A little bit testing 24/7 settings on my 5666MHz chip.
http://www.pohrani.com/t/3W/Bb/3vqQUf6e/4700stable.jpg
On MIVE it needed 1.35v for 4700MHz but on UD4 it needs 1.4v.
LLC Enabled
PLL OV Auto (disabled)
vCORE = 1.4v
VTT = 1.02v
VCSSA = 0.920v
Printable View
A little bit testing 24/7 settings on my 5666MHz chip.
http://www.pohrani.com/t/3W/Bb/3vqQUf6e/4700stable.jpg
On MIVE it needed 1.35v for 4700MHz but on UD4 it needs 1.4v.
LLC Enabled
PLL OV Auto (disabled)
vCORE = 1.4v
VTT = 1.02v
VCSSA = 0.920v
Nice clocks, whats the loaded voltage on that with a DMM ?
1.408v
cpuz shows 1.368v,
DMM is at this point 1.408v? for me is very different, 1.452v cpuz = 1.46v DMM here in terms of offset
you have a large discrepancy
manually set to 1.45v in bios idle cpuz was 1.440v, and 1.416v load
DMM indicated at this point, 1.44v idle and 1.42v load:)
On my UD4 if it shows 1.386 on CPU-Z its usually a tad lower with DMM, but as what Kinzui said at 1.452 cpuz dmm shows 1.46 as well on my board.
Missed my testing?
http://www.xtremesystems.org/forums/...&postcount=359
I've seen that chart and it can be a good reference but some of the numbers dont reflect like that on my system, must be due to the chip/stresstest program/offset used.
I havent made a chart for mine but at the top of my head I recall having readings from CPU-Z/ET6 @ 1.332 and 1.344 at loaded which results in 1.3508v on the caps and auto voltage with +0.080/+0.090 offsets which result in 1.4508 and 1.4600 on the caps.
Talking about voltages, whats better for this board, Normal+Offset or just manual + LLC ? ( Im not sure of normal+offset makes LLC work )
I'll re-test it maybe with F4 BIOS to see are there any improvements in LLC. Have to do it before I RMA it. The LAN port just stopped working few days ago :/
This just blows, I already had about 17hrs of Prime95 Blend running, I left it running over night and when I got up.. bam, BSOD error code 124.. this is just 4.5Ghz with a loaded voltage of 1.35v ( via DMM ) LLC enabled. It survived 50 LinX runs with AVX.. This is so frustrating!
I've probably got a VERY VERY bad chip or im doing something wrong..
That would be great, maybe you can add in between voltages like 1.375, 1.475 and 1.575.Quote:
Originally Posted by Bullet92
what improvements dude, you are setting 1.6 and getting 1.611-1.609. that's tight as a virgin lol
Have you tried just using the computer normally to see if anything else crashes?Quote:
Originally Posted by Demo
There's really no reason to need to have it run for 24 hours with blend unless you're trying to either get an award, or degrade your CPU.
Seriously, just leave it as is and use the computer the way it was meant to be used. If anything ELSE crashes, then you can be concerned. Pass 6 hours of playing bad company 2, survive a few hours of youtube and being idle, and you won't find anything else that will bring the system down, except 18 hours of prime blend ~_~
Anything else is just feeding your OCD, and degrading your chip.
I've never had any crashes with the computer used normally even with 4.8Ghz, so this is how some of you guys do it.. just like 2-4hrs of prime then if its stable let the everyday application decide?
I dont have BFBC2 to test around the CPU.. I think the only CPU intensive game I have is Dragon Age 2 ATM.
Precisely.
What you need to do is LOWER your voltage to the LOWEST Possible point at which you do NOT get crashes or BSOD's (or erratic "lower performance" (due to crc errors)) in benchmarks. This will preserve your processor and preserve the overclock.
You know, my second 2600k has degraded by at LEAST 0.030v just thanks to all the attempts I did to get Prime stable at 5 ghz. Originally it was just FINE at 1.450v in BIOS, 1.404v load (LLC1), and now it will sometimes BSOD even at 1.470v BIOS, 1.416-1.424v load (LLC1). All thanks to repeated priming; temps 75-85C coretemps/realtemp (Tcase temps are about 10C lower).
This even made my formerly hours stable 4.5 ghz 1.270v BIOS (1.236v load) BSOD after while...thanks to the 5 ghz testing...had to set it to 1.275v BIOS (1.248v load). I'm just DONE with prime. Period.
So now I'm just playing Bad Company 2 at 1.405v BIOS LLC2 (instead of LLC1)...which gives 1.428v (most of of the time)-1.440v load (only reaches this during the hardest stress tests, since LLC2=more stress on cpu = higher vrise instead of droop). So far did multiple hours...as long as THAT'S good to go, nothing else will bring the system down. (had a BSOD at 1.400v, none YET at 1.405v; I suspect as long as no more degradation occurs, 1.410v BIOS should be the highest I'll ever need; might need that when it starts getting alot warmer in here in summer).
I'll keep that in mind, so I'll just run bench tests instead like 3dmark06/Vantage/3Dmark11 and proceed to gaming, I just need to find a game to play around that can really put the stress on the CPU, Lately all I've been playing is SC2..
I've been like priming for a 6-7days, its not non stop but mostly just overnight and this is the first time I went past more than 8hrs.
On a side note, even priming at stock/speed voltages would degrade the processor?
If I move to benchmark testing, anything else you can recommend aside from 3dmark series?
And for memory overclocking which iteration of the PI benchmark is most suitable and what tests to run?
Not stock, although a few people have had their processors die for no reason.Quote:
Originally Posted by Demo
But people doing 50 loops of LinX and 12 hours of prime at 1.4v+ is simply NOT healthy for the processor UNLESS you have some serious cooling. When LinX came out, people did 5-10 runs and 20 runs was considered max, and if you did 50, you were simply OCD excessive. Now people casually let it run all night...
Ok, how about for memory overclocking, how do you test it ?
memory overclocking is easily tested by memtest 86+, just to make sure you can make it to windows without trashing the installation/file system, then by 15 minutes of prime blend. If that passes, then you can use the comp and see if it holds up.
Hell I've found unstable memory overclocks in the first 10 SECONDS Of prime blend....
I've tried Memtest86+ and it does not detect anything correctly on my SB system haha, its v4.20 ill go with blend on prime but at stock speeds. Would LinX also degrade the processor at stock speeds?
Prime and linX wont degrade the processor at stock speeds. I think a lot of people don't even realize (or remember) what Linpack was for. It was for QA testing CPU's for binning/stability. But now people want to do it to their cpu's as well. The problem is, Intel doesn't use Linpack to test cpu's running at super high volts. They use them to test the CPU's running at stock speeds, at high heat conditions to make sure the CPU will be fully stable with the retail HSF up to the throttling point. The end users are using Linpacks in ways it was not designed to be used. The fact is, when you run a CPU out of specifications, all bets are off as to what will happen. Basically, the more you overvolt the CPU, the better the cooling has to be if you want to preserve the life of it. Hard to say the same for Prime but you're still pushing chips into conditions they were NOT intended to be in.
Yeah 4.20 doesn't detect the settings correctly, but at least it does work....
LinX testing is out of the window for any OCing that im doing from now on, Prime95 blend might still be used for testing memory OCing but I think I'll revert back to stock CPU speeds if I wanna test memory OCing.. or would the CPU speed also have a great effect on the stability of memory OCing ?
Also I have another 2600K here that that can only post at x47 w/o PPL-OV and has a tad higher VID (1.255) for 3.4Ghz, the current chip I have installed and played around with can only do x49 w/o PLL-OV and has a lower VID (1.240).. I dont know if its worth testing the other 2600K as I wont have it for maybe more than 1-2days more..
How about things like wPrime/HyperPI they dont degrade CPU's right?
None of the programs will degrade chips when run at stock. Fine as long as temps are kept reasonable.
When testing memory overclocks, its always best to keep the CPU at stock speeds to eliminate having to deal with multiple issues at once. Far better to know if the memory is capable of that speed, than to have to deal with a stressed IMC + a stressed CPU at the same time. OC'ing is like a puzzle...gotta know how the pieces all fit in....
But for overclocking, aside from benchmarks and games what other synthetic tests I can run to check for stability ?Quote:
Originally Posted by Falkentyne
Why are you wasting your time with silly unconclusive stress tools... :confused:Quote:
Originally Posted by Demo
If it's daily stable, then it's stable... I play a lot of Bad company 2 and if anything is wrong it will crash to desktop or freeze... I can run 24hrs Prime or LinX for 100 loops...
If the rig is daily stable with your current settings, so be it, even if it's just for minesweeper dlx edition...:up: Your chip will prolly be an average one and needs 1.35Vcore for 4.5Ghz action... if you want one that does higher speeds or want one at a lower Vcore start buying CPUs then till you find a "golden one"...
Why are you wasting your time with silly unconclusive stress tools... :confused:Quote:
Originally Posted by Demo
If it's daily stable, then it's stable... I play a lot of Bad company 2 and if anything is wrong it will crash to desktop or freeze... I can run 24hrs Prime or LinX for 100 loops...
If the rig is daily stable with your current settings, so be it...:up:
http://i1184.photobucket.com/albums/...8/readouts.jpg
There seems to be a +0.010~0.012 difference between the read points that I've indicated with green and blue ( higher read outs on blue ) should I go with the higher read out as my actual voltage?
EDIT:
So I've settled for 4.8Ghz for 24/7 use but I have a slight issue with the voltage, Im using manual + LLC at 1.470 which gives me an idle voltage of 1.458v and loaded voltage of 1.440 is there a way to keep the idle voltage under 1.440 while maintaining the loaded voltage @ 1.440v?
EDIT2:
Would any of the power saving features on the board cause instabilities or lowered performance?
EDIT3:
After enabling all power saving options I get a BSOD when Im only stressing 1 core as it goes back to 4.8Ghz but gives a voltage only good for 4.5Ghz, a good example for this is 32M SuperPI run. my voltage stays at about 1.280-1.380 but I do require 1.404v minimum to keep it stable.
EDIT4:
Is it normal for the PCH heatsink to get really really hot?