AMD FX 8350 OC 5Ghz

D

Deacon_Jones1988

Limp Gawd
Joined
Dec 8, 2012
Messages
159
I was wondering if I could get some input from fellow Piledriver owners on this. I have an FX 8350 right now it's overclocked to 5.0ghz. Strange problem I have here.... I can play games no prob, with no heat issues. (47*C max so far) I have played Skyrim, the Witcher 2, Bioshock Infinite, SC2 HOS, and a few others. I have had not a single hiccup. Except when i do stress testing. Prime95 will only run half of my cores, IntelBurntest either stops, with error message, AIDA64 stops at 3 minutes says due to hardware failure. To me this is strange considering i can game. I know some people I have consulted with previously on overclocking in general wouldn't consider the system stable since the tests won't run. Im running a H100i cooler, my voltages are 1.416v. I only have high temps with IBT, i have seen temps as high 71*C (is this too high for this chip? can't get an official answer on that one.) any feedback would be greatly appreciated. :)
 
Try AMD overdrive built in stress tester see if that fails.

I think you can google the prime95 thingie heard about that before.
 
Gaming does not stress your CPU anywhere near P95 or IBT.

There are many instances where an overclock will be technically unstable, but stable enough to play games because games, especially the ones you're playing, utilize at most 4 cores, and mostly only 2. There's a gigantic difference between 2 to 4 cores not fully loaded and 8 cores fully loaded.
 
^ +1

You'll have to drop your OC a bit for it to be completely stable.
You can also use Povray's latest beta to test stability. The beta is fully multithreaded and has a built in benchmark.
http://www.povray.org/beta/
 
Temps: Monitor package temps, and core temps should not be more than 10*C more. Max package temp is 62*C, max core temp is 71*C. Your board may need fans on the VRM heatsink to run that high with any sort of stability, adding some 80mm fans on top of my sabertooth R2.0 dropped my package temps by almost 10*C.

Bottom line, 1.46V is not enough to be stable at 5.0ghz, even the best of the best 8350's usually take at least 1.500V, 1.530-1.550 is average. My 8320 takes 1.560V but it had a high stock VID of 1.380. H100i may not even be able to handle the heat at 5.0+, I couldn't do it with the Kraken X60. It took custom water to get to that point. If you want to use IBT, download the AVX version, its 2.54 as opposed to the non AVX 2.53. If you get negative numbers as an output it is still unstable even if it doesn't crash.
 
LigTasm said:
Bottom line, 1.46V is not enough to be stable at 5.0ghz,
Click to expand...
Agreed.

...even the best of the best 8350's usually take at least 1.500V, 1.530-1.550 is average. My 8320 takes 1.560V but it had a high stock VID of 1.380.
Click to expand...

Going a little OT here, but is that with Core Performance Boost/Turbo/Whatever its called disabled? I know my stock voltage dropped to 1.3 after I disabled it and rebooted. It seemed to make overclocking easier, at 1.45v the 8120 would do 4.6 GHz and the 8320 will do 4.8. With it on going past 4.5 was always unstable when overclocking via the multiplier. It makes me wonder if that was the reason [H]OCP and a few other reviewers topped out at 4.5 in their Vishera reviews. I have a hard time believing there is a huge amount of variation in what these chips are capable of seeing how my 8120 and 8320 are roughly equal in OC potential (with the 8320's improved efficiency allowing it to go a little further on the same voltage).
 
Definitely take heed of the voltage advice but also be sure to test your memory. Go ahead and try those tests with some other sticks if you have some spare or test them one by one if you don't. It's a very easy thing to eliminate from the list of potential problems. I had similar weirdness on my i7 920 system where games would work fine but Prime and IBT crapped out. I'd also get the occasional random BSoD just putzing around on the desktop not doing anything demanding which was frustrating. Turned out to be the memory even though it would pass quick rounds of MemTest and such, was a head scratcher for awhile!
 
Do those have power saving options in the bios (like speedstep, c1e etc like the intels do) that you can try turning off? My 2600k wasn't stable at anything over 4.3 with those on and had similar issues (not loading all cores during benching). Once I turned them off, I can run prime95 all day at 4.9ghz.

Sorry for not knowing the amd options... I haven't owned one since my stellar opty 165.
 
Turn basically everything off. Cool N Quiet too. 5 GHz might not be possible on your chip and I am one box those that believe if you can't pass OCCT (my personal favorite) IBT or P95, then your overclock is not stable. Mine will run 4.8 all day on IBT but crashes in OCCT so I'm stuck at 4.6 because that's the speed it passes everything.
 
You will need to increase the voltage a little bit at least. However, do you have LLC enabled/ disabled and at what setting is LLC set at if enabled?
 
if the computer is stable for the things you use it for why do you care if it can't stand up to a stress test that doesn't replicate your real world usage?

only time I care about stress testing stability is when I am giving the computer to someone else.
 
if the computer is stable for the things you use it for why do you care if it can't stand up to a stress test that doesn't replicate your real world usage?
Click to expand...

Probably to eliminate the 1 random BSOD per month or something like that..
 
DeathFromBelow said:
Agreed.



Going a little OT here, but is that with Core Performance Boost/Turbo/Whatever its called disabled? I know my stock voltage dropped to 1.3 after I disabled it and rebooted. It seemed to make overclocking easier, at 1.45v the 8120 would do 4.6 GHz and the 8320 will do 4.8. With it on going past 4.5 was always unstable when overclocking via the multiplier. It makes me wonder if that was the reason [H]OCP and a few other reviewers topped out at 4.5 in their Vishera reviews. I have a hard time believing there is a huge amount of variation in what these chips are capable of seeing how my 8120 and 8320 are roughly equal in OC potential (with the 8320's improved efficiency allowing it to go a little further on the same voltage).
Click to expand...

Yes, turbo boost must be disabled for any real overclocking to happen. I can run my chip up to 5.2ghz with Cool 'n Quiet, C1E, C6 still enabled. This brings my idle wattage down from 260W to 190W. APM must be disabled as well or you get throttling, HPC depends on the board itself. As far as the overclocks reviewers got, it is probably mainly due to heat. These chips run extremely hot and it you don't cool the VRM section of the board you most likely will not get anywhere near max OC.

For example, I could barely get stable at 4.6ghz with my Sabertooth board and no VRM cooling. Adding a single 80mm fan on top of the heatsink immediately allowed me to ramp up to 1.575V and find the maximum my chip would boot at and do some stress testing.

Also, Vishera can be picky about FSB/HT/NB speeds. NB must be equal or faster than the RAM, so if you're running a high clocked kit you may have to add some NB voltage to allow you to bump it up.

Some people like to try overclocking them multi only, I have had better success with FSB+multi. I can run 200x25 for 5.0ghz at 1.560V, or I can run 250x20.5 for 5150mhz with 1.545V. However, this takes a lot more tweaking as all of the other speeds get out of whack as well and most new overclockers don't want to mess with it.
 
drescherjm said:
Probably to eliminate the 1 random BSOD per month or something like that..
Click to expand...

Far Cry 3 would BSOD after a couple hours of play. I'd never remember to save and suddenly black screen, thudding noise, restart PC and redo the game. Got old quick. No issues with the other 600+ Steam games, Origin, etc.
 
I have downclocked the cpu back to 4.8ghz. It is completely stable at this clock, with all of the aforementioned stress tests. To begin with what I did was start overclocking at 4.4, for this i incresed the multiplier to x22, left the volts on auto, stressed, passed. Then i increased multiplier to x23 put the volts too 1.375v, stressed for over an hour on Prime95, 10 passes on IBT, 1hr 4mins with AIDA64 no problems, 1hr with OCCT. Then I increased the multi to x24 to 4.8ghz, the volts up to 1.392v. (I was shooting for 1.3875v but my mobo rounded it too 1.392v) stressed with no issues. then after receiving feedback from you guys, i tried some of the things mentioned while at 5.0ghz, speaking of: I reached 5.0 with the multiplier at x25. Anyway, I turned the cool & quiet feature off, the c1e and then began to stress the cpu. I didn't get any BSODs but none of the tests would finish to my liking, OCCT would give me an error message after 2 mins, P95 would shut down anywhere from 1-5 of the cores instantly after starting it, IBT would either give me an error message or my system would hang. So for now I'ma hang back at 4.8ghz, thats a decent overclock in my eyes. But any input is greatly appreciated.
 
Deacon_Jones1988 said:
I have downclocked the cpu back to 4.8ghz. It is completely stable at this clock, with all of the aforementioned stress tests. To begin with what I did was start overclocking at 4.4, for this i incresed the multiplier to x22, left the volts on auto, stressed, passed. Then i increased multiplier to x23 put the volts too 1.375v, stressed for over an hour on Prime95, 10 passes on IBT, 1hr 4mins with AIDA64 no problems, 1hr with OCCT. Then I increased the multi to x24 to 4.8ghz, the volts up to 1.392v. (I was shooting for 1.3875v but my mobo rounded it too 1.392v) stressed with no issues. then after receiving feedback from you guys, i tried some of the things mentioned while at 5.0ghz, speaking of: I reached 5.0 with the multiplier at x25. Anyway, I turned the cool & quiet feature off, the c1e and then began to stress the cpu. I didn't get any BSODs but none of the tests would finish to my liking, OCCT would give me an error message after 2 mins, P95 would shut down anywhere from 1-5 of the cores instantly after starting it, IBT would either give me an error message or my system would hang. So for now I'ma hang back at 4.8ghz, thats a decent overclock in my eyes. But any input is greatly appreciated.
Click to expand...

Vishera hits a pretty hard voltage wall at 4.8. Give it the juice and it will go.
 
I'd like to remind you, that overvolting might reduce your CPU/GPU life span or directly kill it, if the overvolt is too much.
 
Deacon_Jones1988 said:
but will my h100i cool this chip with voltage at 1.5v+
Click to expand...

Only one way to find out ;)

Honestly, it might work if you're in the lower range of 1.5V, say like 1.525V or so. It really depends on your chip and ambient temps, also, seriously consider some fans on the VRM heatsink like I said before. It helps a tremendous amount.

Here is a pic of how I did mine, 2x Lepa D70 80mm fans with a "bracket" I made out of a PCI slot cover.

DSC02817.jpg
 
thanx for the post :) . I was Just wondering where the vrm is located exactly? on my MOBO the ASUS m5a99x EVO Rev. 1.0
 
Deacon_Jones1988 said:
thanx for the post :) . I was Just wondering where the vrm is located exactly? on my MOBO the ASUS m5a99x EVO Rev. 1.0
Click to expand...

There should be a big blue heatsink next to the CPU socket, that will be the VRM section. Between the socket and the expansion slots is the northbridge, the two heatsinks are often linked and those need cooling the most. The one down by the SATA ports is the southbridge which really doesn't heat up too much, air moving through the case will cool it just fine.
 
stealthy123 said:
if the computer is stable for the things you use it for why do you care if it can't stand up to a stress test that doesn't replicate your real world usage?

only time I care about stress testing stability is when I am giving the computer to someone else.
Click to expand...

I happen to be in this school of thought. I'm not wasting 24 hours of my electricity to run Prime95 at 100% when in reality my day to day usage doesn't come anywhere near that.

I run the stress tests for 10-15 mins or a single pass of Prime95 and thats good enough for me.

Haven't seen any BSOD for a long time.
 
daglesj said:
I happen to be in this school of thought. I'm not wasting 24 hours of my electricity to run Prime95 at 100% when in reality my day to day usage doesn't come anywhere near that.

I run the stress tests for 10-15 mins or a single pass of Prime95 and thats good enough for me.

Haven't seen any BSOD for a long time.
Click to expand...

I agree, I'm not knocking anyone's rituals, but I stress a cpu for a bit. (an hour at max) if it goes good im satisfied.
 
daglesj said:
I happen to be in this school of thought. I'm not wasting 24 hours of my electricity to run Prime95 at 100% when in reality my day to day usage doesn't come anywhere near that.

I run the stress tests for 10-15 mins or a single pass of Prime95 and thats good enough for me.

Haven't seen any BSOD for a long time.
Click to expand...

Intel's Linpack (used by tools like Intel Burn Test) seems to be the most popular stress test nowadays. The standard 10-pass test in IBT is quick, if the system is remotely unstable it wont pass.
 
Hi, first post here :)
Well, I got my new hardware this weekend and I been playing around a bit with it, for as much time as I had - nowhere near finished... (All system in my sig, anything missing pls ask)
I had a MSI K9N2 Sli Platinum before with 965BE @3.8 auto / stock volt and 4GB of Patriot 1066DDR2 - and I was just not aware how many setting my new board really has (Formula-Z) oO :D

Anyway, regarding OCCT vs prime95 vs IBT - I find prime95 and OCCT harder on the 8350 than Linpack. With IBT I ran maximum settings at 4.8Ghz and apart from way too high temps it passed. prime95 core 7 and especially 8 gave up quite quickly with same setting. - I guess if either test does not pass it is just not fully stable, maybe ok for gaming but not for everything everything.
(DDR3 @ 800, Digi+ on max pretty much, all CnQ, C1E, C6, APM etc off).

Regarding IBT temps / temps in general - I had max settings max temps on IBT 64C CPU, 63C core - with my finger on the reset button.... I'd be too scared with 71C, especially with my 8350 air cooled.

I got my system (very unstable) to validate 5.2Ghz with CPU-Z (which pretty much does not mean anything, or does it? ...).
Currently, back to 4.5/4.6GHz at with all green-options enables, stock Digi+ controls and auto voltage for everything but DDR3 @800 1.5v, all stable in testing (I don't run prime95 more than an hour b/c imho I think it's overkill and I just don't have the time...)
I will try to get the max out of the multiplier first to then go back and also adjust bus speed, ram, etc.
My goal really is to get a much out of this without having to turn off the CnQ etc b/c I just don't see the value beyond versus longevity and money spend on power.

I think I can get to 4.8GHz stable per multiplier without going too extreme on the Digi+ settings and with green stuff enabled, but I have the feeling that with my air cooling I won't get much further when I want it really stable. Like for me 1.5+ volts on the cpu is just not going to happen for more than a quick test and I won't bother trying 1.55+ volts b/c I'd never use it anyway... I'll try to not go beyond 1.45 (1.475) for daily use, with CNQ etc on.

I'll see if I can find some magical way to go to and beyond 4.8GHz with as little overkills as possible :D
 
kaefers said:
Anyway, regarding OCCT vs prime95 vs IBT - I find prime95 and OCCT harder on the 8350 than Linpack. With IBT I ran maximum settings at 4.8Ghz and apart from way too high temps it passed. prime95 core 7 and especially 8 gave up quite quickly with same setting. - I guess if either test does not pass it is just not fully stable, maybe ok for gaming but not for everything everything.
Click to expand...

It all depends on your particular sample. My [previous cpu] was most sensitive to Windows Media Player visual plug-ins I discovered! My [latest cpu] was definitely IBT sensitive. So, run everything. :D
 
xorbe said:
It all depends on your particular sample. My [previous cpu] was most sensitive to Windows Media Player visual plug-ins I discovered! My [latest cpu] was definitely IBT sensitive. So, run everything. :D
Click to expand...
lol @ WMP - **** program anyway :)
I agree, it depends... For me it's if just one stress test causes instability then imo the overclocking is not just good enough, simple as that. I work from my PC at home as well and can't afford random BSOD or freezes, no matter how rare ;)

I'll do a couple of tests and tweaks now - I'll keep a log and post it here if I find something useful :)

EDIT:
I have a 3 pages log so far, I will need some more work on that though this week to get somewhere nice,
trying to do a 'fail-safe' little step-by-step guide :)
Basically, vDroop issue is massive with me. Leaving CPU Load Line Calibration at Auto I need 1.475v for prime95 4.4GHz, with CPU LLC only at Medium (1st setting) I easily get this down to 1.4000v, same everything.
Of course I won't eventually need 1.40v at all for 4.4GHz (I know that already now) but it just shows the sensitivity massively for manipulating one little setting...
 
Last edited:
Deacon_Jones1988 said:
I have downclocked the cpu back to 4.8ghz. It is completely stable at this clock, with all of the aforementioned stress tests. To begin with what I did was start overclocking at 4.4, for this i incresed the multiplier to x22, left the volts on auto, stressed, passed. Then i increased multiplier to x23 put the volts too 1.375v, stressed for over an hour on Prime95, 10 passes on IBT, 1hr 4mins with AIDA64 no problems, 1hr with OCCT. Then I increased the multi to x24 to 4.8ghz, the volts up to 1.392v. (I was shooting for 1.3875v but my mobo rounded it too 1.392v) stressed with no issues. then after receiving feedback from you guys, i tried some of the things mentioned while at 5.0ghz, speaking of: I reached 5.0 with the multiplier at x25. Anyway, I turned the cool & quiet feature off, the c1e and then began to stress the cpu. I didn't get any BSODs but none of the tests would finish to my liking, OCCT would give me an error message after 2 mins, P95 would shut down anywhere from 1-5 of the cores instantly after starting it, IBT would either give me an error message or my system would hang. So for now I'ma hang back at 4.8ghz, thats a decent overclock in my eyes. But any input is greatly appreciated.
Click to expand...

What does a 8530 @ 4.8Ghz pull in CB 11 in single core? Can you check?
 
Hi,

as promised I been testing pretty much step by step how far I can get with my air-cooled 8350... As a spoiler: No, prime95 stable not to 5Ghz... :/

Although I will mention this below again:
1) I understand that 10ish minutes prime95 is not generally regarded as stable by some/many, but I just don't have the time to test it. Following results are meant as an indication of potential stability, not as set law.
2) I tried to keep it consistent, but after a while I got a bit more trial n error - I hope it is readable.

Especially regarding core #8 (you'll see below :(), I am very happy to hear your suggestions.
Is my cpu / core #8 just sh** ??? Any further suggestions regarding settings ???

CPU tested is:
AMD FX-8350, FD8350FRW8KHK, FA 1308PGN

Enjoy :)

(A) Basics
All settings in UEFI Bios as stock before any further testing changes, except otherwise mentioned (hopefully).
UEFI BIOS
-Turbo Core off
-CPU & PCI Spread Spectrum off
-APM & EPU off
-Xtreme Tweaking on (?) – for what it’s worth…
-Ram 800MHz @ 9-9-9-24 @ 1.5v stock
-Digi+ Auto / Default / Normal (I’ll change this when I don’t get further)
-All Other Auto / Default (clocks, volts, timings etc)
-Misc I have a little fan on the VRM from an old board (M2N Deluxe Wifi)
-Method for testing Asus AI Suite II, Check CPU-Z, HW Monitor Pro, AIDA64, prime95, Intel Burn Test, OCCT; (extra: AMD OverDrive, SpeedFan, CoreTemp)
-Misc II What I don’t show changing doesn’t change from test to test
-Misc III Some test run only for 5/10/20min, means I am trying to figure out if remotely stable or not; if stable, more testing to follow after.

kaefers01
Goal: a basic 4.4GHz prime95 FFT stable overlock via multiplier with as little stressing components / settings as possible
I will eventually test how far I can get my earlier CPU-Z validation: http://valid.canardpc.com/2796112 , while remaining prime95 / stress-test(s) stable only (1) as well as while being stable AND being able to make use of AMD CnQ, C1E etc (2), both with as little stress to e.g. CPU (vcore), without random extreme overvolting or too extreme settings for VRM and others.
Maybe later I will also include BUS Speed HT, NB, RAM overclocking (when I have time).

vcore at 1.37500
Prime95 FFT fail

kaefers02
change vcore from 1.37500 to 1.38125
Prime95 FFT fail

kaefers03
change vcore to 1.38750
Prime95 FFT fail/freeze

kaefers04
change vcore to 1.40000
Prime95 FFT fail

kaefers05
change vcore to 1.42500
Prime95 FFT fail
- Misc Network connection takes ages to connect again after reboot
- Misc AI Suite II gives TurboVHelp.exe error, may need to reinstall

kaefers06
change vcore to 1.45000
Prime95 FFT fail

kaefers07
change vcore to 1.47500
Prime95 FFT pass (20min) @ CPU 52C @ Core ~46C
- Too high vcore for 4.4Ghz
- Other settings likely to affect stability
- Actual readings from HWM / Asus AI Suite II: CPU vcore 1.451 avg
Vcore in AI Suite II much higher than what CPU-Z/ HWM shows

kaefers08
- reinstalled AI Suite II
- disabled “Xtreme Tweaking” in UEFI Bios

change vcore to 1.45000
Prime95 FFT pass (15min) @ CPU 50C @ Core ~44C

kaefers09
change vcore to 1.42500
Prime95 FFT fail
- no idea is “Xtreme Tweaking” in UEFI BIOS makes a difference, but I leave it off for now – no clue what it really does.

kaefers10
change vcore to 1.44375
Prime95 FFT pass (20min) @ CPU 49C @ Core ~43C

Time to change some other settings to see if it makes a difference

kaefers11
change vcore to 1.42500
change CPU/NB Voltage to 1.20000
Prime95 FFT fail/BSOD

kaefers12
change vcore to 1.42500
change CPU/NB Voltage to 1.18750 (stock)
change NB HT Voltage to 1.22500
Prime95 FFT fail

kaefers13
change vcore to 1.42500
change CPU/NB Load Line Calibration to High
others as default (Basics)
- Misc Ambient temp drop (nighttime) from around 22C to 20C in this room
Prime95 FFT fail

kaefers14
change vcore to 1.42500
change CPU/NB Load Line Calibration to High
change CPU Load Line Calibration to Medium
others as default (Basics)
Prime95 FFT pass (15min) @ CPU 50C @ Core ~44C
- Looks like CPU LLC does what is should, it compensates for vDroop and keeps it stable (google, e.g.:
http://www.masterslair.com/vdroop-and-load-line-calibration-is-vdroop-really-bad)
- Disadvantages of adjusting CPU LLC: potential (over-)spikes in CPU vcore IF you already run the CPU at high vcore!
- Next: switch off CPUNB LC, then lower volt, then higher clock…

kaefers15
change vcore to 1.40000
change CPU Load Line Calibration to Medium
others as default (Basics)
Prime95 FFT pass (15min) @ CPU 49C @ Core ~42.5C

kaefers16
change vcore to 1.38750
change CPU Load Line Calibration to Medium
others as default (Basics)
Prime95 FFT fail ‘only’ after 5min

kaefers17
change vcore to 1.38750
change CPU Load Line Calibration to High
others as default (Basics)
Prime95 FFT pass (15min) @ CPU 50C @ Core ~44C
- Note the higher CPU and vcore temps despite lower CPU vcore; that’s due to the higher CPU LLC!

kaefers18
change vcore to 1.38125
change CPU Load Line Calibration to High
others as default (Basics)
Prime95 FFT pass (15min) @ CPU 49.5C @ Core ~43.5C
- Note the higher CPU and vcore temps despite lower CPU vcore; that’s due to the higher CPU LLC!Hi,
 
That’s it for now @4.4GHz; further steps would be higher tweaking of Digi+ controls, lower vcore volts, maybe CPU/NB (and/or NB HT volts, NB volts, … you try yourself if you like 4.4GHz for you already).
Further, you can try as optimized as you can get with lower volts and be prime95 / other stress tests stable, but then check if this still allows you for enabling AMD CnQ, C1E, C6, APM, ERP…. which (may) make sense to you if the PC is idle at times…

4.5GHz+

kaefers19
change vcore to 1.40000
change CPU Load Line Calibration to Medium
change multiplier 200 x 22.5 = 4.5Ghz
others as default (Basics)
Prime95 FFT prime95 crashes

kaefers20
change vcore to 1.42500
change CPU Load Line Calibration to Medium
change multiplier 200 x 22.5 = 4.5Ghz
others as default (Basics)
Prime95 FFT core 8 (my 8350s weakest / most power-hungry core it seems) fails

kaefers21
change vcore to 1.45000
change CPU Load Line Calibration to Medium
change multiplier 200 x 22.5 = 4.5Ghz
others as default (Basics)
Prime95 FFT stable 10min
- vs kaefers07, achieved plus 0.1GHz at minus 0.25vcore, but at / because of Medium vs Standard (off?) CPU LLC

kaefers22
change vcore to 1.43750
change CPU Load Line Calibration to High
change multiplier 200 x 22.5 = 4.5Ghz
others as default (Basics)
Prime95 FFT stable 5min

kaefers22b
change vcore to 1.42500
change CPU Load Line Calibration to High
change multiplier 200 x 22.5 = 4.5Ghz
others as default (Basics)
Prime95 FFT stable 5min

kaefers22c
change vcore to 1.40000
change CPU Load Line Calibration to High
change multiplier 200 x 22.5 = 4.5Ghz
others as default (Basics)
Prime95 FFT stable 10min @ CPU 52C @ Core ~46.5C

kaefers22d ;) - forgot to count...
change vcore to 1.38750
change CPU Load Line Calibration to High
change multiplier 200 x 22.5 = 4.5Ghz
others as default (Basics)
Prime95 FFT core 8 failed again (and shockwave flash crashed, was watching some youtube J )

kaefers23
change vcore to 1.36875
change CPU Load Line Calibration to Ultra High
change multiplier 200 x 22.5 = 4.5Ghz
others as default (Basics)
Prime95 FFT stable 10min @ CPU 51.5C @ Core ~46.5C (fails any vcore below with these settings, core #8)


So, for now, 4.5GHz takes at least 0.1vcore more and the next highest setting for CPU LLC (High) compared to 4.4GHz
Of course, I could and should run more stress testing - if I wanted to keep it there – well, I don’t right now…
Let’s try 4.6GHz next…

kaefers24
change vcore to 1.42500
change CPU Load Line Calibration to High
change multiplier 200 x 23 = 4.6Ghz
others as default (Basics)
Prime95 FFT guess what, usual suspect #8 f’d up again… maybe I should replace core 8 :D

kaefers25
change vcore to 1.40000
change CPU Load Line Calibration to Ultra High
change multiplier 200 x 23 = 4.6Ghz
others as default (Basics)
Prime95 FFT prime95 crashed, started again and this time core 6 failed

kaefers26
change vcore to 1.40625
change CPU Load Line Calibration to Ultra High
change multiplier 200 x 23 = 4.6Ghz
others as default (Basics)
Prime95 FFT PC freeze

kaefers27
change vcore to 1.41250
change CPU Load Line Calibration to Ultra High
change multiplier 200 x 23 = 4.6Ghz
others as default (Basics)
Prime95 FFT core 8…

kaefers28
change vcore to 1.41875
change CPU Load Line Calibration to Ultra High
change multiplier 200 x 23 = 4.6Ghz
others as default (Basics)
Prime95 FFT stable 10min @ CPU 55C @ Core ~52.5C
 
Last edited:
4.7GHz
Upfront – looks like at higher overclocks (at least with my air cooled system, there is a quick limit as to using both high vcore and CPU LLC tweaking, as explained/suggested earlier)

kaefers29
change vcore to 1.45000
change CPU Load Line Calibration to Ultra High
change multiplier 200 x 23.5 = 4.7Ghz
others as default (Basics)
Prime95 FFT fail #8

kaefers30
change vcore to 1.47500
change CPU Load Line Calibration to High
change multiplier 200 x 23.5 = 4.7Ghz
others as default (Basics)
Prime95 FFT fail #8

kaefers30
change vcore to 1.48125
change CPU Load Line Calibration to High
change multiplier 200 x 23.5 = 4.7Ghz
others as default (Basics)
Prime95 FFT PC freeze

kaefers31
change vcore to 1.46250
change CPU Load Line Calibration to Ultra High
change multiplier 200 x 23.5 = 4.7Ghz
others as default (Basics)
Prime95 FFT stable ~4min, but temps getting too high so I manually stopped test, CPU 57C, Core 62.5C

kaefers32
change vcore to 1.46250
change CPU Load Line Calibration to High
change multiplier 200 x 23.5 = 4.7Ghz
others as default (Basics)
Prime95 FFT fail #8 and few seconds later #6

kaefers33
change vcore to 1.46875
change CPU Load Line Calibration to High
change multiplier 200 x 23.5 = 4.7Ghz
others as default (Basics)
Prime95 FFT vs kaefers30 but supposedly less stable still, fails but doesn’t freeze à I’ll try some other settings now

kaefers34
change vcore to 1.46875
change CPU Load Line Calibration to High
change CPU/NB Voltage to 1.20000
change multiplier 200 x 23.5 = 4.7Ghz
others as default (Basics)
Prime95 FFT #8 fails

Going for Glory…

kaefers35
change vcore to 1.47500
change CPU Load Line Calibration to High
change CPU/NB Load Line Calibration to Extreme
change CPU Current Capability to 120%
change CPU/NB Current Capability to 120%
change CPU Power Phase control to Extreme
change VRM Fixed Frequency Mode to 400KHz
change CPU/NB Voltage to 1.20000
change multiplier 200 x 23.5 = 4.7Ghz
others as default (Basics)
Prime95 FFT #8 fails

kaefers35b...
change vcore to 1.47500
change CPU Load Line Calibration to High
change CPU/NB Load Line Calibration to Extreme
change CPU Current Capability to 120%
change CPU/NB Current Capability to 120%
change CPU Power Phase control to Extreme
Change CPU Power Response Control to Fast
change CPU/NB Power Response Control to Fast
change CPU/NB Voltage to 1.20000
change multiplier 200 x 23.5 = 4.7Ghz
others as default (Basics)
Prime95 FFT #8 fails

kaefers36
change vcore to 1.45000
change CPU Load Line Calibration to Ultra High
change CPU/NB Load Line Calibration to Extreme
change CPU Current Capability to 120%
change CPU/NB Current Capability to 120%
change CPU Power Phase control to Optimized
change CPU Power Response Control to Fast
change CPU/NB Power Response Control to Fast
change CPU/NB Voltage to 1.20000
change multiplier 200 x 22 = 4.4Ghz
change CPU Bus Freq to 205x22 = 4.510GHz
others as default (Basics)
Prime95 FFT stable 10min @ CPU 61C @ Core ~59.5C

kaefers37
change vcore to 1.45000
change CPU Load Line Calibration to Ultra High
change CPU/NB Load Line Calibration to Extreme
change CPU Current Capability to 120%
change CPU/NB Current Capability to 120%
change CPU Power Phase control to Optimized
change CPU Power Response Control to Fast
change CPU/NB Power Response Control to Fast
change CPU/NB Voltage to 1.20000
change multiplier 200 x 22 = 4.4Ghz
change CPU Bus Freq to 210x22 = 4.620GHz
others as default (Basics)
Prime95 FFT stable 5min @ CPU 63C @ Core ~61C

kaefers38
change vcore to 1.44375
change CPU Load Line Calibration to Ultra High
change CPU/NB Load Line Calibration to Extreme
change CPU Current Capability to 120%
change CPU/NB Current Capability to 120%
change CPU Power Phase control to Optimized
change CPU Power Response Control to Fast
change CPU/NB Power Response Control to Fast
change CPU/NB Voltage to 1.20000
change multiplier 200 x 22 = 4.4Ghz
change CPU Bus Freq to 210x22 = 4.620GHz
others as default (Basics)
Prime95 FFT stable 5min @ CPU 63C @ Core ~60C

kaefers39
change vcore to 1.43750
change CPU Load Line Calibration to Ultra High
change CPU/NB Load Line Calibration to Extreme
change CPU Current Capability to 120%
change CPU/NB Current Capability to 120%
change CPU Power Phase control to Optimized
change CPU Power Response Control to Fast
change CPU/NB Power Response Control to Fast
change CPU/NB Voltage to 1.20000
change multiplier 200 x 22 = 4.4Ghz
change CPU Bus Freq to 210x22 = 4.620GHz
others as default (Basics)
Prime95 FFT stable 5min @ CPU 63C @ Core ~59.5C

kaefers40
change vcore to 1.43750
change CPU Load Line Calibration to Ultra High
change CPU/NB Load Line Calibration to Extreme
change CPU Current Capability to 120%
change CPU/NB Current Capability to 120%
change CPU Power Phase control to Optimized
change CPU Power Response Control to Fast
change CPU/NB Power Response Control to Fast
change CPU/NB Voltage to 1.20000
change multiplier 200 x 22 = 4.4Ghz
change CPU Bus Freq to 212x22 = 4.664GHz
others as default (Basics)
Prime95 FFT #8 fails

kaefers41
change vcore to 1.45000
change CPU Load Line Calibration to High
change CPU/NB Load Line Calibration to Extreme
change CPU Current Capability to 120%
change CPU/NB Current Capability to 120%
change CPU Power Phase control to Extreme
change CPU Power Response Control to Fast
change CPU/NB Power Response Control to Fast
change CPU/NB Voltage to 1.20000
change multiplier 200 x 22 = 4.4Ghz
change CPU Bus Freq to 212x22 = 4.664GHz
others as default (Basics)
Prime95 FFT #8 fails

So, after a bit of testing it seems, thanks to a useless core #8, that I will have severe trouble to achieve a prime95 stable overclock beyond 4.7GHz, bearing in mind that I can’t go much higher with CPU vcore / CPU LLC as my air-cooled CPU just won’t run cool enough, plus it’s a bit warmer now in this room. I’ll open the window for the tests below ;)
I am writing this on the last settings above, i.e. kaefers41. With all my monitoring tools running, prime95off, my CPU is idling away at now 34degC, the cores are doing pretting much nothing @ ~15.5degC @ 4.664GHz.

A few more (too) ‘hot’ tests:

kaefers42
Room temp lower, mainboard temp idle now 22C instead of 25C
Power draw from socket (Brennenstuhl energy monitor) ~330W for system, HDDs unplugged
change vcore to 1.47500
change CPU Load Line Calibration to Ultra High
change CPU/NB Load Line Calibration to High
change CPU Current Capability to 130%
change CPU/NB Current Capability to 130%
change CPU Power Phase control to Extreme
change CPU Power Response Control to Fast
change CPU/NB Power Response Control to Fast
change CPU/NB Voltage to 1.22500
change NB HT Voltage to 1.22500
change multiplier 200 x 22 = 4.4Ghz
change CPU Bus Freq to 212x22 = 4.664GHz
change HT multiplier (AMD Overdrive) to 12 // 212*12=2544
DDR now @ 850MHz
others as default (Basics)
Prime95 FFT stable 10min @ CPU 62C @ Core ~62C

kaefers43
change vcore to 1.46875
change CPU Load Line Calibration to Ultra High
change CPU/NB Load Line Calibration to High
change CPU Current Capability to 130%
change CPU/NB Current Capability to 130%
change CPU Power Phase control to Extreme
change CPU Power Response Control to Fast
change CPU/NB Power Response Control to Fast
change CPU/NB Voltage to 1.22500
change NB HT Voltage to 1.22500
change NB voltage 1.11250
change DRAM voltage 1.525
change multiplier 200 x 22 = 4.4Ghz
change CPU Bus Freq to 215x22 = 4.730GHz
change HT multiplier (AMD Overdrive) to 12 // 215*12=2580
others as default (Basics)
Prime95 FFT stable 5min @ CPU 61C @ Core ~61.5C
Misc fails #8 @1.4625 vcore after 3min

kaefers44
change vcore to 1.47500
change CPU Load Line Calibration to Extreme
change CPU/NB Load Line Calibration to Extreme
change CPU Current Capability to 130%
change CPU/NB Current Capability to 130%
change CPU Power Phase control to Extreme
change CPU Power Response Control to Ultra Fast
change CPU/NB Power Response Control to Ultra Fast
change CPU/NB Voltage to 1.25000
change NB HT Voltage to 1.22500
change DRAM voltage 1.525
change multiplier 200 x 22 = 4.4Ghz
change CPU Bus Freq to 218x22 = 4.796GHz
change HT multiplier (AMD Overdrive) to 12 // 218*12=2616
others as default (Basics)
Prime95 FFT freeze

[/B][/U]kaefers45[/B][/U]
setting set in UEFI Bios, boot Win7 OK
enable HPC mode
change DDR CL10-10-10-30
change vcore to 1.47500
change CPU Load Line Calibration to Ultra High
change CPU/NB Load Line Calibration to Extreme
change CPU Current Capability to 130%
change CPU/NB Current Capability to 130%
change CPU Power Phase control to Extreme
change CPU Power Response Control to Auto
change CPU/NB Power Response Control to Auto
change CPU/NB Voltage to 1.22500
change NB HT Voltage to 1.22500
change DRAM voltage 1.525
change multiplier 200 x 22 = 4.4Ghz
change CPU Bus Freq to 215x22 = 4.730GHz
change HT multiplier to 10 // 215*10=2150
NB Frequency 4.730/2 = 2.365GHz
DRAM Frequency 800*2.15/2 = 860MHz
others as default (Basics)

AI Suite II:
change CPU Bus Freq to 218
change CPU Voltage to 1.48750-1.5000 volts
CPU now 218*22 = 4796
Prime95 FFT even with 1.5000vcore #8 fails pretty quickly – that is before I would have had to stop test b/c the cpu cooler
just can’t handle that for long…


Final:
Hope nobody feels spammed :)
Sorry for counting errors :)rolleyes:) and any other errors :eek:
 
Not too sure if this if helpful at all... but I'll post it as a final result for now (due to lack of time I can't do more testing):

http://valid.canardpc.com/2803305

CPU: 4719MHz, 19.5 x 242, 1.43125vcore, CPU/NB 1.30, RAM 2266MHz 1.575v 10-11-11-28-45
HT 2671MHz
NB 2671MHz
CPU Load Line Calibration to Ultra High,
CPU/NB Load Line Calibration to Extreme,
change CPU Current Capability to 120%,
change CPU/NB Current Capability to 120%,
change CPU Power Phase control to Optimized,
change CPU/NB Voltage to 1.3000,
change NB HT Voltage to 1.22500

I am sure there is more in there, especially with water cooling.
These settings I tested with prime95 FFT, Blend, OCCT and IBT.
Can't go much higher otherwise my heatsink would melt... :(
Found that when really going for the RAM / HT-Link / NB speeds, other cores start failing as well. Maybe I give back my 8350 and try another one :)
 
No point in pushing HT-Link speeds... It doesn't do anything besides add to instability.

Pushing the IMC (NB) and memory can add performance, though. NB doesn't do as well on BD/PD as it did with Deneb, so 3GHz-3.2GHz is about as much as you'll get on air. With good memory, though, 2400+ is quite doable on air.
 
Thanks, I keep that in mind and try it out when I find some more time, had enough of tweaking for now... waiting for new case / water. Mostly core #8 bugs me out tbh.
I got to 2400 with CL12 but I found with the higher latency the increase in performance is not much at all that's why I left it for now. Also not too sure if this 1.5v ram would do well if @1.65v for 24/7...
 
xxbassplayerxx said:
No point in pushing HT-Link speeds... It doesn't do anything besides add to instability.

Pushing the IMC (NB) and memory can add performance, though. NB doesn't do as well on BD/PD as it did with Deneb, so 3GHz-3.2GHz is about as much as you'll get on air. With good memory, though, 2400+ is quite doable on air.
Click to expand...

With 12-12-12-34 I tried NB a little higher, but as I though before, 2.7ish seems to be the point where I can't go further (my settings are wrong... ?!?).
I get it somewhat stable with 1.4v on CPU/NB but as far as I read 1.3/1.35v and no further.
Nowhere getting near your 3Ghz... :confused:
 
Hi, been a while.
For whom it may concern ;)

I exchanged my AMD FX-8350 (former 1308PGN) to another one, now 1302PGN. Well, while technically older, is the new cpu much much better to OC than the previous one.
4.8 prime FFT stable is dead simple (vcore, cpu llc).
4.9 is a bit more tricky as I need to add just a little VDDA (volt for PLL) to make it happy.
5.0 seems a lot more tricky, it's not fully stable yet, prime95 fft runs but will produce some errors eventually. But I just received my new Mushkin ram (after handing back the Patriot Viper which I strongly suspected to cause instability at higher clocks) and will try what this new ram does/doesnt like.

Pretty certain I will get 5.0 dead stable with this chip - as of now.
For reference, that's what the chip needs according to CPU-Z under prime95 FFT load to not crap up (Ram @ 1600 for testing that was).
Also, changing vcore in Asus AI Suite one tick up/dowm will result in the CPU-Z voltage to stay the same sometimes. It will only go up with the next tick or the one after. Meaning, CPUZ may show 1.404v but it's either a) 1.404- b) 1.404+- or c) 1.404+ by tiny amounts that CPU-Z won't show... So cpu might not be stable with the first setting a) which tends to be between 1.428 and 1,404 but only with b) or c) etc...
Anyway:
4.2 - 1.260v
4.3 - 1.272v
4.4 - 1.308v
4.5 - 1.320v
4.6 - 1.356v
4.7 - 1.404v
4.8 - 1.440v
4.9 - 1.476v
5.0 ~ 1.512v (not stable yet, needs testing)

So the batch / lot where this cpu came from seems alright and I am happy with reaching that! And that's although FA 1302 PGN implies it's older than my previous one FA 1308 PGN... Week 8 must have been a bad week then :]

Anyone care to comment, please do :D
 
Interesting read! Grabbed a FX-9370 to drop into this Sabertooth 990FX r1 board. Seems like it loves voltage to the chagrin of my video cards. With an air cooler installed they are fine, but the air cooler I has clearance problems for my Corsair Vengeance memory. When I install the H100i they overheat in demanding games with the eye candy turned up from the lack of air flow around the CPU socket. I placed a 120mm fan over the VRM's, but that's not enough it seems. Using a Silverstone RV03 case like pictured here. Guess I'll stare at it for a bit until something pops into my head. Any suggestions?

:)
 
cageymaru said:
Interesting read! Grabbed a FX-9370 to drop into this Sabertooth 990FX r1 board. Seems like it loves voltage to the chagrin of my video cards. With an air cooler installed they are fine, but the air cooler I has clearance problems for my Corsair Vengeance memory. When I install the H100i they overheat in demanding games with the eye candy turned up from the lack of air flow around the CPU socket. I placed a 120mm fan over the VRM's, but that's not enough it seems. Using a Silverstone RV03 case like pictured here. Guess I'll stare at it for a bit until something pops into my head. Any suggestions?

:)
Click to expand...

Where did you mount the H100i? I have an 8320 @4.6, 1.45v, and dont have any problem with my vrm in my RV03 using the H100i. I have the H100i mounted in front as an intake, the mother board is a gigabyte ud3 rev.4.
 
You must log in or register to reply here.
Back
Top