Why do AMD 64 CHIPS LOSE THERE OVERCLOCK OVER TIME

[cronic]

Let me start this bt saying that no overclock is guarentteed and results do vary. So here are some of my thoughts after living with the AMD 64 and Nforce 4 for almost a year now.
Since January I have personally owned 7 different AMD 64 CPU
2 3500 Winnie
3 FX-55 cLAWHAMMER
1 4400X2 Toledeo
1 3800x2 mANCHESTER current chip (got my eye on a 170)

With every single chip they have all overclocked wonderful for anywhere from a few weeks to as much as a month or so. Anywhere from (2600-3120) prime stable for at least 12 hours, with no issues. BUT every single one has fallen off after a while. Some worse than others, but all fell off. Usually they fall off between 200-300 Mhz. I understand that cpu will lose some over time, but It seems that these fall off very quick. I know that the p4 is completly different chip design, but my experiences with the older p4 2.4 or 2.8 didn't loose oc so quickly, if at all.(maybe I didn't have them long enogh to fall off). I am not tring to bash anyone, as I like both Intel and AMD equally. I buy what is the fastest at the time I am buying and then overclock it.
I am currently on a quest for dual core 3000Mhz and it has been a challenge. I was wondering if anyone has had the same or different results with there amd. Plese stste what your results have been and more importantly how long have they been running at that speed, did they fall off, and your overall experiences.

[phadeout]

Did you ever run them without a heat spreader? Maybe over time, with a lot of stress, the internal contact dries up or loses consistency?

[Kadaj]

2 3500 Winnie
3 FX-55 cLAWHAMMER
1 4400X2 Toledeo
1 3800x2 mANCHESTER current chip (got my eye on a 170)

all this since january and you're using an XP-120?

[carmelo]

I think I damaged by chip one day by running too much voltage through it. I booted with 1.72 volts on my Newcastle just to see if it would fix my L2 cache always failing. Aftwards when I reliezed the chip had just reached a wall I backed it down to 1.62 and ran my old stable speed of 2.6. Pi would fail in seconds and games would now crash, when they had not prior to this. I am now stable at 2.55. I can't run anything above 2550 cause after awhile somethien ends up crashing. So it's not exactly the same, but I can see if some one is running alot of voltage, after awhile that'll take a large toll on a chip. May not kill it right away, but may make it difficult to reamin at that speed.

[cronic]

all this since january and you're using an XP-120?

I have run the xp-120 as well as a vapochill xe (on the fx-55 and 3500) and am currently running a water chilled cooling system. I have always upped the voltage, but I have been careful to no overdo it. Except with one of the 55 i killed.

[Kadaj]

I have run the xp-120 as well as a vapochill xe (on the fx-55 and 3500) and am currently running a water chilled cooling system. I have always upped the voltage, but I have been careful to no overdo it. Except with one of the 55 i killed.

ahhh. Well, I probaly don't have to remind you that increased voltages over extended periods of time will produce strain on the processor, regardless of core temp to a point.

but my experiences with the older p4 2.4 or 2.8 didn't loose oc so quickly

probaly because they were on the 130nm process, which had not even 1/5th the power leakage problems that Prescott or K8's 90nm process have.

[aicjofs]

I guess the big question is how much Vcore.

If you bumped from say 1.4 to 1.5 for the overclock I would be concerned that it fails so quickly. If however you bumped from say 1.4 to 1.6-1.65 or greater then while still concerning it is not a total suprise.

I have been doing this since the first P2/K6-2's and never had a CPU, AMD or Intel, "slow down" unless I got crazy with the Vcore(which I have done in the learning process, never fried one though ).

If you believe in burn-in, I remember the old school train of thought was to give it as much juice as it could take for the beginning of it's life.

Vcore aside...The times I have noticed a loss in clock speed is when I turn the computer off for a vacation of 2 weeks and the Overclock was right at it's limit. The 1st boot never seems to work, it takes a few days for things to settle back in, but otherwise 24/7 I never notice a loss of speed with a moderate overvolt.

[Eloman23]

Thats kind of interesting. I have seemed to notice just the opposite. I used to have a winchester 3000 (have a 148 opty now), and it's oc got better up until I sold it to my roomate. Originally it would only do like 2.39 with 1.5 vcore (i never go above 1.5 for air ). After like 6 months I played with it again, and it was doing about 2550 at the same vcore. My roomate currently runs it at 2550 ( he needs a ram divider however) at the exact same vcore. I use an xp-90 and heat was never an issue. Only thing I can think of is maybe to much voltage?

-Elo

[fatty]

After I bent pins on my FX53 it started to down clock so I got rid of it before it hit stock speeds

[shuRe]

i know what your talking about cronic. I have run my newcastle at 2.6g 1.65vcore stable on a hyper6 with good temps for about a year, and recently it cant even reach 2.5 stable with any volts.

I dont think its the contact between the core and the ihs because my teams havnt changed, it must be a deterioration of the cpu?

[Ket]

my fav stalwart chip is my winchester 3200+, initally clocked to 2.55GHz, ended up doing 2.65GHz, a memory controller that handles 320HTT+, nothing has ever dropped off with it, its constantly been consistent for about a year now.

[Kadaj]

it must be a deterioration of the cpu?

overclocking always deteriorates the processor. If anyone with an overclocked processor kept it in that state for long enough, it wouldn't even be able to stay stable at its stock speeds, let alone the overclocked one eventualy.

[SPQQKY]

My Winchester has been running over a year with the same overclock, stable as can be. I recently brought it to a new benching high point, so I actually got about 50Mhz more out of it.

[RyderOCZ]

Another good thing to include would be your Vdimm.

Are/were you running BH based ram or similar? 3.0V+ on Vdimm for extended periods?

[johann]

My Xeons have been running a 1.2Ghz (3.6Ghz) overclock for a year, sometimes I have to think hard to remember the stock speed. stock 1.3v running at 1.57v

[rdgrimes]

Thank God that chips burn out, it's a great excuse to buy a newer one.

[Raven.exe]

have been running my s754 3000+ claw @ 2.36ghz @ 1.575vcore for over a year and just recently got it up to 2.44Ghz 1.65vcore. It has not shown any type of degrading clock speeds...

[kimandsally]

Well to me this reinforces that burning in doesn't work, every cpu that I have kept has lost it's overclock over time, at first I thought I wasn't measuring them right but I have had, 4X 3000 Winnies 1X 3200 Winnie 1X 4000 SD 2X FX55 and now a 4400X2 this is in about 9 months every one has had a reduced overclock over the time I kept them, I also killed a FX55, we have a lot in common cronic.

[cronic]

Well to me this reinforces that burning in doesn't work, every cpu that I have kept has lost it's overclock over time, at first I thought I wasn't measuring them right but I have had, 4X 3000 Winnies 1X 3200 Winnie 1X 4000 SD 2X FX55 and now a 4400X2 this is in about 9 months every one has had a reduced overclock over the time I kept them, I also killed a FX55, we have a lot in common cronic.

i didn't think i was the only one who has seen this happen.

[ozzimark]

i've had a 3200+ clawhammer for more than a year now. it's on almost 24/7. i've put it up to 1.8v on air, brought them temps past 90c for kicks, silly things like that. it normally runs at 2.3ghz with 1.55v. that's the speed it could do when i got it, it's the speed it does today. ymmv i guess

[ben805]

it's more than likely the memory controller that were dying slowly, check the vcore vdimm sticky.

[ocZZ]

This happened with my last 3200+ was stable at 2.7 @ 1.65v for about 6 months and then started having problems (bsod/reboots) I put it at 2.65 @ 1.61 volts and it was fine .. but I could never get it back to 2.7 stable no matter what voltage it was running at.

[cronic]

it's more than likely the memory controller that were dying slowly, check the vcore vdimm sticky.

I don't think Thats what is causing my issues. I have never used vx or redline and I have never had the vdimm over 2.95-3.0 as I have always run TCCD on this board, and the cpu has always had the voltage over stock settings. But that is definitly something worth knowing for the future.

[shuRe]

I don't think Thats what is causing my issues. I have never used vx or redline and I have never had the vdimm over 2.95-3.0 as I have always run TCCD on this board, and the cpu has always had the voltage over stock settings. But that is definitly something worth knowing for the future.

i agree, i dont think its the memory controller due to the fact i have only ever run 2.9vdimm through my tccd.

I think its down to the issue that overclocking your cpu increases the rate that it degrades as such. Its probably again that the time of which your cpu can hold a stable o/c is down to its batch and production quality.

[Ket]

CPUs will degrade if not given time to run in from new, weather burn in actually works is neither here nor there, but one thing is like anything though, if you buy a bike and theres no oil on the chain, itll work, but not as smoothly as it could do until it has more oil on it. same kinda princable as with a CPU, from new it runs great, but perhaps not as smoothly as it can do until its run in a bit.