I was just curious, you know how when you overclock, there is the FSB and the multiplier....
Physically what is the multiplier?
Printable View
I was just curious, you know how when you overclock, there is the FSB and the multiplier....
Physically what is the multiplier?
At a basic level you take the FSB speed of your CPU and times it by the multiplier built into the CPU. That gives you your CPU speed. It's how CPU makers effectively make their different CPUs. All but the most expensive CPU's are what's call multiplier up locked - you can lower the multiplier, but not raise it.
This can still be good as if you have blazing fast RAM and a motherboard that supports high FSB's you can raise the FSB and memory speed, lower the multiplier and leave the CPU at the same speed. Of course, if you can raise the multiplier of your CPU that makes overclocking a lot easier as you can leave the RAM and FSB at default speeds and just raise the multiplier to get a quicker chip.
so would a high FSB and a low multiplier be better than an equally clocked CPU, but with a lower FSB and a higher multiplier?
Not necessarily, it depends I think. With a low multiplier, it's you can bump up the FSB to increase the speed of your RAM, whilst kind of keeping the overall processer speed the same (if it doesn't like being overclocked).
On the other hand, if your ram is the limiting factor, you could raise the multiplier, so even though the FSB would be the same and the speed to the RAM is the same so it's not being overworked, the cpu would work that much faster.
However, if the cpu is say 3ghz, it doesn't matter whether that's all FSB or multiplier at the end of the day, because it still has 3 million cycles per second. I think...
I'm just trying to articulate better what Kiro-San said :p
Well although you're right when you say the CPU speed is still the same, because you have a faster FSB and faster RAM you'll actually get a bump in performance. It won't be massive but it'll be there.
ok kool, thanks
The Front Side Bus is the link between the CPU and the northbridge chip, pretty much the same as AMDs Hypertransport Bus, except Intel has opted to keep the memory controller on the northbridge chip as part of a more flexible, modular design.
How it works is, the FSB frequency speed is the rate at which the northbridge communicates with the CPU, the CPU frequency is set by that times it's multiplier setting. These days the memory clock rate is set independently of the FSB, but you always need the FSB to be at least as high as the memory clock for more balanced performance, which should be obvious since the CPU needs to work with other hardware devices via the northbridge chip (which in turn needs to communicate with the southbridge also).
This is why you find overclockers often talking about 'low multipler' stuff, but this only applies if you have seriously silly high performance RAM, but you can get a part with a fairly high multiplier and lower the multiplier via the BIOS if you find the CPU is starting to crack before the northbridge, although this is highly unlikely with the Core 2 Duo, they scale very very well. If I were to get a Core 2 Duo today, I'd still go for the E6600, the multiplier is fairly low, very reasonable cost, lots and lots of fine cache. Although waiting till march for the prices to come down would be wise.
Intel has done a bit of a sneeky by disabling the multiplier lock to allow an increase greater than the default on the Extreme Edition chips so you can get insane overclocks if the northbridge can hack it, even though technically there's no difference between the EE chips and the normal chips.
so why do people with an X6800 just max out the multiplier?? on my EVGA 680i, the max is 60 so they would get 266*60=15.02GHz...
BTW also, what does the "quad-pumped" (266*4) part mean?? better performance?? the new Intel C2D E6x50's are gonna have 1333FSB, (266*5), so will they perform better??
Sorry about all these questions, I just built my new rig (old one was a 1.4 GHz AMD Athlon XP 1600+ with GeForce 2MX 400(:mrgreen:) and im trying to get "in the knowledge"
shout :stop: if im being to question-ey :)
The reason people don't max out the multiplier is that your CPU would be spending all it's time waiting for memory I/O, it's a big balancing act between CPU speed, stability, Northbridge speed and stability, and memory speed and stability.
Intel introduced the "Quad-pumped" FSB back in the early P4 days, basically it's the same idea as DDR memory, but twice as fast.
That could be confusing if you're talking about 'real FSB' speeds :pQuote:
Originally Posted by aidanjt
FSB is an outdated and misleading term anyway, I'll refer to it as CPU clock in my post. The CPU clock is pretty much the pulse at which the system communicates. On modern Intel systems this is 266mhz, and on my 939 socket AMD it's 200mhz. Rather than work at the same speed as this pulse, CPUs can work at their own rate, and cycle say 10 times per 'pulse', thus my X2 3800 is actually a 2ghz chip, with a 10x multiplier on a CPU clock of 200mhz.
RAM can also work at the same time as this pulse, and 1:1 ratio RAM runs at 266mhz for Intel, and 200mhz for AMD. Now because DDR RAM transfers more information per pulse, double the amount in fact, manufacturers give it an 'effective' DDR speed of double the real speed.
The link between the RAM and the CPU is one of the most important communications for performance, so intel 'quad pump' their FSB, similarly to DDR's doubling, giving an 'effective' 1066mhz for the core 2 duo. (AMD use hypertransport, which is a bi-directional, multi-channel bus usually running at 1000mhz).
Anyway, you don't have to run your RAM at 1:1 speeds - AMD system for a long time would run the CPU clock faster than the RAM, so I could overclock my CPU clock to 250mhz, which would give my processor speed as 10x250 = 2.5ghz, but if I ran a 5:4 CPU/RAM divider I could keep my RAM tight and happy at 200mhz.
On Intels you tend to go the other way, so say run your CPU clock at 266mhz and run your ram at 333mhz, on a 4:5 divider. Note that in this case you are really running your RAM faster than the 'FSB', hence my objection to the above quote :p
Because the CPU just can't finish things fast enough. If you have a slow enough CPU clock you might have time to get in more cycles before your next 'tick' from the system, but as the CPU clock increases, you have less and less time in which to complete your task before you need to accept a new task.Quote:
Originally Posted by DevilMayCry42
Pentium 4's used a trick to specifically enable a very high multiplier - put very simply they had a really long pipeline of instructions which passed around the task so that they could quickly take on a new one while still processing the old one - think of it like passing on hot potatoes :p Unfortunately while this did enable really high multipliers and resultingly high ghz speeds (like 4ghz etc.) the amount of work that actually got done per processor cycle was really small - it had only got a small way along the task in a single cycle.
AMD proved it was better to stick with the older system of having doing less cycles, but being able to do more work in each cycles.
Intel learned their lesson, and the Core 2 Duo does even more work again per cycle, hence why a 1.86ghz Core 2 Duo thrashes an AMD at 2.0ghz.
Yes, but not for the reason you think. The new C2D's actually have a CPU clock of 333mhz, so when you 'quad pump' that, you get 4x333=1333mhz FSB. The faster CPU clock means that your overall speed, at the same multiplier, is higher. In fact your RAM will also be faster, if you are sticking to a 1:1 ratio, so there's a boost all around.Quote:
BTW also, what does the "quad-pumped" (266*4) part mean?? better performance?? the new Intel C2D E6x50's are gonna have 1333FSB, (266*5), so will they perform better??
Oh, and you can never be too question-ey, just forgive us if our answers are based on personal understanding at the time and aren't always accurate :D
Yea, do ask questions, you're not the only one that's learning ;).
As far as my overclocking knowledge goes, I left everything to [Auto] in the bios, except for the FSB speed which I gradually inched up to 400mhz, but anymore it doesn't seem to like it (will start at low settings). I think 1.83 to 2.8 is respectable enough without pumping extra voltage in though :). Also, by some miracle it means I can set my ram to run at exactly 800mhz, perfect! :D
so basically, I Shouldn't be bothered if my DDR2 800MHz kinston ram is being reported as 392 MHz in CPU-Z?
BTW i know is probably the wrong place, but does anyone know why CPU-Z reports my E6600 with a varying multiplier?? one minute it's at 9x, the next its at 6x, so the speed drops from 2400MHz to 1600MHz....? im gonna post this in the help section aswell, but i just thought i wud ask you.....
I think it's something called speedstep. If you're only using 10% processor usage to save electricity and the environment :p it lowers the multiplier.
what about the RAM speeds?? they're reported as 400MHz... is that correct, because of the whole dual data rate thingy???
All your stuff on the P4's weakness was quite correct, however the above statement isn't. What I'm about to say next is probably fairly well know but bear with me, it helps setup my point.Quote:
Originally Posted by kalniel
CPU Performance = IPC x clock rate (CPUs Ghz rating)
So if you have a CPU with a nice high IPC value and you bump the Ghz higher that's all good, you get a better performing CPU.
Your statement however doesn't make sense really. IPC is how many calculations your CPU can perform per clock cycle. Clock rate is how quick those instructions get done. The disadvantage of ramping the clock rate is you need more power and therefore you generate more heat, ala those toasty warm P4s. It's two different ways to reach the same goal.
IPC Way = Do more things, but do them slower.
Clock rate way = Do less things, but do them quicker.
This was the historic difference between Intel and AMD. Of course Intel has done a complete 180 now and gone AMDs way but even more extremely than AMD ever did.
The real reason people don't max the multiplier to say, 60, is because there is no CPU on the planet that will do 15Ghz+. Sure, you can get individual transistors that run crazy fast (Intel demonstrated 10Ghz transistors a while back). The other reason is that you get a better performing system when you have your FSB and RAM running quicker along side your nippy CPU. That's why a system overclocked the FSB/RAM way will be quicker than a system overclocked to the same CPU speed, but done the multiplier way. Of course, no everyone has a board that can do high FSB or memory that can do high speeds, but if you have a CPU that is multiplier up unlocked you should really pair it with decent components in the first place.