CPU FAQ Beginners, Unofficial

siliconjon

Distinguished
Mar 12, 2002
346
0
18,780
CPU FAQ #1 [Beginners]
Acronyms and other definitions
------------------------------
BIOS = Basic Input/Output System (this is the operational
configuration for your motherboard)
CPU = Central Processing Unit
ECC = Error Correction Code, or Error Checking and Correction
FCPGA = Flip Chip Pin Grid Array (interface for some Pentium CPU's)
FPU = Floating Point Unit
FSB = Front Side Bus (also referred to as the system bus)
GHz = GigaHertz (billion cycles per second)
MHz = MegaHertz (million cycles per second)
MMX = MultiMedia eXtensions
PPGA = Plastic Pin Grid Array
SECC = Single Edge Contact Cartridge
SECC1 = Slot interface for some Pentium2 CPU's (233Mhz=450)
SECC2 = Slot interface for some Pentium2&3 CPU's (currently up to 1Ghz)
aka = "Also Known As"

Now, this is the beginners' FAQ, so many of you savvies may find this ridiculously easy. But for the ones to whom this may be of assistance,
enjoy!

Q: What do the MHz and GHz actually mean?

A: These numbers literally mean how many cycles, or complete calculations, can be made in one second. The higher the number, the faster the chip...BUT these numbers are most significant when comparing the same types of processors. For example, a 1.6GHz Duron, even though the speed matches, is not as powerful as a 1.6GHz Athlon chip. A Pentium 4 1.6GHz is not as powerful as the same speed Athlon. But a 1.6GHz Athlon would be half as powerful as a 3.2GHz Athlon CPU of the same core architecture. So you can use these numbers as a guide, but only when comparing them to similar chips. These numbers aren't as helpful as they used to be, and they will only become less helpful in the future as CPU architecture makes some radical changes.


Q: Why do the Athlon processors use numbers different from their actual clock speed for their titles?

A: Here's where a little further explanation of the last question comes in. Athlons, due to their different architecture, can out perform a Pentium 4 chip of the same speed. This brought about some problems, since the average consumer knew little more than "the higher the Hertz, the better". So, to bring this difference to our attention, they have used numbers such as 1800+ for a 1533MHz (or 1.53GHz) chip to show that when tested, the chip could compete with an 1800Mhz Pentium 4. To check out one of the many benchmark tests done here on THG, pitting Intel vs. AMD, check out <A HREF="http://www.tomshardware.com/cpu/02q2/020610/thoroughbred-14.html" target="_new">http://www.tomshardware.com/cpu/02q2/020610/thoroughbred-14.html</A>


Q: What are these "second" names on these processors, such as Thoroughbred, Coppermine and all the others?

A: This is where architecture comes into play. Consider "Athlon" the last name of the processor, and "Thoroughbred" would be its first name. Athlon is the family of chips, and Thoroughbred would be a spawn of that family. As the chip changes slightly, the first name will change. The last name will only change when the company feels that the architecture has changed enough so that the chip is no longer anywhere close to its predecessor. When this happens, new chipsets are often required and completely new motherboards are needed to support the processor. The Pentium families are similar, but rather than completely changing their last name they have been merely adding numbers. The Pentium 3 is of a different family of CPU's from the Pentium 4. Changes in architecture that result in a new first name for the chip are often the fsb and die size. Changes in the internal design of the chip itself will result in a new family.


Q: I want to upgrade my processor; how do I pick out a proper CPU?

A: There are three important things to consider when upgrading a CPU: interface, frontside bus speed, and clock speed. You must be sure to match these numbers with your motherboard's compatibility list. If you have a Socket A, aka Socket 462, interface on your motherboard, then you must get a matching processor. If your motherboard supports 100/200 & 133/266 fsb, then you can get any current Socket A Athlon processor. If your motherboard only supports 100/200 fsb, then you must be sure your processor does not exceed these numbers for its frontside bus. With the clock speed, your best bet is to be sure your motherboard will handle that high of a speed. Many motherboards will be compatible with clock speeds higher than originally listed in their documentation, since the chip just wasn't out yet to list. But this isn't always true, so the first place to check is your motherboard's manual. If it does not list the clock speed you're looking for, then contact the manufacturer's website and try to locate any information on the motherboard, and it being compatible with the higher clock processors. If you're still unable to locate an answer, try calling the motherboard mfg, or a retailer/vendor and asking them if the board will take the chip you want, or if they will at least guarantee you can return the CPU risk free if it is not compatible with your motherboard. Often, a BIOS flash is released that will allow a board to use chips it was previously incompatible with.


Q: What is "µ"?

A: µ Stands for micro, meaning it is a number multiplied by 10 to the negative sixth power [10^ -6]. 1 µ is equal to .000001. This number is most often affiliated with the processor's die size. The smaller, the better, when we're talking about the chip's physical size. For further explanation, research electronics fundamentals.


Q: Why is a smaller processor better?

A: Well, the smaller they get the manufacturing down to, the more they can squeeze into a smaller space. Also, when dealing with such high-speed electronics, a smaller size allows for shorter distance in the circuits. The shorter the distance, the quicker the electrons can complete their trip.


Q: What is cache and do I need to be concerned with it?

A: Yes, cache is an important feature of processors. You want more, and faster both. Cache is a very small amount of memory (usually between 128K and 512K) that is used by your CPU to store very important information so that it doesn't need to store and fetch this information from your slower system RAM. It's like your desktop at work. You have plenty of room to file all around your office (hopefully), but the items you are currently working on (or will be working on shortly) are kept on the desktop in front of you because it's closer and faster to access. I think that means I have a couple gigs of cache on my desk at work, because I usually can't see much of the wood finish.


Q: What is the difference between slot and socket processors?

A: This is the interface of the processor. A slot processor, or SECC, will plug into your motherboard in a similar fashion to RAM and PCI cards. These processors can look a lot like a cartridge to a gaming console. It will be a single edge card that will slide into a single slot, and often have retention mechanisms to hold it tight. A socket processor, or PGA (pin grid array), is the more popular these days; it consists of a lot of pins underneath, with the chip and bridges showing on top. And then there are different types of slot and socket processors. Socket A is an Athlon or Duron chip with a 462-PGA, Slot A is an Athlon or Duron chip with a SECC interface. Slot 1 is a Pentium 2 and Pentium 3 interface, FC-PGA is a Pentium3 and Celeron socket interface with a 370-PGA. Some Intel processors have a Slot2 interface, and there are some with a FCBGA (flip chip ball grid array, much like PGA, but short rounded ball like interface versus long skinny pins). Socket 478 is the Pentium4 PGA interface. These interfaces are not interchangeable. Though there are some adapters that will convert a FC-PGA to a Slot1.

Here is a good article on CPU's from the one and only THG:
<A HREF="http://www.tomshardware.com/howto/00q2/000412/" target="_new">http://www.tomshardware.com/howto/00q2/000412/</A>


I'm not really human, I just play one on TV
 
feel free to make it official, or append it to the official one if you wish. This is free for THG to publish as they wish, so long as a reference to it's creator is given.

It is also incomplete, but I mentioned some of the more important facts.

I'm not really human, I just play one on TV
 
Bye Bye FAQ...the powers that be say "Be Gone!" They don't want you!

If you find this usefull, post a thx to keep it around...have any questions you would like added? Post em or PM them to me, and I will throw it in there.

TatA!

I'm not really human, I just play one on TV
 
Problem is we cannot keep bumping it up. FatBurger currently has good contacts with Fredi, who granted him the RAM forum FAQ. Instead of bumping, PM it to Fat, then Fredi will stick the FAQ on top, which can never anymore be gone no matter how many topics are created after it. It'll stay on top.

--
:smile: Intel and AMD sitting under a tree, P-R-O-C-E-S-S-I-N-G! :smile:
 
it is only an assumption but are you the webmaster?


:smile: i like toasted cpus but not AMD-inside. :smile:
 
it is only an assumption but are you the webmaster?


:smile: i like toasted cpus but not AMD-inside. :smile:
 
i only read a bit of it and i spoted some rather large mistakes for example;

"Why is a smaller processor better?"

your answer is not wrong, but certainly not right, smaller process is better because it means less distance for electrons to travel which means lower resistance which means lower voltage needed which means less power needed which means less heat generated and we all know what that means (i know i do because my overclocked tbird 1400 must be using about 80watts!!!!!!)

but great idea though, with a bit more input it could be really good.



I need a 1.5 Ghz Athlon + 512mb ddr ram to write emails......honestly
 
Well, for one, this is beginners...for two, I did say:

"Also, when dealing with such high-speed electronics, a smaller size allows for shorter distance in the circuits. The shorter the distance, the quicker the electrons can complete their trip."

But that is a nice quick rundown of the advantages you gave there. Though, they are not exactly accurate themselves...the chip's architecture will make more of a difference there (vs. die size). But that stuff gets complicated...and belongs in electronics FAQ, or advanced CPU FAQ.

What else do you disagree with?


I'm not really human, I just play one on TV
 
Well, were's that FAQ project you mentioned some weeks ago?

A little slow, are we? Only took an hour to put this together...what in the heck are ya's doing?

Left, Left, Left, Right, Left...

I'm not really human, I just play one on TV
 
Q: Why do the Athlon processors use numbers different from their actual clock speed for their titles?

A: Here's where a little further explanation of the last question comes in. Athlons, due to their different architecture, can out perform a Pentium 4 chip of the same speed. This brought about some problems, since the average consumer knew little more than "the higher the Hertz, the better". So, to bring this difference to our attention, they have used numbers such as 1800+ for a 1533MHz (or 1.53GHz) chip to show that when tested, the chip could compete with an 1800Mhz Pentium 4. To check out one of the many benchmark tests done here on THG, pitting Intel vs. AMD, check out http://www.tomshardware.com/cpu/02q2/020610/thoroughbred-14.html
just a minor detail...the athlon xp 1600+ is supposed to mean that it will perform at the same level as the older athlon (t-bird) at 1.6 ghz....not a P4 at 1.6 ghz (either way...the t-bird will still outperform the P4 at the same speed)...
at least thats what i heard...this makes more sense...for legal reasons...

😱 <b>Who fixed <font color=red>ATI</font color=red>'s leaky faucet??</b> 😱
 
Alright, it's time to get rude. If you want to disagree, how about an educated argument? Taken from AMD's own FAQ:

"Q: What do the 2200+, 2100+, 2000+, 1900+, 1800+ and 1700+ numbers mean?

A: These are model numbers. AMD identifies the AMD Athlon XP processor using model numbers, as opposed to megahertz, such as the 2200+, 2100+, 2000+, 1900+, 1800+ and 1700+ versions. Model numbers are designed to communicate the relative application performance among the various AMD Athlon XP processors. The AMD Athlon XP processor 2200+ can outperform an Intel Pentium® 4 processor operating at 2.2GHz on a broad array of end-user applications. "

This can be found at the following url:
<A HREF="http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_756_3734^3876,00.html#24363" target="_new">http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_756_3734^3876,00.html#24363</A>

For a hardware forum, there seems to be a shortage of expertise around here. That took an entire 30 seconds to locate.

<font color=red>01010011010</font color=red>
 
Who said I'm working on an FAQ?

Am I supposed to be? I let this one die, since you did not seem interested in my help in the matter...so why waste my breathe, eh? And surely you haven't been working on it all this time!? No human is that slow.

<font color=red>01010011010</font color=red>
 
Oh, I'm sorry. I didn't read it close enough to realize you were talking to pr497. My bad, I'll go back to watching Intel and AMD play chess against each other :redface:

<font color=blue>Hi mom!</font color=blue>
 
<bump>

Man, this forum has quite the traffic flow. This thing was burried 6' under.

But just to offer my hand in getting some official FAQ up and posted. And again, there is much more than this to be included, but I'm not going to waste my time if it's not going to be locked into the head thread for easy review.

<font color=red>01010011010</font color=red>
 
Ok, I'm gonna nitpick here.

A: These numbers literally mean how many cycles, or complete calculations, can be made in one second. The higher the number, the faster the chip...BUT these numbers are most significant when comparing the same types of processors. For example, a 1.6GHz Duron, even though the speed matches, is not as powerful as a 1.6GHz Athlon chip. A Pentium 4 1.6GHz is not as powerful as the same speed Athlon. But a 1.6GHz Athlon would be half as powerful as a 3.2GHz Athlon CPU of the same core architecture. So you can use these numbers as a guide, but only when comparing them to similar chips. These numbers aren't as helpful as they used to be, and they will only become less helpful in the future as CPU architecture makes some radical changes.
Somewhat true but the explaination is....lacking. A CPU clock cycle is not 1 complete calculation. A CPU's clock cycle can be defined in may ways. Think of it as a "pass". In the olden days, a CPU did only 1 thing at any given time. It could first fetch the instruction from memory, then decode it, then execute it, and then release it. Each would take several passes through the CPU (with each pass using different parts of the CPU) to do. Each "step" would be a single clock. At times, it would take somewhere around 100 clocks to finish a single instruction. Then came pipelining. Instead of the CPU only working on 1 thing at once, it would work on multiple things and using different parts of the CPU at once. Each time an instruction goes from one stage to another (until it gets to the very end) that is a "clock". Now, in modern hyperpipelined CPU's, a single step no longer does just 1 thing. Rather, multiple steps are used to do 1 thing (such as decode). This means that the "GHz" can be high when you divide the process up into many steps, but each step would do less. The processor would still be doing as much work per "clock" because although it is working on smaller parts, it is working on twice as many parts (in a 20-stage vs 10-stage, the 20-stage would do half as much stuff in each stage, but since there are 20 stages, can work on 20 things at once). The problem comes from data dependencies. In a 20-stage pipeline, the results from 1 instruction takes 20 cycles to complete and another instruction may depend on the results of the previous instruction. So you have to wait till that instruction finishes. This brings about idle-time in the processor where "MHz" is wasted. The CPU can predict how the instruction would turn out and start the instruction, but if the prediction was wrong, time was wasted and that instruction would have to be started again. Suffice it to say, this hurts the average work per "MHz" somewhat (although in the P4, a lot of other things hurt it as well), however the benefit would be that you can make the CPU run at a lot faster clockspeed (and since it does have the potential to produce the same IPC as a 10-stage design), and also make the processor as a whole run faster. All you have to do is make refinements to the core to optimize for data dependencies.

A: This is where architecture comes into play. Consider "Athlon" the last name of the processor, and "Thoroughbred" would be its first name. Athlon is the family of chips, and Thoroughbred would be a spawn of that family. As the chip changes slightly, the first name will change. The last name will only change when the company feels that the architecture has changed enough so that the chip is no longer anywhere close to its predecessor. When this happens, new chipsets are often required and completely new motherboards are needed to support the processor. The Pentium families are similar, but rather than completely changing their last name they have been merely adding numbers. The Pentium 3 is of a different family of CPU's from the Pentium 4. Changes in architecture that result in a new first name for the chip are often the fsb and die size. Changes in the internal design of the chip itself will result in a new family.
Not quite. "Thoroughbred", "Palomino" and "Northwood" are all code-names. They're names the company uses internally when refering to a core. "Athlon" and "Pentium" are commercial names. They're the names the company uses when selling these chips. Two completely different cores can have the same commercial name (Palomino and Thoroughbred were both called Athlon and the upcomming Clawhammer will be called Athlon as well). There is no rule as to when a commercial name should be changed. It's just marketing.
 
Each time an instruction goes from one stage to another (until it gets to the very end) that is a "clock".
So you're saying that one of these stages is not in itself a complete calculation? They're designed to be complete in their own designation. I consider those stages to be calculations, because, well, that's what they are! Lots of little logic gates arranged to perform multitudes of operations/calculations as efficiently as possible. Of course other words can be used to explain the process. That's a nice big speal, but I was aiming for some simplicity, being both FAQ and that for beginners. It's a conceptual thing, not a step by step explaination for your Thesis (though even for that, your explaination is lacking about 20 more pages of technical details). And ya went a bit of topic, as well. But hey, it's a forum! We are allowed to nitpick, flame, rant, scribe, and yes, even troll...though many may not like them all.

As for the rules with the naming...you don't see the pattern so far? Sure, they could change their patterns, but have they? And the Clawhammer has not yet been released on the market, so they haven't "officially" chosen a name, eh? And from what they've been saying they're going to adobt a suffix or prefix, much like Intel, to distinguish it from the previous family of Athlons.

Anyway, I think it's a good explanation for those beginners out there wondering what the MHz really is in the PC's they're looking at. Your paragraph is...for a completely different question.

<font color=red>01010011010</font color=red>
 
Back in the days, when each pipeline stage was used to do one whole task, that may be true. But in modern hyper-pipelined MPUs, tasks are split between stages. On the P4, for example, the 20-stage integer pipeline means that the decoding stage is split into 2 or so stages (can't exactly remember) and so are the fetch and issuing stages. The only stage that does 1 complete thing is the execution stage and that's only 1 out of 20 stages in the integer pipeline. So no, they do not do 1 calculation. They don't even do calculations. Only 1 stage out of 20 in the integer pipeline does calculations, all the other stages are either issuing or fetching or decoding, etc. some are simply there just as relays (2 of them I think) to just take the signal and send it again.

As for the naming. Intel doesn't consider "Northwood" or "Willamette" suffixes or prefixes. Intel uses the "a" and "b" suffixes to indicate a different in models. "Northwood" and "Willamette" are internal code names for the core. They were never meant for commercial use. Of course, being the enthusiaste community, we get news a lot earlier than normal people and at the time we get them, only the internal code-names are available. Oh, and btw, if you look on AMD's roadmap, you'll see "Athlon" in there and below the name you'll see "Clawhammer based".<P ID="edit"><FONT SIZE=-1><EM>Edited by imgod2u on 07/10/02 02:38 PM.</EM></FONT></P>
 
AKAIK this forum has never has a sticky. But it REALLY REALLY needs some stickies. The 1st post would be a great one.
Good job Silicon.

Benchmarks are like sex, everybody loves doing it, everybody thinks they are good at it.