Q6600 O/C question - Is 400mhz FSB x8 CPU multi or 333mhz FSB x9 CPU multi better?

Which is the better overclock? Has anyone compared them in benchmarks?

I'm getting my Q6600 system this week and am reading up on my options.

I've found that 3.0-3.2 is a sweet spot for air overclocking. I've got PC-6400 RAM (800mhz) but at first I'm going to try to get away with the Intel stock cooler. My case has tons of cooling on it. 4x80mm, and 1x120mm + another 120mm in the PSU.

Let us know about the stock cooling in a couple of weeks, I ran a quad at 3.2 on stock for a couple of days but did not have time to really watch and monitor the system with heavy loads before having to turn it down to 2.8 and give it to the customer.

Ok, no noone in the whole world with a computer has ever thought of that or done the benchmarking.,


It depends, mainly on the bios and the MCH.

Old theory would make the 8x 400 1:1 the correct answer without even needing to benchmark because all the system timings are in synch eliminating wait states for memory access etc.

However now-a-days Intel has done some interesting things with system memory access and also the MCH will change system memory access latencies based on the FSB. Basically it "changes gears" to allow high FSB speeds without croaking.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3239

So to be sure you have to run the benchmarks yourself, or you can read this and do what it says. I am fairly certain the information is correct as MCH "straps" are nothing too awful new and I understand basically what is going on. But for some reason when I try to make sense of the article the way the information is presented confuses me. Thats just me, or the 2 pots of coffee. Anyway the article explains the REASON that 9 x 333 may be better because the MCH will use tighter latencey settings (Trd) because the MCH is not as stressed by a high FSB

There are a lot of free memory benchmark programs but it is probally best to just run superpi with a fairly large setting for the number of places 1M. and compare the results at both settings and you will quickly see if there are any big difference between the two settings.

Also download Memset and see what your Trd is.

Finally got everything hooked up.

I can run at 3.0ghz with 333mhz FSB and 667mhz RAM using a 1:1

or I can run at 3.0ghz with 333mhz FSB and 800mhz RAM using a 1.2


But for the life of me I can't get anywhere near 400mhz FSB to test it. My CPU overclocks and runs stable all night long on prime 95 with default voltage at 3.0, but I can't get it to boot up at 400mhz fsb even if I drop the CPU speed all the way to the lowest x6 multiplier for testing purposes.

Guess I won't be the one to input the answer here...... ;(

9x 333mhz is better as you are not stressing the rest of the system.

That's most likely because your Northbridge might need more voltage. I know mine does in order to get the magical 400fsb. Try that. It's the MCH setting.

"magical 400 fsb" huh? What chipset are you talking about? see sig.

My Q6600 is running at 460x8 right now.

My board hits a wall at 480. Seems to be a bios timing issue. Maybe lack of a higher bootstrap from what my experimenting has shown.

I just wish I could get 520 fsb so my RAM would be running at just about its max completely stable speed.

And for your answer. 400x8 1:1 memory divider is gonna give you a lot higher memory throughput than the 333x9 will. Upping the memory speed above the fsb speed brings almost no gain whatsoever on an Intel setup.

Some people increase MCH, ICH, VTT, and DDR voltage together to squeeze more out of the FSB. Not sure why it helps, but it does for some, although logically MCH should be the most important one. I bumped my MCH 2 notches from default, ICH +1, VTT +1 to get my best final settings. Also try turning off all power saving, speed reduction, etc. settings if you want to run lower multi with higher FSB. For some reason on my IP35-E when I leave on C1E, EIST, etc, even when I reduce the multiplier it FORCES it back to default max multiplier so I was not able to go 8X400 with those turned on since it forced it back to 9X400.

I highly doubt 8x400 is slower than 9x333 in most apps however, even with differences in system memory latencies - you're giving up memory bandwidth and 200Mhz raw CPU speed. A more even comparison would be 9X355 vs 8X400 since at least the CPU speed would remain constant and potentially you would be trading off memory bandwidth with system memory latency. At least for encoding, I can almost guarantee the raw CPU speed 8x400 will beat 9x333.

I ended up going for 9X370 with the speed/voltage reduction since it didn't make any visible difference from 8X400 with some prelim benchmarks I did that I cannot recall, but the main things I do are video encoding so I remember testing it with Cinebench and some Xvid encoding, which may be completely irrelevant to you. Using a higher multiplier and lower FSB also saves a tiny bit of energy since most of the time when the CPU isn't being heavily loaded it will run at 6X333 and a lower voltage. (3ghz to 2ghz, vs 3.2 to 2.4 - 33% decrease vs 25% decrease although this is prolly only a few cents a day difference

skizzled said:

That's most likely because your Northbridge might need more voltage. I know mine does in order to get the magical 400fsb. Try that. It's the MCH setting.

Click to expand...

I haven't messed with any of the voltages outside of RAM and CPU and I didn't increase the CPU over 1.24...

I might try the MCH setting...what's the safe range on all of these???

Overclocking the Northbridge is a new concept to me - I'm not new to overclocking ---tried it way back in the pentium 100 to 133 days, then failed on my k6-2 450mhz (lousy chip wouldn't overclock even 33mhz) , then o/ced my duron 600 to 900, then my tbird, then my p4 1.6, then my p4 2.4, then my p4 3.0, then my p4 3.2, then my be-2400, now my q600

This is the first generation I've heard of overclocking the northbridge though

I just got done playing with it a bit.

I CAN get to 400mhz FSB on your advice -- but I have to increase my MCH a LOT....
from 1.25 to 1.81 to get to 400mhz without a prime95 error within a couple minutes( I can finally see windows load about 1.6), and though I've not tested I suspect the 1.81 mch voltage might need to be upped another couple notches to get to an overnight stable prime95 torture test.

it only goes to 2.x ??? is 1.81 too high for the MCH to leave running indefinately? I brought it back down for the night as I'm not in the mood to damage any of my components for a few extra percentage points of speed. The northbridge is too hot to touch for sure.

My CPU seems to need about 1.3 volts at 3.2, but it'll run at the stock 1.2 volts at 3.0

Sorta seems like my setup has a happy 3.0, but doesn't want to do much after that.

BTW I did have to disable the EIST function, but not the other - I think the EIST function is the one that controls the speedstepping and thus the rebellious jump to a x9 multiplier. The other one just lowers the voltage when it can to save power.

I ran future mark 06 once at 3.0ghz FSB 333, 1:1 ram, RAM = 667 --- got 12026 marks

Ran future mark 06 once at 3.0ghz FSB 333, 1:2 ram, RAM = 800 --- got 11977 marks



So I guess the answer is keep your RAM at 1:1 unless it benefits you to overclock your processor.

I need better cooling to be able to overclock my processor so I guess I'll just stay at 3.0 with a 1:1 RAM divider, and underclock my 6GB (2x1 & 2x2) of PC6400 RAM

Archaea said:

I ran future mark 06 once at 3.0ghz FSB 333, 1:1 ram, RAM = 667 --- got 12026 marks

Ran future mark 06 once at 3.0ghz FSB 333, 1:2 ram, RAM = 800 --- got 11977 marks



So I guess the answer is keep your RAM at 1:1 unless it benefits you to overclock your processor.

I need better cooling to be able to overclock my processor so I guess I'll just stay at 3.0 with a 1:1 RAM divider, and underclock my 6GB (2x1 & 2x2) of PC6400 RAM

Click to expand...

BTW - I tried 1:25 on the RAM divider but windows was unstable with default voltages everywhere except RAM set to 2.1volts. My RAM is Corsair XPS and Patriot PC 6400. Timings are set to 5-5-5-12