fsb, fsb clock, core clock...?

[shmeggegie]

well, I guess I just need someone to direct me to a site or page that explains all the terms in this post's subject heading. It'd be great if I could understand a paragraph like the following:

"Facts about the Pentium 4/2400: according to Intel's plans, this is will be the last CPU to have a 100 MHz FSB clock. Future P4 processors will run exclusively on an FSB clock of 133 MHz, which actually results in a slower core clock speed. After all, the multiplier lock will continue to exist. The Pentium 4/2533 with 133 MHz FSB will be launched at the beginning of May, with a fixed multiplier of 19.0. With respect to marketing, the faster system clock will even be sold with 533 MHz FSB, where the data is only transferred using the QDR process (rising and falling edges of the signal, double the amount with DDR)."

which is from the "The Final Battle:
P4/2400 vs. Athlon XP 2100+ (2400+)" article posted here. Actually, just about every term in there I need a definition for, and how they interact with one another. anyway, any help is greatly appreciated. I don't know much about hardware, but I aim to change that and knowing this would be a big help in that direction.

 

[shmeggegie]

p.s. - what's a troll? I ask because of the signature of the week. really hoping I'm not one.

 

[lhgpoobaa]

lol no
your a newbie

a troll is someone who is annoying, someone who who likes making trouble, is biased and cannot/willnot se any other points of view, and generally spreads crap whereever they post.


Anything i think of as 'Decent' is unlikely to ever become 'OEM'

 

[Guest]

ok,
The speed(mhz) of a processor is derived from two things. The fsb (front side bus) and the multiplier. The front side bus is the speed at which information is sent to the processor, the multiplier is the number of cycles a processor will do between each burst of info from the motherboard. Hence 133*19 = 2533mhz.

To make matters more confusing the fsb can be different for the processor and the memory. Untill recentlt they were the same. The "memory clock" can be different. In the cas of the sis645 motherboards the memory clock can be set at 100,133, or 166mhz DDR. The fsb can be set at 100 or 133mhz.

To be honest I don't really understand to mechanics of DDR(double data rate) ie 133mhz DDR is called 266mhz and uses pc2100 ram and QDR(quadruple data rate). I look at it as the ability to transfer more information. DDR usually is equivalent to between 10-30% more performance than SDR(single data rate) depending on the processor. Right now QDR is limited to Rambus Ram.

I can try to answere other question if you have secific ones.
Also many other members are more knowlegable than I.

May Fortune Favor The Foolish

 

[eden]

I too found this par to be very confusing, and still don't get the way the guy wrote it.
From my POV, the P4 2.4GHZ is the last of the 400MHZ FSB saga, moving on to the next 533MHZ FSB series. I dunno what the heck they meant about the multiplier, I assume since there is higher FSB, lower multiplier will be needed and again the possibilities for OCing are easier too. But I don't get what they meant by lower core clock!

--
For the first time, Hookers are hooked on Phonics!!

 

[chuck232]

Maybe they meant if you used that clock multiplier on a 100MHz QDR bus, it would be slower than the ones they have now..... I dunno maybe you have to read between the lines. I gotta admit these reviews and stuff are pretty technical and if you don't know your stuff, all you can really do is read the conclusion and hope that you made the right decision when it comes to that purchase.

 

[Guest]

Oh yea,
100mhz QDR is 400mhz according to intel.(133 QDR = 533)
Just like 133mhz DDR is 266mhz according to AMD.(and Intel)


May Fortune Favor The Foolish

 

[shmeggegie]

so then, the P4 fsb is (at the moment) 100 mhz qdr, coming out to roughly 400 mhz, right? To fully use this, though, you'd need qdr RAM, namely RAMBUS. AMD uses ddr at 100, and 133 mhz, comin out to 200, and 266 mhz. (Hence the names of the via chipsets, I gather.) Further, using ddr RAM on a p4 restricts what kind of performance you get from it (due, I'd guess, to bottlenecking.) so that you're not getting peak performance from it, whereas the 333 mhz chipsets and accompanying ddr-333 ram for amd chips aren't really providing you with the bonus you think they are because the chip only runs at 133 anyway (more bottlenecking issues, I think.) Am I totally off here, or do I have it about right?

Now, my next questions, to make sure I understand this. Where do these mhz numbers come into things? You know, P4 2400mhz. and AthlonXP 2100+ (I know the Athlon name is to say what P4 it is roughly equivalent to (or surpassing of).) and stuff like that. Is that just the fsb multiplied by the multiplier? further, what's the multiplier lock mentioned in the quote I posted above? How does that affect performance? These questions are largely directed at the guy who said he'd answer any more specific questions I may have had before, but anyone is welcome to answer, really. Thanks in advance.

 

[Quetzacoatl]

The future solution to RDRAM is possibly dual-channel DDR for the PIV. Chances are they'll be seen this summer. Too bad Nvidia released the dual-channel for AMD even though they can't use it yet!

"When there's a will, there's a way."

 

[AMD_Flinkster]

A locked multiplier is a multiplier that cannot be changed - i.e. it is built-in from the factory and cannot be altered (with ease anyway).

Intel Chips have had locked multipliers for a very long time, so there's no way of increasing the core clock (core clock=FSB*multiplier) other than to increase the FSB of the CPU. But with AMD Athlons , their locked multiplier can be "unlocked" so to speak by closing all the L1 bridges(little pins on the CPU itself exposed and not covered or anything) . Previously with Athlon Thunderbird (TB or Tbird) which had ceramic packaging, this could be easily dont by a pencil , because graphite conducts electricity. But with teh XPs/MPs , it can also be done but its a bit more complicated. Read about unlocking XPs/MPs in the CPU section.

Please bare in mind that overclocking the FSB not only results in higher core clocks, but increases the RAM , PCI(33mhz) , and AGP(66mhz) buses proportionally as well. Thats why you cant go very high with FSB overclocking except if the board supports individual clock speads. All this says that unlocked CPUs can attain higher overclocks usually. To get the maximum overclock, watercooling is used, along with increasing the CPU voltage (to be able to run higher clocks with stability), increasing the FSB, and increasing the multiplier(if it were and AMD CPU with an unlocked multiplier).

I know this seems lenghly, but I remember when I had the same questions and I kept wondering.

So there , take all the answers you want !

Any mystery devised by mortal mind can be solved therewith - Holmes

 

[shmeggegie]

Thank you, that was a great answer, and helped me very much. I have one more, if you don't mind, though. I was reminded of it when you wrote "AthlonXP/MP." I read somewhere that you can put two AthlonXPs in a dualie AMD board. If this is the case, what's the difference between them and the MPs? Is there a reason to get AthlonMPs, and not just two XPs?

 

[AMD_Flinkster]

Athlon MPs were designed to work in SMP (symmetric dual processing). Athlon xps might do in some cases on some motherboards, but it neither officially supported nor endorsed by AMD.

Any mystery devised by mortal mind can be solved therewith

 

[bikeman]

To be honest I don't really understand to mechanics of DDR(double data rate) ie 133mhz DDR is called 266mhz and uses pc2100 ram and QDR(quadruple data rate). I look at it as the ability to transfer more information. DDR usually is equivalent to between 10-30% more performance than SDR(single data rate) depending on the processor. Right now QDR is limited to Rambus Ram.

Explaining DDR is not that difficult: In most cases, data transfers from one chip to another are done with a certain timing. For that timing to be equal to both chips, one of them, or maybe aven a third chips provides the other(s) with a clock signal. This signal consists of a block signal, which could be interpreted as a series of 0-s and 1-s : 010101010101010101... In a regular system, the sending chips sends it's data to the other when, for example, the signal changes from 0 to 1. A 133 MHz clock signal would imply 133 million data transfers per second. A DDR-system doubles the data transfer rate by not only sending data on transitions from 0 to 1, but also when a 1 becomes a zero. A 133 MHz signal gives 266 million data transfers per second.

QDR buses are harder to explain. Especially since I only learned how they work recently, in <A HREF="http://forumz.tomshardware.com/hardware/modules.php?name=Forums&file=viewtopic&p=507501#507501" target="_new">this</A> thread somewhere else on this forum, a little below half of the posts. I (bikeman) ask a question to Girish, which then replies quite elaborately. If you understand what he sais, you get how the P4's QDR-bus works. I had to read it twice, too. And I summarised in order to check wether I got it right.
Anyway, according to what I looked up on the internet (I love Google), there is a different type of QDR bus. That type is used in the AGP 4x-bus. There, both the chips generate their own internal clock signal, based on the 66 (or 33?) MHz PCI clock, by multiplying it by 4. They keep it in sync with the original, 'low' frequency signal, so they don't mess up the data. AGP 8x, which is about to be used in videocards, does the same, but rises the speed another two times.

I hope this is not a too messy post, and if anybody doesn't get what I posted, or if I made mistakes, please tell me so! I'm eagre to learn, and eagre to correct myself by learning more! (Is eagre actually an English word?)

Greetz,
Bikeman

<i>Then again, that's just my opinion</i>

 

[endless_n00b]

Please bare in mind that overclocking the FSB not only results in higher core clocks, but increases the RAM , PCI(33mhz) , and AGP(66mhz) buses proportionally as well. Thats why you cant go very high with FSB overclocking except if the board supports individual clock speads.

This is a very good point that doesn't get enough attention IMO. I just read about someone that took his 1.6A to 2.4 and it posted, but wasn't stable because of a PCI card. Basically he couldn't take it any higher than 2.1 because of his sound card. Think about that. I was like "WTF?!" when I read that. What does sound have to do with overclocking? But your explanation above about the FSB affecting the RAM, PCI, and AGP buses explains it. My question though, is was it really his pci card or his mobo that was the cause? Put simply, are there any pci cards I should AVOID!?


----
A newbie is only a newbie for as long as you allow him to be.
-Anonymous Veteran

 

[Willamette_sucks]

Dear Biliam82, DDR ram is not 10-30% more effective, it is 100% more effective, and this is easy to understand, because it is a fairly simple idea. Normal SD ram only recieves/sends/stores, one time per clock. 133MHz means it does this 133 time per second. DDR266 runs at only 133MHZ, but it recieves/stores/sends 2 times per clock cycle, making the EFFECTIVE rate 266MHz. It is 100% faster than normal SD ram. Basically the ram just does stuff on the rising AND falling ends of the clock cycle, instead of just in the beginning. And QDR, i believe, is not used with RAMBUS RAM, that actually runs at 800MHz(someone check me on this), QDR is used with the FSB speed of the pentium four. 100mhz x 4 = 400 mhz.

The first LAN I went to was at a PETA convention. They booted me after I shot a zombie in HL DM!

 

[FatBurger]

RDRAM runs at 400MHz DDR, so an effective 800MHz.

<font color=blue>If you don't buy Windows, then the terrorists have already won!</font color=blue> - Microsoft