. AMD, IBM, and Intel are the leaders in general purpose processors (some would argue that IBM is not a general purpose processor, but that’s another story). Every micro processor needs what’s called a memory controller. The memory controller is a system on the chip that will control memory addressing. Without the memory controller, your computer would not be able to use any type of RAM that’s on your computer (Video RAM, System RAM, etc).

Most people that use Windows as their base operating system will be using 32-bit registers on their processor. To most people, this really means nothing. Well, this number has a lot to do with how much RAM you can use as well as other architectural changes. The people that use the 32-bit operating system will only be able to use about 4048 GB (2^32 bytes) of RAM in their system (Or about 4.25 Million Bytes. I’ll cover why a MB (megabyte) does not equal 1 million bytes some other time). That’s total RAM mind you (Video RAM, Physics RAM, System RAM, and any other type of RAM on the system). We are rapidly coming to the mark where 4 GB is not enough memory for a power user. Intel, AMD, and IBM have a new architecture that uses 64-bit registers. This allows for computers to use 17592 PB (Petabytes). This is a huge amount of memory. Ed told me an interesting fact the other day. He said that the human brain can, in theory, store up to 100 TB (terabytes) of memory. Well, a computer with 17592 PB can then store over 170,000 human brains. Can you fathom being able to load your brain into a computer so you could see any image that you remember; you could see any equation you’ve had in the past; etc? That’s over 170,000 human brains. Computers have come a long way in these last few decades.

Let’s recap for a second. The Memory controller gives the CPU control over the total system RAM. The 32-bit CPU can call on up to 4 GB of total system memory while the 64-bit CPU can call an ungodly amount of total system memory. Now that the CPU has an idea how much RAM it can handle, the CPU has to support the speed of the RAM. The memory controller takes care of this for the CPU as well. This is where Intel and AMD differ (I will stop mentioning IBM because I don’t know enough about their processors to make a comparison). AMD built their memory controller on its CPU. Intel took a different route. Intel put their memory controller on a device that’s called the North Bridge. The North Bridge controls the system PCI buses as well as other buses.

Having the memory controller off the CPU itself is a good and bad thing. It’s good because if a new type of RAM comes out, you just buy a new motherboard (which is the cheaper solution in most cases). The new motherboard will have the new memory controller onboard so the CPU will know how to use the new type of RAM. The bad part is that it slows your memory access time. It’s an ok solution, but I think AMD has the better idea, with a little bit more cost per upgrade involved. AMD has their memory controller built into the CPU itself. Unlike the Intel chips, if you want to upgrade to the latest and greatest RAM type, you will have to upgrade your CPU, Memory, and Motherboard. This will become very costly. AMD CPUs are rather cheap now, but the new technology is always expensive. The good thing is that there is almost no latency when accessing the RAM. Yay, I hate waiting! Then with AMD’s Hyper Transport system, the AMD CPU can directly access the RAM on a very fast bus which will tern into later latencies in dual channel memory solutions. To upgrade an AMD system to the latest and greatest RAM specs, you will have to spend about $700, while you would only speed about $400 on an Intel based system.

You guys still with me? Haha! Now that we have the fundamentals down, we can start learning all about how RAM works. Yes, we haven’t even scratched the surface on computers yet. I’m sure everyone has used excel before. Will RAM works almost like an excel spreadsheet. Your operating system will say it needs the information saved under column D, row B. If you were in excel, you would look at your columns until you found column D and drag your finger down until you hit row B. Well, the RAM is the same thing in a sense. Your computer requests a piece of information located at some memory address, then your RAM sends that piece of information back to the CPU. Piece of cake, right? Well, we’re not done with RAM yet!

Have you ever ordered RAM before? If you have, you will see something on the RAM that looks like this: Timings: 2, 3, 3, 6 (DDR timings) or Timings: 4, 6, 6, 12 (DDR2 timings), or Timings: 8, 12, 12, 18 (The new DDR 3 timings). These numbers actually mean something. I’m not going to go into much detail here. You only really need to know that lower numbers here are better. As you can see, the DDR3 timings are just very large, meaning slow in this case. These timings key to how fast the RAM can move data, so the slower the timings, the slower the RAM can send information to the CPU.

Now that you know everything about timings, we’ll finish up our little discussion about RAM with the RAM bus speed. You can calculate this number pretty easily. For any computer made within the last 4 years, your core system oscillator runs at a speed of 200 MHz. This is going to be your RAM’s base speed. Your RAM has a system built in that will essentially force your RAM to send out more information that the core system says you can. Since the RAM runs off a separate bus as the core system oscillator, you can achieve this without much of a hassle. I’ll explain why in a later blog. In terms of DDR, this increase happens at a rate or 33 MHz. DDR means double data rate, so for DDR RAM the speed step increases happen in doubles, so the first calculated step would be 66 MHz. Your base speed is 200 MHz + 66 MHz for the step increase for a total of 266 MHz. Now, this is just the first rung of DDR RAM. You can also have a higher grade of RAM that will run at speeds of 333 MHz (200 + 133 (calculated step X2)) and 400 MHz (200 + 200 (calculated step X3)). When DDR2 came out, we saw and even bigger increase is RAM speed. Literally what DDR2 means is two times the double data rate. So the base multiplier will be X4 instead of X2. For DDR2, you will have the base step at 33 MHz at a X4 multiplier for a total of 132 MHz for the first calculated step. Also, your base speed will be increase by a factor of 2, making your new base speed 400. So for the tier one of DDR2, your speed will be 400 MHz + 132 MHz (rounded to 133 MHz) for a total of 533 MHz. Your tier two and three would look like this: 667 MHz (400 MHz + 267 MHz) and 800 MHz (400 MHz + 400 MHz). DDR3 is the same way as DDR2 in a way. DDR3 just means three times the double data rate. This will make your base calculated rate: 198 MHz (rounded to 200 MHz). I got this by taking the base rate (33 MHz) and multiplied it be the double data rate, and then multiplied that by the DDR’s rate increase (3 in this case). The base rate will be increased by a factor of 3 as well. The base speed is now 600 MHz (200 MHz * 3). Tier one for DDR3 would be 800 MHz, tier two would be 1000 MHz, and tier three would be 1200 MHz. This is some incredible speed.

Now that you know a bunch about RAM, you can now understand why DDR3 is not a practical solution for right now. Even though you have a super high speed at which your RAM can operate, your RAM can only manipulate RAM at a certain rate (controlled by the timings). The problem with DDR3 right now is that they can’t control the data that fast. The slow down the timings to make sure there is no loss of information. This makes the RAM more stable and reliable. What good is RAM if your information is lost in the process? I’m sure this will be fixed with new innovations. It’s the same thing that happened to DDR 2 when it first came out. High end DDR was still better. Over time, DDR2 became the winner. Now that DDR3 is out, we’ll see how things play out and see when DDR3 memory becomes as good or better as DDR2.

Sorry for the very long post today, but this is something I wanted to discuss, and it’s not as simple as a few loose definitions. Well, it can be, but I want to be difficult! I’ll fill everyone in on video cards and CPUs tomorrow. It will be a much smaller most next time! I swear :)

Until next time!

Leave a Comment

You must be logged in to post a comment.