Too easy? is 100c too much?

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

GuitarDaddy

Lifer
Nov 9, 2004
11,465
1
0
Originally posted by: taltamir
ok this is odd. i thought it was the C1E (lowers the voltage and the multiplier when idle).
But i see now that this is not the case. Even when set at a solid 3.6 ghz 9x multi with no change I clearly have a voltage drop when I go from idle to max load, and then the voltage goes back up when I idle again.
it is detected as 1.213 idle and 1.211 load. Why would it do that?


thats vdroop and thats very good if it only droops .002, almost non-existant, you must have a board that allows disabling load line calibration?
 

SimMike2

Platinum Member
Aug 15, 2000
2,577
1
81
Originally posted by: taltamir
100 degrees Fahrenheit = 37.7777778 degrees Celsius

So yea, definitely C.

this thread was NEVER about extreme overclocking. It was about how easy it is to OC (even to the extreme) with the automatic voltage adjustments on the gigabyte board, and wondering if there is a price to pay... to which I was told yes, that it over voltages like crazy and will burn out my chip if I let it continue.

Anyways, aftering failing OCCT at every voltage level between 1.225 and 1.25 I decided to use larger jumps at first and then lower it, going to 1.275. Stable so far. Showin up as 1.2v on CPUz, which means that when it was detecting 1.52 it must have been closer to 1.6 (the mobo allowes up to 2v.)

Anyways, so far it is stable at at 1.275v @ 3.6ghz with the EIST and C1E turned off (i left the thermal safety thing on though).

Core temp tells me it is 105C before it engages, so i never actually hit it before, even at the highest settings I had I only reached 103c for a moment.

I am now getting 40-45cc idle and 58-60c load. Pretty nice at a solid 3.6ghz.

I am only gonna aim at 3.6 simply because 400fsb is such a nice round number, and it leaves my ram at exactly DDR2-800 v1.8 5-5-5-15.


I am using:
Coretemp 0.99 to check temp
OCCTPT to stress test CPU
Memtest+ 2.01 to test ram.
CPUz to check my voltage level.

Your temp and voltage settings are almost identical to my older Gigabyte DS3965. Auto voltage gives too much voltage, even at stock settings. The only thing I do different is set my voltage for around 1.265. With this setting, my core voltage at idle is around 1.22, according to CPU-Z. I can run at 3.6GHz no problem with this setup.

I run at 3.6 for exactly the reasons you describe, I don't have faster memory and I don't want to overclock the memory. 3.6 is the sweet spot for many people, especially people with Gigabyte boards because this puts the memory right at DDR800 speed.

 

taltamir

Lifer
Mar 21, 2004
13,576
6
76
I actually do have faster memory, I am undervolting it. (to JEDEC spec 1.8v, all the manufacturers are releasing memory that requires overvoltage to operate at required speeds)

Thanks for explaining it guitar daddy.

Originally posted by: DarkRogue
I could be wrong, but I was told EIST (SpeedStep) lowers multiplier and C1E (Enhanced Halt State) lowered voltage. I somewhat remember being clarified on this, but I can't remember what it was though. Try disabling both EIST and C1E and seeing if it still happens.

Maybe, the anandtech article i just read refered to EIST as being the same as speed step... however my bios specifies that EIST is used to allow drivers to change voltage and clockspeed on the fly (aka, gigabyte dynamic energy saver), and that C1E enhanced halt reduces voltage and multiplier. And then there is a "thermal protection 2" settings that reduces it if temperature goes above "threshold".
It wouldn't surprise me if gigabyte misnamed them.

Since vdroop happens regardless of EIST and C1E settings, is there a reason to actually disable them? that dynamic underclock could end up saving me some power...
 

BlueAcolyte

Platinum Member
Nov 19, 2007
2,793
2
0
They might reduce stability, but if you stress-test with them on and it passes, then you should leave them on.
 

VirtualLarry

No Lifer
Aug 25, 2001
56,341
10,044
126
Originally posted by: taltamir
I literally got out of bed, turned off the alarm clock, walked over to the computer, turned it on, and started ocing.
That's the spirit!

 

VirtualLarry

No Lifer
Aug 25, 2001
56,341
10,044
126
Originally posted by: taltamir
ok this is odd. i thought it was the C1E (lowers the voltage and the multiplier when idle).
But i see now that this is not the case. Even when set at a solid 3.6 ghz 9x multi with no change I clearly have a voltage drop when I go from idle to max load, and then the voltage goes back up when I idle again.
it is detected as 1.213 idle and 1.211 load. Why would it do that?

vdroop under load
 

kpo6969

Member
Jul 31, 2007
89
0
0
I have the GA-P35-DS3L and E8400 clocked @ 3.6
When I 1st tried the auto settings the bios set it @ 1.3 for 3.6
My board has a vdrop of .046
This is what I run it at (if it helps anyone out) for 3.6
1.262 in MIT
1.236 shown in bios
1.216 in cpuz
 

Idontcare

Elite Member
Oct 10, 1999
21,118
58
91
Originally posted by: schneiderguy
So you think I should do it manually at a lower voltage then? Any other tool I can use to check the voltage?

yes, 1.52 volts is way too much :Q

I wouldn't put 1.52 volts on my 65nm Q6600's (on air) let alone dare to think of torturing a 45nm chip with such electric fields on those even smaller geometries. Ouch :(