Serious, sudden Core freeze on CPU & GPU!

[wingclip]

Hi folks, it's important to note that I've been successfully running my system since 5-8-16 so I know the system can run as I describe below and maintain a very strong stability level. Ergo, this issue isn't due to some 'new' application or change I made or recently tried in the past few days.

The problem I have, which is severe, began about two days ago:

It was then that I realized my CPU core speed suddenly dropped to .800MHz and wouldn’t go any higher regardless of what I changed in the UEFI and/or associated software, (Like Windows 7 “Power Settings”).

I found that my GPU core speed had the same problem! It will not go any higher than 500MHz and the monitoring software I have, (GPU-Z, CPU-Z, HW Monitor, just to name a few), indicates the maximum load on the GPU never gets higher than 26%.

The following are the basic specs:
O/S: Windows 7 64-bit

/MSI Z97 MPower MAX AC
/liquid-cooled, i7-4790K [Devils Canyon; default clock is 4.Ghz]
/MSI R7970BE 3GB [Default clocks: 1050MHz Core & 1375MHz Memory]
/G-Skill Raptor 2 x 8 @ 2400MHz (has not been affected)
/Corsair RM1000 PSU
/4 internally mounted SSD's @ 500GB each
/One 1TB Seagate SSHD connected via USB 3.0 port on the board's I/O panel

What I found so far:
I removed & tested the 1000-watt PSU with a multi-meter. The 12V pins would only supply 7.42 volts. (The RM1000 is a single 12V rail PSU). The 3.3V pins would only supply 1.72v and the 5v pins didn’t supply any power at all!

I replaced the RM 1000 PSU with my other desktop’s, 700W OCZ PSU. I opened to the mobo’s UEFI and got every indication all was correct and normal again! The CPU frequencies & voltages displayed, were all correct figures from the default speeds, on through to the 4.8Ghz O/C profile speeds.

NEVERTHELESS, after only a few minutes, (while I was still in the UEFI ‘hardware monitor section), the CPU frequency began to fluctuate. Before I knew it, the CPU core speed dropped to .800 MHz again and stayed there for about 10 or 15 seconds. Then it would suddenly jump back up to the proper frequency for about 5 seconds, and then drop back to 800MHz again.

That cycle repeated itself over and over and I noticed another strange anomaly as well; In the same UEFI’s “CPU” frequency & voltage monitor section, I can also see the CPU temperature. The CPU temperature display would rapidly decrease to as low as 28C, and then back up to 44C, (at 4.7Ghz, 44C is about normal for this O/C profile).
Each time the temperature dropped, so did the CPU’s core frequency and voltage!

ONCE AGAIN, (same thing, different PSU):
It didn’t take long before the CPU frequency no longer went higher than 800 MHz again. The GPU core frequency also maxed-out at 500 MHz; I was back where I started…

Even though the PSU volt tests I did on the RM1000 indicate that the power supply is dead, I still can’t be certain what was to blame for this entire problem. I was sure the RM1000w PSU was the problem but now it may be possible that it was a “victim”.

• Did the PSU fry and then permanently kill the CPU, GPU, and or Mobo? ALTERNATIVELY, did one or more of the hardware components initially cause the PSU to fry?

• If so, which one?

In fact, I still have the OCZ PSU powering the system, which I’m using to type this post with. I don’t know if I’ve harmed that power supply since I’ve connected it to this system! (Not yet, anyway). The RM1000 behaved the same way before I replaced it with this OCZ.

I did test the OCZ 700 for voltage output BEFORE I replaced the RM1000 and the OCZ tests showed that it was putting out the proper voltages beautifully. I’m going to test it again when I finish this post.


MY NEXT MOVE:
My mobo and my GPU both have dual Bios switching capability. My next move is to switch the mobo bios to #2, and then flash the bios to the latest version, (which was the version on the bios switch #1). The GPU has the last known updated bios, (Tahiti), so I’ll use that same version in position #2.

If that doesn’t bring any positive results, (and I seriously doubt it will), I’ll have to switch out my GPU with another graphics card installed in the other desktop.

Before I do that, I’m hoping someone may recognize the symptoms I described and offer some suggestion I haven’t tried yet. I’d rather not try my other graphic card if at all possible because I don’t know what may be causing this problem.

I may wind up killing another Graphics card and still be no closer to the solution.

My greatest fear is that I killed the mobo, CPU and GPU due to a bad PSU! The fact that the problem continues in spite of a different PSU does not mean that the initial problem wasn't the PSU. The 16GB of 2400MHz RAM seems to be unaffected.
Thanks,
Rich

 

[the54thvoid]

If two different PSU's replicate the same fault it could well be a faulty mobo. Look for any visible signs of damage on the mobo components and perhaps even look for any signs of a 'short' maybe a stray screw or the like causing issues. If it isn't the mobo, it could be the mains power is erratic - i've heard that happening before to folk in these forums from a bad domestic power delivery?

 

[Hood]

What about AC power from the wall/power strip/UPS? Low voltage from the mains could cause these problems. Try checking to make sure you have 110-120 VAC at the PSU plug at the times when clock speeds drop. Not sure, but this problem could have killed your RM-1000.

 

[wingclip]

If two different PSU's replicate the same fault it could well be a faulty mobo. Look for any visible signs of damage on the mobo components and perhaps even look for any signs of a 'short' maybe a stray screw or the like causing issues. If it isn't the mobo, it could be the mains power is erratic - i've heard that happening before to folk in these forums from a bad domestic power delivery?

Thank you,
"If two different PSU's replicate the same fault it could well be a faulty mobo." Yes, I'm afraid of that possibility... The mobo is brand new although I mistyped the date I first started using it. In fact, I installed that board and the i7-4790K around the first week of March 2016.

The biggest concern of course, is which one killed whom? The PSU or the Mobo? Or, can it be possible that the CPU and/or GPU are the culprits? I'm leaning toward the PSU because it was the weakest link in the system.

I found out too late, that the Corsair RM series, (and particularly the RM1000), had a very bad reputation. As for the CPU and GPU, they may NOW be behind the issue, (assuming the mobo isn't), as a result of the PSU.

In fact, I suppose the mobo and the two processors could all be victims at this point but the G-Skil RAM isn't showing any symptoms...

I'm not sure if that's a hint that the mobo is OK or maybe the RAM simply doesn't get whacked in the same way as the others.

(The PC mobo is a modern marvel to me, (seriously). It absolutely puts me in a state of 'awe' when I even begin to try and understand how it was conceived, let alone how it operates!)

I'm going to do a full visual inspection of the mobo ASAP! That hadn't occurred to me.
Thanks again and I'll update results if anything changes, good or bad.
Rich

 

[hat]

I have the suspicion that what you're experiencing with the ocz PSU is normal power saving features. You may want to install an application such as Prime95 in order to stress the CPU, and watch CPU-Z for core clock speed. If it doesn't ramp up to 4.7GHz (or whatever you may have it set to) as expected, then you may want to check with Coretemp or Realtemp and see what the temps are and what the cpu load is at.

Try clearing your CMOS. Use the jumper or pull the battery, don't just reset settings in the CMOS setup. If that still doesn't help, have a look at your windows settings again. Try seeing how it behaves in a Linux livecd. If none of this helps, it's possible you got a bum board now, possibly more. There's no way to know for sure, until you check each one by one. PSU connects to everything, so everything is at risk.

 

[wingclip]

What about AC power from the wall/power strip/UPS? Low voltage from the mains could cause these problems. Try checking to make sure you have 110-120 VAC at the PSU plug at the times when clock speeds drop. Not sure, but this problem could have killed your RM-1000.

Thank you,
Yes I run a Cyber-power 1500VA which can handle a 950W draw. I checked the APU's software report and it is maintaining a steady 118v/119v AC supply. It doesn't show any power fluctuations in the last two weeks and I also ran the test procedure on it to be sure it is running correctly. Unfortunately, the fluctuations of the CPU no longer occur, as I had mentioned, the processors simply stay at 800MHz and 500Mhz (CPU & GPU respectively).

It just so happens that I bought that APU when I bought and installed the mobo and CPU. Prior to that one I had a Cyber-power 1000VA which now handles my other desktop and peripherals.

I'm hoping that if I killed anything, (and I'm pretty sure I did), it would have been the RM1000. It's the least expensive to replace and it's under warranty. However, I don't know how well Corsair will honor that end of it. If they do fix it, I'm selling it immediately after I get it back.

I'm buying another OCZ PSU to replace it and of course it'll be about the same power rating as the Corsair.
Thanks again,
Rich

 

[wingclip]

I have the suspicion that what you're experiencing with the ocz PSU is normal power saving features. You may want to install an application such as Prime95 in order to stress the CPU, and watch CPU-Z for core clock speed. If it doesn't ramp up to 4.7GHz (or whatever you may have it set to) as expected, then you may want to check with Coretemp or Realtemp and see what the temps are and what the cpu load is at.

Try clearing your CMOS. Use the jumper or pull the battery, don't just reset settings in the CMOS setup. If that still doesn't help, have a look at your windows settings again. Try seeing how it behaves in a Linux livecd. If none of this helps, it's possible you got a bum board now, possibly more. There's no way to know for sure, until you check each one by one. PSU connects to everything, so everything is at risk.

Thanks hat,
I'm going to try some of those suggestions for sure. However, I over-rode all the OCZ and Windows power-saver facilities. I did that in the UEFI and then in the O/S settings when I booted to the desktop. The two processors should have stayed at their maximum clock speeds with or without a load.

Even so, I still tried to push them by using a high processor demanding program. Google Earth, which I mentioned in the initial post, is set to all the highest quality levels and all the attributes like; state & county borders, state and public park labels, etc., always draw my processors to their max frequencies.

Especially when I zoom up to barely 1000 feet above ground level and then give the planet a hard spin. But neither processor went beyond the 800 MHz, (CPU), and 500 MHz (core of the GPU). In fact, the GPU-Z monitor reported that the highest load on that card during the 'spin' was 26%!!

LOL, I've done tests like that before, (for different reasons), and it always pulled a 100% load on both processors...

I'm going to give that Prime 95 a try after I do a few other things like you had suggested. I forgot that I didn't jump and reset the CMOS since I installed that OCZ. Also, I am about to switch the mobo to the #2 bios position and then flash the update. I'll do the same for the card.

I'll update as soon as I can.
Thanks again,
Rich

 

[hat]

If GPU load was only 26% that probably means that Google earth wasn't even taxing your GPU. Try a more intensive test. You say you have a 7970 right? You must be a gamer then? Fire up a GPU intensive game and check what happens.

It's possible something bad could have happened in Windows, too. As a last resort before replacing hardware, try reinstalling Windows.

 

[wingclip]

If GPU load was only 26% that probably means that Google earth wasn't even taxing your GPU. Try a more intensive test. You say you have a 7970 right? You must be a gamer then? Fire up a GPU intensive game and check what happens.

It's possible something bad could have happened in Windows, too. As a last resort before replacing hardware, try reinstalling Windows.

"If GPU load was only 26% that probably means that Google earth wasn't even taxing your GPU"
I must thank you for chasing me on that G Earth test. As of yet, I don't know why I didn't pull a load and/or hit a max core frequency when I did that because it always was able to hit the max core clock fequency I had set it to in the past with that quick test.

However, as I was writing back to you to assure you that the G-Earth test would've been enough to bring my "max Core clock speed" to the default setting, I remembered that GPU-Z has a "PCIE-Express Render Test" option that I know would pull something out if the card was any good.

Incredibly, (and for the first time in two days), the GPU hit the default core frequency, (1050 MHz which I had it set for at this time), and pulled a 50% max load!!!

That test isn't as aggressive as one like Prime 95 but it didn't have to be in order to prove to me that the card is not dead!!

So thank you very much for the "push"! It made me look at that consideration again. I wouldn't have tried it that way and I was nearly convinced I had lost that card.

OK, so I just finished eating lunch and now I am seriously considering a software issue POV. I'll try not to go into the long detailed story version but; I will say that this entire problem happened 2 days ago when I installed a "Alpha-Test" software being developed by the company I write user manuals for.

It so happens that one of the beta testers posted an anomalie involving his Graphic card's behavior after the installation of that same software. I think he said his card fans cranked up as if the processor was under a load. Yet his monitoring panel never indicated it was even trying to do anything 'hard'. He then said his card began to do some "strange things".

He posted that in mid June and when I read it two days ago, I emailed him for more details but he never replied. The problem is that the same day and week, about 20 other Beta testers installed that Alpha program, (I'm not at liberty to say what it was), and no one else experienced any problems.

In fact, everyone posted how great the Alpha release looked and behaved. The person who posted the problem, was never answered and he never did a follow-up post. It's very possible that he wound up killing a lot of the days that followed, trying to do what I now am trying to do!

I've worked with that Beta tester on other, similar projects and he usually answers anything I ask about within a few hours. I know that it could have been a coincidence when it happened to him and I've no better proof of any connection when it happened to me.

Nevertheless, I removed the installed Alpha program the same day and within an hour of installing it because I was afraid it may have had something to do with my new problem.

Until you mentioned it, I all but forgot about the possibility of it being a software issue. Frankly, I'm not entirely convinced it is because I just can't see how the software could have damaged the PSU. Then again, I'm the farthest thing from a computer expert that you can get. I know how to build them and have been for 7 years but that's it.

I'm still going to switch to "Mobo Bios #2" and may even do the same for the GPU but I have restore points set going back as far as 4 weeks and I'm likely to try that before I consider a O/S re-instillation.

The O/S is on a "semi-dedicated" SSD but the process would still place me in for a lot of work in many different ways. So I really rather not go through an O/S re-installation unless it's my last resort.

Thanks again,
I'm shutting down the computer now so I can inspect the mobo and then do a bios switch procedure.

I'll update as soon as I can.
Rich

 

[phanbuey]

Software can't really damage components but it can make you think that you have unstable hardware by locking up your machine and insta-shutting it off.

I had an issue similar to this some time ago, and it turned out to be a Chrome plug in... would hard lock the computer if I played a game and had chrome open in the background... I swore i thought the videocard was dying.

Can you take a look at the event viewer?

 

[phanbuey]

go to start>administrative tools> event viewier > windows logs > System... that should tell you if there is something heinous going on.

 

[phanbuey]

if you don't find anything, try to turn off C0 suspend state in the Bios under processor managemnet - turn off all of the C-states and speedstep, and see if the CPU still drops.

 

[hat]

It's possible you have boogered up software as well as a bad PSU. They both can have similar effects, but that doesn't mean those separate issues have to be connected.

You've got to stress your system more to see how it performs under stress. You wouldn't use a semi to haul a wheelbarrow would you? Give that rig something heavy to haul (p95, GPU intensive game) and see if it really ramps up.

 

[wingclip]

if you don't find anything, try to turn off C0 suspend state in the Bios under processor managemnet - turn off all of the C-states and speedstep, and see if the CPU still drops.

 

[wingclip]

@hat & Phanbuey:
Thanks for the suggestions and info! (some things such as the CO and C-states as well as speedstep have been disabled from the beginning, but thanks!)

I’m eating lunch and I have a conference with the boss just after I’m done so I’ll check the event monitor ASAP.

However, should I look for the anomaly on and around the day I first noted this problem, (early July 4th)?

The fact that I can’t get more than 7.08 volts to the system could mean that this anomaly would be evident every day since, huh?

I don’t run Google Chrome when I “fly”, (I’m a 7+ year, dedicated flight-sim hobbyist, if I hadn’t already mentioned that somewhere). However, the add-ons I use absolutely apply information received in real-time online.

UPDATE so far:
Sorry I took so long getting back. In the last 24 hours, I was able to confirm that when the Corsair RM1000v is disconnected from the computer system and bench tested with a DVMM: It should not give a 7.07-volt reading from any of the connection ports, let alone the 12v pins. (The 3.3v pins gave me 1.74v as well).

That was confirmed via by a Corsair Tech-rep via telephone yesterday. The PSU has been RMA’d

Unfortunately, the damage had been done and is even worse than I thought. Late yesterday, I found that one of my four internally mounted, 500GB each, SSD’s was simply dead…

It’s a 7-month-old Crucial CT500MX100 drive and doesn’t show up on any of the computers I’ve tried. Regardless of a direct system USB connection, or SSD docking station with and without 12v power supplied, no show.

A call to the Crucial tech rep got me someone who claims Crucial incorporates a “security chip” in their SSD’s, which will cause the drive to shut down just as I had described. He claims that when the drive’s security-related parameters are met and triggered; the drive will automatically shut down and become invisible to the O/S, regardless to the connection method.

He said others have had the same symptoms and he’s never had anyone fail to ‘resuscitate’ the drive using the “special instructions” he sent me via email.

I truly doubt this and even more, the “magical fix” that he insists I try. He claims the “instructions” he sent should be executed completely, at least 4 or 5 times, assuming that any attempt before that didn’t bring back the drive.

I received the instructions a short time ago and though I’m very doubtful, I have nothing to lose in trying them.

This is bad because the drive was a dedicated program source and storage point for all my work-related data, (I work from home). It just so happen that I scheduled a backup session for that drive (due to run Monday of next week)… (of course).


Also, I must agree that your instruction to stress the processors is something I need to try, for sure. Doing that short PCIE Render stress test yesterday definitely proved to me that all may not be entirely lost. I’m going to download and install P95 for sure.

As of this moment, I know that my 700w OCK is putting out the proper voltages without fail. Unfortunately, the HW Monitor CPU-Z, (to name 2), clearly show that the CPU will not go any faster than .800MHz and the highest the CPU core voltage will go is .726v, (even though I tried core voltages as high as 1.400v).

Interestingly enough HW Monitor indicates what present Main Voltage trio, (3.3, 5, & 12.), are doing. No matter what I change, the 12v line will not go above 7.082v. The other two, (3.3 & 5v), are reporting normally, no problems.

However, the “IA” voltage does seem to always report the voltage I set it to via the UEFI, (plus any I/OA Offset if I included one).

The MSI Z97 MPower MAX AC has a DVMM test port on the board. I did use this port when testing the voltages and found the same thing, CPU Core max voltage was .724/.726.

A mobo Bios switch to position 2 didn’t help and I’m going to run a system restore to a point two days before this happened.

The reason it takes me so long to execute each of the options and troubleshooting steps I want to try, is the additional work and the day-to-day stuff and errands I must keep up with at the same time. Then there’s the filling out of RMA info forms & PO searches, follow up calls… (sigh, I really wish this didn’t happen)
Thanks to you all,
I am definitely going to pursue these suggestions and I hope to get back with some good news.
Rich

 

[Frick]

You cannot trust software with voltage readings. So it was good you checked out the PSU with a DMM, but you should have left it plugged in. If the 5V line was at 0 the SSD/HDDs wouldn't work, and I'm quite sure the system wouldn't start with the 12V at ~7V.


Have you seen anything in the event log?

 

[basco]

i think you have software problems or some settings you miss.
or like your graphiccard working again with setting it under load properly
you are focusing too hard on voltages . like Frick said it should not start with 7volt on 12v lane.
and your readouts are under load ?
because your prozessor voltage for idle is fine and your using this machine since 2-3 days.
sometimes its easier + quicker to take a old hdd or small ssd and install fresh(takes max 20 min) to see if some software or other is interfering with your system

 

[eidairaman1]

Reset that UEFI to default, bear in mind todays motherboards have power saving functions for clock speed, so it will only ramp up when it is needed. My rig I turned off all the power saving functions because I prefer my cpu to run at 5.1GHz at all times.

 

[wingclip]

Hi again, Folks!
First, I want to thank all of you for helping me out on this problem. I work from home and the work I do requires a CPU core speed to run at a minimum of 4.7Ghz, (usually, more than that).

I had a “breakthrough”, (I don’t know how big just yet), thanks to all of your suggestions and I apologize because I don’t have time right now to answer each of you. I will say that it’s only because of all your input that I hadn’t spent money I don’t have, buying the entire contents of my case!

The “breakthrough”:
Apparently, I managed to make the CPU core voltage draw the maximum I had set it for in the UEFI in this particular test session and that is the first time it happened in 5 DAYS! It happened while I was cropping a screenshot of the HW Monitor panel in Photoshop. While the PS program was running, I also had the Google browser, MS Office Word, and Dragon 13 Naturally Speaking open as well.

In fact, if I did it correctly, I have the screenshot of the full HW Monitor panel and a cropped shot of the monitor's voltages attached. Take a look.

Now this is not normal by any stretch, (because the processor was setup via UEFI & Win 7 Power Control, to stay at max with or without stress), but it definitely made me more open to the possibility of a certain kind of software issue.

One that I hadn’t considered… DRIVERS! IF the PSU started this entire problem, could it be possible that the sudden voltage change that resulted in a dead PSU, happened at the same moment I executed the O/C profile to begin?

On July 4th, it was right after I switched to the 4.8GHz O/C profile that I became aware of the lack of processor speed and response. That’s what started this menagerie.

Of course it would be a “one in a million chance” coincidence, but I’m 60 years old and seriously; may I get struck by lightning, hit by a truck, and fall over cold-n-dead, if I’m lying when I say I’ve seen these “one in a million” chances happen far more often than “1 in a million” times”!

I must confess that I did not run a dedicated stress-test program on the CPU. I felt if the CPU frequency increased during the test, it would only prove that the voltage and frequencies are not behaving normally.

I mean for 4 months, I ran the CPU O/C profiles I created that took the GHz & Volts up anywhere from 4.7Ghz to 5.0GHz, depending on which I selected at the time.

In addition, profiles that were set to 4.8GHz/volts or higher, was always teamed with UEFI and Windows Power Control settings in order to maintain their maximums.

I was very used to seeing my monitors report speeds and voltages at maximums for both processors even when idle and no stress applied.

If you can see my point, you may understand I didn’t feel I should use up time just to prove the CPU could maybe reach set voltages when under extreme, and by my POV, atypical, stress applications. It wouldn't explain why I had to do that in the first place and that’s what I needed to know.

ALL THAT SAID, (yeah, I know), when I saw the CPU actually reach the set voltage while running the way I most often and normally do, well that was a different story in my mind. This indicates to me that it doesn’t require an immense and unusual amount of stress to get the rated volts and speeds.

I have to go and between the time I have left tonight and when I resume tomorrow, this is the plan:

1. I’m going to uninstall all MSI and Intel utilities that I have, (even though most are turned off).
2. Round up all the latest MSI and Intel Drivers to one folder
3. Remove, turn off and/or disable all the MSI and Intel Drivers that I can, including the chipset, (I don’t know if that’s possible, but…)
4. Shutdown
5. Switch the Bios switch over to the 2nd position
6. Clear the CMOS
7. Startup to the UEFI, (I’ll be sure I’m in as near to the default state that I can be for a boot to the O/S short of drive settings such as switching to AHCI and PCIE external Graphics, etc.)
8. Install via “M-Flash” the latest Bios version, (actually, I’m not so sure of that. I may be better off updating to the version just before the latest one in order to force changes in anything that may otherwise want to stay the way they are.
9. Boot to O/S and install Chipset and whatever other drivers I can and/or need.
10. Reboot to the O/S
11. Then after all the drivers are confirmed and set, cross my fingers and Reboot to the UEFI Post and see what I have

Thanks to all again and I will post the results as soon as I can. Thanks for your patience because I can only get to the thread once or twice per day while I’m trying to figure this out. I will check this thread in the morning first thing in case I may have overlooked something that someone noticed.

Rich

This is a quick rundown of some, (definitely not all), of the things I’ve tried before I got that voltage reading:

Things like:
• Resetting the UEFI & CMOS (at least 4 times, minimum)
• Switching to another BIOS and flashing it and trying all over again
• Bringing it down to one stick of RAM and trying it in all 4 slots
• Starting the system and testing the power while all the fans and peripherals were unplugged with the exception of the cooling pump for the CPU
• Unplugging and removing the power to all drives with the exception of the operating system drive.
• And because my MSI motherboard has the capability of controlling and monitoring the voltages directly from the motherboard workbench testing ports, I was able to remove the GPU and try to increase the core voltage of the CPU while monitoring it with the DVMM.

Please keep in mind that those aren’t the only things I’ve done but perhaps they were some of the “key” troubleshooting moments.