AMD 8350 Overclock issue

[Gealii]

I've had my pc built for a couple months now and had never thought about overclocking until i went to liquid cooling the cpu. That being said i have been able to overclock my 8350 to 4.4GHz (not sure why bios is showing 4.61GHz) being stable on prime95 for half an hour with temps never going over 50c.

The main issue i'm having is that i believe i need to up the voltage on my cpu to stabalise it at a higher clock. I think i need to change the cpu Vcore in my bios, but it is set to Auto and i can not figure out how to change it.

Am i trying to change the correct voltage and how would i be able to do it?
BTW specs are in the sig

 

[Johan45]

You probably just have to select it and switch to manual. Give this a read Iit'll help you help us. Is that a revision 3 board?

 

[Zebodog]

Use the "+" and "-" keys to change values.

Watch for throttling on that board if it's a rev 3. Fire up a blend test in Prime 95 and watch your CPU clock; I wouldn't be surprised if you see one or more cores drop speeds.

 

[Gealii]

thanks for the replies so far
I do have the rev 3 mobo
I am still looking over the link Johan45 gave and am now going to try the tip Zebodog gave

EDIT: thanks Zebodog that changed the voltages just as i wanted them to

 

[caddi daddi]

gealii, that link johan45 put up is still a very good read and applys very well to the 8350.

 

[Gealii]

yea the link was very helpful, so far I've managed a 4.7GHz prime95 stable, 4.8 froze after 15min changed some settings and it will get tested again in the morning also I'm hoping to reach 5.0 by tomorrow night in which a much longer prime test will be in the order due to 5.0 being my goal

 

[Mandrake4565]

Gealii, just watch your temps 70 on the cpu and 60's on the package. The jump in Cpu V core from 4.8 to 5.0 is known to be very large, unless you have a very good cpu. I know in my case it was a big jump.

 

[Gealii]

Gealii, just watch your temps 70 on the cpu and 60's on the package. The jump in Cpu V core from 4.8 to 5.0 is known to be very large, unless you have a very good cpu. I know in my case it was a big jump.

Well I'm stable at 4.8GHz rite now but watching temps i reached 65C during prime95 testing according to hwmonitor is this the temp that i should leave in the 60s? and stop going higher until i buy better cooling fans

 

[Zebodog]

^^^^

I had the same problem when I ram my UD3; take a look at your 12V reading.

HW Monitor provided questionable results on my 990fx-ud3.

 

[Mandrake4565]

We usually find the cpu socket temp will be 8-20c higher then the package temp, depending on the case. I would be leaning towards what Zebodog said, it may not be an accurate temp reading.

 

[RGone]

However that CPU "package" temp of 65c is likely accurate and beyond the 60c we have been setting as a cut-off point for package temps. We have heard of those wanting to fudge that package temp higher but a hihger package temps, IS a temp I will not tell a user to go for. With the cooling I have the Package temp seldom even breaks 50c so I know I don't worry but once it gets beyond 60c it is there because the package or core temp is warm and voltage and current is going to the cpu.

It would be an awesome piece of silicon to be stable at 4.8Ghz with only a mini-loop like a H100i doing the cooling. Well the radiator in and ice chest maybe. Hehehe. And then that wall or step at about 4.8Ghz that takes what seems a ton of Vcore to bust on thru to 4.9 and 5.0Ghz is there for at least 99% of us so far. Prime 95 Blend mode for two hours can be a difficult undertaking when 4.7 and 4.8Ghz are reached. So I would pay very close attentioin to the readings of the "latest" version of HWMonitor. and I would UNinstall any other programs that can compete with each other for polling time for temps and voltages. Like AiSuite for one.
RGone...

 

[Gealii]

However that CPU "package" temp of 65c is likely accurate and beyond the 60c we have been setting as a cut-off point for package temps. We have heard of those wanting to fudge that package temp higher but a hihger package temps, IS a temp I will not tell a user to go for. With the cooling I have the Package temp seldom even breaks 50c so I know I don't worry but once it gets beyond 60c it is there because the package or core temp is warm and voltage and current is going to the cpu.

It would be an awesome piece of silicon to be stable at 4.8Ghz with only a mini-loop like a H100i doing the cooling. Well the radiator in and ice chest maybe. Hehehe. And then that wall or step at about 4.8Ghz that takes what seems a ton of Vcore to bust on thru to 4.9 and 5.0Ghz is there for at least 99% of us so far. Prime 95 Blend mode for two hours can be a difficult undertaking when 4.7 and 4.8Ghz are reached. So I would pay very close attentioin to the readings of the "latest" version of HWMonitor. and I would UNinstall any other programs that can compete with each other for polling time for temps and voltages. Like AiSuite for one.
RGone...

I do have the latest hwmonitor and basically what i took from this was without a better cooling solution i should stop while I'm ahead and after an hour of prime95 temps never went over 57 @ 4.7 so i mite just bump it down to there til i get more cooling

 

[Mandrake4565]

Gealii, these chips are heat mongers, unless you plan on going for a full custom water setup you are probably going to run into a heat wall before you have issues with a mhz wall. I can do prime blend 2 hours @ 4.7 with 1.46 v which has proven plenty stable for my needs. Going higher for my chip is where it really needs me to step on it's neck and make it squeal with the v core. I have mine stable 2 hours prime at 5.1 though I never run it there for anything other then benching. Unless you are planning on benching it you may find that 4.7 is plenty fast.

 

[Gealii]

Gealii, these chips are heat mongers, unless you plan on going for a full custom water setup you are probably going to run into a heat wall before you have issues with a mhz wall. I can do prime blend 2 hours @ 4.7 with 1.46 v which has proven plenty stable for my needs. Going higher for my chip is where it really needs me to step on it's neck and make it squeal with the v core. I have mine stable 2 hours prime at 5.1 though I never run it there for anything other then benching. Unless you are planning on benching it you may find that 4.7 is plenty fast.

Custom water cooling may come around January, but for now i have to stick with the h100i. I guess you could call it benching as i currently have the fastest pc out of my IT friends, and i want to keep it that way only willing to put 1500usd more into the rig not counting the custom water cooling, so any improvement to bump up even momentarily is worth it

 

[caddi daddi]

4.7? 4.6 is a very good 24/7 clock for air or a all in one.
most of us on big water don't run much more than a 4.8 24/7 clock.

 

[Mandrake4565]

Like Caddi said even if you go big water the amount of Cpu V required for most chips to run 24/7 at 4.8 or above, stable really doesn't make sense.

 

[RGone]

Like Caddi said even if you go big water the amount of Cpu V required for most chips to run 24/7 at 4.8 or above, stable really doesn't make sense.

That is pretty darn accurate. C_D was so used to big Vcore that he pulled the same stunt I did on the FatLady and had to move back to CHV to clear my head. The Bulldozers you could bulldoze but the Visheras need a little more gentle hand on the way up in Vcore and then

The image below is the reason most of us say that the big numbers are little more than e-pen stuff and hard to support hour after hour and day after day.

 

[ShrimpBrime]

I would think the All in one would do better than 4.6ghz for a daily clock. But this also depends on the pick of the Cpu. Mine was running a stock cooler, modified Fan 4.4ghz 1.4v NP. It was within temp limits. Anything past 1.4250v seemed to get hard on the ol' air cooling.

From benching these processors, I've found nothing but mhz helps the darn thing.

For example....

If you ran the FX-8xxx with all but one core off, making it a single core, then dig out an old FX-55 and run them both at the same clock speed, they'll bench pretty close to each other.

So @ OP, Push that darn liquid cooling for all it's worth. If your going to overclock.... Do it right.

 

[Johan45]

I would think the All in one would do better than 4.6ghz for a daily clock. But this also depends on the pick of the Cpu. Mine was running a stock cooler, modified Fan 4.4ghz 1.4v NP. It was within temp limits. Anything past 1.4250v seemed to get hard on the ol' air cooling.

.

The AIO I have is a decent unit but my chip takes alot of power to get it moving. Mandrake (on water)was saying 1.46 for his 4.6 OC Mine is at 1.474 and as I go up in Hz the voltage need increases dramaticaly. I'm sure that the heat plays a part in the extra neede volts. But at nearly 4.9 I needed 1.58v to remain stable and my temps were at 63°c. I have to admit that Tt AIO i'm using does a really good job for what it is. So the trade off of volts and heat for the extra speed really isn't worth it for 24/7 use.

 

[Mandrake4565]

1.46 for his 4.6 OC

Correction 1.46 for 4.7
I have to agree with you with the "cold" statement, you can definitely use less Cpu v when the chips are cooler. When I was making a Cinebench run @ around 5.4 this winter I had my office down to 40f and I was able to use 1.59 v core. My max temps were only 61c I tried re running it recently with the weather hotter now, room temps are 70f I couldn't even get off a run at 5.3 with 1.59 cpu v with out it crashing and my temps were getting into the mid 70's C.

If you ran the FX-8xxx with all but one core off, making it a single core, then dig out an old FX-55 and run them both at the same clock speed, they'll bench pretty close to each other.

Shrimp in single core benches my 955 @ 1 ghz lower will destroy my 8350 in most instances.