Gaming Overclocking my Athlon II X3 425

[person66]

Hello all, I am having some troubles with overclocking my AMD Athlon II X3 425 cpu. I can get it to anything below 3.4GHz easily (right now it is at 3395.4MHz with 1.340V and I have been running prime95 for an hour or so without any problems) but as soon as it hits anything above 3.4GHz, it becomes very unstable. Even going from 3395.4MHz to just over 3400MHz and it cant run a prime95 test for more than a minute without one of the cores getting an error. I have tried increasing the voltage all the way up to 1.440V, but it just seems to be more unstable. The temperatures are fine, they never get above 46°C, and I am new to this, so I am not sure what could be causing the issue.

I am fine with the CPU as it is now, but I just want to know why a few MHz more causes it to become so much more unstable.

Also, is it normal for the Prime95 workers to finish their tests at different speeds?

 

[Scorpei]

Different speeds is fairly normal unless they finish A LOT different. As for 3,4GHz (just below) thats a fine result, seeing as the 425 comes from 2,7GHz if memory serves. Thats 25%.... did you try to unlock a core / L3 cache?

Anyway it could be the chip has a drop there. My e4500 has the same thing at 3GHz (stock 2.2). 2,99GHz is fine, anything above and it isn't stable (on air). The voltages around that part also aren't very interesting, it ran with fairly low at 2,99 and to achieve anything above would require a LOT of power (and thus, cooling). In general though I have the rule of thumb that as soon as you notice voltages really need to go up, you're better of at a lower setting. Required voltage tends to hit a wall and go exponential at some points on a chip and usually the gain isn't that great when compared. For example my i7 870 runs fine at 3,9GHz without much voltage applied (1,2V), to reach 4,1GHz it needs roughly 1,3-1,35 (I never pushed it), while running stock is 1,2V (or when voltaged manually ~1V. So the first GHz required .2V where the last 200 required nearly the same.

So in the end, it's normal. My i7 has yet to show me that cliff but gets a little hot for my taste to try, but my e4500 clearly had a 3GHz barrier. The thing has been running for at least 5 years now without any problems in a 24/7 setup so all in all I'd call that a good result .

 

[marcus134]

oc roof can be caused by various things,it may be caused by the cpu itself being unable to reach higher clock at this temperature, the mobo vrm (voltage regulation module) may be at their limit, the chipset may be at it's limit (can be corrected with additional cooling and voltage), your ram may be unable to support higher clock or the latency needs to be slacked or the ram voltage needs to be higher.

You have to remember that electronics component have nano-scopic variations, and despite those variations they'll operate normally within manufacturer norms but as soon as you go out of those norms ( like overclocking), each piece may react differently. Meaning that oc'ing two identical configuration, you won't get the same results.

 

[DCG]

Try running prime for 8 hours, most likely you will see a core failing around 5 hours.
Or run LinX for 20 rounds with all memory ( I advice you have page file on or els it will kill your results ).

With LinX you'll have to monitor your pc with cpu-id ( or somerhing like that ) and kill LinX if your cpu gets too hot ( don't know the temp for an athlon though.

 

[person66]

OK, thanks both of you for answering my question

I am perfectly happy with it as it is now, I was just curious as to why it wasn't working if I made it any higher.

As for Scorpei's question of if I tried to unlock the 4th core, yes I did, but windows wasn't getting past the glowing orbs startup screen (it just froze up halfway through the animation), so I disabled it again. I didn't know you could unlock the L3 cache, how would I go about doing that?

 

[exangel]

I use a stress/bench test called OCCT which uses Prime95 but generates realtime charts that it prints out (PNG) and halts the test when the first error is detected (or when the test is finished)
Haven't used Prime95 in a long time personally, I recommend OCCT. Using this will allow you to see your voltage & temperature history, including at the first error point.

In overclocking contests, they usually require a MINIMUM 5 hour (OCCT CPU, large data set) run completion as proof of stability (usually in addition to other benchmarks); in finding stable settings for my system, I fiddled til I got a stable 5 hour, re-tested at 8 hour, then lastly at 12 hour, if i felt that should be the final setting.

I had to reduce my overclock significantly when I upgraded to two dual channel DDR1866 kits (same make/model) from one kit (2x4 -> 4x4). Either my motherboard or one of the new matched modules doesn't even like performing at (DDR1866) standard (my own CPU's standard dictates 1600MHz RAM speed with the FSB at 200). (My system)

Running the OCCT/Linpack test may cook your (AMD) CPU (if overvolted) though, it's way more intense than the standard test. I don't recommend trying it unless you enable overheating alarms/auto shutdown &/or get water cooling (and if you didn't mention it but you're hitting 46c on water cooling, you have almost definitely found your limit).
I once was running my CPU around 80c for a good while before I glanced up and realized my autoshutdown/thermal alarm settings had been reset in my BIOS without my noticing!

edit: oops, an omitted word