nows that definitely settling if you'd had since day 1 and those speeds and has only lost 10Mhz lol , nice chip :up:
Printable View
Well I lost about 0.03v vcore on my better 2600k for sure. Was perfect at 4.5 ghz 1.236v for two months, but after repeated prime tests (between 1-3 hours, multiple times) at 5 ghz, the entire CPU degraded across all ranges, so that 4.5 ghz now requires 1.260v for 12 hour prime blend, and 5 ghz requires about 1.440v (not priming there anymore). Started noticing more problems after using LLC2 to keep my idle voltages low, by having BSOD in BC2 main menu (but not while in a game) or idle on desktop; originally had some BSOD in BC2 in game at 1.40v bios (1.428v load), raised to 1.410v bios for 1.440v load, in game bsod's stopped but idle/main menu bsod's happened.
After noticing these were all 0x124's, and noticing that when I tried priming at 4.5 ghz at 1.236 (which used to be stable) to 1.254v load, the crashes were *always* 0x124's (while on my first, heavily degraded 2600k, the crashes were ALWAYS 0x101's, even at 1.3v load). Thought something was strange, so i raised up the PLL to 1.86v, and this greatly delayed how fast the BSOD's happened (1.236v was BSOD in 5 minutes, with PLL at 1.86, it was able to go about an hour and 10 minutes before BSOD). 1.242v bsod'd after about 2 hours with 1.86v pll (instead of 45 minutes without it). 1.248v with 1.88v PLL lasted five hours then gave a roundoff error, instead of a BSOD. 1.260v passed blend 12 hours without raising PLL at all.
Seeing that raising PLL helped (though didn't fully eliminate) the crashes here, and since 0x124's were happening at 5 ghz in BC2, too or desktop, I raised PLL to 1.86v, dropped the vcore from 1.420v BIOS (which also stopped the idle bsod's, as that was about 1.440v in the main menu), to 1.410v and passed 1.5 hours of main menu BC2 just sitting there, and some more time idle on desktop, with no 0x124. so yeah there was still degradation that happened, but I was able to compensate for part of it with the PLL boost. While I still need 1.440v for full game stability, 1.428v is now fine with lighter load thanks to raising the PLL. Pretty sure trying 1.40v bios will still cause crashes, since the vcore change at 4.5 ghz also relates to the vcore increase at 5 ghz (that's why I do my prime testing at 4.5 ghz--to prevent more degradation, and I can find it much safer that way). BTW, after doing the PLL 5 ghz main menu testing, went back to 4.5 ghz again and passed yet another 12 hour blend at 1.260v. So feeling happy, swapped out the CPU for the first 2600k that I can abuse at high vcores at 5 ghz again.
BTW, might be able to be fine without raising PLL and thus having idle voltage slightly lower than load voltage at full speed (from using LLC2 on the 2nd chip), by using offsets to have the CPU downclock when idle...tested on the first 2600k and works quite nicely...5 ghz at 1.5v full load, and C1E making it downclock when idle or light load, while clocking fully at 1 thread full load (as long as C3/C6 are turned off; seems like C3/C6 prevents vcore/clocks from rising fast enough when only 1 thread is fully loaded...I had a 1 thread game (Crossfire) just BSOD in the freaking game lobby in windows XP, when using C3/C6. No BSOD when I disabled C3/C6). Offsets plus C1E or EIST (which do the exact same thing, except EIST is controlled by software and C1E by hardware) definitely work nice to avoid having high idle voltages and dealing with excessive vdroop from LLC1, or not having enough idle voltage with LLC2...
well after 0x124's can be tricky , I base those more i memory problems , to test I take memory to lose as i can , and ramp up processor to full speed if doesn't get 0x124's , a subliming somewhere is causing you hell , the ud7 set sub-timing very hard and unsuitable for priming , you think its proc every time . .
.5 ghz at 1.5v not bad considering its load :) screw it ,there's warranty on these babies
to much prime can be bad at high vcore yes , but at 1.254v shouldn't degrade or chip is bad to begin with , imagine all the people encoding blowing thyre chips ? because overclocked a bit . . . some chips where just built to die ,
I think my chip has settled finally cause I primed all last night and had good and bad results" bad was memory " and she holding , when she starts getting bsod and needs more juice , i reset ud7 and she back to normal 14.40 load 14.90 bios
I should add that Ive never primed for more than a half hour. I think that is the killer for degradation.
last prime95 is more more strong that old version
are u sure that u use always the same version?
Would running the memory from 1600Mhz to 2133Mhz with a Vtt of 1.080 and vDIMM of 1.520 somewhat also degrade the chip faster?
I'm running a VTT of 1.07 with vddr @ 1.65 24/7 100% load for some time now and I can not report any kind of degradation. Also no ill effects from 1.46 vcore.
I see a major snowball effect again... people are getting scared... :rofl:
1) Some CPUs degrade, though they are rare. Some withstand the heaviest abuse with loads of high speed rams at ridicilous volts...
2) Also keep in mind that some users not always tell the true story; Degradation of two CPUs in a row... I wonder... :shrug:
3) You guys all love the loadline... coz it makes you prime, run Linx at that precious 1.295Vcore. Ever imagined that bad manufacturer implementation ( soft or hardware wise ) or usage of that setting could create spikes/overshoots etc... there's more than meets the eye... ever wondered why there's Vdroop/drop ?
4) And by all means keep on stress testing , it's good for you and your CPU... :up:
5) I forgot especially raise the PLL to 1.9 ish region !!! It makes the game even better...
Running two 2600K at 4.5 at 1.35Vcore (no loadline) with 4 and 8Gb 2133Mhz Gskill rams at 7-9-7-21 1.67Vdimm... no Vcio adjusting required... games and all my apps run without a hickup... maybe I could aim for lower Vcore, but that's not my interest...
Dont forget that people are always trying different timings and different bios so there is way to many variables.
good call, ambients , humidy matter too... it's not always the CPU either...
w/o LLC I need to set the bios offset to something like +0.090 to get a 1.360v load and that idles at about 1.410v. with LLC Id only need +0.030 to get 1.360v but I dont have the tools to see how much overshoot it does.. Maybe you can give some specific insights to this instead of pointing to a tech documentation ?
No. High Temperature and High Vcore are the "killer for degradation"
Prime simply runs codes to calculate prime numbers, it only uses the CPU for what it was designed to... :eek:
If people choose to run Prime at higher temperature than recommended or at higher voltages than recommended, that this the problem.
Some people say its perfectly fine to idle at high voltages something like 1.500v since the power draw in idle is very minimal and its just as good as a 0.960v idle and let Vdroop bring it down to 1.400v when the cores are loaded since its actually the power draw that can degrade the CPU and not voltages alone..
Something like 9000v at 0.01amps will have the same effect of 9v at 10amps ?
According to the loadline tables on the spec sheets, 1.52v VID is about 1.48v idle. What someone could do is set a 1.52v in BIOS, NOT use loadline calibration at all (since using LLC goes outside of specifications), make sure their board isn't bugged that forces LLC on you (MSI boards force vrise LLC even if its disabled in the BIOS), then leave the cpu completely idle so its around 1.47 vcore for an extended time, and not have any load put on it (say, like, a day or two), and then go back to their full prime stable setting and see if there has been any degradation. That should put that to rest.
Btw, is it me or have things gotten slightly hostile lately? I saw this way back during the pentium 4 and core 2 days, when some people said degradation was a myth...and I proved them both wrong, when the SNDS/GNDS reports came flooding in later....
Even degraded my X6800 with 1.55v, which everyone said was safe....
No one understood back then the difference between VID and VCC (after vdroop)...and the whole loadline article on anandtech came to the complete wrong conclusion about the voltages...(example: when people said max "voltage" on core 2 quad was 1.362v, this is wrong. That's the max VID, and VID is the voltage target sent through the BIOS, before LLC is applied. The max vcore depends on the load applied to the processor. Removing LLC removes the loadline slope, which helps the chips overclock more, but also goes out of the VRD specifications. Obviously, you're going to be fine if you use 1.35v with LLC, since you would still be within electrical specifications (not for THAT voltage/vid, but if you had used a higher VID without LLC). But 1.52v with LLC for 1.52v load? That's like using a 1.65v VID with no LLC. And max VID is 1.52v not 1.65v ....
Just saying...something to think about.
Not quite to that scale since most things have much much lower breakdown voltages but as far as power goes thats true.
Voltages alone CAN degrade the CPU. A semiconductors breakdown voltage is dependent on temperature and doping (this isn't a variable for us so ignore it) so it is possible to damage a chip at idle with very high voltages but it can happen at lower voltages during cpu loading because you also increase the temperature due to power dissipation. I have no clue at what voltage it happens at in these chips though. I have been running 1.5V through quite a few 2600k's with no degradation.
Basically an Idle voltage of say 1.5000v is still harmful even if it drops down to 1.380v when fully loaded?
I think this says alot.
If an experienced overclocker knows what they are doing, there should not be any degradation. This has always been the case. If you have "proved" people wrong in the past, then you have "proved" that you are an inexperienced overclocker. Period.
How can experience actually prevent degradation, Im missing something here.. Say both have the exact same hardware down to the atomic level. The other one is a world champion overclocker and the one just some random newbie who followed the 1.5v settings. Same temps and all.
A couple months ago it was Winter and colder. Stress tests will pass at higher speeds with colder temps. It gets warmer out, temps go up a little, and now you need a little more voltage to pass stress tests since they arent operating as efficiently. I usually just drop 100-200Mhz off my CPU speed once it starts getting warmer out.
Yes degredation happens, but not if you are using resonable settings.
^^ Care to explain this please?
Here, I'll invalidate your argument right now:
Person A: 1.65v through a Northwood is perfectly safe, just don't go above 1.7v.
Person B: Uses 1.65v 3.81 ghz, WITHOUT a vdroop mod, vcore drops under load to 1.52v from vdroop. Uses high end cooling, max temps: 60C.
Person B: CPU slowly degrades, after two weeks of having to lower mhz, person B's CPU can only run at 3.70 ghz at 1.65v.
Person A: you incompetent OC'er!!
....
....
1 year later:
Tons of reports of degraded CPU's and a few chips no longer stable at stock, when run between 1.6v-1.7v.
So you're still calling person B an incompetent overclocker, just because his CPU degraded, right? Just because person A's cpu did not degrade at 1.7v?
Sigh...I simply don't understand some people on these forums....I really don't.
@Admin: (nevermind)
No need to leave, just keep the Intel spec sheets in the back of your mind. No need to wave them around each time as mostly anyone cares about them.. You are overclocking and you dial in the voltages, you shoudl be aware of the risks, and take note of high temps... just coz the Intel rates 100°C doesn't mean 90°C load is still good for your CPU. I keep all my Intel CPUs for daily ops around max 70°C LinX or Prime95 loaded.
But also keep in the back of your mind that Intel binned these CPU for their rated speed, extra performance can not be warranted, if you overclock you are running a risk of damaging hardware... degredation is not a myth, but's not a certainty either... some CPUs can withold laods of volts, heat,.. with ease. Others go a bit down in performance or need a slight up voltage after a while... or are ready for hardware heaven :p
There's more than just the CPU that can degrade in performance... loads of factors play a part (see post 84-85-86 to name just a few)... and really man a 0.03 loss, are you really worried about that ?
What's the Intel tech spec on CPU PLL ? Every reknown overclocker here says to keep PLL sub 1.9 --> that's for +5.6Ghz action, you up the PLL for 4.5Ghz to 1.86 to avoid BSOD... Seems something is wrong with the way you got things dialed in... or something wrong with the board/bios/ram combo... (I'm starting to run for the big Dino monster that will come out of his cave soon) eg I run 1.7PLL at 4.6 daily, PCH at 0.95 (old B2 board) On the Asus boards also lots of BSOD at idle with the C states enabled and voltage offset used. I disabled the crap and it runs fine now ( though at idle at 4.6 at 1.35Vcore )
Have you tried another bios, think Sergio still uses F4 bios on his UD7 as it's the best suited for him ( but I could be wrong )
Because it never happened to them thats why they blame other for not doing things right whats funny is that they keep spouting that every chip is different.. someone needs to practice what they preach!