8350 Overclocking help

[Sintezza]

try Aida 64 for stress testing.

 

how did you applied your thermal paste?

[BigChimp97]

try Aida 64 for stress testing.

 

how did you applied your thermal paste?

My 30 day trial with Aida has ran out

[WaitForMeImOnDialUp]

i would keep trying, something must be wrong i bought my 8350 in october and hit 4.8 stable and my motherboard wasnt as good as yours.ibt will raise temps way higher than aida or occt. try occt.

 

on your IBT, do you have your stress level set to "standard" ? and "threads" set to all ?

 

i dont mean to sound like a retard, but do you have Windows power management set to "performance mode" ?

[D3MON]

4.4Ghz is... very dissapointing. Might aswell leave it at stock and let it turbo boost upto 4.4 tbh

OK here's my guide (had a terrible headache earlier so couldnt summon the strength to post allot.)

I highly doubt its the limit of the chip (though i think i reached it with my 8120 but i digress)

this all assumes that you have epu turned off, coolnquiet is fine to leave on, turbo must be off and ram at rated clock speed.

monitoring

monitor voltage with cpu-z idle at desktop and then run cinebench and see how it v-droops, also use core temp to monitor temps, dont worry about idle temps its peaks you wanna keep an eye on.

try 1.35v in bios at a 21x multi for 4.2ghz

boot  into windows and see what it is idle, if its 1.375 or something just assume its 1.35 and cpu-z is overeading by 25mv (ive spent long enough tinkering with fx's just go with me)

starting tests

run cinebench and see if the voltage droops, if it drops to say 1.325(ish) then reboot/bios/loadline calibration and bump it from medium>high or high>very high.. reboot and test idle/cinebench load voltages again. adjust as needed.

assuming your voltages under load are pretty much what you set in bios run cinebench with core temp and see what your max temps are.

if your temps are above 70*c then you're obviously running too much voltage or you have a leaky chip.

what i do in those^ scenarios is start around a lower clock speed like 4ghz and see how low i can get the load voltage to be and how it effects temps

so maybe a 20x multi at 1.3v, boot/cpu-z/cinebench and see what teh actual load voltage is and again adjust voltage as nessesary, the higher the default voltage the more effect loadline has, so 1.3v on ultra loadline might be 1.35 under load, however 1.4v on the same ultra loadline might reach 1.475v under load so monitor it!

clocking up.

so you have a starting point where you have a voltage set, that stays at that point under load (give or take) and doesnt die before finishing cinebench AND is udner 70*c under load. (passes cinebench, well get into stress testing later)

now download AMD Overdrive, go to the section where you can adjust the multiplier and bump it up one to say 21.5 4.3ghz

run cinebench and see temps, if windows aero greys out or it crashes reboot but with that same 21.5 multi and just run cinebench again, if it still crashes or greys out you've hit a voltage limit and it needs to be raised, IF it passes cinebench fine with no issue then its possibly amd overdrive being picky and all multi bumps should be carried on in bios from here.

if raising the multi doesnt cause any issues and it still completes the test bump it again, keep bumping the multi until it greys out or crashed.

if it does crash tehn go into bios and try 1.375v, boot into windows, check idle voltage and then cinebench load voltage, adjust LLC as nessesary and monitor temps again.

if that still fails try 1.4v in bios, boot, idle voltage/cinebench voltage, you get the idea...

if that fails try 1.425 but you MUST check the load voltages, too much vrdroop can cause itt o fail, too much voltage gain due to LLC can cause extreme temps and vrm throttling, whenever you hear somone getting low benchmark scores despite high voltage and high clock speeds-throttling.

slowly slowly catchy monkey

by bumping the multi and only under a failure increasing the voltages WHILE checking vdroop under load AND temps you should be able to pass 4.5ghz

i'll reiterate my point that on big power hungry cpu's like fx8's and 2011 sockets going for every last mhz is usually a waste of time, power, silicon degradation plus cooling can become a problem.

I would set it to either 22.5 or 23 multi and try 1.4v (under load)

once you get it to pass cinebench at these speeds run prime 95 blend and watch temps, if it goes over 70*c thats not good and the voltage (and maybe overclock) will need to be dailed back, but lets see first.

let me know how you get on. and ill guide you once i know what you've done

[-rascal-]

What are your other BIOS settings (would help A LOT if you can share these info)?

 

CPU/NB Voltage?

CPU VDDA Voltage?

 

What is your CPU/NB Frequency set to?

What is your HT Link speed set to?

 

CPU Load Line Calibration?

CPU/NB Load Line Calibration?

CPU Voltage Over-Current Protection?

CPU/NB Voltage Over-Current Protection?

CPU PWM Phase Control?

 

 

You may have indeed got a piece of the silicon wafer where it can't overclock very far...but don't give up yet.

 

 

TIPS

1. As you increase your overclock, the CPU/NB voltage needs to be increased as well. Not JUST the Core Voltage.
2. When you are stress testing, do you notice your Core Voltage dropping lower than what you have set in the BIOS? If so, you need to adjust your Load Line Calibration.
3. Bumping the CPU-VDDA voltage up a little bit can help compensate for voltage drooping under load.
4. As mentioned before, DISABLE APM (Application Power Management) Master Mode, and ENABLE HPC (High Performance Computing) Mode.
5. DISABLE Cool'n'Quiet and C1E for the time being; you can re-enable them back afterwards if you wish.
6. AM3+ FX chips typically can't run stable when the CPU/NB Frequency is much higher than 2700MHz or 2800MHz. Setting it to 2600MHz or dropping it down to 2400MHz won't have much performance impact.
7. The HT Link Frequency should match the CPU/NB Frequency (or just a step below) for optimal performance.
8. If your RAM may be causing your overclock to be unstable, bump the voltage up slightly (i.e. 1.5V --> 1.55V).
9. Print-Screen (F12) to save a screenshot of the BIOS to a USB drive is VERY handy

 

I may be one of the luckier ones with a better overclocking FX-8350, as I'm able to do 4.8GHz with 1.4V (4.9GHz with ~1.45V though). I had increase the CPU/NB and tinker with the Load Line Calibration settings.

 

You ideally want to use the Load Line Calibration setting(s) where the voltage will stay as constant as possible to what you've set in the BIOS.

If you voltage droops down to low under load, then your overclock will fail -- not enough voltage to keep stable.

If your voltage is too high, then you'll be using more voltage than needed, as a result, excessive heat.

 

 

You may need to bring your voltage up to 1.45V ~ 1.48V for your particular chip, but...only testing will hail it down to the decimal point. Otherwise, here are my personal BIOS settings that I used for 4.8GHz. (mind you these screenshots are almost a year old, so they are not the most up-to-date, but they can be useful, nonetheless)

 

NOTES:

• You can leave SB and NB voltage at stock...I was fiddling with some of there things at the time, so I had those over-volted as well.
• You typically shouldn't need 1.300V for the CPU/NB for 4.4GHz !~ 4.6GHz. I even ended up just using 1.25V for 4.8GHz / 4.9GHz.
• Stock CPU-VDDA is 2.5V. I was using 2.6V, but ~2.55V is usually enough. 
• Choose the Load Line Calibration settings that will make the voltage as stable as possible FOR YOUR PARTRICULAR CPU

 

 

 

 

 

 

 

 

@Priller, you got anything else?

[BigChimp97]

OK here's my guide (had a terrible headache earlier so couldnt summon the strength to post allot.)

I highly doubt its the limit of the chip (though i think i reached it with my 8120 but i digress)

this all assumes that you have epu turned off, coolnquiet is fine to leave on, turbo must be off and ram at rated clock speed.monitoring

monitor voltage with cpu-z idle at desktop and then run cinebench and see how it v-droops, also use core temp to monitor temps, dont worry about idle temps its peaks you wanna keep an eye on.

try 1.35v in bios at a 21x multi for 4.2ghz

boot  into windows and see what it is idle, if its 1.375 or something just assume its 1.35 and cpu-z is overeading by 25mv (ive spent long enough tinkering with fx's just go with me)starting tests

run cinebench and see if the voltage droops, if it drops to say 1.325(ish) then reboot/bios/loadline calibration and bump it from medium>high or high>very high.. reboot and test idle/cinebench load voltages again. adjust as needed.

assuming your voltages under load are pretty much what you set in bios run cinebench with core temp and see what your max temps are.

if your temps are above 70*c then you're obviously running too much voltage or you have a leaky chip.

what i do in those^ scenarios is start around a lower clock speed like 4ghz and see how low i can get the load voltage to be and how it effects temps

so maybe a 20x multi at 1.3v, boot/cpu-z/cinebench and see what teh actual load voltage is and again adjust voltage as nessesary, the higher the default voltage the more effect loadline has, so 1.3v on ultra loadline might be 1.35 under load, however 1.4v on the same ultra loadline might reach 1.475v under load so monitor it!clocking up.

so you have a starting point where you have a voltage set, that stays at that point under load (give or take) and doesnt die before finishing cinebench AND is udner 70*c under load. (passes cinebench, well get into stress testing later)

now download AMD Overdrive, go to the section where you can adjust the multiplier and bump it up one to say 21.5 4.3ghz

run cinebench and see temps, if windows aero greys out or it crashes reboot but with that same 21.5 multi and just run cinebench again, if it still crashes or greys out you've hit a voltage limit and it needs to be raised, IF it passes cinebench fine with no issue then its possibly amd overdrive being picky and all multi bumps should be carried on in bios from here.

if raising the multi doesnt cause any issues and it still completes the test bump it again, keep bumping the multi until it greys out or crashed.

if it does crash tehn go into bios and try 1.375v, boot into windows, check idle voltage and then cinebench load voltage, adjust LLC as nessesary and monitor temps again.

if that still fails try 1.4v in bios, boot, idle voltage/cinebench voltage, you get the idea...

if that fails try 1.425 but you MUST check the load voltages, too much vrdroop can cause itt o fail, too much voltage gain due to LLC can cause extreme temps and vrm throttling, whenever you hear somone getting low benchmark scores despite high voltage and high clock speeds-throttling.slowly slowly catchy monkey

by bumping the multi and only under a failure increasing the voltages WHILE checking vdroop under load AND temps you should be able to pass 4.5ghz

i'll reiterate my point that on big power hungry cpu's like fx8's and 2011 sockets going for every last mhz is usually a waste of time, power, silicon degradation plus cooling can become a problem.

I would set it to either 22.5 or 23 multi and try 1.4v (under load)

once you get it to pass cinebench at these speeds run prime 95 blend and watch temps, if it goes over 70*c thats not good and the voltage (and maybe overclock) will need to be dailed back, but lets see first.

let me know how you get on. and ill guide you once i know what you've done

Wow... thats the longest reply ive ever had! Thanks for so much info will definitely follow this and get back to you, to see how successful i was!

[BigChimp97]

What are your other BIOS settings (would help A LOT if you can share these info)?

 

CPU/NB Voltage?

CPU VDDA Voltage?

 

What is your CPU/NB Frequency set to?

What is your HT Link speed set to?

 

CPU Load Line Calibration?

CPU/NB Load Line Calibration?

CPU Voltage Over-Current Protection?

CPU/NB Voltage Over-Current Protection?

CPU PWM Phase Control?

 

 

You may have indeed got a piece of the silicon wafer where it can't overclock very far...but don't give up yet.

 

 

TIPS

1. As you increase your overclock, the CPU/NB voltage needs to be increased as well. Not JUST the Core Voltage.
2. When you are stress testing, do you notice your Core Voltage dropping lower than what you have set in the BIOS? If so, you need to adjust your Load Line Calibration.
3. Bumping the CPU-VDDA voltage up a little bit can help compensate for voltage drooping under load.
4. As mentioned before, DISABLE APM (Application Power Management) Master Mode, and ENABLE HPC (High Performance Computing) Mode.
5. DISABLE Cool'n'Quiet and C1E for the time being; you can re-enable them back afterwards if you wish.
6. AM3+ FX chips typically can't run stable when the CPU/NB Frequency is much higher than 2700MHz or 2800MHz. Setting it to 2600MHz or dropping it down to 2400MHz won't have much performance impact.
7. The HT Link Frequency should match the CPU/NB Frequency (or just a step below) for optimal performance.
8. If your RAM may be causing your overclock to be unstable, bump the voltage up slightly (i.e. 1.5V --> 1.55V).
9. Print-Screen (F12) to save a screenshot of the BIOS to a USB drive is VERY handy

 

I may be one of the luckier ones with a better overclocking FX-8350, as I'm able to do 4.8GHz with 1.4V (4.9GHz with ~1.45V though). I had increase the CPU/NB and tinker with the Load Line Calibration settings.

 

You ideally want to use the Load Line Calibration setting(s) where the voltage will stay as constant as possible to what you've set in the BIOS.

If you voltage droops down to low under load, then your overclock will fail -- not enough voltage to keep stable.

If your voltage is too high, then you'll be using more voltage than needed, as a result, excessive heat.

 

 

You may need to bring your voltage up to 1.45V ~ 1.48V for your particular chip, but...only testing will hail it down to the decimal point. Otherwise, here are my personal BIOS settings that I used for 4.8GHz. (mind you these screenshots are almost a year old, so they are not the most up-to-date, but they can be useful, nonetheless)

 

NOTES:

• You can leave SB and NB voltage at stock...I was fiddling with some of there things at the time, so I had those over-volted as well.
• You typically shouldn't need 1.300V for the CPU/NB for 4.4GHz !~ 4.6GHz. I even ended up just using 1.25V for 4.8GHz / 4.9GHz.
• Stock CPU-VDDA is 2.5V. I was using 2.6V, but ~2.55V is usually enough. 
• Choose the Load Line Calibration settings that will make the voltage as stable as possible FOR YOUR PARTRICULAR CPU

 

 

 

 

 

 

 

 

@Priller, you got anything else?

Woah thanks man, will defo have to use this aswell!! Make a little mix of this and the other guys help

[BigChimp97]

OK here's my guide (had a terrible headache earlier so couldnt summon the strength to post allot.)

I highly doubt its the limit of the chip (though i think i reached it with my 8120 but i digress)

this all assumes that you have epu turned off, coolnquiet is fine to leave on, turbo must be off and ram at rated clock speed.

monitoring

monitor voltage with cpu-z idle at desktop and then run cinebench and see how it v-droops, also use core temp to monitor temps, dont worry about idle temps its peaks you wanna keep an eye on.

try 1.35v in bios at a 21x multi for 4.2ghz

boot  into windows and see what it is idle, if its 1.375 or something just assume its 1.35 and cpu-z is overeading by 25mv (ive spent long enough tinkering with fx's just go with me)

starting tests

run cinebench and see if the voltage droops, if it drops to say 1.325(ish) then reboot/bios/loadline calibration and bump it from medium>high or high>very high.. reboot and test idle/cinebench load voltages again. adjust as needed.

assuming your voltages under load are pretty much what you set in bios run cinebench with core temp and see what your max temps are.

if your temps are above 70*c then you're obviously running too much voltage or you have a leaky chip.

what i do in those^ scenarios is start around a lower clock speed like 4ghz and see how low i can get the load voltage to be and how it effects temps

so maybe a 20x multi at 1.3v, boot/cpu-z/cinebench and see what teh actual load voltage is and again adjust voltage as nessesary, the higher the default voltage the more effect loadline has, so 1.3v on ultra loadline might be 1.35 under load, however 1.4v on the same ultra loadline might reach 1.475v under load so monitor it!

clocking up.

so you have a starting point where you have a voltage set, that stays at that point under load (give or take) and doesnt die before finishing cinebench AND is udner 70*c under load. (passes cinebench, well get into stress testing later)

now download AMD Overdrive, go to the section where you can adjust the multiplier and bump it up one to say 21.5 4.3ghz

run cinebench and see temps, if windows aero greys out or it crashes reboot but with that same 21.5 multi and just run cinebench again, if it still crashes or greys out you've hit a voltage limit and it needs to be raised, IF it passes cinebench fine with no issue then its possibly amd overdrive being picky and all multi bumps should be carried on in bios from here.

if raising the multi doesnt cause any issues and it still completes the test bump it again, keep bumping the multi until it greys out or crashed.

if it does crash tehn go into bios and try 1.375v, boot into windows, check idle voltage and then cinebench load voltage, adjust LLC as nessesary and monitor temps again.

if that still fails try 1.4v in bios, boot, idle voltage/cinebench voltage, you get the idea...

if that fails try 1.425 but you MUST check the load voltages, too much vrdroop can cause itt o fail, too much voltage gain due to LLC can cause extreme temps and vrm throttling, whenever you hear somone getting low benchmark scores despite high voltage and high clock speeds-throttling.

slowly slowly catchy monkey

by bumping the multi and only under a failure increasing the voltages WHILE checking vdroop under load AND temps you should be able to pass 4.5ghz

i'll reiterate my point that on big power hungry cpu's like fx8's and 2011 sockets going for every last mhz is usually a waste of time, power, silicon degradation plus cooling can become a problem.

I would set it to either 22.5 or 23 multi and try 1.4v (under load)

once you get it to pass cinebench at these speeds run prime 95 blend and watch temps, if it goes over 70*c thats not good and the voltage (and maybe overclock) will need to be dailed back, but lets see first.

let me know how you get on. and ill guide you once i know what you've done

Ok, so i got it upto 4.8Ghz with cinebench at 1.4 volts and it didnt crash. However i went over to prime and two cores failed. So i upped the voltage progressively upto 1.45, i even dropped the multiplier to 235 and its still failing. What should i do next?

[BigChimp97]

This is what is happening every time on Prime, even at 4.6Ghz

[BigChimp97]

OK here's my guide (had a terrible headache earlier so couldnt summon the strength to post allot.)

I highly doubt its the limit of the chip (though i think i reached it with my 8120 but i digress)

this all assumes that you have epu turned off, coolnquiet is fine to leave on, turbo must be off and ram at rated clock speed.

monitoring

monitor voltage with cpu-z idle at desktop and then run cinebench and see how it v-droops, also use core temp to monitor temps, dont worry about idle temps its peaks you wanna keep an eye on.

try 1.35v in bios at a 21x multi for 4.2ghz

boot  into windows and see what it is idle, if its 1.375 or something just assume its 1.35 and cpu-z is overeading by 25mv (ive spent long enough tinkering with fx's just go with me)

starting tests

run cinebench and see if the voltage droops, if it drops to say 1.325(ish) then reboot/bios/loadline calibration and bump it from medium>high or high>very high.. reboot and test idle/cinebench load voltages again. adjust as needed.

assuming your voltages under load are pretty much what you set in bios run cinebench with core temp and see what your max temps are.

if your temps are above 70*c then you're obviously running too much voltage or you have a leaky chip.

what i do in those^ scenarios is start around a lower clock speed like 4ghz and see how low i can get the load voltage to be and how it effects temps

so maybe a 20x multi at 1.3v, boot/cpu-z/cinebench and see what teh actual load voltage is and again adjust voltage as nessesary, the higher the default voltage the more effect loadline has, so 1.3v on ultra loadline might be 1.35 under load, however 1.4v on the same ultra loadline might reach 1.475v under load so monitor it!

clocking up.

so you have a starting point where you have a voltage set, that stays at that point under load (give or take) and doesnt die before finishing cinebench AND is udner 70*c under load. (passes cinebench, well get into stress testing later)

now download AMD Overdrive, go to the section where you can adjust the multiplier and bump it up one to say 21.5 4.3ghz

run cinebench and see temps, if windows aero greys out or it crashes reboot but with that same 21.5 multi and just run cinebench again, if it still crashes or greys out you've hit a voltage limit and it needs to be raised, IF it passes cinebench fine with no issue then its possibly amd overdrive being picky and all multi bumps should be carried on in bios from here.

if raising the multi doesnt cause any issues and it still completes the test bump it again, keep bumping the multi until it greys out or crashed.

if it does crash tehn go into bios and try 1.375v, boot into windows, check idle voltage and then cinebench load voltage, adjust LLC as nessesary and monitor temps again.

if that still fails try 1.4v in bios, boot, idle voltage/cinebench voltage, you get the idea...

if that fails try 1.425 but you MUST check the load voltages, too much vrdroop can cause itt o fail, too much voltage gain due to LLC can cause extreme temps and vrm throttling, whenever you hear somone getting low benchmark scores despite high voltage and high clock speeds-throttling.

slowly slowly catchy monkey

by bumping the multi and only under a failure increasing the voltages WHILE checking vdroop under load AND temps you should be able to pass 4.5ghz

i'll reiterate my point that on big power hungry cpu's like fx8's and 2011 sockets going for every last mhz is usually a waste of time, power, silicon degradation plus cooling can become a problem.

I would set it to either 22.5 or 23 multi and try 1.4v (under load)

once you get it to pass cinebench at these speeds run prime 95 blend and watch temps, if it goes over 70*c thats not good and the voltage (and maybe overclock) will need to be dailed back, but lets see first.

let me know how you get on. and ill guide you once i know what you've done

sorry for all this spam, but i've managed to get it stable at 4.5 Ghz on prime with a voltage of 1.4. If i try to make the clock any higher even at 1.45 volts it isnt stable. I havent got a clue why but any help would be great