PC shuts down when RAM is at 2400mhz

[hanlinbian]

Hey guys I need your help,

I've had this build for about 5 months now, it has been running quite well until recently.

Specs:
Asus Maximus Hero VII Z97 MB
i7-4790k with Corsair H110 CPU Water Cooler
32Gb Corsair Vengeance DDR3 @2400 mhz
2x GTX980 SLI
EVGA SuperNova G2 1300W PSU

Symptom:
1. The computer shuts down a few minutes after any intense game starts playing.

2. In windows event log viewer, I see 2 instances of "kernel-power" with event ID 41 logged in the past hour, I haven't figured out what this means but I think it might lead to a clue.

3. After research and trouble shooting it myself, it looks like the problem only occurs when the RAM is set to its labelled frequency (2400 mhz as they are supposed to run at), when I set DRAM frequency back to "auto" in the BIOS, the motherboard sets it to 1333 mhz and games run for hours fine.

4. Temps are fine, not even close to overheating.

Strangely enough, I believe the RAM has been set to and running at 2400 mhz for as long as I've had the PC and the problem did not occur until recently as I mentioned.

The only thing I remember doing before the problem started to occur is that I changed the direction of some chassis fans to get a better airflow in the case, for which I had to unplug and re-plug some power connectors, but that really doesn't seem to be the issue because like I said, things are fine when the RAM is set to 1333 mhz. But maybe I'm wrong, so I'm throwing this out just in case.

Please help and much thanks in advance!

 

[commandosupremo]

Since you are able to run it at 1333 but not 2400, I would conclude that it is properly seated, but that is worth checking since you did open the case and move some stuff around. Then make sure that all of the timings are correct for 2400MHz as well as the voltage. If you are just changing the frequency and not the other settings (or not using XMP to change the settings) then that would be my guess. This doesn't explain why it would have changed in the first place but that's how I would diagnose it.

 

[hanlinbian]

Since you are able to run it at 1333 but not 2400, I would conclude that it is properly seated, but that is worth checking since you did open the case and move some stuff around. Then make sure that all of the timings are correct for 2400MHz as well as the voltage. If you are just changing the frequency and not the other settings (or not using XMP to change the settings) then that would be my guess. This doesn't explain why it would have changed in the first place but that's how I would diagnose it.

Hmm I've never messed with RAM timing or voltage before, so frankly I don't really know where to start if that's the case, could you give me some directions? Also, I have tried resetting CMOS and leave everything at default except for RAM clock speed, problem persisted.

 

[commandosupremo]

So without a product number I'm guessing that you have this kit: http://www.newegg.com/Product/Product.aspx?Item=N82E16820233606

So as you can see the numbers 11-13-13-31 are the timings for the kit. If you go to the bios you should see a setting, probably under overclocking stuff where you can enable XMP. This should set the timings automatically for the kit as well as the voltage (which is 1.65V). Else you will have to do it manually. This should be detailed in the manual for your motherboard, but basically you need to find the following settings: CL (cas latency), tRCD (row address to column address delay), tRP (row precharge time) and tRAS (row active time) and set them to 11,13,13 and 31 respectively. They will be under memory timings or DRAM timings or something similar. This video demonstrates how to do this but not on your motherboard or RAM (but it should be very similar): https://www.youtube.com/watch?v=dih6rEeDYcQ.

If you haven't done this before, it's very likely you probably weren't running the RAM at 2400MHz from the beginning. I've never had a motherboard correctly auto detect my RAM. If you are setting everything to auto by resetting cmos, then upping the frequency, you are probably staying at the default voltage and timings which work when the RAM is running at 1333MHz but not at 2400MHz.

 

[hanlinbian]

So without a product number I'm guessing that you have this kit: http://www.newegg.com/Product/Product.aspx?Item=N82E16820233606

So as you can see the numbers 11-13-13-31 are the timings for the kit. If you go to the bios you should see a setting, probably under overclocking stuff where you can enable XMP. This should set the timings automatically for the kit as well as the voltage (which is 1.65V). Else you will have to do it manually. This should be detailed in the manual for your motherboard, but basically you need to find the following settings: CL (cas latency), tRCD (row address to column address delay), tRP (row precharge time) and tRAS (row active time) and set them to 11,13,13 and 31 respectively. They will be under memory timings or DRAM timings or something similar. This video demonstrates how to do this but not on your motherboard or RAM (but it should be very similar): https://www.youtube.com/watch?v=dih6rEeDYcQ.

If you haven't done this before, it's very likely you probably weren't running the RAM at 2400MHz from the beginning. I've never had a motherboard correctly auto detect my RAM. If you are setting everything to auto by resetting cmos, then upping the frequency, you are probably staying at the default voltage and timings which work when the RAM is running at 1333MHz but not at 2400MHz.

So I tried to set the XMP profile in the BIOS, it did automatically set the timing to the correct numbers and voltage labelled on the RAM kit (btw it is the kit you pointed at from newegg). It did not work, problem still existed. I ten tried to set XMP profile to manual and adjust the timing and voltage myself, or crank the voltage up and down a little, ended up with no luck either.

During follow-up investigation I also see the same issue if the CPU clock multiplier is set to 46, even when RAM is running at 1333 mhz, yet I haven't done enough testing to say whether that is part of the problem. I will test it more today.

From time to time, I kind of suspect I got a bad PSU but without an extra one to test with I don't know, and seeing it working fine when I'm not overclocking RAM or CPU, I feel a little hesitant about spending anothe 150 bucks or so on a new PSU.

 

[commandosupremo]

Okay, if increasing the CPU multiplier causes the same issue then I too would suspect either the power supply or maybe the CPU itself, but that is much more unlikely. I would check and see if you could just swap a PSU with a friend real quick and see if it boots or swap the RAM and see if it boots, then you could check it without switching out the PSU.

We had a machine in my lab that had a flaky PSU and it caused similar problems. Sometimes it would just turn off and you had to change settings until it magically came back on or it would fail under high memory and CPU load. Maybe you could get EVGA to check out your power supply, I doubt it's that old.

 

[hanlinbian]

Okay, if increasing the CPU multiplier causes the same issue then I too would suspect either the power supply or maybe the CPU itself, but that is much more unlikely. I would check and see if you could just swap a PSU with a friend real quick and see if it boots or swap the RAM and see if it boots, then you could check it without switching out the PSU.

We had a machine in my lab that had a flaky PSU and it caused similar problems. Sometimes it would just turn off and you had to change settings until it magically came back on or it would fail under high memory and CPU load. Maybe you could get EVGA to check out your power supply, I doubt it's that old.

I have an old corsair 700W PSU lying around, do you think that unit will support my rig? It's non-modular so if I do try it I have to take the whole rig apart 3 times, can be frustrating you know, but now you've suggested the PSU might be the cause, I would probably have to check that out sooner or later. And much thanks to you for helping me continuously.

 

[commandosupremo]

No problem, I would try it only if you don't have other RAM laying around to test first. I wouldn't put it in there too nicely, just plug it in see if it boots with XMP enabled. If you are worried about the power, just remove one of the GPUs.