Prime 95 shutdowns even when under clocked.

[druppes]

Hi,

I have just built my second pc and want to overclock it to around 4.6ghz. I started by testing it out at clock and quickly found it would have a system crash about 20 minutes into Prime 95 blend (with no Bsod or anything it just reset the pc). I then checked out the bios settings making sure everything was on default and I noticed that the Vcore was at around 1.45v on the auto setting so I moved it down to 1.25v. Prime 95 still crashed the system around the same time again.

My cpu temps are pretty average with the cpu around 60C degrees in speedfan with a max around 68C. Idle temps are around 35C and water temp is around 28C-30C with ambient temp around 23C

I have a cpu/gpu water loop setup. I tested the GPU properly before setting the water up and it performed fine but I never tested the cpu much as I didn't trust the stock cooler.

I have tried under clocking the cpu to 3.8ghz and its just as unstable on prime 95.

I have tried memtest and it passed and just tried intel burn test but only had time for a couple of passes and no crashes so far.

So far my only ideas on what it could be is dodgy bios settings somewhere, considering vcore setting was so off there could be more stuff i just don't know. Or overheating as my case has pretty bad airflow due to a preliminary watercooling setup not having much airflow so all the water cooled stuff is great but everything else is probably a bit hot especially the mosfet as the heat sink is pretty hidden from any airflow. Also I did a pretty bad job of applying the thermal paste to the cpu (went for the spread it over first as apposed to the pea in the middle style).

I'm a little worried to keep stressing the computer and causing all these crashes so wanted to seek advice early. If it where just crashing prime 95 I would try a lot more with different stress tests and settings etc but could be another 25 system crashes before I got close with my trial and error approach.

I'm new the pc building and my first venture was very simple so I may have done something wrong along the way. I could have done something as simple as missed a driver or two although I think I installed all the ones I needed. I haven't updated the bios as I was a bit intimidated by the idea of messing it up.

My system specs are :

Gigabyte GA-Z97MX-Gaming 5 Motherboard
i7 4790k
GTX 780
500GB Samsung 840 Evo SSD
Corsair Vengeance Pro Series Silver 16GB DDR3 1600MHZ 9-9-9-24

 

[tomkis90]

Try a diffrenet stress testing program because prime95 sometimes does that to haswell cpus.
OCCT (Linpack test) , Aida64, Intel burntest works great.

If that doesn't help lower your Ring Ratio/Ring Frequency to 4.2ghz or 4.1ghz or increase the Ring Voltage by just a bit, that stopped me from crashing.
( I got to stable 4.7 ghz wtih 1.256V )

EDIT : also change the ram frequency to 1333mhz while overclocking and testing for stabilly, after the cpu is stable you can go ahead and mess with the ram.

Also be sure to choose the soliuton that helped you so that another people can see it!

 

[druppes]

Try a diffrenet stress testing program because prime95 sometimes does that to haswell cpus.
OCCT (Linpack test) , Aida64, Intel burntest works great.

If that doesn't help lower your Ring Ratio/Ring Frequency to 4.2ghz or 4.1ghz or increase the Ring Voltage by just a bit, that stopped me from crashing.
( I got to stable 4.7 ghz wtih 1.256V )

EDIT : also change the ram frequency to 1333mhz while overclocking and testing for stabilly, after the cpu is stable you can go ahead and mess with the ram.

Also be sure to choose the soliuton that helped you so that another people can see it!

Thanks so much for the great response. It gives me hope to hear it might just be prime95 (I was thinking that was very unlikely) I have just done some IBT testing and its very stable at a simple 4.2ghz @ 1.2V. I ran furmark at the same time and although my water temps shot up the system was still stable. This might be a fluke so ill try some more in a bit.

I'm guessing I'm ready to start overclocking now so I will push it up 4.6 @ 1.3 and work down from there with Voltage, and hopefully get a nice safe stable overclock there, happy not to reach higher for now.

Will let you know how it goes. Thanks for the help

A minor question, How can you run intel burn test more than once? My current process is waiting for one pass then clicking run again. I put in 20 times but it seems to be doing way too quickly to be running 20 times (takes around 5 minutes)

Also I will try OCCT as well to make certain its stable once IBT passes a test.

I had already set the ram to 1333 when I first couldn't pass Prime95.

 

[tomkis90]

Allways glad to help.
What I would suggest is do one run of IBT at every step of oc. like if you increase the multiplayer by 1 do the IBT stress test if you pass it go back to bios and adjust the multiplayer or voltage more and when you're done do a few hours of OCCT because sometimes even when you pas IBT on the high setting you fail at OCCT after 10 or 20minutes.

 

[druppes]

Allways glad to help.
What I would suggest is do one run of IBT at every step of oc. like if you increase the multiplayer by 1 do the IBT stress test if you pass it go back to bios and adjust the multiplayer or voltage more and when you're done do a few hours of OCCT because sometimes even when you pas IBT on the high setting you fail at OCCT after 10 or 20minutes.

Well I found that on standard mode it goes very fast when I just turned it up to max settings it took a lot longer per pass. Still no problems on highest settings. I does however produce a fair amount of heat with my maximium been 81C and average 60C which seems a bit hot for my 4.5ghz @1.26v (new oc). However as I said i'm running air through medium density radiators with the cases stock fans that run at 400 rpm lol. Also my thermal paste was so unbeliveably bad, I just spread it out with my finger, there must be so much air in there lol. Now I have a stable system I will get some NF-F12s and some better thermal paste and was thinking of getting a new mosfet cooler as the one at the moment is getting no airflow what so ever, poor thing.

I did look online and I am surprised I hadn't found sooner that prime95 isn't optimal for haswell yet. Thankyou so much, could have gone through a lot of pain without your help.

 

[tomkis90]

I have the same cpu and at voltage of 1.264v I can do 4.7ghz it crashes after two or three hours rendering so If I would bump the voltage to 1.27v it would be stable.
Try setting the clock speed to 4.6ghz at the same voltage it should work just fine.
DO NOTapply thermal paste with your finger its not good for you and for you cpu.
I have Noctua Nh D14 it already comes with good noctua thermal paste and I dont even go past 70C at 4.6ghz.

 

[druppes]

Okay, yeah it was probably a pretty bad idea to use my finger to spread the paste. On the plus side it tasted pretty good when I licked my fingers clean .

So I'm going to grab some new thermal paste with the order of fans. thinking arctic silver 5 but then read its not actually that good and so I'm going for Prolimatech PK-1 which is meant to be one of the best. Also got some cleaning liquid. Best to do it properly this time.

Ow and the system is defo stable at 4.5 ghz so really really happy. Ill try 4.6 will let you know if it works.

 

[tomkis90]

Cool!
And I'm really hoping that licking the finger was just a joke haha.

 

[druppes]

Haha yes it was a joke.

I'm currently just tweaking the 4.5ghz to see how good my cpu is. I am at 1.22v and dropping still very stable. Will try 1.21 next. My temps are fairly high at the moment but I don't want to change the thermal paste till I take apart the water loop in a couple of days for optimization. Do you think the finger spreading is safe for a few days while I wait? And is it okay to stress test with it and also is 85C max (on max IBT settings) too high for now? I know in the long term its a bad idea but for a a few tests is it okay or should I just use my pc normaly till i get better thermal paste application on. Sorry for all the questions, very off topic from first post now hehe.

 

[tomkis90]

Haha yes it was a joke.

I'm currently just tweaking the 4.5ghz to see how good my cpu is. I am at 1.22v and dropping still very stable. Will try 1.21 next. My temps are fairly high at the moment but I don't want to change the thermal paste till I take apart the water loop in a couple of days for optimization. Do you think the finger spreading is safe for a few days while I wait? And is it okay to stress test with it and also is 85C max (on max IBT settings) too high for now? I know in the long term its a bad idea but for a a few tests is it okay or should I just use my pc normaly till i get better thermal paste application on. Sorry for all the questions, very off topic from first post now hehe.

There is NO need to hurry, your cpu can wait!
Its better to wait than to damage something but you should be just fine.
Also I would recommend watching this video on how to apply thermal paste

 

[druppes]

Great, I will be careful till i get some proper thermal paste application, OCCT does run cooler thankfully. I am at 4.6Ghz with Vcore of 1.21 and had a BSOD after a good 30minutes of OCCT so I guess I just need a little more Voltage and i'm good to go for now. Will probably just go 1.23 for stability and avoid all the testing of voltage in between for now. Thanks so much for all the help.

 

[tomkis90]

No problem.
You can post here if you need any more help.

 

[druppes]

Hi Tomkis,

I would like to let you know that this is now 100% solved. I have Re-seated my CPU using the rice sized technique in that video. It has lowered my temps by 15 - 20 degrees!!!! Didn't expect such a huge change. This has allowed me to properly stress test the system and everything is running perfect, I have done 20 passes on IBT on extreme with a max temp of 70 and average of 55 (@4.6Ghz).

I hope other new overclockers will find this thread or others about Prime95 not working optimally with the new Haswells.

Thanks,

 

[tomkis90]

Nice!
If you want you can push it to 4.7ghz because 70C is still very nice temp under that load.

 

[druppes]

As you suggested I just tried 4.7 and can't get it to work,

I should probably create a new thread now haha but thought id ask you first,

To push to 4.7 I raised the Vcore a fair bit from 1.22 - 1.3 thinking I could then move the vcore down step by step untill I get unstable then move it back up and thus have a good overclock.

However when raising my vcore up my system cant take more than about 3 seconds of any stress test at first I thought it was the new 4.7ghz but I tried raising the Vcore @ 4.6ghz and what was a perfectly stable overclock also dies within 2-3 seconds of any stress test.

My setting where ( I lowered the uncore from 4.2 to see if that helped)

4.7Ghz - Vcore=1.3 - uncore=3.9 - CPU ring = 1.180

The ring ratio although I have changed i'm not sure if its actually been changed as I can't/don't know how to monitor it and the stock value is pretty low.

These are pretty much all I have changed expect for some recommended power savings stuff to be turned off. I havn't touched the Vrin overide (or whatever its called) its currently at 1.800V Also I cant find the uncore voltage in my bios. My Ram is set to low settings at multiplier of 1.33 and the voltage is uped slightly from 1.5 to 1.55.

The temps in the 3 seconds before shut off look to be 45-55 degrees.

Any ideas? I'm thinking it has to be a lack of some voltage that is also required to be highered as the Vcore is increased.

I am actually pretty happy at 4.6 but would be nice to know I have the option to raise it by more.

 

[tomkis90]

I have the same problem with 4.8ghz I just cannot get there. as soon as I stress the cpu even with 1.31v it just crashes after like a minute so I can not really help you with that. sorry.

 

[druppes]

Ow well, Yeah its a problem with stressing it with higher Vcore rather than the higher GHz,

I will try the opposite to the normal approach and raise the GHZ and then keep raising the Vcore until I can get it stable.

Any advice where to start with Vcore? 4.6 is stable at 1.215-1.22. So i guess i can start with 2.23.

 

[druppes]

Actually went for another approach, Will stick at 4.6 and raise Vclock until I get instability.

 

[tomkis90]

Ow well, Yeah its a problem with stressing it with higher Vcore rather than the higher GHz,

I will try the opposite to the normal approach and raise the GHZ and then keep raising the Vcore until I can get it stable.

Any advice where to start with Vcore? 4.6 is stable at 1.215-1.22. So i guess i can start with 2.23.

If you somehow where amble to apply 2.23v to your cpu it would probably explode.
Max safe voltage for haswell is around 1.3v.

 

[druppes]

Hehe meant to say 1.23 yeah would be bad use 2.23 if at all possible.

Incase you where interested I have managed to get it working stable on both 4.7@1.25 and 4.8@1.31

One of the problems I had was some bad bios settings, the thing that made the most difference was changing CPU VRIN LOADLINE CALIBRATION from normal to extreme. It instantly made everything 10 times more stable. I also turned off all the suggested power saving options. 4.8ghz crashes OCCT after maybe an hour and half whereas I still have yet to crash anything on my 4.7Ghz clock so I went with that for better stability.

It is also working perfectly for hackingtosh which is a suprise.

The only negatives are if i start raising the uncore frequency above 4.0ghz it starts getting unstable very quickly but i guess I can live with that.

Thanks once again for all the help.

 

[tomkis90]

Maybe I need to find this CPU VRIN LOADLINE CALIBRATION myself, because I never changed it to anything and I can't get 4.8ghz stable even at 1.31v.

But your results are fine and I'm happy for ya!