Overclocking 8350 with sabertooth 990fx R2.0 and...questions

[slick nick]

Hello everyone...I have been snooping around for a couple of years and have only been an active member for a short while. I have done lots of searches and lots of reading but have been unable to answer a few questions. First off, I am only playing with cpu oc'ng atm, I will mess with the ram and gpu later on...one thing at a time. I started oc'ing using

I was able to reach a stable 4.4 ghz but since I wasn't sure of some of the settings I set it back to 4.2 ghz until I learned more.

First off, what is the difference between oc'ng using the cpu ratio and/or the cpu bus frequency. It seems that some change the ratio and others change the bus freq. What is the benefit of oc'ng with one or the other?

I started with Ram freq at 1600, NB freq at 2200 and HT link at 2600 Mhz. When changing the FSB freq. the 3 previously mentioned settings change (as they should). What I want to know is how much should I allow them to increase before I lower their frequencies. Also I have seen some guys leave the NB and HT at 2200 and others set them both to 2600? Any reason for the different preferences?

I am using prime 95, open hardware monitor and cpuz. My goal is to reach 4.5 or 4.6 Ghz and keep at 60 C or less. I do a lot of video rendering, gaming, etc. so I will have it running for hours or days. I once ran it for almost 3 days straight while rendering. Is 60 C too high of a temp to run for a long time or should I limit myself to lower temps at max stress? Currently at 4.2 Ghz and max stress from prime95 I reach a max socket temp of 50 C and max cpu temp of 39 C. If anyone has any other advice, feel free to chime in.

Thanks,
Nick

 

[slick nick]

So, I was able to overclock it to 4.6 Ghz using default cpu (1.3) and NB voltage (1.2). I ran a stress test for about 7 hours. For the first 3 hours it ran fine. CPU temp never went higher than 54 degrees Celsius and the socket temp never went higher than 64 degrees Celsius. I fell asleep and about 4 hours later I checked it and the chip and socket temp were about 3 degrees Celsius higher each and 2 of the processors were shut down due to failing the stress test. Is this considered decently stable or not? When video rendering, I have never been able to repeat the stresses which prime95 is capable of. Also, are these temps too high or should I try to lower them? 4.5 Ghz was easily reached. I ran a stress test for about 16 hours and everything passed. My cpu temps never went higher than 53 degrees Celsius and the socket never went higher than 63 degrees Celsius.

I was able to OC the NB frequency to 2.3 Ghz...not much of an increase...but when I increase it to slightly over 2.4 Ghz, I fail my stress test. Is the NB frequency dependent on the NB voltage and the CPU voltage or just the NB Voltage?

Edit-On a side note, I rendered a video which took about 2 hours with the cpu OC'd to 4.6 GHz and the cpu temp never went higher then 47 degrees Celsius and the socket peaked at 51 degrees Celsius.

 

[d1nky]

7 hours of a stress test is stable to me, and if you can do heavy rendering without fail then even better. Temps were slightly to the high side, but i doubt you'll hit them in normal usage.

What cooler are you using?

CPUNB is a core independent of the others, when overclocking this you are overclocking the memory controller also. Just like the other cores you will need to adjust the voltage. But remember that this core is the most delicate and gets hot a lot faster, it is also located near the other cores so vcore/cpunb voltage is a balancing act of heat and voltage.

I used to run my FX8350 with 2.6ghz cpunb 1.35v / 2820 HT 1.23v / cpu 5.88ghz 1.52v / 2ghz ram
-But this was on a full custom loop which could handle the heat for several days of crunching.

To your other queries, FSB is useful as tweaking your found overclock to a higher speed and allowing other multipliers higher speeds Or speeds you wish.

For example, 23 x 200 = 4.6ghz (stable) you could increase the FSB three or four notches until the edge of stability. which means faster ram, HT and CPUNB. (26x 205 = 4.71 ghz) ETC

 

[slick nick]

7 hours of a stress test is stable to me, and if you can do heavy rendering without fail then even better. Temps were slightly to the high side, but i doubt you'll hit them in normal usage(When I rendered a video for 2 hours with it OC'd at 4.6 Ghz and the max chip temp was 46-47 degrees Celsius and the socket was about 51 degrees Celsius. Are these temps considered high? During normal use, I never saw higher than 33-35 degrees Celsius and it stays in the 20's for the most part).

What cooler are you using? CM Hyper 212 Evo with second PWM fan...it's in my specs

CPUNB is a core independent of the others, when overclocking this you are overclocking the memory controller also. Just like the other cores you will need to adjust the voltage(Which voltage? The CPU manual V or the CPU/NB manual V?). But remember that this core is the most delicate and gets hot a lot faster, it is also located near the other cores so vcore/cpunb voltage is a balancing act of heat and voltage.

I used to run my FX8350 with 2.6ghz cpunb 1.35v / 2820 HT 1.23v / cpu(This is the cpu/NB voltage right?) 5.88ghz 1.52v(what do these refer to?) / 2ghz ram (what was your cpu ratio and FSB?)
-But this was on a full custom loop which could handle the heat for several days of crunching.

To your other queries, FSB is useful as tweaking your found overclock to a higher speed and allowing other multipliers higher speeds Or speeds you wish.

For example, 23 x 200 = 4.6ghz (stable) you could increase the FSB three or four notches until the edge of stability. which means faster ram, HT and CPUNB. (26x 205 = 4.71 ghz) ETC(I'm not sure I understand this part...What's the difference between having an overclock of 4.6Ghz with 20 x 230 or 23 x 200. I realize that when you increase the FSB, the Ram Freq, the NB Freq and the HT freq also increase but you can manually change those frequencies to make them closer to default. I have also read that OC'ing the HT is useless and creates instability so to leave it at 2.6 Ghz or lower which is default...?)

By the way...thanks for your input.

 

[slick nick]

So I set the FSB back to 225 for a 4.5 Ghz clock speed which had a been stressed by prime95 with the first setting (small FFT's...etc) for about 16 hours with no issue. I was reading up on other stress tests...and I decided to stress the settings with the blend test on Prime95. It failed after only about an hour...what could cause this? CPU ratio 20.0, FSB 225, Ram 1.5 Ghz (rated at 1.6Ghz), CPU/NB Freq 2240 Mhz, HT 2600 Mhz, VCPU 1.32, VCPU/NB 1.2. I am trying to keep the voltages at default to keep the heat down so I realize that I can fix a lot of these issues by increasing the voltages. I have read over several places that OC'ing to 4.5-4.6 Ghz is possible with default voltages...just trying to replicate those results.

Thanks,
Nick

 

[slick nick]

Does anyone know how to fix the blend test failure when OC'ing? It fails the blend test at about 1 hour but passes the stress test when ran for more than 14 hours.

 

[MxPhenom 216]

Does anyone know how to fix the blend test failure when OC'ing? It fails the blend test at about 1 hour but passes the stress test when ran for more than 14 hours.

Blend stresses your memory too. So something involving your memory being unstable is likely causing it to fail.

 

[slick nick]

Blend stresses your memory too. So something involving your memory being unstable is likely causing it to fail.

My default frequency for my memory is 1600 Mhz. I didn't overclock the RAM at all. When it failed the blend test, I reduced it to around 1300 Mhz and there was no change in blend test failure. Are there any other settings that would be affected? How can I make the memory stable?