I just finished building my gaming rig and I am currently in the process of over clocking and tweaking all the settings. One of the programs I am using to torture test my CPU is Prime95. I did the Auto tune that came with my Asus motherboard and the clocks it put it at failed after 5-7 minuets of running Prime95. My question is how "real world" are Primes torture tests? Will SC2 or other CPU heavy games crash my system?
My Rig: AMD Phenom 2 x4 965 @ 4.3ghz on 3 cores (Stocks is 3.4 on 4 cores) ASUS M4A88TD-V EVO/USB3 8gigs of Gskill 1666 XFX ATI Radeon HD 6870 Corsair TX650 Corsair H50 Liquid Cooler
Also is it worth it to OC to 4.3 only on 3 of the 4 cores? Or is lowering it and being on 4 cores better?
It is realistic as a stability check but the temperatures are not realistic because it puts the processor at 100% full load which no game or non-synthetic benchmarking software will be capable of doing.
If it fails prime95 in only 5 minutes than it is likely to fail in your daily usage as well.
Prime95 is realistic, especially for SC2. As time goes on, the newer games will stress your system more and more until you start to reach the higher end of your CPU's processing power. Take the findings as a sign that you have an unstable overclock.
On that note, be careful with the "autotune" function. I've yet to hear of anyone that was especially pleased with their results (i.e. either clocked too high or too low), and since failure could lead to a fried cpu, I think it's be much safer in the long run to manually oc your cpu yourself.
4.3 GHz is pretty high for Phenom II. That's also very high for automatic OC parameters--they'll tend to set the voltage uncomfortably high in general, not to mention for such high clock rates. Also, sometimes the CPU will be stable in a stress test yet be unstable in real-world loads. Passing something like prime95 is only a first step.
I got the same CPU as you do and 4.3 is crazy ... if you check the internet most only stay stable on air at around 4 EDIT: Oh u use liquid cooling ... dunno how much u can get but with that :D
I did overclock mine to 3.8 and Prime95 lets my CPU get so hot that my mainboard clocks it down after 5min+ since my casecooling is rly bad XD
But it runs very stable ... since no "normal" programm uses 100% usage (even sc2)
Overclocking by this margin will fry your system and you can defintly take hardware damage from it. Please for gods sake read some over clocking tutorials and get some knowledge before you do something without having a clue !!!
with 3.4 Ghz stock clock I tink 3.7-3.8 is a safe overclock range if you have the appropiate heatsink.
On May 15 2011 19:13 Holy_AT wrote: Overclocking by this margin will fry your system and you can defintly take hardware damage from it. Please for gods sake read some over clocking tutorials and get some knowledge before you do something without having a clue !!!
with 3.4 Ghz stock clock I tink 3.7-3.8 is a safe overclock range if you have the appropiate heatsink.
Him going to 4.3ghz is perfectly fine for the phenom II architecture, but for him to "fail" probably meant he BSOD'd or something... which means he should lower his voltage...and he's probably going above 1.500 volts... or something crazy... like 1.7-1.8... ( very quick cpu death if ran for some time...) or he's overheating... which he didn't tell us his temps either...
he could run multiple instances of prime 95...
you could also run the intel burn test
He should only be manually setting it in the BIOS for much better precision and only slowly raise it to 4.3 doing multiple tests in between like 3.5 / 3.6 / 3.7 / 3.8 / 3.9 / 4.0 / 4.1 / 4.2 / 4.3 with tests at approximately 8 hours or longer...
As long as your temperatures remain in the manufacturers specs there is nothing to worry about. 4.3GHz will never fry your system, too high temperatures will. So don't listen to guys who can't apply any logic to what they say but just repeat what they read on forums. Prime95 is probably good for testing because it doesn't really stress out the cores, but more at a level that a very CPU-heavy game would do. If you want to see how your temperatures rise under full CPU-stress try out any software that uses LinPack (f.e.Intelburntest, Lynx) as others already stated, it's a completely different level to Prime95.
if you want a rock solid machine you should be able to run prime95 for atleast 24 hours. if its erroring after 5-7 minutes thats not a huge deal unless of course its crashing your machine or crashing in game.
Basicly prime95 is detecting very minor calculation errors in the cpu. If your programs aren't crashing you're within the acceptable fault tolerance and should be fine.
Its probably also worth mentioning that yes, overclocking does reduce the life span of the cpu. However the chip have a life span of millions of hours and will be obsolete long before its end of life.
Overclock is not with out its draw backs. For example you may be able to increase the ram/cpu bus to make that portion of the machine faster at the cost of leaving the less time for other devices like the southbridge(harddrive) to talk to the main data pipes.
In short, if your cpu has 25% more data throughput and it causes a reduction of 50% throughput for your harddrive. Your machine would actually be slower in real world usage. I'm not saying this is your situation since I dont know exactly how you have it configured but this type of situation is quite common and often overlooked.
FYI, The performance of a machine should be measured in data throughput, not simply ghz. With out running a few benchmarks I cant say with 100% certainty but I'd wager 4 cores at 3.4ghz out performs 3cores at 4.3ghz.
I've been up all night, hopefully this all makes sense to ya.
For high overclocks temperatures are not the biggest problem, electron migration is. Even if you stay below the temperature threshold your cpu is going to take damage over time when you use excessive voltages. 4.3 ghz on a Deneb with autosettings is asking your mainboard to apply excessive voltages, because it is a high overclock and autosettings never work well. Good cooling may limit the amount of damage electron migration causes, but the higher the voltage the less effect cooling has. The effect of electron migration does not scale linearly, it scales exponentially.
And you should mention that "temperatures" does not only apply to your cpu, your mainboard will be stressed a LOT more with high overclocks and you should really watch your voltage regulators. Cheaper boards sometimes have not sufficient cooling for a high overclock over an extended period of time and die, especially if you use watercooling, because you lack the airflow of your cpu-fan.
@ zengrafx Modern desktop platforms use unidirectional point to point-connections, no bus system.
Cluson that is indeed the cause of the errors. Hyper-transport is a unidirection point to point connection ,thank you for correcting me.
its still worth mentioning there will still be a bottleneck just in a different place. At 4.3 GHz I’d bet he has already come to it.
If you'll indulge me
The inter-core data rate is the weak link only 533 on the phenom-ii 465 Start there. Setting your memory to 1067(2000gtp) is going to be in your best interest. More overclocking headroom less excess voltage and and less heat on the CPU and north-bridge.
Now set the NBfrq to 2400, and add 11 MHz incrementally to the CPU frequency (211mhz). Set the HTfrez to 1800mhz, or 2200 if you have 6gb SSD drives. This would be an optimal ratio with the best efficiency and least impedance/heat while still filling the available (5600gtp) of Hypertransport3.0.
A black edition chip would give you the best possible overclock here by allowing you to increase the multiplier or the CPU freq would continue to be stable past 211mhz up until roughly 229mhz at which time you would need to adjust the Nbfreq to 2600 and the diminishing returns start to become exponential
and thanks that sure helped kill that last hour of work lol