Refresher

[freesoft 2000]

Hi everyone,

I have some rather ridicalous questions so bear with me for a while.

The last time i did overclocking was five years ago was when overclocking was simply configuring the jumpers on the motherboard and had nothing to do with the bios. Well now they all have to do with bioses for overclocking.

Here is my question for example for a 3.2Ghz intel pentium 4 chip with fsb 800mhz the fsb parameter in the bios is set to 200mhz thus using the below formula

200mhz times multiplier = cpu speed

If i were to use a 1.8Ghz amd chip with the fsb of the chip 266mhz the fsb parameter in the bios is set to 133mhz thus using the below formula

133mhz times multiplier = cpu speed

Basically my question is why is that for the intel chip to be rightfully configured the internal bus speed of the intel chip has to be divided by 4 and that of amd has to be divided by 2. Why is this so or am i missing something.

For example see the below article

Understanding System Memory and CPU Speeds - A layman's guide to the Front Side Bus (FSB)-Best Computer Online Store Houston Buy Discount Prices Texas-Directron.com

This is how they calculated the system clock

800MHz (AMD64): 200MHz clock
1066MHz (Pentium 4/LGA775): 266MHz clock

It seems that both of them are also divided 4, i thought that usually because AMD does only double pumping, thus it should only be divided by 2 or have i misunderstood it somehow?

Also if using AMD sempron, it's architecture which i assume is also double pumping that means that i divide by 2 and not 4. Am i right or have i misunderstood that also?

A really detailed explanation would really be helpful

Any help is greatly appreciated

Thank You

Yours Sincerely

Richard West

 

[zeus]

Ive had to read your post a few times because I didnt really get what you didnt understand.

I think the main reason overclocking has moved to BIOS is so you can adjust the FSB 1mhz at a time. You used to be able to do it in windows with those little programs so why not in BIOS. No confusion as to which clock generator your using then either. Manufacturers have added things like crash free BIOS too which helps reassure yourself when you start pushing your luck! It would be nice to be able to have both but I only saw that on a board which had dip switches just as skt478 came out.

The reason as to why some AMD chips have a x2 multiplier and Intel has the x4 is because they are different design chips, from different manufacturers who had different directions for the chip in general.
Intel got it into thier heads that a fast clock speed was the way to go (more the intel marketing team I think) AMD designed their chips so that clock speed wasnt everything and managed to get slower clocked chips to perform better than the equivalent pentium 4's.
It was the K7 range of AMD cpus which was in competition with the early P4's and they chose to use only a double data rate transfer with the northbridge. All in all the K7s beat the P4s outright. Intel hoped for 10ghz out of the netburst design which they must have assumed would out perform whatever AMD could offer. I think they got 3.4ghz or 3.6ghz something like that. Then came CPUs like the Extreme Edition, nicknamed the emergency edition because the P4 was lost!

Given that it was apparently the marketing teams from intel that pushed the designers to make fast chips I reckon it was because they played on many peoples lack of understanding of how they work. So many people would walk into the shops, see the PCs with intel chips had far bigger numbers and so "lets buy that one". AMD kinda gave it back by giving their chips names... the XP2000+, which is 1.6ghz. Its better to see "2000" than "1600" when looking for a fast PC!

AMD K7s maxed out at 2.2ghz!! It was the XP3200+ and it had a 200mhz (x2 so 400mhz) FSB. By which time intel also had 200mhz FSBs still with the x4 multiplier. As its all about communication with the northbridge that also had to match these pumped FSBs. No doubt through collaboration, there weren't any chipsets which support more than a double pumped FSB. The "pumping" is the same as DDR RAM and its common knowledge you cant use DDR on a motherboard which only supports SDR (the different plugs kinda give it away but its essentially the same SD ram)

Intel have stuck to the x4 multiplier and have increased the the FSB up to the point where its now at 266mhz. The K8 AMD processors (Athlon 64s and Semprons) have had the multplier bumped up to x4... its just gone with the times.
To be honest its a bit different with the K8 AMDs. They dont actually have a FSB because there is no memory controller to transfer to. The memory controller of the northbridge is now built into the CPU itself and the CPU corresponds with the memory controller via HTT (hyper transport) You can get AMD cpus with a "fsb" of 2000mhz!

 

[freesoft 2000]

Hi everyone,

Ive had to read your post a few times because I didnt really get what you didnt understand.

Sorry about that

To be honest its a bit different with the K8 AMDs. They dont actually have a FSB because there is no northbridge to transfer to.

In the case for the AMD K8s' then how would one calculate the processor speed? Does that mean that now all k8s' are also divided by 4 because they are all do quad pumping like intel chips?

Any help is greatly appreciated

Thank You

Yours Sincerely

Richard West

 

[freesoft 2000]

Hi everyone,

No one knows this??

Richard West

 

[zeus]

I worded my first post wrong.... The K8s DO have a FSB it just that the memory controller doesnt use the FSB because its built into the cpu.
Most motherboards will set it out so that it will look the same as a normal cpu, memory controller and ram setup.

In a normal system theres a PLL Clock generator (which will produce the clock speed the FSB uses) which supplies a frequency to the northbridge.
The northbridge will then distribute the frequency around the PC. One of the places the frequency will be given is the memory controller. The memory controller then administers data between the ram and the cpu. It just so happens that the memory controller is on the northbridge.

Think of the memory controller being in between the cpu and ram.

RAM----MEM CONTROLLER----CPU
............./NORTHBRIDGE
......................|
......................|
.................CLOCK GEN

The frequency from the clock generator to the northbridge is the small clock number (ie 100mhz, 133mhz 200mhz etc) Then between the RAM and Mem Controller you have features like DDR (Double Data Rate x2) and also between the CPU and Mem Controller you can get QDR (Quad Data Rate x4) hence RAM running at 400mhz and CPU FSB 800mhz all from a 200mhz FSB. 200x2 for the RAM and 200x4 for the CPU. As you probably know the CPU frequency is the frequency from the clock generator combined with the CPU multiplire. The large FSB is just the speed the CPU corresponds with the northbridge.

The AMD K8s are different because the memory controller isnt on the northbridge, its on the CPU. There is still a northbridge because the northbridge on a traditional setup controls the AGP bus and also distributes the clock from the PLL Clock Generator to the southbridge (which controls the PCI bus, USB, Onboard sound, LAN, ATA inteface etc.)

An AMD k8 setup may look like....

RAM----CPU/Mem Controller
............HTT
..............|
..............|
............NORTHBRIDGE-----Clock Gen


The northbridge will take the clock frequency to the HTT part of the CPU. Along this transfer the frequency can be multiplied by a number called the HTT mulitplier making it possible to have a 800mhz transfer from the Northbridge to the CPU as a whole. I think the 754 AMDs can have a x4 HTT multiplier giving 800mhz transfer and the 939 AMDs can have a x5 HTT multiplier giving 1000mhz trasfer. This is also DDR so it goes up to 1.6ghz and 2ghz Northbridge to HTT transfer. I dont know what multiplier the AM2 AMDs can have but I think the maximum HTT specified by the HTT consortium is 2.4ghz which would make it possibly a x6 multiplier. This is why people say the AMD K8s have a 2000mhz FSB.
There is then a muliplier on the CPU which acts upon the HTT frequency ie the 200mhz. (HTT is whatever the clock generator is, the multiplier only applies to the transfer between the HTT part of the CPU and the Northbridge) If you have a x12 mulitplier on the CPU you will have a 2.4ghz CPU. ie 200mhz x 12 = CPU. Hyper Transport (HTT) is always what the clock generator is, remember its only the transfer from the northbridge to the HTT which is multiplied.

Its because the clock generator is still the source which is multiplied by the CPU multiplier that motherboards can still have the traditional options in the BIOS.
The hardest bit to grasp with HTT is the memory speed.

The memory controller is running at the same speed as the CPU frequency, so in this example so far its running at 2400mhz. Data between the memory controller and CPU is also 2400mhz. But because the memory controller is running at 2400mhz the RAM will be also... not to mention its gonig to be DDR so thats 4800mhz! Which is silly for RAM!
Ignoring the DDR bit (so our RAM=Mem Controller speed, 2400mhz) there is then a multiplier applied to trasfer rate between the RAM and memory controller. I think at this point its more a ratio because we are going to need to divide the RAM speed by a number.
The ratio is a ratio of the CPU multiplier...
So if we were to use a 1:1 mem ratio the RAM would run at 200mhz. 200mhz because we are using a x12 CPU muliplier so the sum will be 2400mhz divided by a 1:1 ratio of 12... which = 12 so the ram will run at 2400mhz/12 = 200mhz. Being DDR2 the final RAM speed will be 800mhz. 800mhz because its DDR so thats a x2 multiplier but also because two cells on the RAM can be used at once so this effectively double the DDR speed, hence 800mhz.

But what if we had say DDR2 667 and not DDR2 800.
The ratio of the CPU multiplier will not be 1:1 anymore. A 4:5 ratio will be better because a 4:5 ratio of 12 (the x12 CPU multiplier) is 15 (12/4 x5 =15)
So from the 2400mhz memory controller speed or CPU speed whichever you see easiest to grasp, divided by 15 is 160mhz.
Again its DDR so that makes 320mhz and its also going to be DDR2 so that makes 640mhz. Hence DDR667 isnt actually 667mhz when you have a 2400mhz CPU. Only the 3ghz AMD K8s will allow 666.68mhz RAM frequency.


In summary you find the CPU speed from the clock generator multiplied by the CPU multiplier... just like its been since the Intel 4004! Its these multipliers which make life difficult!
Or its the HTT transfer speed divided by the HTT multiplier and multiplied by the CPU multiiplier.

You find the RAM speed by multipling a ratio of the CPU muliplier by the clock generator frequency.
And instead of having a FSB transfer data between the Memory Controller to the CPU it is done by HTT at the same frequency as the CPU frequency.


Though in the past it was possible to say you find the CPU from the FSB multiplied by the CPU multiplier. Remembering the FSB was the transfer from the Memory Controller to the CPU and this isnt really a bus by traditional definition on the K8s because they are on the same chip. Dont get me wrong, it is a bus its just not a bus like a FSB. The transfer frequency is the same as the CPU frequency. Hyper Transport isnt new, its been around for ages and AMD have only just jumped on board. Intel haven't yet.
HyperTransport - Wikipedia, the free encyclopedia

 

[freesoft 2000]

Hi everyone,

200mhz because we are using a x12 CPU muliplier so the sum will be 2400mhz divided by a 1:1 ratio of 12... which = 12 so the ram will run at 2400mhz/12 = 200mhz. Being DDR2 the final RAM speed will be 800mhz. 800mhz because its DDR so thats a x2 multiplier

You mean 400Mhz right and not 800Mhz right?

In summary you find the CPU speed from the clock generator multiplied by the CPU multiplier[/QUOTE]

So basically correct me if i am wrong but this is the formula for processor speed

Processor speed = System clock * multiplier

So basically to find out the system clock using intel chips, i just divide the chip fsb by 4. Am i right on this?

And for AMD i divide the fsb by 2 to find out the system clock but this is only applicable to chips before AMD64. Am i right on this?

And for AMD64 i divide the fsb by 4 to find out the system clock just as the same as intel chips now. Am i right on this?

One more thing if i am using celeron chips, do i didved the fsb by 4 as well to find out the system clock as celeron are intel chips or am i wrong.
If i am wrong what is the correct way to do it?

Any help is greatly appreciated

Thank You

Yours Sincerely

Richard West