Question Why is Broadwell x99 Asus ROG MB overclocked now with BSODs?

jakester100

Honorable
Feb 16, 2019
11
0
10,510
Situation: Overclocked PC was fine, but then got BSODs resulting in functionality only when not overclocked and added voltage given to CPU and DRAM.

Background: PC used primarily for League of Legends while also running Twitch, Spotify, Steam and Discord. Was set for 13% overclock using Asus UEFI preset on the EZ tuner with XMP on.

Then began getting BSODs, almost always IRQLs; intermittently almost daily. So, updated all drivers and still had problems. Ran Memtest86 and got a few error with all 4 sticks. Running sticks separately and the errors went away. Decided to up voltage on DRAM. Still got a few errors. Upped voltage on CPU and errors on Memtest86 went away.

Tried using the UEFI to do overclocking, which basically set memory to XMP. BSOD returned. Set it back to the solution that worked above and now it is functional again…I think.

CPU when taxed runs no hotter than 65C. Board temp sensors all look fine but one at 127C which is either close to the heat exchanger or busted.

Questions:
  • What might be wrong if anything?
  • Should I let it be or is there something that needs replacing?

Hardware
Rog Strix X99 MB with bios 1902
Intel i7-6850K Processor 3.6Ghz
CORSAIR Hydro Series H100i
MSI GeForce GTX 1080 Ti Founders Edition
Corsair Vengeance LPX 32GB (4x8GB) DDR4 3200 C16 for DDR4 Systems – White
Samsung 960 EVO - 500GB NVMe - M.2 Internal SSD, with 50GB free
Samsung 850 EVO 1TB 2.5-Inch SATA III Internal SSD
MSI SuperNOVA 850G2 850W PS
Acer Predator XB1 XB271HU bmiprz 27" 16:9 IPS Monitor
Sony BRAVIA KDL32EX523 32in TV

Bios set with no overclocking. My only changes:

CPU voltage 1.30v
DRAM voltage 1.35v


Thanks!
Jason
 
What is your VCCSA voltage set to? This is voltage for the IMC inside the CPU. This works in conjunction with DRAM voltage. Have an ASUS X99 Deluxe w/ 5820k build myself which was OC'd to 4.4Ghz. Now it's back at stock for a family member. This voltage is picky with X99. There is a good "range" for stability. Being to low or to high can cause issues. Also, what about CPU input voltage and cache voltage?
 
  • Like
Reactions: jakester100
Thanks for looking into it. I can't check yet as my kid is in the middle of a campaign...lol, but I'll let you know soon. I think that systems agent is at default, but let me check and hopefully get back to you within the hour.
 
No hurry, have some things to work on myself. I had an offset of +.143v for agent voltage. I believe stock is around .868v or so. XMP adjusts this, but it's based on averages. My particular CPU didn't like XMP default voltage, and after a bunch of research at the time, I figured out where to start adjusting. It's actually how I got signed up to these forums. My CPU voltage was 1.325 manual voltage 44X multi all-core, and cache was 1.25v 40x. CPU current capability 140% DRAM current 130%. DRAM @ 2666 with tightened timings at 1.35v. Ran this way for about 3 yrs. 1.3v on CPU was considered to be about max safe, but pushed a bit to the 1.325v myself. CPU runs fine today as daily driver.
 
Ok, here are the numbers...still wondering why after almost two years it stopped like being overclocked.

CPU core: manually set to 1.3V
CPU cache manually set to 1.3V
CPU systems agent (auto, running at .872V)
CPU input voltage (auto, running at 1.776V)
PCH I/O voltage (auto, running at 1.5V)
 
Hi there again. It's possible that degradation has set in, but hard to confirm that with certainty. Being that your SA voltage is stock at .872, I'd try bumping that up a bit for XMP and running latest bootable Memtest86 again. I ran mine at about 1.06v after the offset. Again, this is the voltage tied directly to the memory controller. Leave DRAM at the 1.35v setting. CPU voltage of 1.3v is fine. What speed were you running cache at? I needed 1.25 for 4.0Ghz on there. I also had input voltage upped to about 1.9v. I left PCH voltage at AUTO 1.5v, no need to change that from my own experience. If I remember correctly, I had a similar situation, at least with SA voltage, where on 2 occasions either months or even years apart, I had to go back and make some adjustments for this voltage for boot stability. However, my issues were not BSODs, I would have trouble during cold boot and a rare occurrence during warm boot. All this being said, with your Memtest86 issues using 4 sticks and voltage adjustments, I feel that the BSODS are memory or memory controller related.
 
  • Like
Reactions: jakester100
Ok; so here's the latest. Got some memory errors.

I made the voltage changes above.

Ran memtest. Got an error on test 4 in the first pass. Took out two sticks. Ran it again for 1 pass and no errors. Put in the other two sticks. Ran test with all. Got an error this time on test 5 during first pass. Took out two different sticks (making sure to put the remaining two in the right slots). No errors on 1 pass.

So it could be memory controller or maybe a problem with third and 4th slots?
 
Ok. The is strange. Memory slots themselves can indeed have issues or fail. Worked a a PC for a friend recently, and 2 slots failed on an MSI Z97 based board after plenty of testing. Any changes to PC or moving cooler around? Uneven mounting pressure on these types of CPUs is know to cause flaky RAM slot problems. Have you tried XMP off running at 2133 only with 4 sticks?
 
  • Like
Reactions: jakester100
OK. Everything normal without XMP and at 2133, then slots themselves are fine. You may end up having to run fewer sticks and retesting 1 at a time to see if there is a flaky stick. DRAM voltage can be safely set to 1.4 if needed for DDR4.

If 2 sticks seems to have issues, likely problem is elsewhere. 2 DIMMs going bad in same kit would seem highly unlikely. I would leave OC CPU settings off and test RAM only using XMP settings when possible to narrow down the culprit.
 
Ok; bummer. With XMP on, my son had a BSOD. So I ran memtest86 again and on the 2nd pass with 4 dimms got two errors. So, now I'm running it with two dimms to see if I can isolate it to a dimm or to the controller.
 
Thanks for checking in. Basically I was never able to isolate to any one dimm. It only appears with 4 and then not even all the time. And only an error or two.

So...I decided to slowly up the clock speed manually. 3.6-3.7-3.8-3.9 up to 4.0. Each time running a stress test and no errors. And did a modest 20Mhz on the GTX 1080i.

Then last night he got two IRQL BSODs. One after several hours of play. Then other an hour after that.

He plays League of Legends on one screen and then has twitch etc on the other.

Sheesh...

...so today's options:

  1. Part and sell of the PC and just redo with a Z390/i9/RTX 2080 set up.
  2. Buy a Lenovo/Dell gamer that's pre-built, but keep the costs down.
  3. Swap out the memory and board with a Z390/i9.

Jason
 
Sorry to hear the issue is still there. Certain about buying new? Your 1080 Ti is very similar to 2080, though Ray-Tracing and DLSS are slowly picking up steam. I can vouch for Z390/i9 combo being a great improvement vs X99 in added performance and lower power consumption. You do lose quad-channel memory and considerable quantity of PCI-E lanes though.

I can also vouch for Ryzen if that matters to you. I can provide details for both if needed. Feel free to PM me if needed.
 
Time to maybe run the ram at 3000 or loosen your timings. Maybe even just use 16 Gigabytes to help the imc. Run the cpu at stock or just a mild overclock. I would always suggest to do everything in the bios and forget about doing it in Windows.
 
Time to maybe run the ram at 3000 or loosen your timings. Maybe even just use 16 Gigabytes to help the imc. Run the cpu at stock or just a mild overclock. I would always suggest to do everything in the bios and forget about doing it in Windows.

Thanks for the advice.

I might cut down the ram. Want to over clock but only after I can make sure it can make it at least 10 days without a BSOD.