I'm no expert (just look at the several posts I've made), but I do know that FSB stands for Front-side bus. It's a system bus between the CPU and the RAM and maybe other things too, I don't know.
July 9, 2001 9:39:05 PM
Hey,great stuff Utopia. This is exactly what us newbies, strangers need! We need some detailed correspondence now and then. Its easy for you guys, but for us that sit back and try and work it out we thank you.lol
~OMG IT nearly melted, lucky I had the extra fan *inside* ~
Basically, the motherboard operates at a specific frequency, and adjusts the electrical signals accordingly. This is where the Front Side Bus comes from. Now, on a modern motherboard (currently being made), this is either 100 or 133, only exception being Macs.
The ISA bus is designed to run at 15MHz, I believe. I'm not totally sure about that one. The PCI bus is 33MHz, and the AGP bus is 66MHz. That's assuming you haven't overclocked the busses. More on that later.
Now, with the system bus at 100 or 133, that is how fast the CPU interfaces with the motherboard, as well as the RAM interfacing with the MB. That's where the PC100/PC133/etc. comes from. It's what the RAM is guaranteed to run at. RAM can be under or overclocked to run at different system speeds, but not all RAM is good at overclocking. My generic RAM won't run even 2 MHz above that, although most name brand RAM will run a fair amount above it.
Now, you've been wondering about 200/266/400 bus speeds on Tbirds and P4s, I'm sure. That is basically double- and quad-pumped. What that means is for every clock cycle (100 or 133 million times a second/MHz), it executes twice or four times. Therefore, it's not a TRUE 200/266/400 bus, but it acts effectively the same (in theory, since there are no true 200/266/400 busses out there).
Now we come to Rambus. Rambus does not actually run at a quad-pumped 100MHz or 400MHz. It's dual-channel, which means it operates basically like a RAID 0 set of hard drives. The two drives/sticks of RAM are used as one, with the data split up, and written simultaneously to each, boosting the speed by double (again, in theory, it doesn't work perfectly effectively, although relatively close). That's why you MUST have multiples of two sticks of RDRAM in a P4 system, and why the benchmarks for the P4 SDRAM reference board were so crappy. No dual-channel (I don't think).
One of the advantages of the nForce chipset is dual-channel SDRAM/DDRSDRAM, which is a first.
Now, SDRAM and DDR-SDRAM (Synchronous Dynamic Random Access Memory and Double Data Rate SDRAM), operate the same way. DDR-SDRAM basically executes twice every clock cycle, like the Tbird processors.
Still with me?
If you have a Tbird 1.2/266 (like me), then it's a 133 bus. Putting in PC100 will give you problems, since either you have to overclock the RAM (not always possible) or underclock the processor (make it a 1.2/200 or something similar). Either way, it's not a good choice.
Now, when you overclock the FSB, you also overclock the ISA/PCI/AGP busses (with exceptions). There is a divider in use, so that with a FSB of 100, your PCI is 33 and AGP is 66. Most motherboards will offer you choices such as 133/33, 135/34, etc. You're therefore changing the ratio between FSB/PCI/AGP so that the speeds they're running at is more of a difference. MOST PCI cards cannot go above around 38 or so, with cards being both better and worse than that.
The AGP bus is overclocked at the same time, although most modern video cards will overclock a good amount more than most PCI cards. ISA cards won't overclock very much, if at all.
CPU clock speeds...have you ever wondered what exactly makes a 1.2 a 1.2? Well, I'm going to tell you anyway. That number is what you get when you multiply the FSB and the CPU's internal multiplier. My 1.2 has a 133(266) bus, and a multiplier of 9 (well, I bumped it up to 10, but factory it's a 9). So, 133x9=1200, or 1.2 gigahertz.
What is that multiplier of 9? I'm glad you asked. The earliest processors (real early) didn't have that. The whole thing ran at the FSB (then no more than 25). Needless to say, there was no realtime 3D rendering going on. As the math in programs became more complicated, most operations could not be completed in a single clock cycle. It took several clock cycles (that went at 25 million per second) to complete a computation. There had to be a way to change that.
So, Intel decided that with the 386, there should be an internal multiplier. In other words, even though the motherboard was sending information 25 million times a second, the processor was doing the math at 50 million times a second. However, it was sending it back to the motherboard at 25 million per second (the FSB). Remember, the reason for doing this was because the math required of the processor was more advanced (as required by more graphic-intensive programs). Because of this, the processor itself stopped becoming the bottleneck (to a certain extent).
Ok, I've gotten a bit too far into processors, so I'm going to stop now. If I realize I left something out, I'll add it.
And when someone sees I've said something dumb, let me know
My Athlon can beat your Ferrari off the line.<P ID="edit"><FONT SIZE=-1><EM>Edited by FatBurger on 07/10/01 08:47 AM.</EM></FONT></P>
Holy crap, I didn't realize how long that was until after I posted it.
My Athlon can beat your Ferrari off the line.
July 9, 2001 10:48:31 PM
Couple things Fat... The AGP bus is 66MHz not 89. And the PCI,ISA,AGP Busses get there speed from the divider. (100MHz FSB 1/4 Divider = 33MHz). So you can overclock and still be in specs with some speeds. The ram can also be underclocked so underpar ram can be used in a faster system.
Not I, that was bigpete591
But I think what he was saying is if you have, say PC133 that's supposed to be Cas2, but it won't run that way because it's poor quality, you can always back it down to either Cas3 or PC100 to make it work. You should ask him.