Prime95 crashing during Blend (no errors during memtest86) - help me stabilize RAM (Ryzen 7 setup)

Blarzor · Jul 27, 2017

Ryzen 7 1700 @ 3.7Ghz (1.272V)
Gigabyte GA-AB350-Gaming 3 with F7 bios
G.Skill Ripjaws V 32GB 3200Mhz 16-18-18-38
https://www.gskill.com/en/product/f4-3200c16d-32gvk

What I've tried:
1. X.M.P. enabled keeps rebooting my PC and I had to clear CMOS to make it work again.
2. Now I'm using these settings with X.M.P. disabled (this settings have been told to me by a member of Gigabyte Forums).

2666Mhz. 14-14-14-34 @ 1.35V Command rate: 2T

These settings seem stable during normal use but not in Prime95 Blend test.

3. I've tried using Command rate 1T before but it crashed my PC during Blend test. Now that I've set it back to 2T, only Prime95 crashes.

4. Above mentioned timings did not show any error during 7 hour memtest86 test. They crashed PC ONLY during Prime95 Blend test.

Can you guys help me stabilize this RAM please?

TechyInAZ · Jul 27, 2017

Increase your SOC voltage a tad.

Blarzor · Jul 27, 2017

Are you talking about Dynamic VCORE SOC (DVID) or Vcore SOC?

In my BIOS, I have only Dynamic VCORE SOC (DVID). (apparently that is an offset or something in Intel language)

TechyInAZ · Jul 27, 2017

hmmmm, i'm not sure. I hope this helps: http://www.overclockers.com/amd-ryzen-overclocking-guide/

Blarzor · Jul 27, 2017

Hmm, unfortunately that doesn't tell much about that particular setting. Even the motherboard manual doesn't say anything about it...

TechyInAZ · Jul 27, 2017

Blarzor :

hmmm, I'll call in a few other peeps and see if they can help out.

Seanie280672 · Jul 27, 2017

Raise Dynamic VCORE SOC (DVID) to +0.180 this should put your over all SoC voltage at about 1.100v which is what AMD actually recommend it to run at.

Also, which version of Prime95 are you running, use 26.6 http://www.mersenneforum.org/showthread.php?t=15504

Are you monitoring voltages and temps with anything ? are you getting any vdroop ? unfortunately and this is one of the reasons I dont recommend Gigabyte B350 boards to anyone, they proper skimped, you dont even get loadline calibration (LLC) in the bios to stop voltage droop.

Something else to note, is ive just noticed you are running 2 x 16gb sticks of ram, these are double sided, and has the same effect as running 4 x single sided sticks of ram, it puts a lot of strain on the memory controller, so for now, dont expect to get any higher than 2666mhz, bigger isnt always better.

Blarzor · Jul 27, 2017

Seanie280672 :

Ok before I apply the settings you told me to, I'd want you to check the context of my whole overclock so far so you know what I'm exactly dealing with. Here are the photos of my bios:

Voltages (I already just visually applied your number for now):
http://i.imgur.com/XCUrqu7.jpg

CPU + RAM:
http://i.imgur.com/TSzEef1.jpg

CPU advanced settings:
http://i.imgur.com/bjnHhh5.jpg

Memory advanced settings:
http://i.imgur.com/TXJFdtq.jpg

Channel timing controls (same settings for A and B channel):
http://i.imgur.com/eACxCNP.jpg

CPU-Z Info:
http://i.imgur.com/DI7MXjj.png

I was using latest Prime95 28.10. May I ask why is the older version better?

I'm using Aida64 to control temperatures. Sorry I don't know what Vdroop is.

So after checking the pictures of my bios. Am I OK to insert the +0.180 to Dynamic VCORE SOC (DVID)?

Seanie280672 · Jul 27, 2017

Blarzor :

Yes good to go almost, just a couple of things, disable core performance boost as you are manually overclocking, disable AMD Cool & Quiet, leave the rest as it is.

Your CPU NB Volatge in AIDA64 should now read about 1.100v ish, give or take a couple of digits: http://imgur.com/g0ygJ2T

Older version of Prime as the newest version, your CPU doesnt support the coding completely, AVX instructions etc, this could be why its crashing.

Vdroop on your board isnt controllable from what I can see, theres no LLC in the bios, Vdroop = Voltage drop on the vcore, so lets say your set it to 1.25v in the bios for example, launch AIDA64 and monitor volatges and temp, but when you fire up prime, the volatge drops to lets say 1.225v, thats vdroop, voltage droop, normally board manufactures build a feature into the bios that you can adjust to stop this from happening, it locks in the voltage to a constant flow, zero droop, but for some reason, Gigabyte decided not to add it on there B350 boards.

Blarzor · Jul 27, 2017

Seanie280672 :

Blarzor :

Yes good to go almost, just a couple of things, disable core performance boost as you are manually overclocking, disable AMD Cool & Quiet, leave the rest as it is.

Older version of Prime as the newest version, your CPU doesnt support the coding completely, AVX instructions etc, this could be why its crashing.

Vdroop on your board isnt controllable from what I can see, theres no LLC in the bios, Vdroop = Voltage drop on the vcore, so lets say your set it to 1.25v in the bios for example, launch AIDA64 and monitor volatges and temp, but when you fire up prime, the volatge drops to lets say 1.225v, thats vdroop, voltage droop, normally board manufactures build a feature into the bios that you can adjust to stop this from happening, it locks in the voltage to a constant flow, zero droop, but for some reason, Gigabyte decided not to add it on there B350 boards.

Alright, thanks for information!

Will try it now. I also updated my post with CPU-Z information if you want to check it out btw.

Blarzor · Jul 27, 2017

Ok I just applied those settings and I'm currently running Prime95 Blend test as we speak. For some reason I don't have CPU NB Voltage data in my Aida64.

http://i.imgur.com/PDruVmm.png

Seanie280672 · Jul 27, 2017

Blarzor :

Might be on that list as SoC voltage, its a tad bit high, 1.14v not dangerously though, so no panic, max is 1.20v, when you get a min, drop it in the bios to +0.150v instead.

Also try this version of AIDA64 instead, I know it will only be the 30 day trial, but they update it all the time: https://www.aida64.com/downloads/NWVjNmYyM2Q=

Blarzor · Jul 27, 2017

Seanie280672 :

Well 2 bad news. Blend test was stable for 30 minutes and I closed it, then I decided to try Small FFTs test with new settings and PC crashed. I guess it's because of extra voltage. My CPU diode temperature was reaching 85 celsius before the crash. With DVID on auto my Small FFTs test never crashed PC. Prime95 crashed once even with the old version during Small FFTs test as well.

This was also the first time that my Windows 10 crashed in a way that I've seen the actual message without PC actually straight up rebooting.

I think that +0,180 brings too much heat. What should I try?

Seanie280672 · Jul 27, 2017

Blarzor :

Seanie280672 :

Well 2 bad news. Blend test was stable for 30 minutes and I closed it, then I decided to try Small FFTs test with new settings and PC crashed. I guess it's because of extra voltage. My CPU diode temperature was reaching 85 celsius before the crash. With DVID on auto my Small FFTs test never crashed PC. Prime95 crashed once even with the old version during Small FFTs test as well.

This was also the first time that my Windows 10 crashed in a way that I've seen the actual message without PC actually straight up rebooting.

I think that +0,180 brings too much heat. What should I try?

Bring it down to 0.150v

Also, which cooler are you using ? those temps seem very high for such a low voltage.

Search

Prime95 crashing during Blend (no errors during memtest86) - help me stabilize RAM (Ryzen 7 setup)

Blarzor

Distinguished

TechyInAZ

Titan

Blarzor

Distinguished

TechyInAZ

Titan

Blarzor

Distinguished

TechyInAZ

Titan

Seanie280672

Estimable

Blarzor

Distinguished

Seanie280672

Estimable

Blarzor

Distinguished

Blarzor

Distinguished

Seanie280672

Estimable

Blarzor

Distinguished

Seanie280672

Estimable

TRENDING THREADS

Latest posts

Moderators online

Share this page