Instability on new X99-SOC Champion build

[2dfx]

I'm trying to track down a system instability issue on my newly built pc. Specs are as follows:

X99-SOC Champion
5820K
32GB g,skill Ripjaws V F4-3200C16Q-32GVK 32GB kit
2x Fury X
950 Pro m.2
1300W PSU

Mersenne prime blended is crashing. Win 10 is unstable and resets randomly at full load. Here's what I think the trouble is and why:

1. Bad memory controller on CPU (I bought it used)
2. Incompatible memory with the board - unlikely, see edit

For the first reason - with the entire kit of RAM installed or in 2 channel, Prime crashes within 5 mins. With 1 stick installed and only 1 stick, Prime runs for about 4-5h then crashes with a segmentation fault.

Now for the second reason - with all of the above considered, Memtest86+ 5.01 will run for 24h+ without throwing any faults. I doubt the memory itself is bad which leads me to suspect these two scenarios. Also, this particular kit is NOT on Gigabyte's QVL for memory.

Am I down the right path here? System is bone stock, XMP is enabled. Thanks all!

EDIT - went out and purchased a kit on Gigabyte's QVL...same result.

[dewot]

Re: Instability on new X99-SOC Champion build

Hi 2dfx,

That is a pretty nice board - I am thinking your problem may be you need some extra voltage on the memory controller, or the ram, or both.
Used to be a more common problem, but sometimes with all your memory slots populated, you need extra voltage to stay stable.

Passing memtest for 24 hours is a strong indicator that the CPU, memory controller and ram are all working together fine at your current
memory timings and speeds. Memtest is only testing small segments of ram at one time though - it does not test all memory chips at once.
Meaning the memory controller does not really get "loaded up" in memtest, so voltage stability issues with the controller may not appear.
Also, because memtest is only working one ram chip at a time, your ram voltage may be stable enough to pass under memtest, but when you
are using all ram chips at once, your ram voltages may need to be increased for stability.

You can find segmentation error on wikipedia - it could be a programming problem in your prime, but I think that would have been reported
and fixed by now, as the 5820 has been out for awhile. Still, I would download the latest OCCT and test using that. It will not heat the CPU
as much as prime testing, but it should help to see if it may be an issue in prime. Test at all stock clocks. Under OCCT, I have found the
following is generally true:

BSOD, code 124- not enough vcore
BSOD, code 101- not enough voltage on the memory controller, or possibly other internal CPU voltages
Screen freezes (no BSOD)- Ram problem. More ram voltage, or reduce ram speeds.
"Error detected" message in OCCT- not enough voltage on the memory controller, or possibly other internal CPU voltages

Again, those are general guidelines, not hard and fast rules.

Google "Intel Processor Diagnostic Tool" and download that to test your CPU. It will test a lot of things including the memory controller.
However, even if you pass the Intel test, the increased memory controller voltages may be necessary for stability with a fully loaded board.

Finally, I recommend downloading Driver Booster free (or something similar). Watch the install screens carefully - it is pitching other stuff to
support its freeware status. Even tho you are on Windows 10, not all of your Intel INF files and drivers are necessarily the latest (and
Windows 10 is not 100% stable yet). I just built a Z97 / I5-4690K system with Windows 10 Pro. Driver Booster found several INF file
updates including the DRAM controller, PCIE controller, AHCI controller, and several others. Those INF files tend to fix very specific issues. But who knows, if it updates your DRAM controller, there may be an issue there.

Sorry, that post got to be pretty long. Hope that helps!

[2dfx]

Re: Instability on new X99-SOC Champion build

Sorry, that post got to be pretty long. Hope that helps!

No trouble at all, very detailed and got me thinking about variables I had missed.

Bumped the Vmem to 1.35 with the same result. It is still more unstable the more sticks I put into it.

Intel CPU Diagnostic passes.

EDIT - NVM about OCCT voltages. Gigabyte's hardware monitor and the BIOS itself are reporting the correct voltages.

[dewot]

Re: Instability on new X99-SOC Champion build

Good news on the CPU! Just to be sure - on the latest bios? Did you do a CMOS reset to clear settings made by the prior owner?

Increase the system agent voltage by .25 or .30 and see if that helps stabilize things. You can go as high as .50. That is the memory controller voltage. Not positive, but I think base system agent voltage is 1.0 volt.

Check memory timings - the GSkill specs say tested timings were 16-16-16-36-2n @ 1.35 volts.

Tried to keep it a little shorter this time )

[2dfx]

Re: Instability on new X99-SOC Champion build

Good news on the CPU! Just to be sure - on the latest bios? Did you do a CMOS reset to clear settings made by the prior owner?

Increase the system agent voltage by .25 or .30 and see if that helps stabilize things. You can go as high as .50. That is the memory controller voltage. Not positive, but I think base system agent voltage is 1.0 volt.

Check memory timings - the GSkill specs say tested timings were 16-16-16-36-2n @ 1.35 volts.

Tried to keep it a little shorter this time )

Running F21a which as of this morning, is Gigabyte's latest for this board. CMOS clear has definitely been done multiple times.

I have boosted the memory voltage to the module's spec of 1.35V and manually set the timings to 16-16-16-36 as per the modules themselves. It still crashes.

For reference http://gskill.com/en/product/f4-3200c16q-32gvk that link is the memory kit I'm using. I'm going to see if I have a "system agent" setting in BIOS I can change. For now I'm re-testing at 17-17-17-39.

[dewot]

Re: Instability on new X99-SOC Champion build

You could try running your memory at 1600 (half speed) to see if your board is stable at that speed. I would guess that it would be. That would more or less confirm memory is the cause of the instability.

Here is an article on overclocking Haswell E. You are not overclocking the CPU, but your memory situation is somewhat like overclocking - you are using all your memory slots, and that DDR4 is very high speed stuff. So your memory controller will probably need extra voltage to get stable.
This link explains system agent voltage pretty well (note they are using an Asus board - Asus sometimes uses different names for various voltages in the bios than Gigabyte):

Intel Core i7 5960X Extreme Edition S-spec QFRA CPU Overclocking Guide

Page 3 explains voltages, but there is also some pretty good info in the rest of the article. Gigabyte does call it system agent voltage in the Z97's. When I overclock a Z97 board I automatically go +.20.

[2dfx]

Re: Instability on new X99-SOC Champion build

You could try running your memory at 1600 (half speed) to see if your board is stable at that speed. I would guess that it would be. That would more or less confirm memory is the cause of the instability.

Here is an article on overclocking Haswell E. You are not overclocking the CPU, but your memory situation is somewhat like overclocking - you are using all your memory slots, and that DDR4 is very high speed stuff. So your memory controller will probably need extra voltage to get stable.
This link explains system agent voltage pretty well (note they are using an Asus board - Asus sometimes uses different names for various voltages in the bios than Gigabyte):

Intel Core i7 5960X Extreme Edition S-spec QFRA CPU Overclocking Guide

Page 3 explains voltages, but there is also some pretty good info in the rest of the article. Gigabyte does call it system agent voltage in the Z97's. When I overclock a Z97 board I automatically go +.20.

Loosened the timings even further to 18-18-18-40 with no improvement. Boosted agent voltage by 200mV. Actually resulted in the system being more unstable, so I reverted that change.

Finally, I did as you suggested and tried the system at 1600 and even 1333MHz. Still with the same result. Prime will either fail entirely, or it will have 1-4 workers fail. Can't be memory.

I think at this point I'm looking at a defective CPU or motherboard.

[dewot]

Re: Instability on new X99-SOC Champion build

If the CPU passed the Intel test, pretty good chance its ok. I do see on Newegg reviews several people having memory issues.

You might open a ticket with Gigabyte - they are usually pretty good at supporting those high end motherboards. That board should be new enough to be under warranty, but I'm not sure about Canadian warranty rules for consumers.

Good luck!

[2dfx]

Re: Instability on new X99-SOC Champion build

If the CPU passed the Intel test, pretty good chance its ok. I do see on Newegg reviews several people having memory issues.

You might open a ticket with Gigabyte - they are usually pretty good at supporting those high end motherboards. That board should be new enough to be under warranty, but I'm not sure about Canadian warranty rules for consumers.

Good luck!

Read the reviews...essentially the same problem I'm having. No memory works...instability...hanging on POST code 50.

Opened an RMA with Gigabyte. Disappointed the board was DOA out of the box.

[dewot]

Re: Instability on new X99-SOC Champion build

That was surprising when you couldn't get it stable at 1600 - any modern baord will do that speed. I did see one of the reviewers say he took one stick of ram and moved it from one slot to the next until he found the slot that produced a code 50 when he booted the board. Maybe you have a bad memory slot?

Anyway, should be a nice board once you get it up and running!

[Vinster]

Re: Instability on new X99-SOC Champion build

try an older bios. new isn't always better.

Vin

[jaggerwild]

Re: Instability on new X99-SOC Champion build

Set Un-core toX33 and leave it there, don't ask!