denali6194
New member
I heard somewhere that if all dimms are filled that your OC suffers. Is this true?
how many sticks??
- Thread starter denali6194
- Start date
cbiltsouris
New member
sort of Ive heard somthing about that ur ram might oc a little lower? then again I could be COMPLETELY wrong aswell
S_I_N
New member
correct oc'n works best dealing with just a single dual channel (or triple) being used. The main reason being matching sticks. Now if you bought a quad channel kit where all 4 sticks are matched then it might ignore this issue. I just know from the past whenever I added more sticks my oc would fail and I'd have to redo it
denali6194
New member
"might"...
I had 4 identical sticks of mushkin but never thought to try taking two out. I'm going for 2x4G in my next build just in case
thx though
S_I_N
New member
"might"...
I had 4 identical sticks of mushkin but never thought to try taking two out. I'm going for 2x4G in my next build just in case
thx though
4 identical sticks is not the same as MATCHED. Matched means they were hand picked (binned) to run well with each other. This is also why there a tad more expensive to procure.
S_I_N
New member
true then there is that as well. the extra load on the bandwidth.
This is all somewhat dependant on the board/CPU. For instance, many LGA 1366 systems have been working fine with all the DIMM channels full, but on the other hand many LGA 775 systems were rubbish with all the slots filled.
The northbridge on many Nvidia chipsets was very weak (think the 790i ultra), which couldn't get stability with all the DIMM channels full. On the other hand, if the memory controller is on the CPU the overclock may be much better. Ultimately hmmblah is right, that there is extra strain on the memory controller, but this may not affect your overclock much.
It comes down to what the overclock you find acceptable is. If you absolutely need every last Hz out of your machine, then yes going with half occupation of channels (i.e. 2 or 3 sticks installed) will be better, on the other hand if you can sacrifice a little overclock for more memory you may have better real world system performance.
The northbridge on many Nvidia chipsets was very weak (think the 790i ultra), which couldn't get stability with all the DIMM channels full. On the other hand, if the memory controller is on the CPU the overclock may be much better. Ultimately hmmblah is right, that there is extra strain on the memory controller, but this may not affect your overclock much.
It comes down to what the overclock you find acceptable is. If you absolutely need every last Hz out of your machine, then yes going with half occupation of channels (i.e. 2 or 3 sticks installed) will be better, on the other hand if you can sacrifice a little overclock for more memory you may have better real world system performance.
SmoulderingCircuits
New member
It's not the memory controller or any other component being put under stress, it's merely basic math that the more variables you add to an equation the more chance there is of a low probability outcome being the outcome; namely, that two sticks of ram are two potential failure points for an overclock. Four sticks of ram are, however, four potential failure points for an overclock. It's similar to why unibody chassis are stronger than those assembled from sections: the unibody has only one weakspot wheras every section of the assembled chassis is a weakspot. Every core in a processor is a potential failure point, every memory chip, everything. By using minimal components, so ram included, you minimise the probability of failure.
What you say is true, but the controller does most certainly have an effect. You might be able to use two sets of memory by themselves at a high frequency, but when you add them all together the controller can no longer do those higher frequencies with all the channels populated. It's not the memory crapping out, it's the controller.
SmoulderingCircuits
New member
I'd say from my experience that you're more likely to have one of your four or six DIMMS go belly-up because it couldn't remain stable while the other three or five could than you are to clog up the memory controller but things may well be different using 2133Mhz> modules
denali6194
New member
good to know. mine weren't matched just meant that they were all the same momdel, brand, etc. anyway I'm going with a 8G (2x4G) Mushking kit with 7-8-7-24 timings if i remember right. thanks for all the info!
S_I_N
New member
yeah but AMD is to easy to fool
nm I see lol ur cranking the volts on it look at my second cpu-z validation link in my sig
FragTek
New member
yeah but AMD is to easy to foollol no seriously nice job. what volts is ur cpu at??? lol you just multi oc'd AMD is so much better with a FSB oc combined with multi
nm I see lol ur cranking the volts on it look at my second cpu-z validation link in my sig
I see no difference between FSB mix or straight multi, bandwidth is almost identical.
I can boot in to windows just fine @ 4.4ghz under 1.5vcore as well but it's not prime stable.
S_I_N
New member
have ya benched the 2 different ways??? when I did mine I saw an increase in scores. casue ur also oc'n the ram at same time as well as everything else dependent on FSB not just the cpu. gotta watch the ram though as with mine I had to keep it close to stock speeds while bumping fsb
FragTek
New member
Well that's the compromise without spending a fortune on RAM, the faster the FSB the looser the timings have to be unless you're dropping the ass out of the memory multiplier also killing performance.
The great debate is tighter timings vs. faster clock speed. I say it's a pretty much a wash.
I haven't done much real benching between the two methods just comparing the GFlops from IBT while doing some stability testing, both ways seem fairly identical.
The great debate is tighter timings vs. faster clock speed. I say it's a pretty much a wash.
I haven't done much real benching between the two methods just comparing the GFlops from IBT while doing some stability testing, both ways seem fairly identical.
SmoulderingCircuits
New member
SmoulderingCircuits
New member
Same here, then 4.5 won't even boot.
