I've got a question about why I when to adjust the FSB speed vs adjusting the CPU multiplier when overclocking.
I'm clear on what the two settings do to the hardware, but I'm not sure which adjustment would be better for performance.
I've noticed a lot of people (I'm referring to E6600s here) drop their CPU multipliers to hit high CPU OCs, but it seems to me that the extra FSB speed required to compensate for the decreased CPU multiplier would simply put extra stress on the mobo and possibly the memory.
Now I understand that if my FSB is 266 and I'm rocking some sweet low latency PC-6400 DDR2 I'll need to make some adjustments to either my FSB or my memory multiplier to get my memory running at its rated 800mhz and get the full bandwidth that I paid for.
But my noob-senses are telling me that E6600 at a 9x stock multiplier cranked to 400-fsb is the ideal setting as I can run my memory 1:1 at its lowest timings, whereas if I drop my multiplier to 7x I'd have to run @ 514fsb and then have to lock my memory at a lower speed or potentially run looser timings / put more stress on the mobo. (do high FSBs put more stress on the NB/mobo? or is CPU clock the more important variable)
Anyways, I'm sure I could figure it all out when my new system arrives but thought I'd farm it out for an expert opinion first.
I'm trying to wrap my head around this OC thing so bear with me.
People drop CPU multipliers so they can increase the FSB and get more bandwidth. To run PC2-6400 at rated speed you need 400 FSB but it isn't easy to get the e6600 to 9x400=3.6GHz. You'd need a good cooler for that. So most people drop their multiplier to 8 so they can run the DDR2 at full speed and the CPU at a fair 3.2GHz.
So yeah to get the most out of your memory you're best to drop the CPU multiplier and raise the FSB or just tighten the timings.
I chose to keep the CPU multiplier at 9 and run the memory below its rated 800MHz. Because it's going slower than specs, I could tighten the timings. This gives a little less bandwidth than running the memory at 800, but most regular tasks won't need the more bandwidth anyway so it isn't a sacrifice.
If u get a board with the 650i or 680i chipset they support setting the fsb and memory frequencies idependently. I have an E6400 @ 3.0, 375 x 8, and 2x1GB OCZ platinum @ 800 MHz 4-4-4-12 1T on MSI P6N-SLI platinum. Haven't tried to oc the memory yet.
actually, 650i and 680i boards do not set the memory and fsb clocks independantly, they just use a very very large variety of fancy dividers and multipliers to get a very close speed. the only board (and chipset) i know of that does allow true memory independence is the dfi lanparty x3200 (rd600), this is a great board but is quite complex for a beginner, also making the FSB higher give good theoretical performance increases it actually contributes rather little to the speed of the system over all. personally i think i would only go to FSB-1600 (400mhz).
FSB overclocking is the most common method of overclocking, as almost every
CPU allows you to adjust this value. As mentioned in the section above, the Intel
Core 2 Extreme X6800 is the only processor model that you can increase the
multiplier value. This means that the only way to overclock a lower Core 2 Duo
model’s clock frequency is through the FSB. The following procedure explains how
to do this:
1. Enter the BIOS by restarting your computer and pressing DEL during POST.
2. Next navigate to Advanced Chipset Features FSB & Memory Config
3. Go to the FSB – Memory Clock Mode and you see three options:
a) Auto – this is the default setting to let the BIOS determine the FSB and
MEM clock values
b) Linked – this mode allows you to change only the FSB bus speed; the
MEM bus speed adjusts accordingly.
c) Unlinked – this mode allows you to independently change the FSB and
MEM bus speeds.
Also to keep in mind with 965 chipset, 400mhz is often unstable because it is the limit of the memory 1066 memory strap.
And also dropping the multiplier actually OC the northbridge further.
The FSB is a bi-directional physical data bus. The processor communicates with the rest of the computer through that data bus. Because of this it is ideal to have the highest possible FSB with the lowest possible muliplier because the higher the FSB frequency, the wider the bandwidth and the faster it can communcate with the rest of the computer. However, I have tried both raising the multiplier to lower the FSB, and lowering the multiplier to raise the FSB, and have seen little, if any, difference in scores on benchmarks such as 3DMark and PCMark. Where you do see a gain in performance is not neccesarilly in gaming, but in synthetics. A higher FSB will significantly raise your score in benchmarks such as SiSoft Sandra, Everest, and encodings.
Here's a full analysis by our very own Graysky, take a read. The basic conclusion: won't make much real difference. Therefore, a higher multi is better because you'll need lower FSB, and therefore lower volts, which then means lower temps, etc.
Take a read, he posted in a number of other forum sites too but I'm too lazy to track them down.....but they all contain the same info anyways:
Proof is right on here. 3 years ago, raising the FSB made much bigger improvements than just upping the multiplier. However, with todays FSB speeds much higher and being double and quad pumped already, the increased bandwidth does not seem to net you near the gains it did a few years back when stock buss speeds were 66, 100, or 200mhz.