• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Is odd multi more stable?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Just a nickname

Member
Joined
Aug 28, 2010
Hey,

I read somewhere that odd multi were more stable than even multi. Is that true?
If yes then why? :-/

I'd like to bring my computer to a stable 4.4GHz with the lowest voltage possible but I've only used x20 multi so far and I need a lot of VTT to stay stable. I could use the x21 multi but I need to keep the intel speed step enable to use it and I fear it could bring some instability (thus requiring more voltage?)...

So two question:
-Odd multi = ++stable?
-Intel speed steep enabled require more voltage to remain stable?

Thanks for your help :burn:
 
Id like to know the WHY part too. I have heard that, and experienced it myself. Some say you can even use less voltage, but that I believe is a personal experience more so than a rule...at least I havent experienced that.

I dont (and its reccomended not to) use Speedstep when overclocking as it potentially cna be a source of instability for whatever reason, so I have no idea. I would imagine not though as its idle and lowers the multi/voltage so I wouldnt see a need to increase it.
 
I see but having speedstep enabled let me get the x21 multi... If only I could find a way to use that multi without speedstep!

I will do some test like 191x21 and see if I can use lower voltage. Then I'll try higher frequency :rock:

Note: The purpose of this is because my ram do not really like voltage and to keep it low, I need to be as close as their rated speed so lower BCLCK is better.
 
Try enabling the 21x multi and disabling EIST, C-State, Speedstep. It should lock it at 21x. This worked for many boards Ive owned (both P55 and X58).
 
Ahh, good call Doz, I believe thats what I did on the same baord (well, I didnt have the SLI).
 
Yep I tried that but I don't have access to the x21 multi. It is enabled when having turbo on. Like, x20 is selected when I have turbo on but the real multi is x21...
Example
190x21
BCLCK 190
Multi 20

If I select the multi 19, the frequency go from 190x21 to 190x19 :rain:

-----------

So far the results show that it is more stable with the odd multi simply because the bclck used is lower (my theory)... I need lower skew value and the system is still stable for a good 30mins with lower vcore (however, I'll probably need the same settings to stay stable).
 
Select 20 and leave it on... and see what happens. I think if its lower than stock it wont lock.

Yes, thats logical that its more stable with a lower bclk, but that doesnt isolate if thats the multi or not. Why, at only 200bclk are you messing with skews? Stock was fine for me even up to 220 (which was a solid bclk wall for me).
 
At 200x20 I was stable previously with the setting on my sig when I had an old BIOS (without the skew control option). When I flashed to the newest, I wasn't anymore stable and I could only be stable with skew on 200ps (that option was added later on the BIOS)...

So I had two choices, bumping the BIOS or bumping the skew... I like to have the lowest voltage possible so I bumped the skew.
 
i have to use speedstep to use higher than stock multi and have never had an issue. if the multi dropping bothers you set the windows powwer plan to performance and the muli will stick (i do this when benching)
 
it might depend on what os, on vista i believe it's in the screen saver settigs under power plans (i'm on my phone so i can't check) set it to performance and the multi will stick
 
nvm it works... just takes time to switch of mode so I tought it didn't work.

You :rock:

BTW: I managed to be stable @ 200x21 1.3vcore & 1.175 vtt while it takes me 1.325vcore and 1.225 vtt @ 210x20
 
thats a bclk thing, not a multi thing. Its tough to test this as clockspeed and bclk are effected which effects the voltage.
 
Many have found odd multis are more stable than even, and I never heard any satisfactory explanation as to why either. Several of us on xtreme did charts with different multis, and usually I would need about .035 more vcore or 5-7 notches higher vcore to get same mhz with even multi vs odd.

For example, my i950, for 4.4 prime stable,
23x192 = 4.4, I need 1.34 vcore cpuz load.
24x183 = 4.39, I need 1.38 vcore cpuz load.
21x210 = 4.4, I need 1.34 vcore cpuz load.

I got 4.6 linx 20 runs stable with HT on using 1.42 vcore cpuz load, using 23x200. Using 24 multi, I cant even get 4.5 stable with 24x187, even using vcore much higher.

For benching, I can do 5ghz spi 1m stable at 23x218 w/ 1.55vcore.
However, using 24 multi, max I can get is 204 BCLK or 4.9ghz.

Same with i940, the 22 multi was worthless, I could bench higher mhz with 21 multi.

Almost all HWBot high scores are with odd multis, even with cpus where the even multi would be highest multi.
 
Many have found odd multis are more stable than even, and I never heard any satisfactory explanation as to why either. Several of us on xtreme did charts with different multis, and usually I would need about .035 more vcore or 5-7 notches higher vcore to get same mhz with even multi vs odd.

For example, my i950, for 4.4 prime stable,
23x192 = 4.4, I need 1.34 vcore cpuz load.
24x183 = 4.39, I need 1.38 vcore cpuz load.
21x210 = 4.4, I need 1.34 vcore cpuz load.

I got 4.6 linx 20 runs stable with HT on using 1.42 vcore cpuz load, using 23x200. Using 24 multi, I cant even get 4.5 stable with 24x187, even using vcore much higher.

For benching, I can do 5ghz spi 1m stable at 23x218 w/ 1.55vcore.
However, using 24 multi, max I can get is 204 BCLK or 4.9ghz.

Same with i940, the 22 multi was worthless, I could bench higher mhz with 21 multi.

Almost all HWBot high scores are with odd multis, even with cpus where the even multi would be highest multi.

:confused:
Wow now that's strange... I bet it's a conspiracy. The odd number are trying to take the control of the universe! :shock:

I'll try to check (and understand) out how the cpu multipliers work. Might explain some things.
 
Back