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

[O/C]Downcore That CPU

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

sno.lcn

Senior2 Member
​
Joined
Nov 21, 2005
Location
Atlanta, GA, USA
Downcore That CPU
by Dolk


In the later parts of Summer 2009, I was in an overclocking competition. Someone was talking about lowering the number of cores so that they could increase the overall speed of the CPU. This made sense, sort of.

Let's say you have a Phenom II 955 BE clocked at its max water-cooling overclock, which is roughly 3.8 GHz. The current voltage setup for the CPU is 1.55v on the CPU with a 2.7 VDD. The CPU-NB is at 1.35v with a Frequency of roughly 2.8 GHz. If you were to disable a core so that the CPU was now a hypothetical “720 BE”, you would think less voltage is required to get the same speed.

In theory this makes sense, but in reality it really didn't work out this way...
Click to expand...


Click here to continue reading.




Discuss this article below. If you are interested in contributing to the front page (www.overclockers.com), please feel free to contact me.


Thanks,
sno.lcn
 
Last edited:
Hmm, what word can I make up... I WANT A WORD OF MY OWN!!!! ROFLMAO! :)

BUMP for a great article!!!!!!
 
If a single core is holding back your overclock, than wouldn't disabling that core enable you to overclock just a little higher?

Makes sense to me.:shrug:
 
Rather unlikely also for different cores to perform that differently, considering they were all manufactured together from the same material at the same time.
 
But thats just what its doing AFAIK... I have a feeling since AMD's architecture is 'native' quad core, and Intel (Penryn/Yorkfield) are two duals together, this is why one would experience that result on the intel side.

Its quite common to see benchers disable two cores (I cant bring myself to say downcore :p) to push for higher clocks. Bounce that off Deanzo/MIA/Gautam. Maybe its just for intel??? I dont recall specifically to restate it here as a fact though. (I did that for Summer FW for CPU intensive apps to get more clocks, but unsure aside from my thoughts above as to why) :shrug:
 
Last edited:
Dont alot of ppl do this on LN2 to get a max super pie times? mostly with intel like u guys have said, but if u go to Xtremesystems.org and look in the extreme cooling section u will c this in a lot of screen shots and A LOT of suicide shots for max speed

Duke
 
Hmm, well written as usual Dolk I am not understanding hte article though.

I do not understand why you chose to run multi-threaded benchmarks as proof that downcoring does not provide OC benefits :s

Also the voltage being lowered for lesser cores is not accurate, the current is what would be lessened. And also TDP of the chip
IE you have 4 12v fans running. If you remove 1 fan, you still need 12volts to power the other three, the difference is the amp draw is reduced by 25%.

With AMD and shared L3 cache, I would be interested to see if disabling cores, effects the remaining cores ability to access the cache.


Also where you able to OC any higher with less cores?
 
well current is also directly related to voltage and resistance, so disabling cores would in theory LOWER resistance in the cpu as it is not going 2 the other 3 cores so thus lowering the amps? (unless each chip is considered on its own "loop" i guess if u will) I guess it all depends on how the power is split up for the cores

Duke
 
Dukemurmur said:
well current is also directly related to voltage and resistance, so disabling cores would in theory LOWER resistance in the cpu as it is not going 2 the other 3 cores so thus lowering the amps? (unless each chip is considered on its own "loop" i guess if u will) I guess it all depends on how the power is split up for the cores

Duke
Click to expand...

It's also possible that the path 'around' the disabled cores is roughly equal to that path through said cores.
And if this does create a path of less resistance then it would be roughly equal to lowering the VCore correct?
Then based on that info, it would only lower temperatures?
That's just IMHO, based on what little science I know.
 
well i think it would b rather hard to make a trace (your path) replicate the resistance of a CPU as it is constantly changing depending on what gates/resistors r open or closed (why u get more heat at 100% vs idle) but TBH i really have NO damn idea lol, the thing is idk if it just turns voltage off 2 that core altogether or if like u side it just goes on an alternate path, i would say that is what makes the diff (and wold also determine if other cores could access the turned off cores cache) but like i said im just stabbing at fog here....in the dark no less!!

Duke
 
I would guess the only time it helps to "downcore" an AMD is if the motherboard is having trouble delivering smooth power at high vCore. If the motherboard is adequate to provide plenty of power then there's nothing to "add" to the power delivery and, therefore, no gain in the overclock. In other words, if the CPU has reached it's limit no amount of playing with it, short of cooling it down more, is going to help.

However, if the motherboard is running short of power at higher vCores then reducing the power draw by "turning off" other cores might help. Testing would need to be done using a motherboard model known to have voltage problems on the high end.


Edit:
There might be an exception to this if CPU cooling is being taxed. If some cores are turned off then there would (theoretically) be less heat produced, which could allow for more speed and/or vCore to be applied assuming there is insufficient cooling to start with. If you've got plenty of CPU cooling then you probably wouldn't see much gain from fewer cores.

In a nutshell, "downcoring" might help if you've cheaped out on your components ... :p
 
Last edited:
Each core should require the same V (to a point) regardless of how many cores. The A draw will increase, however that being said, V(A)=W so therefore just by cutting cores theoretically you should reduce heat as you are reducing the A draw. Now the tricky part. Do 2 cores at X voltage drawing y volts create higher temps or just more heat than 4 cores with the same power draw and hence requiring better cooling to dissipate said heat?

I look at it this way 4 2500 btu heaters will get it no hotter than 2 but it will get hot faster and be able to maintain the heat better. If it sounds wrong do this, measure the temp in front of a heater in a small room after 1 hour then add a heater to the room. After another hour the total room temp may feel hotter but the temp in front of the heater will not be much different if any different.
 
Last edited:
You guys are pretty close to my overall theory. (Its been awhile since I have done this article so I may be a bit rustic on this topic). If I remember correctly, one of my theories I was proving was that even though a Phenom II 550 has 2 cores, it will require the same amount of voltage for the same OC as a 940 or a 955.

My other theories that I proved was that you hardly gain anything from lowering the number of cores, even stability was the same. It is thus proven that even if the cores are disabled by the AMD or by you, that they are still active. (Thus applying the "Enable Core ACC" Theory). But one thing I did not test was the VDD required by each of these CPUs and their cores. If you have a 3.8 ghz 940, you need a VDD of 2.7v to help stabilize the OC. But if I disabled two of its cores, would the VDD be the same? Something I did not bother to test and should have.

Note to self while I still have my 550, test theory of VDD applied to down cored CPUs.
 
You must log in or register to reply here.

Similar threads

U
Delidding Zen 4 Ryzen, 7950x etc
2
Replies
27
Views
2K
User5566
U
N
CPU unstable no matter what I do, and it's getting worse
2
Replies
25
Views
1K
mackerel
mackerel
U
Upgrading to AM5, G.Skill vs Kingston RAM?
Replies
2
Views
577
Woomack
W
J
pump controller??
Replies
3
Views
309
Nebulous
Nebulous
mackerel
Relative IPC and efficiency of some Intel CPUs in Cinebench R15
Replies
3
Views
193
mackerel
mackerel
Back