testing stability with z97 and 4790k...

beekermartin · Dec 11, 2014

I recently installed a z97 motherboard and 4790k. I immediately ran into BSODs using the ram from my x58 rig. It was stable with 2 sticks of 2gb but with 4 sticks I would get random BSODs. I decided to get new ram and now have 2 x 4gb of g.SKILL 2133 ram installed. I have everything at stock settings in the bios except for xmp for the ram and I manually set the voltage to 1.6 volts because that is what the ram is suppose to be set for with haswell.

The problem I am now having is hard crashes and reboots using p95 blend test. My temps are fine. Highest core hit 71c. It just crashes and reboots. I tried the same thing using my old 2x2gb of 1600 ram and it does the same thing. I ran memtest on both sets of ram and they both passed.

Some people have posted in my other threads that I shouldn't be using p95 with haswell. I am using the newest version, 2.85 I believe. I've always used p95 as one of the stability tests with all my prior builds.

I ran 3mark11 this morning and it passed 3 times. I hit 14823 with my 780 set to 1254 max boost and 7000 ram. So it is stable running 3dmark.

What other programs should I use to test stability with haswell? I would like to overclock the 4790k but I don't want to do that until I know it is stable at stock speeds.

Vexillarius · Dec 11, 2014

What does the RAM's spec sheet say about voltage? The crashes could be because you set the voltage to 1.6V if that's not the voltage determined by G.Skill.

1.65V works just fine with Haswell. Intel doesn't officially support anything above 1.575V, but higher voltage RAM shows up in plenty of mobo QVLs and it works absolutely fine. Intel doesn't support it because it's technically overclocking, not because it would damage anything.

EDIT: Intel has its own stress testing utilities. Try Intel XTU.

beekermartin · Dec 11, 2014

It is rated for 1.6 for haswell and 1.65 for older processors. I already tried 1.65 and that didn't help.

What is strange is I got the exact same hard reboots with my old known to be good old ram when I only used 2x2gb. So I don't think the ram is the issue.

Vexillarius · Dec 11, 2014

Ah I understand, my bad.

Have you tried putting the modules in different slots?

Honestly, if it's passing XTU/Linpack/OCCT/etc as well as Memtest and you don't get random BSODs or other crashes or data corruption then it's probably related to Prime, not to your system.

beekermartin · Dec 11, 2014

I will try the programs you mentioned: XTU, Linpack and OCCT.

I still find it strange that P95 just hard crashes and reboots??? I read it has something to do with the new instructions that Haswell has. I am still doing research about it.

beekermartin · Dec 11, 2014

I have tried the new ram as a single stick and with both installed but I always used slot 1 and 3. I had the same issue with 4 sticks of 2gb installed. They all passed memtest so I doubt the slots are the issue.

Vexillarius · Dec 11, 2014

I'm still not entirely sure what the problem with Prime95 and Haswell is. I know the AVX instruction set causes a 0.1V bump in VCore if voltage is set to anything but manual, causing skyrocketing temps (this is on Intel, it's not Prime's fault). Prime is also just incredibly hard to get stable on Haswell, unrealistically so, even for a synthetic.

CompuTronix · Dec 11, 2014

Use P95 version 26.6 - http://windows-downloads-center.blogspot.com/2011/04/prime95-266.html

It's explained in here: Intel Temperature Guide - http://www.tomshardware.com/forum/id-1800828/intel-temperature-guide.html

beekermartin · Dec 11, 2014

I read in another thread to use version 26.6.

What I don't understand is why the version I am using, the newest 28.5, crashes and reboots but my highest core temp has only hit 73c. So it isn't crashing because of temps.

CompuTronix · Dec 11, 2014

Haswell has AVX, AVX2 and FMA3 instruction sets which greatly accelerate math calculations. P95 after version 26.6 forces AVX, AVX2 and FMA3 code to run on your FPU. If Vcore isn't high enough, your processor will crash.

beekermartin · Dec 11, 2014

CompuTronix :

Even at stock settings??? I believe I am hitting @1.22 or so according to CPU-Z.

beekermartin · Dec 11, 2014

I'm sorry to bump this up but should I be failing p95at stock 4790k settings????

CompuTronix · Dec 11, 2014

No, but just to say we did, have you tried 26.6?

beekermartin · Dec 11, 2014

I haven't yet but I will when I have time. I still find it hard to believe that the newest version is so tough that a stock cpu can't handle it.

beekermartin · Dec 12, 2014

I am now running 26.6 blend test while I go to bed. Hopefully I Have something to add when I wake up.

beekermartin · Dec 12, 2014

P95 26.6 blend test ran for over 13 hours with no errors or crashes.

Is that enough of a stability test to say the new ram is stable at 2133mhz and 1.60 volts???

Vexillarius · Dec 12, 2014

beekermartin · Dec 12, 2014

I am running IBT now at maximum memory. One core has hit 85c after the first run. I guess IBT can really produce some high temps!

CompuTronix · Dec 12, 2014

That's shown here:

Shown above from left to right: Small FFT's, Blend, Linpack and Intel Burn Test.

CT :sol:

beekermartin · Dec 12, 2014

How does this look?

Is it ok for me to try overclocking now? I might actually try to find the lowest stable vcore voltage first just to see if I can lower it from the auto setting.

CompuTronix · Dec 13, 2014

Please run P95 version 26.6 Small FFT's for 10 minutes and let's see what your Core temperatures are.

Also, please let us know what your ambient temperature is.

If your cooler is the H60 shown in your signature, please keep in mind that many big-air coolers are better, so don't expect good overclocking results.

beekermartin · Dec 13, 2014

My highest single core temp when I ran P95 for 13 hours straight was, I believe, 73c. I am at work now but I will run it for 10 minutes when I get home. My ambient temps were 64 degrees F when I ran the test.

I disagree with you about the H60. In push/pull with a good air flow case it is an excellent cooler. It is within a degree or two of the H80 and that is within three or four degrees of the H100. I understand there are air coolers that can equal or beat it slightly but I am not going back to huge tower coolers. I have had excellent results using the H60 and I don't plan on changing it anytime soon. Maybe I just got a "good one" but I am very happy with it. If I switch to a case than can handle a larger radiator, like the H100 has, I will upgrade the cooler then.

My highest core temp running IBT was 87c. I ran it twice for ten runs with memory set to maximum and both times core 2 hit 87c. The rest of the cores were lower. Vcore was @1.212 most of the time but it would go slightly higher sometimes. I plan on running it again and watching the Vcore closer.

I think my first step is going to be to set the max speed to 4.4 for all cores and try lowering the voltage until it becomes unstable. That way I can find the lowest possible Vcore for the stock 4.4 speed. Once I know that I will start to overclock to find the sweet spot between voltage and max speed. Since it is so fast at stock settings I don't expect a big difference between stock and the max overclock I achieve. The difference between 4.4 and 4.7 is not gonna make a big difference, if any, for most of what I use my computer for. With that said I paid for an unlocked processor so I mine as well tweak it a bit.

CompuTronix · Dec 13, 2014

beekermartin :

You said that you ran Blend, which is a cyclic workload that yields fluctuating temperatures, and is not appropriate for thermal testing. Please run Small FFT's, which is a steady state workload that yields steady temperatures, and is appropriate for thermal testing. You only need to run it for 10 minutes.

beekermartin :

Thank you for this information. Since your ambient is 64F which equals about 18C, and Standard Ambient is 22C, then whatever your Small FFT's temperatures are, you need to add 4C to correct your test temperatures to what they would be at Standard Ambient. Intel's Thermal Specifications are based on standards. Also, you must remember that your Core temperatures won't look very good come summer, unless you maintain a 22C environment for your rig.

beekermartin :

Respectfully, I was not disputing your decision processes or reasoning for purchasing your computer components. It's not my place to criticize anyone's rig. I know that 22 nanometer processors are extremely difficult to cool when overclocked. I see a lot of these threads, so there's a few things that I feel obligated to point out. There was nothing more to it.

beekermartin :

87C at 18C ambient becomes 91C at 22C Standard Ambient. However, if you refer to the SpeedFan Chart above, it shows that IBT's fluctuating workload peaks at about 110% which is 5 to 6C higher that P95 Small FFT's.

Also, as you overclock and increase your Vcore and clock frequency, if you keep an eye on "Power" or "Watts" in Core Temp or HWMonitor, you'll see that Core temperatures scale up with Watts.

CT :sol:

beekermartin · Dec 13, 2014

Thanks for the excellent responses CT!

Your right. I did run blend. I will run the small FFTs for ten minutes and post my results.

What is the max individual core temp you think is safe while running IBT? I understand ambient will have an effect on that. I monitor temps constantly. I don't believe there is any game that will come close to hitting the temps IBT does. I just want to be able to test stability using something like IBT so I know the worst case scenario.

I did some research and I have come across a few threads about people hitting 100c! with Haswell processors.

CompuTronix · Dec 13, 2014

beekermartin :

I recommend only P95 version 26.6 Small FFT's for "Thermal Testing", and look for an "average" Core temperature in the mid 70's when corrected to 22C Standard Ambient, as described in the Temp Guide. Any and all popular "Stress Testing" utilities should be used to determine if your rig is stable, and may yield Core temperatures in the 80's. Any games or apps should also be used to test for stability.

beekermartin :

These threads commonly involve the stock cooler, improperly seated coolers, P95 versions later than 26.6, or the FPU test on AIDA64. For 22 nanometer processors on big-air or decent liquid cooling, if 1.25 Vcore is not exceeded, then all other factors being equal, unless someone is making an obvious blunder such as testing at high ambient temperature or with limited ventilation, then popular stress testing utilities shouldn't produce extreme temperatures into the 90's.

CT :sol:

testing stability with z97 and 4790k...

Distinguished

Reputable

Distinguished

Reputable

Distinguished

Distinguished

Reputable

Intel Master

Distinguished

Intel Master

Distinguished

Distinguished

Intel Master

Distinguished

Distinguished

Distinguished

Reputable

Distinguished

Intel Master

Distinguished

Intel Master

Distinguished

Intel Master

Distinguished

Intel Master

Share this page