AMD FX6300 OC guidance, problem with prime95?

[Mark.]

Hi everyone, im new to the forum and to overclocking, i have spent a bit of time reading up on what needs to be done to achieve a safe and stable overclock and i have begun experimenting today. I have just built this pc and i was curious of its capabilities, its a fairly low end pc in some peoples eyes but its what il be using for the next few years probably so id like to try make the most of it

so far i have the clock speed at 4.3ghz, tested with prime95 for just over 2 hours and 30 mins, temps reached 57c at one point but was mostly around the mid 40c area the whole time. I encountered "Cannot initialize FFT code, errcode=1005,Unable to allocate memory. One possible cause is the operating system's swap area is too small." with worker 4 about 1 hours 30 mins into the test but didnt realise at first. I have searched around and people seem to think this is an error with 32bit versions of prime possibly?

Im currently running windows 7 ultimate 32 bit, no activation code or SP1 etc, its just temporary untill i get my 64bit disc

MOBO > Asus M5A97.
CPU & Cooler > AMD FX6300 & Zalman CPNS10x heatsink/2 fans.
MEMORY > 1x8gb Kingston Fury HyperX Black 1600mhz.
POWER SUPPLY > BeQuiet 730w Bronze PSU.
VIDEO CARD > GTX280 =P.
HDDs/Optical drives > 750gb seagate 7200rpm sata
SOUND DEVICE > Onboard
O/S > Windows 7 Ultimate 32bit, not activated or updated
Case > Antec 900, 200mm top case fan, 2 front intake fans 120mm, 1 exhaust fan 120mm

Im using this time to try get some experience with overclocking before i upgrade to 64bit windows, im trying to build a rough template for me to work with when i OC on the new operating system.

Im wondering if anyone can help me with the problem i encountered running prime95 and also have a look at these HWM and CPU-Z screenshots to tell me if they think voltages and clock multipliers look correct etc. Screenshots taken are with Prime95 stopped!



Any input would be appreciated, thanks for your time

 

[Johan45]

Everything looks OK to me, which P95 test were you running? FFts or blend? It's likely just a bit more voltage is needed if it was that far into the test the CPU and board start to get warm that's when things like that usually need just a touch more.

 

[RJARRRPCGP]

Never saw Prime95 give me that error message before... Prime95 is not fussy with RAM allocation like Linpack is.
If swapping is required, Prime95 will just wait until Windows is done swapping.
Even with an unstable system, that error message is unlikely.

While OTOH, Linpack requires that the RAM be pre-allocated before it can use the RAM.

Ever saw a not-enough-memory error message when you select maximum with Intel Burn Test? That's why.

 

[Mark.]

Thank you for the quick response!

I was running the program in blend mode, ive not yet tried any other tests, once i seen worker 4 had stopped i stopped the test completely as i didnt want to potentially damage the system. None of the workers actually reported any problems, just number 4 stopped working 1 hour 30mins into the test, but it reported no errors.
motherboard temps didnt go beyond 29c the whole test and all voltages on the board seemed stable.

I did a quick google search of the error and i seen people had the issue when using 32bit windows mostly, there didnt seem to be any explanation why but i didnt dig too deeply.

How much would you recommend i potentially increase the cpu voltage by? and in what steps, 0.25? more? Would it also be worthwhile trying to add to the multiplier more to try get the highest clock speed for the voltage used? Ive never overclocked a machine before so im not sure how much headroom i have left with multipliers,voltage and also core temps. I want it to run its best with the current air cooled setup, if 4.3ghz OC is the most i can get then il be happy with that

Thanks again for your time : )

 

[RGone]

Torture test your CPU with Prime95 Installation And Running Tests

There maybe minor changes in the GUI of Prime 95 but the what to do is still as valid as ever to get to where most of us test.

The third image at the link above shows Run a Torture Test and has a small dot next to the In-place large FFTs. Change that to Blend (tests some of everything) and then click on the OK button to begin testing as the majority of us do our testing.

By the way the first two images of that link are spot on in general to get to where you then setup as shown in image #3.

By the way we need to see not only the HWMonitor capture done while doing P95 Blend mode but also the CPUz > CPU TAb, Memory Tab and the SPD tab so can verify ram settings. Luck man.

RGone...

PS: this is link to 32bit version for download.
http://download.mersenne.ca/gimps/p95v285.win32.zip

 

[Mark.]

im currently running prime95 again, i upped the multiplier by 0.5 taking me to 4.4ghz, 22 on the multiplier, i didnt change any voltage settings just to see if the error comes again, if it does il roll back the multiplier to 21.5 again and add slightly more voltage? Im running 32bit prime95 and im currently doing a blend test.

This is how HWM and CPU-Z look, i also have task manager running to ensure the CPU is under full load

http://prntscr.com/7sdyhi

another screenshot after 1 hour of prime95

http://prntscr.com/7seedc

 

[RGone]

@ Mark. Please attach screen captures to the forum so that all can view them and also so that "later" the images are still available down the road when your outside forum links are dead and gone.

I looked at first outside capture shown and according to what I see the ram should be 10, 10, 10, 30 & 39 according to the Jedec settings shown in CPUz SPD Tab. You have the ram tighter than the specs and can cause a problem when trying to get stability. Such ram timing tightening can have the same effect as overclocking the ram which is best 'tested' after cpu stability is assured and not have stability in jeapordy with both cpu speed and possible ram issue.

RGone...

 

[Arbiter Odie]

Hello, Mark! Welcome to the forum!

Could you please try something for me?
Put the CPU back to stock, and see if Prime95 fails again. That will rule out processor instability.
Then, use the Custom setting for Prime95, and make it look like this:



And see if the error happens. This will not test your RAM at all, but it will properly stress the CPU.

Thanks!


EDIT: Also, do what RGone said about the RAM timings, we want to rule that out too.

 

[Mark.]

Here are the CPU-Z tabs and HWM tabs requested, this time running the test Arbiter Odie suggested however i forgot to turn the clock speeds back to standard spec.. The test has been running for over 2 hours 30mins so far with no issues yet, highest CPU temp so far was 58c and Motherboard is 27c. I had also noticed my RAM timing had changed, it was 10,10,10 with the first few overclocks but has since went down to 9,9,9. Im using D.O.C.P mode when changing multipliers and voltages etc, doesnt this setting allow RAM speeds to adjust themselves? Should i go into BIOS and do everything manually instead? How do i take screenshots of my BIOS setting to upload for people to see here?



Thank you all for the quick responses : )

 

[Johan45]

For now I'd set the ram to 10-10-10-30-39 as it suggests in jedec #5. Your temps look good, it's the CPU and package temps you need to keep an eye on. CPU max 70c package max 62c. It's likely the ram that was throwing the error in your last set of tests. Kingston typically usues tighter subtimings from what I have seen so having the mains at CL9 could have thrown it off. Using DOCP isn't likely to work since the ram doesn't have an XMP profile for the mobo to work from. So if it were me I'd do it manually and set the ram timings/voltage myself as well ast the OC.

 

[Mark.]

Here are some screenshots of my settings in BIOS, I ran prime95 in custom mode with FTT's 2048 for over 4 hours whilst internet browsing etc and encountered no problems, CPU temp was 60c at its highest and package was 51c for the most part.
What would be the best way for me to change the ram speeds back to jedec 5 settings, 10,10,10? I have a feeling i may have set my CPU/NB Voltages a bit too high?

Thanks for all the help so far

 

[Arbiter Odie]

You did great. Thank you for the extra screenshots

I've taken one of the pictures, and modified it slightly.



That is where the manual settings for the RAM live.

Set the first three options to "10"
Set the fourth option to "30"

Save, reboot.
Try the blend test again.

Oh, yeah. Just set the CPU/NB voltage back to auto. I don't know where the safe limit is, and it probably isn't helping anyways.

As an amusing side note, I thought your pictures were from my bios, and I was wondering how on earth you'd managed to take screenshots from my computer Our clock speeds are the same, and the bios is almost identical between our boards.


EDIT: for spelling error and bad formatting

 

[RGone]

@ Mark...

...When you wrote this I ran prime95 in custom mode with FTT's 2048 for over 4 hours whilst internet browsing etc and encountered no problems ; it is okay as far as it goes. We and I suggest you run the P95 in blend mode so that ram and some of the connecting busses are also tested. If you run a particular size and one as small as 2048, it all fits in the cache of the cpu and nothing else gets tested. Not so good in the long run. Well not for an overall test for a form of stability.

By the way AMD says CPU_NB is good for up to 1.45V with watercooling and to 1.35V for air cooling so no problem with 1.3V. The faster I push my cpu; it requires some additional CPU_NB to allow the buss the ram talks over to be stable. I usually run 1.285V for CPU_NB and there are some in here that have to run a full 1.35V to the CPU_NB to get DDR3-1866 ram to run at just stock speeds. So no there is not a number "X" for CPU_NB voltage, no more than there is a one size fits all for CPU voltage. But if you are not running hot then 1.3V on the CPU_NB will hurt nothing and hedges against ram issues as ram speed rises and CPU speed also.

RGone...

 

[Mark.]

Ive just been running the test in blend mode for 3 hours and 49mins and received the same error the first time i tried to use prime. Ive added a screenshot for you all to see, the last time it was worker 4 that stopped, this time its worker 1. I rolled back the voltage on the CPU/NB from 1.3 to 1.225 which is where it was set originally. Ive changed my memory back to 10,10,10,30,39 and added 0.25 vcore power, still with 22 multiplier @ 4.4ghz.

any ideas?

 

[Arbiter Odie]

I know what to do.

Mark, have you ever seen Memtest86+?
http://www.memtest.org/

This is a memory test tool. You boot to it from a disk, or a usb drive. It stress tests your RAM like nothing else-- "blend" doesn't even come close. Download this, and run it right before you go to bed. Remember to use f2 mode so it goes faster!

This will tell us for sure if prime95 is the problem. And honestly? I would use this tool for memory tests, and prime95 for the cpu (using the custom setting), and not bother with blend. It's more annoying because you aren't booted into the windows environment, but if you do it right before you go to bed, it's not too horrible.

Make sure you get at least a few passes before you pronounce it stable. I had the test fail on me two days ago (I'm tinkering with my rig again), and it took three passes to happen!

So use memtest tonight, and tell us how it goes


EDIT: RGone, what do you think of his VRM section? I only have experience with beefy ones, not the 4 phase ones.

 

[Mark.]

Yes, i had looked at memtest86 the other night but didnt download it, i downloaded it last night and tried to boot it from disc as i was going to bed but it wouldnt work.. I have a feeling its my disc drive thats the problem as the RPM doesnt seem consistent at any time =P . Il try boot it from USB instead tonight and see how i get on with that. For now i wont be doing anymore prime95 tests until i run memtest86 and rule out any ram problems, i would hope its not the ram as its one of the only things i bought new for this build haha.

I would like to know how you get on with further overclocking on your machine Arbiter Odie, im currently at 4.5ghz on 1.34v however without running any prime95 tests yet, although it does boot easily and im running it for general browsing and emails just now. Of course thats by no means a stable overclock but im curious as too how much can be done without going over 1.38vcore in BIOS.

Also, my voltage seems to add more to itself if needed, for example, i changed core voltage to 1.36 in BIOS although when i set it to that it seemed to sit more around 1.4 on idle and 1.45 under 100% stress on HWM and CPU-Z, when i went back into the monitor in the BIOS it would also read higher voltage than had been set but would still be 1.36 in the Advanced settings, seems to be between an extra 0.0250v and 0.0350v of movement give or take depending on the situation.. Because of this i need to factor in an extra bit of power when i set the cpu vcore, So to run at around 1.38 i set the BIOS to about 1.34 and it seems to make up the difference on its own? Its a bronze 80 plus BeQuiet 730w PSU, i know its not groundbreaking stuff but should i expect very slight voltage increases and drops from it?

When i had the vcore at 1.36 and it rose to 1.45 during prime95, the pc didnt seem to happy at all.. It heated up alot more than previous tests and generally seemed to run slower.. im currently at 22.5 on the multiplier @ 1.34 vcore in BIOS, also currently reading 1.344 in HWM with a highest reading of 1.368, it will reach 1.38 if a ran prime i imagine...

Excuse the long winded post, Thanks for your help

 

[Johan45]

The memtest disc likely didn't work since it needs to be a bootable ISO disc. You need to use some ISO burner software, there's many of them out there for free.
The voltage fluctuations are normal since you likely still have all the LLC settings on auto.

 

[RJARRRPCGP]

Looks like a Prime95 bug or a Windows bug. I would look for bug reports. It may be a buggy kernel hotfix.

 

[Arbiter Odie]

Mark, did you burn the iso to the disk properly? Do you know how to do it with a linux iso? It's like that.

My overclocking is with a zambezi chip, yours is a vishera. They aren't exactly the same. Yours will probably clock higher than mine, at a lower voltage. Also, my chip is a lemon so there's that factor too. If you want my guess, your 4.5 GHz clock is probably attainable, but you might need to bump the voltage a little. We'll see.

You mentioned the voltage slippage; that's a function of the LLC



Basically, most (all?) motherboards have trouble delivering the right amount of power when the processor is going full tilt. The LLC tries to help with this, by causing the board to overvolt by quite a bit, in an attempt to keep the voltage at, or above, the value you set.
Try turning it off, and see what happens when you run prime. It should be quite interesting.

And yeah, when the voltage goes crazy high, the computer will get hot, heh. I've also experienced the perceptibly slower operation with stupid high voltage, but I don't know what causes it. It might have to do with overheating... I dunno.

 

[Mark.]

Thank you for your replies,

I do have daemon tools but the thought didnt occur to me until i tried the disc, ive since put memtest86 on a usb stick and it runs no problems. I only made 1 pass in the test though just to see if the program ran ok. I didnt encounter any problems with 1 test but ive yet to leave it over night.

Ive since returned the BIOS setting back to default as i was planning on adding another 8gb stick of ram, i hadnt put my first stick in the optimum slot for dual channel when i built the system anyway so i changed it over and ordered another stick of ram. I still need to install 64bit OS and buy a new GPU aswell so i may hold off overclocking again untill these things are dealt with.

I will try disable the LLC function when i overclock the machine again although now that you have explained it to me id imagine thats why i might be able to run a lower voltage in BIOS and still achieve 4.4 or 4.5ghz. I didnt test 4.5ghz much but the 4.4ghz clock had 2, 4 hours sessions on prime95 and a few 2 hour sessions too, the only problem i ran into was the one i originally made the post about.

Hopefully il be able to overclock again soon