My computer shuts off completely while gaming

[Mrblack761]

I have the results from prime 95. I did notice my cpu undervolting and underclocking often during the stress test. But I have amd cool and quiet disabled in the bios. Anything else I should turn off?

 

[Aquinus]

Before I reset my overclocks im gunna run prime95 for 1-2 hours and I will come report back with results and screenshots.

It can still pass Prime95 but not be stable under every condition. I say reset to stock because it rules out the overclock 100%. This isn't about holding on to your clocks, it's about figuring out the problem. If the problem still exists after reverting to stock, you can put your overclock back but for the sake of troubleshooting, just set it to stock. I make recommendations for a reason. You can't rule it out until you ditch it just for the sake of testing. If you don't get rid of it, you'll never know if it's a problem.

 

[Mrblack761]

It can still pass Prime95 but not be stable under every condition. I say reset to stock because it rules out the overclock 100%. This isn't about holding on to your clocks, it's about figuring out the problem. If the problem still exists after reverting to stock, you can put your overclock back but for the sake of troubleshooting, just set it to stock. I make recommendations for a reason. You can't rule it out until you ditch it just for the sake of testing. If you don't get rid of it, you'll never know if it's a problem.

Well sir,You were right and I was wrong. The attached pic is temps and clocks after I restored my cpu to base clocks. The pic was taken while I was in the middle of a long race in project cars. I am happy to say I did not face an unexpected shutdown or undervolting and underclocking, and I was able to finish the race. Cpu temps stayed at around ~38c for the entire race. However, if it was my cpu, why did my old psu die?

 

[Aquinus]

However, if it was my cpu, why did my old psu die?

Coincidence. Correlation doesn't always imply causation. AMD CPUs are known for sucking down a lot of power, so are their GPUs.

Consider this, my 3820 and 390 overclocked can suck 550-watts from the wall. AMD CPUs are known to eat a lot of power, even more when overclocked. The 390X eats even more power than may 390 does, so you very well could be pushing your PSU to the limit with your configuration. That MSI board also doesn't have a very good VRM for driving that 8c behemoth so, that could be the source of a lot of losses and issues. Rosewill PSUs are also garbage in comparison to Corsair (in my personal opinion.) I think you simply pushed the machine too hard for the hardware driving it.

 

[Mrblack761]

Coincidence. Correlation doesn't always imply causation. AMD CPUs are known for sucking down a lot of power, so are their GPUs.

Consider this, my 3820 and 390 overclocked can suck 550-watts from the wall. AMD CPUs are known to eat a lot of power, even more when overclocked. The 390X eats even more power than may 390 does, so you very well could be pushing your PSU to the limit with your configuration. That MSI board also doesn't have a very good VRM for driving that 8c behemoth so, that could be the source of a lot of losses and issues. Rosewill PSUs are also garbage in comparison to Corsair (in my personal opinion.) I think you simply pushed the machine too hard for the hardware driving it.

Ten minutes ago I was talking to my professor (who is also comp tia a+ certified) and I explained to him the situation. He said that it was odd how my old psu died after 6 months of use. And how this also happened on my new psu. He said another possibility is the electrical component in my house. more specifically my room :/

 

[tabascosauz]

Ten minutes ago I was talking to my professor (who is also comp tia a+ certified) and I explained to him the situation. He said that it was odd how my old psu died after 6 months of use. And how this also happened on my new psu. He said another possibility is the electrical component in my house. more specifically my room :/

I highly doubt that the PSU is the issue here, Mrblack.

AMD AM3+ CPUs have a habit of throttling not on the basis of the CPU overheating, but to protect the VRM. You'd expect that there are protections in place to protect the VRMs that are feeding a 120W-220W CPU, and there are...for the most part. Good boards throttle a bit, then shut down when conditions are way too serious.

Pcars is such a shitty game for AMD hardware, but it could be that your OC'd 8320 is doing something wacky under the game. Dial back your overclocks. Don't be so reluctant, you won't die from having a bit less freq.

I thought it was the PSU, but the throttling in Prime95 was a red flag. Check your VRMs.

And for everyone who thinks that only shitty motherboards throttle, think again. Shitty 970 blue-colored MSI boards don't throttle, so they blow up. This behaviour happens on good boards, to save their MOSFETs. Even the 990FXA-UD3 and UD5 can do this. I am fearful that despite all the good features that have been made known about the 970 Gaming (modern audio, matte black PCB, apparently decent VRM), it could just be a blue MSI 970/990FX board in a different suit. And those things have some pretty horrible reputations.

 

[bug]

ram generally doesn't cause a hard power reset unless the sticks are on fire ...

Actually, faulty RAM will cause resets, that's the first symptom. But the OP says the system shuts down and that's not something that could (easily) be blamed on the RAM.

 

[Aquinus]

Ten minutes ago I was talking to my professor (who is also comp tia a+ certified) and I explained to him the situation. He said that it was odd how my old psu died after 6 months of use. And how this also happened on my new psu. He said another possibility is the electrical component in my house. more specifically my room :/

I think your professor is full of crap.

 

[yesyesloud]

Gigabyte offers considerably better VRM sections on their 970 chipsets (ud3 for example) but you are best off getting a Gigabyte 990 chipset, those are generally better suited for the FX. Or Asus, but that generally means paying premium for nothing at all except the brand name.

I lost a ga-990fxa-ud3 to an overclocked fx-8350 (4.7GHz). It will generally handle clocks up to 4.5-4.6 stable, so I guess he'll be fine with it anyway (@4.4GHz).

I can push it to any clock on my Asus Crosshair Formula-Z though. Seriously, there's no comparison.

I have the results from prime 95. I did notice my cpu undervolting and underclocking often during the stress test. But I have amd cool and quiet disabled in the bios. Anything else I should turn off?

Answering this particular question, which was also a major one for me when a got my first fx-processor/motherboard: if you want to reduce voltage fluctuation, you mainly have to fiddle with Load Line Control (LLC) settings - current will vary a lot less depending on what you pick.

 

[Vayra86]

I lost a ga-990fxa-ud3 to an overclocked fx-8350 (4.7GHz). It will generally handle clocks up to 4.5-4.6 stable, so I guess he'll be fine with it anyway (@4.4GHz).

I can push it to any clock on my Asus Crosshair Formula-Z though. Seriously, there's no comparison.


Answering this particular question, which was also a major one for me when a got my first fx-processor/motherboard: if you want to reduce voltage fluctuation, you mainly have to fiddle with Load Line Control (LLC) settings - current will vary a lot less depending on what you pick.

Of course, when you compare a board with 1,5X the price difference it will be more stable. But in a similar price bracket Gigabyte generally offers better value.

 

[darkangel0504]

You should check your PSU.

 

[Jborg]

You should check your PSU.

He already swapped to a new CX 750....

 

[Aquinus]

Answering this particular question, which was also a major one for me when a got my first fx-processor/motherboard: if you want to reduce voltage fluctuation, you mainly have to fiddle with Load Line Control (LLC) settings - current will vary a lot less depending on what you pick.

It may not droop but what you can't see is the voltage spikes when load changes which is what LLC is supposed to help reduce exposure to your CPU from. Droop is to ensure that it doesn't overshoot voltage too much. By jacking LLC up, what you get is something like this:

This is the kind of behavior you'll see with LLC functioning normally, apply vdroop as current draw increases.

This is the kind of behavior you'll see with LLC overridden and essentially turned off.

This happens because when CPU load changes from fully loaded to some level of idle, the current draw drops. Since VRMs don't react instantly to the change in demand, there is a brief spike while the VRMs compensate for the reduced current draw. The same is true going from idle to fully loaded because it "bounces" up and down until it levels out. If the OP has an already struggling VRM, turning LLC or reducing its effectiveness could fry the CPU. I highly suggest not doing it in this case.

 

[yesyesloud]

If he picks the right LLC and voltage until he can keep the lowest possible current stable enough for the current clock (4.4GHz may not need much), he's basically following the protocol for AMD FX overclocking.

The default LLC isn't always the best one depending on the board model. Look:

 

[Aquinus]

If he picks the right LLC and voltage until he can keep the lowest possible current stable enough for the current clock (4.4GHz may not need much), he's basically following the protocol for AMD FX overclocking.

The default LLC isn't always the best one depending on the board model. Look:

At least I'm trying to be helpful and help the OP from not destroying his hardware. That image and your cavalier attitude isn't exactly helping your argument. The motherboard has a half-crappy power delivery system and he's using a CPU that can suck down a lot of power. So it's risky to do what you suggest as the overshoot could be a lot larger than it would be on a board with better VRM circuity.

 

[Atomic77]

hmm I don't know what it could be but it sounds like there is bad wiring somewhere.

 

[yesyesloud]

At least I'm trying to be helpful and help the OP from not destroying his hardware. That image and your cavalier attitude isn't exactly helping your argument. The motherboard has a half-crappy power delivery system and he's using a CPU that can suck down a lot of power. So it's risky to do what you suggest as the overshoot could be a lot larger than it would be on a board with better VRM circuity.

shoz.

I agree with you that a high load line calibration may harm the chip(set) (depending on current, as it will use the cpu voltage setting as baseline and will only spike to a certain limit above it), but I diverge in lowering it would hold the same effect.

In my rather long experience with FX chips, the main problem (since PSU was ruled out) is solely VRM, as already pointed out.

I had exactly the same problem with my ga-990fxa-ud3 rev 1.1 but I thought it wasn't the board because of its advertised specs. I imagined I was pushing its limits anyway (LLC at Very High, 1.42V fluctuating to 1.5, cpu @ 4.7GHZ). Never happened on lower LLCs and 4.6GHz though.

He's definitely found out his VRM limits as to cpu current.

So what to do in order to maintain hardware health without underclocking (at least too much)?

Fine CPU temperature and good PSU: checked. I bet his board default LLC is terrible and it's probably "helping" shut itself down by overheating the VRMs with some crazy voltage spikes.

Just try to keep LLC at "regular/normal" or rather lower but don't disable it (try googling fx oc llc off).

 

[Dieinafire]

This has been only happening in one game in particular. (Project Cars) This also happened to me a few days ago. On the prior incident my old power supply completely died. So when this incident happened again on a brand new psu it leaves me baffled.

cpu: AMD FX 8320 @4.4ghz
gpu: MSI R9 390X
ram: HyperX 2x8gb DDR3 1866
mobo: MSI 970 gaming motherboard
old psu: Rosewill 630w psu green series (now dead)
new psu: Corsair CX750M

Warning this pc may catch fire do to extreme temps and must shutdown for your safety. With your specs you must have to have the fire department on stand by. Scary

 

[Mrblack761]

At least I'm trying to be helpful and help the OP from not destroying his hardware. That image and your cavalier attitude isn't exactly helping your argument. The motherboard has a half-crappy power delivery system and he's using a CPU that can suck down a lot of power. So it's risky to do what you suggest as the overshoot could be a lot larger than it would be on a board with better VRM circuity.

shoz.

I agree with you that a high load line calibration may harm the chip(set) (depending on current, as it will use the cpu voltage setting as baseline and will only spike to a certain limit above it), but I diverge in lowering it would hold the same effect.

In my rather long experience with FX chips, the main problem (since PSU was ruled out) is solely VRM, as already pointed out.

I had exactly the same problem with my ga-990fxa-ud3 rev 1.1 but I thought it wasn't the board because of its advertised specs. I imagined I was pushing its limits anyway (LLC at Very High, 1.42V fluctuating to 1.5, cpu @ 4.7GHZ). Never happened on lower LLCs and 4.6GHz though.

He's definitely found out his VRM limits as to cpu current.

So what to do in order to maintain hardware health without underclocking (at least too much)?

Fine CPU temperature and good PSU: checked. I bet his board default LLC is terrible and it's probably "helping" shut itself down by overheating the VRMs with some crazy voltage spikes.

Just try to keep LLC at "regular/normal" or rather lower but don't disable it (try googling fx oc llc off).

At least I'm trying to be helpful and help the OP from not destroying his hardware. That image and your cavalier attitude isn't exactly helping your argument. The motherboard has a half-crappy power delivery system and he's using a CPU that can suck down a lot of power. So it's risky to do what you suggest as the overshoot could be a lot larger than it would be on a board with better VRM circuity.

shoz.

I agree with you that a high load line calibration may harm the chip(set) (depending on current, as it will use the cpu voltage setting as baseline and will only spike to a certain limit above it), but I diverge in lowering it would hold the same effect.

In my rather long experience with FX chips, the main problem (since PSU was ruled out) is solely VRM, as already pointed out.

I had exactly the same problem with my ga-990fxa-ud3 rev 1.1 but I thought it wasn't the board because of its advertised specs. I imagined I was pushing its limits anyway (LLC at Very High, 1.42V fluctuating to 1.5, cpu @ 4.7GHZ). Never happened on lower LLCs and 4.6GHz though.

He's definitely found out his VRM limits as to cpu current.

So what to do in order to maintain hardware health without underclocking (at least too much)?

Fine CPU temperature and good PSU: checked. I bet his board default LLC is terrible and it's probably "helping" shut itself down by overheating the VRMs with some crazy voltage spikes.

Just try to keep LLC at "regular/normal" or rather lower but don't disable it (try googling fx oc llc off).

Yeah after extensive testing, it seems that my vrm's are getting extremely hot (To the point where you don't wanna leave your fingers on it for more than three seconds). I checked my bios and there seems to be no way to adjust LLC on my particular motherboard. I was running fine oc'd at 4.4ghz with my R9 270X. But ever since I got my R9 390X these problems seem to have popped up outta no where. I'm getting a multimeter tomorrow to see how much my computer is drawing from the wall. Other than that I'm just gunna have to leave it at stock clocks until I get a new motherboard. Or put back in my 270X :/

Ironically, I was going through some more testing(Prime95 and Valley benchmark) at stock clocks and it still seemed to shut off. I ran both programs at the same time for 30 mins and got some interesting temps. My CPU would go up to 49C then downclocks to 1.4ghz to get temps down to 29C then it repeated in that pattern. Also a certain temp on my motherboard got up to 99C then goes back down to 94C when the cpu underclocks and undervolts. I'm assuming this is my vrm's. It was then after I turned off prime95 (with no errors or mistakes) that my computer shut off. Thus losing my unsaved screenshots of the test.

 

[Mr.Scott]

Can that board. It cannot handle an FX. Bunch of people already told you. I'm just reiterating that they're correct.
MSI and FX is a bad mix. Make it a cheap MSI, and the mix can force you into a fire extinguisher situation.

 

[ThE_MaD_ShOt]

Your board is the problem, nothing else. Leave it at stock clocks and you will be fine. That board just doesn't have the strength to handle a oc'd fx8 for lengths of time.

 

[Mr.Scott]

Your board is the problem, nothing else. Leave it at stock clocks and you will be fine. That board just doesn't have the strength to handle a oc'd fx8 for lengths of time.

I've seen them throttle at stock also.

 

[Mrblack761]

Can that board. It cannot handle an FX. Bunch of people already told you. I'm just reiterating that they're correct.
MSI and FX is a bad mix. Make it a cheap MSI, and the mix can force you into a fire extinguisher situation.

I've seen them throttle at stock also.

Your board is the problem, nothing else. Leave it at stock clocks and you will be fine. That board just doesn't have the strength to handle a oc'd fx8 for lengths of time.

Well I can't argue, I'll leave it at stock clocks and get an Asus board around black friday. Any recommendations with my specs guys?

 

[yesyesloud]

Your new gpu is more cpu demanding, that's why your system is shutting down on gaming now.

Fine tuning (lowering in this case) LLC, if doable, would only give you some hope in stock settings, it seems.

It's unfortunate that stock clock won't even get past some stress testing without powering off.

A new board is the answer, of course. But bear in mind most games won't stress your cpu to prime95 thresholds (99-100%). Not that it won't be throttled though. If you're only gaming, stock settings may suit your needs, very basically.

As for new motherboard suggestions, any cheap 990fx (gigabyte is great as mentioned) will handle FX cpus @ clocks up to 4.6GHz flawlessly. Don't buy expensive models if you don't intend to clock fx cpus faster than that.

 

[Solaris17]

I still see no core temp. Only package temp and core utilization. does AMD not support this or something?