Overclocking AMD 8350.

[Jborg]

Hello all,

Heres my specs first of all:

ASUS Sabertooth 990FX R2.0
Gigabyte GTX 970 G1 Gaming Superclocked (+130 Core clock, +330 Memory Clock from stock)
AMD FX 8350 8 Core @ 4.3GHz (Stock cooler)
Corsair Ballistix Elite 2x8GB DDR3 1866
Corsair R300 Carbide
EVGA Supernova G2 750W

So I currently am running the processor overclocked to I believe 4.2Ghz or 4.3 GHz, everything seems to run well. One thing I did notice while benchmarking in Unigine Valley: Somehow by upping my Overclock to the right point on my video card, along with a sweet spot with the CPU, I got a score of 2600, where as before I got like 2100.

My question going forward: I am looking to get this 8350 running stable @ 4.6 or 4.7 Ghz to maximize potential performance. I do have a HYPER EVO 212 Fan sitting here ready to install. But I wanted to know what was the best stress testing program for the CPU, (I know of Prime 95) I do have Prime95 installed I believe, However I do not know if its the correct version or what not and am not sure how to use it. (I did try it out once and in like 20 seconds I was already getting signal my CPU was at 65 celcius.

Any tips suggestions?

I do have my my MOBO settings to DOCP and I have a custom XMP ram profile set up. Also I set all the CPU functions to the higher setting for overclocking.

Any tips would be appreciated. Thank you

 

[neatfeatguy]

Prime95 and Intelburn Test (ITB) are what I use when checking stability and temps of an OC.

 

[ThE_MaD_ShOt]

Also better cooling can't be stressed enough. The stock cooler is for proc running at stock speeds.

 

[Jborg]

Yeah, when I get home from work today, I am going to install this EVO 212. I will proceed to post some screen shots after that.

With the addition of this new proc fan? Will I be able to achieve the desired 4.6-4.7 ghz? (Guess it will take some testing huh!) Going to mess with this stuff more when I get home, and will also be able to post some screen shots.

 

[Jborg]

Bump - Should I start higher with like 4.5 or 4.6ghz? and work down? or opposite? and how long to run Prime95 and on what setting?

 

[m0nt3]

No, a 212 will not be enough cooling for 4.6-4.7 and honestly it is not worth it anyway, unless you are wanting to do some quick benchmark score. For my chip at least anything past 4.5 requires a good bit more voltage, and even with my heat sink it is at the limits for 4.5 (almost 70C socket temp and about 61C core under Prime95 @ 1.35V with LLC @ Ultra i think), will have to double check on that. the performance over stock will be negligible for any game really, There is no observable difference from my 8320 @ stock (3.5Ghz) with 1866 mem. The only time I could tell a difference is if I let the BIOS default the memory speed and timings, which it defaults mine to 1333 with 11-11-11 timings and the most noticeable difference was in BF4 DX11 API where my minimums were quite a bit lower with the slower memory settings. The extra voltage, cooling, and power draw really makes it not worth while. Unless you are just trying to learn about it, then go ahead, but you will need more than the EVO 212.

The longer you run Prime the better, IDK if it is my chip, or just issues with P95, but I have observed that the same settings my go prime95 for 15 hours and fail or fail at around an hour. But monitor temps closely.

 

[Jborg]

Thank you for the input. And honestly I know I may not get a a whole lot out of the CPU OC. However I do want to learn about as you mentioned.

So realistically, the number I am sitting at now is about the highest I can get safely with air cooling?

It seemed odd however, I am not sure if it was from the CPU oc, or the video card OC. But my score increased by 500 points when I OC'd the CPU to 4.3 from 4.1Ghz in Unigine Valley. I know this is GPU based, but how did I get that much of an increase from such a small OC?

-Edit - Also what temp in Prime95 should i start worrying. Because when It hit 65 and a message appeared, that alone freaked me out. Lol

 

[m0nt3]

The higher scores usually from from higher high's in the frame rate. if you still have the benchmark results, valley will tell you the minimum, average, and maximum frame rate. The variance will probably be within 3 FPS on the high end and almost not noticeable on the low. So for actually game play, no noticeable difference. Voltage is going to be the real driving factor in your temps, much less from clock speed alone. you will need to find the voltage limitation of your cooling solution. Run prime 95 for an hour while monitor temps closesly. What program are you monitoring your temps with? Thanks.

 

[Jborg]

I am using HWMonitor to monitor idle temps, and using AMD Overdrive to monitor temps under load.

 

[m0nt3]

Ok, I will verify this when I get home, but HWMonitor with list, I think they are called CPUIN 0-2 and CPUIN1 is the socket temp (I will verify this in about an hour when I get home). The temp listed under AMD FX 8350 will be your core temp. What do your temps look like now?

 

[Jborg]

Idle temps with stock fan.

Temps 30% into rendering a 3 minute video on 1080p in Sony Vegas(Still stock cooler):

Brb, swapping to evo 212. Will test temps again.

 

[Jborg]

Idle with EVO 212 installed

Recording same video with same settings:

Just a quick reference test for temps... Pretty significant tho... 20 degrees lower under load

So now that that is installed.....

What to set this too? And what to up?

 

[eidairaman1]

Check out fx oc club. By the way uou can even ask fullinfusion on it (my rig has no os in it)

 

[m0nt3]

Under Temperatures the CPU line is the socket temp. Blend test is what I do, small FFT will stress the FPU more and cause high temps, i choose blend to test the mem controller as well (CPU/NB speed in BIOS). Do you know what your LLC (Load Line Calibration)is set to?

 

[Jborg]

Figured this is the fastest way to figure out what all this should be set too, did do a little research and set it up as is:
Let me know if i should change anything.

 

[m0nt3]

Looks Good. Did you run Prime 95 yet and check your temps?

 

[Jborg]

Will now, just wanted to make sure everything was good in the BIOS before starting that.

Testing that now, thank you!

How long should I run it? 10-20 minutes?

 

[m0nt3]

at least an hour, as long as your temps are ok

 

[Jborg]

Ran it for 10 minutes, Results:

Will have to test it longer tommorrow or something, however it was not crashing before when running at 4.3 with the stock cooler, I just could not run prime with a good temp.

How are those temps for 10 minutes?

-Also, how do you monitor the temps when you run it for that long?

 

[m0nt3]

The temps will raise higher after longer use, but decent so far, core temp, from what I read the rule is 62C and 70C for the socket. Short periods beyond that wont hurt, but sustained could lead to degradation.

 

[Jborg]

K so I think I am stuck at 4.3Ghz... it runs stable, have not had a crash with it. When I tried 4.4 and ran Prime the socket temp was at 71c and CPU temp was 62c after 10 minutes of running prime. (was going to run it for an hour but didnt want to go any longer with those temps.

So I guess I am just gonna have to stick with 4.3Ghz.


However, I really only game mostly.. sometimes use Photoshop and Vegas. Is it possible to run a OC at like 4.4 or something If I would only be gaming?

Prime seems like it is alot more than I would actually be using - Not sure was just wondering.

 

[m0nt3]

Prime is certainly more than normal use, that is why it is used for testing overclock stability. I know BF4 can get the CPU nice and toasty as well, not like prime 95 though. When you went to 4.4 did you change the voltage or do you still have it on auto?

 

[Jborg]

Prime is certainly more than normal use, that is why it is used for testing overclock stability. I know BF4 can get the CPU nice and toasty as well, not like prime 95 though. When you went to 4.4 did you change the voltage or do you still have it on auto?

I left all the voltages stock, only increased the multiplier by increments of 1. (or 0.5)

Computer never crashed on 4.4Ghz, just seemed to hot after 10 mins.

 

[m0nt3]

If you are trying to test temperature over a short run, use Small FFT instead of blend.

 

[Jborg]

I will give this a shot on 4.4Ghz later today when I get home.

Thanks for the input, its greatly appreciated.