unusual crashing

sadkkf wrote: My specs again: X99 Extreme4 BIOS v3.40 Samsung 850 EVO SSD (boot drive) Windows 7 Pro G.SKILL TridentZ 288-pin SDRAM DDR4 3000 Model F4-3000C14Q-32GTZ (4x8GB) MSI Radeon R9 390 Gaming 8GB PCI Express 3 Intel i7-5820K NOTHING is OC'd. I've *never* OC'd anything with this build. The disk read error is somewhat intermittent. Yesterday, for example, Windows gave some errors, blue screened, reboot and gave the error. I ran MemTest86 and found nothing. Reboot it manually, then Windows automatically ran ChkDsk and that found a lot of errors and booted fine right after. I was rendering a 3D scene for work, went to bed, and over night the PC crashed and rebooted again. Right now I'm seeing the disk read error again. The error is intermittent, but getting more frequent. I bought the parts for this late August last year and probably didn't see the first problem for a few months. Now, the PC is unusable. I've checked the power consumption and it's never gone over 300 watts with an 850W power supply. CPU and case temps are well below limits, too. All I can think of is RAM unless there's something not right about this board.

According to G.SKILL, your memory model, the F4-3000C14Q-32GTZ, is shown as compatible only with Intel Skylake 100 series chipset boards, only Z170 chipset boards are listed. Click on the QVL tab on this page, and then the little grey triangle next to each mother board manufacture:

http://www.gskill.com/en/product/f4-3000c14q-32gtz

It's frustrating that with only two platforms (Intel X99 and Intel 100 series chipsets) currently supporting DDR4 memory that compatibility should be a problem. But some of the CPU voltage parameters related to the memory controller (part of the CPU) are different between these two systems. Those are the VCCSA and VCCIO voltages.

Usually, Z170 compatible memory needs more VCCSA and VCCIO voltage than X99 compatible memory. You should have VCCSA and VCCIO voltage offset settings in the UEFI/BIOS. You could try using a positive 0.1V additional offset voltage to stabilize your memory. Why it passed memtest fine is curious.

If your memory is not over clocked, it should be running at a speed of 2133. Is that what your memory speed is currently?

Next, the main problem seems to be the corruption of your OS installation, given the disk read errors you continue to have.

Another fix for a corrupted Windows installation is to run the sfc command. Start a Command Prompt (Admin) window, and type in sfc /scannow

That will take a while to run, just let it go and it will tell you what the results are. It is possible it may find errors that it is unable to fix.

You mentioned your 850W PSU, but did not specify the model.

Right now I can't even boot from the Windows DVD. No matter what option I try (Safe Mode, Last Known Good Configuration...) all of them want to reinstall Windows.

If I choose the option to repair, it tells me the version of the Software Recovery Software is not compatible with the version of Windows installed. It is the same disc I used to install it originally. Does this mean the recovery doesn't recognize the installation after all the patches and service packs?

wardog wrote: Disk read errors, to me anyways, tells me there's something fishy with the current OS install, the drive or that drives power and or data cables. A write error means something altogether different IMHO. That's I'd be looking at the processor or chipset. Have you scanned the drive with the manufacturers test utility? IIRC I've read that the Anniversary upgrade of 10 borked some previous OS utils. And thanks parsec for coming in and mentioning the two voltages of the X99 that need tweaked when using 100 chipset XMPs on an X99. BTW, the Cat has exited stage left. Time for the mouse to play. Look at that thread for updates coming soon. SuperPi at 5.1 but I had to stabilize the IMC a pinch to get there with 32GB. I now have my old Sammy Green sticks at hand. Hoping 5.2 and above after lessening the IMC load.

I've always thought a DRE was from the OS, too. The thing is, I've had this problem with the original HDD installed. That was an older drive from my old system so when I got the error I thought the drive needed replacing. That's when I bought the new SSD.

Now it's happening with the SSD. It could be the OS because I used the same disc to install both. I doubt the drive is at fault.

This is also Win 7 Pro, not 10, so the anniversary update didn't affect me.

To recap what I've done...

Built the original config with an old HDD. Replaced the HDD with an SSD. Replaced the SATA cable. Updated the BIOS.

The only things I haven't changed are the RAM and MOBO. I just ordered new RAM from the official compatibility list and will install that when it arrives in a few days.

When it does arrive, I'm going  to try to find a different install disc, hopefully one that includes some service packs so I don't have to update everything all over again.

Hm...just noticed I did replace a video card during all of this, but don't recall if any of these other issues happened before then or not. Any thoughts on that?

wardog wrote: What make, model, and size is your PSU and how old is it? Applies to/depends on your above answer: PCIE_PWR1. Above the topmost PCIe slot directly in the corner created by the PCIe slot and the 4 mem banks is a female 4-pin molex that will provide power to the bus to stabilize it. BUT, remember, as PSU's age they inherently loose performance and fall out of spec. Out of spec ie: drooping rails, less overhead, generate more heat causing internal rise of resistance leading to overall poor performance, etc etc.

Those count towards you saying a graphics card was changed. That 390(x) req's a stout PSU rated at 750-800W. And we're not talking an inexpensive PSU either. Not for that GC and X99 board/proc. No no. No Sir.


Edited by wardog - 07 Nov 2016 at 1:09pm

sadkkf wrote: Still, I was monitoring the power usage and it *never* went over 300 watts. Could this mean it's just not giving enough power?

Well, depends. How may rails, as the older ones seemed to have 2-4 separate rails. T'was the "thing back yonder.

If a PCIe rail, if there is in fact one on the PSU and not one large rail to tap from, if that rail is weak or isn't simply providing the output your GC needs/requires, then yes.

Overall draw is one thing, but the kicker is is if it's a multi rail and this particular rail feeding the GC that just wasn't up to powering the new hungry 390 in the first place. Then you'll have issues. This particular issue yuu have now, well, can't say.

Yet with this new system, X99 and 390, you want to consider an older PSU being sub-par and look at it from that vantage point. I am, that's why I asked of the PSU.