We're in hardware, and that is going to be computer hardware. What are we going to start out with? Examining the block diagram of a practical computer system based upon a microprocessor will be useful in understanding its operation. A system level block diagram is illustrated at the right.
Here we have a block diagram of a basic computer system, and this will be the subject of this particular section, and we're going to be looking at, for–, we have a microprocessor here, then we have a system controller, this is usually referred to as a chipset, internal memory, this is usually referred to as RAM, and then cache here, that would be both level 1 and level 2 cache. Over here we have the video controller, video card et cetera, then the hard disk controller USB circuits, then down here we see the Ethernet circuits, power supplies, ISA bus, PCI bus, BIOS et cetera. This will the subject of this particular section.
The major functional blocks in a computer system communicate via a number of buses, and so there are three major buses represented in the diagram, so here we see the AGP, the PCI, and the ISA bus. PCI will be the first one we'll mention. This is referred to as the Peripheral Component Interconnect and it is a bus of moderate speed. Usually, this is about 33 megahertz, that we will look at this more in the next slide. Then there is the ISA, which is the Industry Standard Architecture. It is a much older than the PCI, but also provides for a standard communications half and so ISA has been around. Actually, ISA was around from the very beginning of PCs. It is beginning to fade from the newer systems. Most of them are going to PCI and USB as the primary buses.
Then there is the AGP, which is an acronym for Accelerate Graphics Port, this would be right here. This provides high-speed video information between the system controller and the video controller. For a little bit closer look at the standard busses, I have a table built here, and we're going to look at ISA, PCI, AGP. First of all, the ISA bus, it can be 8 or 16 bits. What that means is that you have a bus and if it's eight, it means that when data is moved on this particular bus, if it's 8-bit wide, it means that there are eight lines, and so what that means is that we could have a byte of dataset. We had 11001011 and we wanted to move this byte on the bus. We would move it in parallel one on a given clock cycle. Here is one clock cycle.
Then this entire byte moves on the bus at one point, and so it would be moved around the system board to different devices, that 8 bits would be moved simultaneously. That's in the case or actually, that was on the first PC, which was an 8-bit machine, and then along with the area, and this is really dating the system, but along came the 286, which was the mighty 16-bit machine. You can do the same thing, only did it with 16 bits. The clock speed for these systems is about, for the ISA is 8 megahertz, but technically, I believe it's about 8.3, but for our purposes, we'll just say 8. You're moving, in this case, either one byte, or you're moving two bytes in this case, and this is in recent years. Usually, ISA is a 16-bit. This is really old, went back to eight that–, but if you do the math here and you figured out, “How much data is actually moving?” We're moving 2 bytes at a time on the bus and the bus is moving at about 8 megahertz, so 8 x 2 would give us a 16-megabyte throughput. This is often the way that we evaluate buses, it's what is the total throughput, how much data can we move through this bus?
In this case with ISA, we're looking at 16 megabytes per second. There actually is an error. Your text is really quite good, but there is an error here. It says 1 to 2 and actually, it should be 8 and 16. I think they're getting confused with the bus width and they weren't thinking in terms of a transfer rate.
The PCI bus came along, and the standard PCI bus is 32 bits wide, and it operates at 33 megahertz. That gives us an effective throughput of 133 megabytes per second, which is quite an improvement from this. We get that number from,- if you think about 32 bits, that's going to be 4 bytes, 8 bits per byte and then times 4. We would be moving, we'd have like this, we'd have 32 lines and we'd be moving at a much faster speed, this 33, actually it's 33.333, and that makes this math actually work a little better, but if we took this frequency times four, then that's where we come up with this number. This is our effective throughput for the PCI bus.Before a PCI, ISA was the primary bus that was used for video, and note that there's quite a difference here in speed, that this is going at 16 megabytes, where this jumps all the way up to 133, so that's why when the PCI video cards, it was quite an improvement over the ISA performance with video. Along came AGP, and before I jump to AGP, I will note here that there are some specialized, PCI bus is oftentimes in servers. They will have an advanced bus that will actually be 64 bits wide, and sometimes it will go as fast as 66 megahertz. This would make for a much faster PCI interconnect, but most commonly it's 32 and 33 megahertz, 32 bits.The AGP came along, and this was the Intel, and actually, both of these were Intel inventions. The AGP moves data at 32 bits on the bus, same as PCI, so you'd have 32 lines and moving 4 bytes at a time. Notice the speed, the speed is twice that of the standard PCI bus, and so because it is twice as fast, then the throughput is actually going to double. We had 133 coming out of our PCI, but the AGP jumps up to 266.For purposes of video, when you think about the changes that have taken place, went from 16 megabytes to 133 all the way up to 266. This actually represents what we would call AGP1 when AGP first came out and your text mentions going up to actually 1 gigabyte, but actually, the speeds have gone higher than that. Let's jump to the next slide, and this is some information.
Actually, I got this from Wikipedia. Even though Wikipedia isn't the most accurate source, this is quite accurate in concerns to AGP. cLet's take a quick look at this, AGP1, it started out with 32 bits operating at 66 megahertz resulting in, this was our throughput at 266 megabytes per second, and it doubled from the 133-megabyte bus, which we saw on the PCI. The next rendition was, the clock doubled, so AGP2 came along, and again, 32 bits at 66 megahertz, but they clock it a little bit differently and they were able to double this speed, so it went from 266 now to 533.This was the AGP 2X series of cards, and then it went to 4X, which went from 533 and doubled it to 1066, and that's where your text leaves off. Actually, AGP is in 8X, and so in 8X, there is a doubling of that up to around, it's the same, 2133 megabytes, which is 2.13 gig. This allows for some very serious bandwidth in regards to video. It makes possible many of the intense graphics that we see today that come from high-performance video cards.This was a quick intro to the AGP, and we looked at the standard buses that you see in PCs, and we just basically introduced the subject, and we'll continue on with other hardware in the next presentation.
Video Lectures created by Tim Fiegenbaum at North Seattle Community College.
In Partnership with NXP Semiconductors
In Partnership with Mouser Electronics
by Jake Hertz
by Darby Hewitt