OC'ing FX-6300 Tonight...

[Johan45]

Back up and running. Running P95 now...I'll post results later.

Thanks for the help

So what was the problem??? Hey I've left the HDMI cable out before and tried for 10 minutes before I realized WTF I had done!!

 

[Sun Tzu]

I think one of the fans was sitting directly on a stick of RAM, which was causing the PC to not fully POST. I don't have my mobo speaker plugged in to hear the beeps, but I'm almost positive that's what it was.

My temps dropped significantly. I left the house with P95 running, and I was stable at 4.55Ghz, with a Vcore set to 1.4. My temps were hovering around 55-56.

My guess is the temps will stay around there. How high can I go Vcore-wise?

 

[Johan45]

If you're quoting the Package temps as 55°c then about 7° more, did you lower you CPU_NB voltage?? That should give you a few more degrees. I thought you said it was at 1.3v which is too high for stock NB clock. Try 1.2 and see if it's still stable.

 

[Sun Tzu]

I dropped it to 1.2, but I bumped the nb speed up one notch, too. Does it make much of a difference?

 

[Johan45]

You'll know when you try to stabilize it. Typically 1.2 v is good up to around 2400 but it depends on the processor and the ram you're using. My 9370 will run over 2600 with 1.23v, auto is 1.45, which is ludirous

 

[Sun Tzu]

Damn. I came home and two cores error'ed out about 25 minutes into the stress test. Should I be seeing this when I'm only setting my Vcore to 1.4?

 

[Johan45]

Ok you dropped the NB voltage to 1.2 and raised the frequency, so try raising the CPU_NBv to 1.23 and see what happens. You do only have the 8 Gb of memory in the system right?

 

[Sun Tzu]

I don't mind dropping the NB frequency back down to 2k. I have no idea what kind of speed bump I'm getting with that anyway.

Yes, I only have the 8 GB of RAM in there right now.

Would it be better to leave the NB frequency up, and just bump the voltage, or keep the voltage at 1.2 and drop the frequency back down to stock?

 

[Johan45]

What my intention was, initially was to see how far you could drop the CPU_NB volts and stay stable if you're after performance set it to 2400 and 1.25v. I just felt that 1.3V to the CpU_Nb was just way too much.

 

[Sun Tzu]

Okay. Got it.

If I'm error'ing out in P95 after about 20-25 minutes, is that directly related to the CPU, or can it be CPU_NB related, too?

How high can you safely go Vcore wise? 1.45?

 

[Waza]

CPU/NB voltage has pretty high variance between different setups. I run mine @2400Mhz 1.1V and haven't had any problems with it and have seen that some people have problems at stock speeds if they try to go bellow 1.4V. Finding your own chips sweet spot could be worth the trouble if you plan to fiddle a lot with different settings.

Okay. Got it.

If I'm error'ing out in P95 after about 20-25 minutes, is that directly related to the CPU, or can it be CPU_NB related, too?

How high can you safely go Vcore wise? 1.45?

It can be cpu/nb related but you would have to test it yourself. And I wouldn't recommend raising the vcore unless your temps are under control. But under proper cooling some people run up to 1.55 vcore. 1.45V is pretty good limit to set for 24/7 use but then again lower is always better

 

[Priller]

My 8350 will do 2400 NB and HT Link at 1.18 volts. With high FSB then it takes 1.3+ volts to run 220+ without an issue. It's actually quite weird that down low it almost doesn't matter but almost any touch in FSB and it needs more CPU/NB voltage.

 

[Sun Tzu]

Hmm...interesting. Would it be better to drop my FSB and start OC'ing with the multiplier? My multi is set to either 17.5 or 18 right now (I forget which).

 

[Waza]

My 8350 will do 2400 NB and HT Link at 1.18 volts. With high FSB then it takes 1.3+ volts to run 220+ without an issue. It's actually quite weird that down low it almost doesn't matter but almost any touch in FSB and it needs more CPU/NB voltage.

I witnessed same kind of behavior when I ran some tests with FSB overclocks.
Other than very slight boost in memory performance I found no advantages to OC FSB over multiplier.

 

[Sun Tzu]

Frustration.

I can't get this thing above 4.4 without it failing P95 and/or rising above 62C. When I bump the Vcore up to 1.4, my package temps go higher than my socket temps.

Ugh.

 

[Sun Tzu]

Settings adjusted slightly. I haven't tried lowering the vCore yet. I'm torn between dropping the vCore, or bumping the multiplier. This whole process has been...well...difficult.

 

[Priller]

Frustration.

I can't get this thing above 4.4 without it failing P95 and/or rising above 62C. When I bump the Vcore up to 1.4, my package temps go higher than my socket temps.

Ugh.

My core temps are always higher then my socket temp.

I also take 62C with a grain of salt because I dont think that is the maximum it can do unless you are talking in sustained operation(server). So I have no issues taking mine up to 78C for short durations. My CPU/board combo throttles at 80C core temps regardless of condition. Both my CPU and mobo have waterblocks and I've tested this down to -5C ambient and 10-13C system temps.

 

[Priller]

I have absolutely no idea what the voltage was doing. I did notice that the CPU fan kept alternating between 0 RPM and the normal RPM. Sorry...I don't know enough about this stuff to give you strong feedback.

Here's where I sit now. I can probably drop my Vcore slightly to see if I can drop a few degrees. I'm done for the night though. What did you want me to try?

Try:

CPU Vcore 1.400 volts.
CPU/NB 1.18-1.225 or 1.250 volts
set ram timings and voltage with stock speeds.
LLC to one under extreme
FSB: 200
Multi: 22
HT link and NB Freq multis to X12. Should give 2400mhz for both.

Then try your testing again at 4.4 or 4.3Ghz

 

[Blaylock]

I hope you didn't miss understand me earlier. I didn't mean that the package should never be higher than the socket. Just that it's unusual. As long as your socket is under 70c and your package in under 65c you're golden. As far as multi OC vs. FSB OC, Multi is a good quick way to OC and to find you max CPU freq. FSB OCing is a bit more frustrating because everytime you raise your FSB your oCing like 4-5 things at once and when hell breaks loose it's much more difficult to ID. For now I would stick to Multi OC. Find you max CPU overclock. Then when you feel like tinkering some more we can address that. Right now 55c on socket and package is great, you still have about 10c headroom.

 

[RGone]

Settings adjusted slightly. I haven't tried lowering the vCore yet. I'm torn between dropping the vCore, or bumping the multiplier. This whole process has been...well...difficult.

Not sure I understand the concept of deciding between raising the multiplier or dropping the Vcore. To cure what? If you raise the multiplier then the cpu speed will rise and likely need more cpu voltage. And if I were going to lower the Vcore, I cerainly would not raise the muitplier.

Unless I am just not reading what you are writing, then no wonder the whole process has been a little 'er difficult.

If you have errors in P95 Blend mode at "X" speed it is most likely the voltage to the cpu is too low. I would say probably 90% of the time. IF you get bad hot when you try to raise the Vcore enough to support the cpu speed you set...then you need to slow the overclock down.

In general overclocking works this way. Raise cpu speed > raise Vcore to support the raised cpu speed. > Test with P95 Blend or similar to determine some form of stability. > If not stable at "X" speed setting then raise the volts to the cpu "provided" so-called safe temps have not been reached. At each increment of overclock it is rinse and repeat until safe temps or volts are exceeded.

FX temps we today consider reasonable for 24/7 use are <70c on cpu/socket and <62c for the core/package temps. I would say that for 24/7 use I would personally not like to have to use more than 1.515Volts to the cpu. So you juggle speed of cpu upward and all the time balance the heat against the speed and ability to run P95 Blend mode.
RGone...