Elementary question regarding OCing.

NoobyDoo

Senior member
Nov 13, 2006
463
0
71
If a OC is stable with one FSB x multi, will it also be stable at the same speed with other FSB x Multi combos ?

For eg, I was playing around with a e5200 (M0) + Giga G31M-S2L + DDR2-800. I first tried 3 G with 240 x 12.5. Was OK, as expected. Then I tried 300 x 10, 333 x 9. Did not work. Infact the mobo did not even hold the BIOS settings ( they would revert to 200 x 10 & 200 x 9 ).

The memory was always set at 800 using the MIT memory multiplier.

EIST, C1E were enabled.

Why ?

 

Billb2

Diamond Member
Mar 25, 2005
3,035
70
86
There is a limit to how high the FSB can go on a particular motherboard.
Note that unless your memory is unlinked, you are also overclocking the memory when you change FSB.

In that case you don't know if you are hitting a FSB limit or a memory limit.
Standard practice is to overclock each component separately to learn the limit of each.
 

Udgnim

Diamond Member
Apr 16, 2008
3,662
104
106
your mobo probably can't handle the FSB increase to 300+

if you haven't done so, you can increase the NB voltage to try to hit higher FSB speeds. would recommend trying to google what NB voltages other people have used to OC your mobo with.
 

VirtualLarry

No Lifer
Aug 25, 2001
56,226
9,990
126
Just because your mobo handles N freqency, doesn't mean it will handle that freqency at a different FSB.

In my case, my mobo + CPU will handle 8x400, no problem. Stable for 24hours. But if I choose 7x400, I get freezes. But 8x350 is also fine. Very strange to me.
 

NoobyDoo

Senior member
Nov 13, 2006
463
0
71
The Giga G31M can "Supports Intel® Core? 2 multi-core and 45nm processors to FSB 1600 (O.C.) ".

The FSBs I used were not higher than 333.

 

Ben90

Platinum Member
Jun 14, 2009
2,866
3
0
Originally posted by: NoobyDoo
The Giga G31M can "Supports Intel® Core? 2 multi-core and 45nm processors to FSB 1600 (O.C.) ". [/i] ".

The e5200 can " Supports FSB 800"
and to answer the opening question... no
 

The Sauce

Diamond Member
Oct 31, 1999
4,738
34
91
I think that it has to do with straps. If that even applies anymore these days. Been a few years since my last build. But as I recall the northbridge would run reliably within certain FSB ranges and unreliably within others having to do with the strap settings. That may be the issue. Hence, you could overclock superbly at, say FSB 251 but not at all at 250 because the divider on the chipset would suddenly reset to a higher level. The trick was to know what the strap settings were for your mobo and use that info to play around with your settings.
 

Spikesoldier

Diamond Member
Oct 15, 2001
6,766
0
0
Originally posted by: VirtualLarry
Just because your mobo handles N freqency, doesn't mean it will handle that freqency at a different FSB.

In my case, my mobo + CPU will handle 8x400, no problem. Stable for 24hours. But if I choose 7x400, I get freezes. But 8x350 is also fine. Very strange to me.

Seconded. A e2160 on MSI P35 Neo2F/R I had for the longest would go from 1.8 to 3.0 (9x333) rock solid all day, and did that from day 1 until I replaced it with a q9550.

drop down to to 8x375 to get a better mem/fsb overclock, no post. this was with my first memory which was 2x1gb, even before I went 4x1gb.

the same mainboard is now overclocking a q9550 (8.5x441) with 4x2gb new memory. much more capable, but still rock solid than its predecessor.

To get that FSB pushing four sticks, I had to give the MCH 1.5v. Its uncharacteristic of me to push voltage to anything, especially to the NB, past an arbitrary 10%. I'm feeding it a 20% overvolt in this case, from 1.25 to 1.50. I'm sure I can only get away with this because of my cooling.

I think that the bootstrap theory is sort of correct, perhaps it has to do with the TRC memory timings, as the FSB scales, I believe this timing positively correlates with it.

I tested a e8400 on my abit ip35-e on a testbench me and a friend set up. even though the e8400 was a C0, we got it to 4.0ghz no problem. used very basic 2x1gb pc800 1.8v that looking back, seemed like diamonds in the rough, with a very sketchy power supply as well. I used the following rough, calculated, untweaked settings, and got a stable orthos run for over two days at 9x445.
mch:1.35
vtt:1.37
vcore:1.375
vdimm:1.9

I also tested the ip35-e with my e6420 after the e8400@4.0 run. got a max again, rough and stable oc of 3.4 with 2x1gb. when i throw my 4x1gb in there, it seems like 3.0 is the max i can get stable once more. im sure that the stress from running four sticks is holding the oc back.

keep in mind the e8400 is a 1333mhz fsb processor from the box, like my q9550 - and that the e2160 is a 800mhz chip like yours. I also have a 1066mhz chip (e6420), it tops out near the same wall as the e2160, I believe this was because they were early conroes, which are 65nm. The steppings are e2160/L2, e6420/B2, q9550/E0 (45nm).
 

Schmide

Diamond Member
Mar 7, 2002
5,581
712
126
Originally posted by: NoobyDoo
The memory was always set at 800 using the MIT memory multiplier.

I think if you set your memory with the MIT to ddr800 on a 266 fsb chip, you're basically setting the memory divider to 3:2. So when you raise you fsb to 333, your memory is running at 1000. Try setting your MIT to ddr533 which for your processor is 1:1, then when you bump your processor's fsb to 333 your memory will be running at ddr667. (and fsb 400 would equal ddr800)

You really don't receive any improvement running your memory over 1:1 ratio anyways.