Hey all,
So I just did a quick stability test with my 8350 OC'd at 4.5 Ghz using prime95 and amd overdrive (2 seperate tests). After 20 minutes, it peaked at 62c, which is for sure pushing it. I'm currently at 1.39325v, and I am using a Hyper 212 evo air cooler. Yes, using a $30 air cooler is risky for something like this, but I was wondering if this would be considered stable enough. Would it be worth lowering the voltage more? Any help would be appreciated. Thanks!
First of all, ditch Prime, it gives too much voltage than is required to test stability.
Use Aida64 and OCCT, give those a go and report back with your temps. :D
I can vouch for aida64. Its a very nice tool. Try manually setting the voltage in the bios. Also, try taking the side of the case off for testing. Just to make sure its got a cool ambient temperature.
not bad for a mild oc and only having a hyper 212 not bad at all.
Okay, just tested Aida64... oddly, it only peaked at 54c after stress testing it for 20 mins. Why would the difference be so huge? Thanks, gonna do OCCT now.
the difference is that aida 64 does not make the cpu draw more power while being ran. prime 95 makes the fx family use more power and as a result overheat. for me even with a noctua nh d14 and 2 5k rpm fans for cooling i over heat with prime 95 at 5 ghz and 1.45 volts.
with prime, I do support the fact that it idioticly stresses the cores to the point of insanety.
But the comments that prime "makes the cpu draw more power" I do not understand. With a good configured VDROOP its you that decides how much the voltage the CPU will draw. I read alot of sources that claim this, but as I already configured my digital phases, I cannot seem to replicate those symptoms.
With my sabertooth the voltage of the CPU stays dead in the water when running Prime95. and no cores fail, at least up to 30 minutes as I cba to let it run longer...
for example at 4.7ghz;
1.465v idle, 1.464v running blend stress test. I dont really see the spike there.
AVX can request more amperage from the motherboard's power management, more amperage = more wattage usage even with the same voltage.
Oh,
So actually my phases are pushing the voltages back up So I don't notice the difference? Sounds possible to me..
Well I didn't know that to its full extend. Thanks for explaining.
Amperage has nothing to do with voltage, Amperage is the actual current being applied to a circuit, voltage is the energy potential of a circuit, higher voltage gains you the ability to pump higher amperage through a circuit.
Wattage is the multiplicative of both amperage and voltage.
Just for reference you may want to open up an article on Electron theory and Ohms law. I'm not going to explain High school grade electrical science on the internet.
OP? How you makin' out eh?
Calm down man. No need for that last sentence.
With unregulated phases voltages go down when lots of Amps are being drawn from the motherboard. (its called a vdrop) So it DOES have something to do with each other. The motherboard has to regulate this, thats what phases are for in the first place.
I was thankfull for your input at first. I wasn't talking down on you. I have absolutely no idea what that last sentence is about.
Here I am with an FX-8320, OCed to 4.640GHz at 1.5V with a CoolIT Domino CPU cooler, though I still need a bit more tweaking on the voltage and LLC. Yes, the VCore is pretty high, but that's what happens when you lose the silicon lottery.
I use different stress testing programs, including Prime95, OCCT, AIDA64, and IntelBurnTest. But another program I've been using lately is called y-cruncher: A Multi-Core, optimized Pi calculating program that is ridiculously much faster and efficient than SuperPi or PiFast. It has a built-in stress test using 4 different algorithms, and supports AVX instructions in the latest versions.
http://www.xtremesystems.org/forums/showthread.php?221773-New-Multi-Threaded-Pi-Program-Faster-than-SuperPi-and-PiFast
Whether or not you decide to use y-cruncher is up to you. It can be just as stressful as Prime95, if not more, at certain times. I mostly use it to stress out my RAM and the IMC. Your mileage may vary.
I've seen my CPU cores go as crazy-high as 80C [+/- 65C-70C on the CPU socket] while encoding with Handbrake, just to see how really comfortable this CPU is with very high VCore voltages [1.60V], and so far, it's still alive...lol! But please, don't try that at home.
I say with your current temps, you still got JUST a little bit more breathing room, but yeah...lower voltages and temps wouldn't hurt since AMD chips like the cold. See how far you can go with 1.375V and a slightly higher LLC to prevent VDroop/VDrop. If you're stable at that voltage, then you might have a golden CPU.
I'd say if your machine is still alive and kicking without errors for at least an hour then you should be good. For me, I usually run a stress test for about 12 hours overnight just to be absolutely, completely sure. If it can pass that, then the overclock is good to go.
heh, yeah, I use y-cruncher to test out the CPU speed of hired virtual servers. (it has good linux support)
Did not know it had a windows version (didn't bother I guess)
And yes, with such high voltaged comes alot of heath. Altough 80C is unavoidable, yup, dont try this at home.
How did you disable throtteling? I got it disabled on my sabertooth, but not on my gigabyte board.