I would appreciate some thoughts on an issue I have been having. First a system breakdown:
Gigabyte Z68A-D3H-B3 rev 1.0
Patriot G2 Series 1600MHZ 4GBx2
OCZ Agility3 SSD
HIS IceQ 6870HD
Crosair Enthusiast series 650M PSU
Hyper-212 Plus Cooler
I'm no expert when it comes to enthusiast computing, but I have an EE degree so I think I'm somewhat competent with the subject.
The system was custom built from a local PC store (Memory Express). On the weekend I was starting to get BSOD's due to memory pagefault errors. After running memtest86+ I determined that there was a faulty memory stick. So I pulled it out and got my store to replace the memory sticks (bought an in store warranty).
So after I put in my new memory, I ran memtest86 and everything was OK. Then I logged into windows and tried to run a couple of Intel burn test runs. It will fail with no over clock, stock bios fail safe settings. Fail is certain on high stress, and random on standard test. I also get random program crashes. But no BSODs as of yet.
So I went into the BIOS to look at some settings. I noticed my Bclk has a lot of jitter. It will fluctuate by about 0.05Mhz. Is this normal? I also get strange errors in the BIOS, such as trying to set the 4th core to a manual turbo setting will reduce all of the cores back to auto. Flashing my bios (To rev13, and back to rev11) did not fix these issues.
I also noticed that when I changed my memory profile to XMP it would only change the DRAM voltage to the setting value (profile voltage 1.65V, default 1.5V), and leaves the controller Voltage (Vqt I think?) at the default of 1.04V, not 1.25V as specified in the XMP profile.
I did some research on the internet and there is a known issue with the antec 300 case regarding ground loops in the front panel headphone jack (they connect the ground to the usb ground and to the case ground). I have had issues where plugging in headphones will cause my PC to hang, and more scary is when plugging in a usb device touching the metal casing will cause the entire PC to abruptly shutdown. There is also a dry hum audible in my headphones when no sound is playing. Clearly a classic ground loop.
I had overclocked my PC modestly, putting it up to 3.8GHz, no turbo. I also did not adjust the voltages from the default settings since it ran stable under intel burn test and prime95.
What do you guys think? A blown voltage regulator? Bad CPU PLL? Do you think it is plausible an ESD event could damage my hardware like this? If not what would you recommend trying for debugging. Before I take my whole PC back to the store I want to do my due diligence.
It's normal for the BLCK to fluctuate a little like that. You'r PSU might be faulty, try testing with another if you have a spare PSU with sufficient wattage for your system lying around (anything 400w or greater should be enough). If possible, you could also try testing this PSU in another system. Your motherboard could also be faulty, at least some of these problems could be caused by a faulty motherboard.
Well, it seems that no one else is posting, so I'll see if I can help any further. Check your motherboard for any damage and/or discoloration on both sides of the board. If there are any damaged components or discoloration on the board, then you know that it could be causing some problems.
Like I said on the PSU, it can be trouble shooted by testing it on a different desktop, if you have another desktop available (even if it's an older desktop, so long as it uses less power than what the PSU is rated for).
With XMP not working properly, I am suspicious of your motherboard and RAM, more so the motherboard than the RAM, especially if it passes memtest. Try memtest again (try it with both modules at once, and then again with only one module, testing each module once. Sometimes, when you test more than one module, it passes even if one or more modules does not work, so testing with each module by itself is as important as testing with both at once.) for a while and see how that goes. If it shows errors, then the memory might be faulty (although if the PSU and/or motherboard are faulty, they can also cause memory errors).
If there is also a ground loop caused by the case, then replacing the case is paramount. I don't know for sure if the case caused it, but it seems to be a likely suspect in this.
This is kinda difficult to troubleshoot because for all we know, multiple components could be damaged, not just one or even just two, and several components can cause the same problems. I fear that you might need to completely or almost completely replace this system.
Hopefully, someone with more experience will come and be of more help than I am. Good luck.
There's a voltage setting in the BIOS (I think that it's related to the memory controller) that when raised to 1.15v or 1.2v, fixes the problem with Sandy Bridge CPUs dying from 1.65v RAM. Sorry, but I can't remember it's name. The problem with Sandy Bridge CPUs and 1.65v RAM is supposedly caused by the greater than .5 discrepancy between the RAM and this setting and decreasing the voltage gap between this setting and the memory to less than .5v is supposed to solve the problem completely.
Of course, this only matter if the memory is being run with a voltage much higher than 1.5v (1.55v to 1.57v or so is supposedly the official limit), so if the memory is 1.5v, then no problem. Also, running at 1.65v without fixing the above mentioned setting generally does not cause any problems other than decreased CPU longevity, stability problems are usually not caused by this at all. Had this killed the CPU, then the system would probably not work at all, let alone let you into the BIOS.
Where did I say 1.65v ram would kill the cpu or cause instability?
I did not say that you did. You said that using 1.65v memory on Sandy systems is debatable (a completely correct statement), so I expanded on the subject and mentioned that it does not causes stability problems because most of the problems here seem to be related to stability, so I was saying that even if the RAM was at 1.65v, that probably wasn't a cause of most of Protazoid's problems, if at all.
Yeah I know 1.65v memory would not cause his problems, what I said was 1.65v memory running at 1.5v may not run right, especially if he's running the timings and speed that it's designed to run at 1.65v.
For instance if you have ddr3 1600 cas 9 @ 1.65v, you cannot necessarily run it at ddr3 1600 cas 8 @ 1.5v, you would want to drop it to ddr3 1333 cas 9 or maybe even cas 8 @ 1.5v.
Lower CAS numbers takes high voltages because the lower a latency timing is, the lower the latency, meaning the higher the performance... 1333 Cas 8 is slightly faster than 1333 CAS 9. The same is true for any other frequency too.
I see what you meant about the memory. Yes, if it's rated for 1600MHz x-x-x-y timings at 1.65v and OP is trying to run it at those performance specifications, except at 1.5v, then it could cause stability problems, such as some of the problems witnessed here, until it was brought down in performance specs to achieve stability at the lower voltage of 1.5v.
the first step with the ram is check the mb vendor list of tested memory. if not there then go to the ram vendor site and check with the intel z77 chipset that come up a lot of vendors have updated the web page and changed models of there ram sticks. a lot of the mb venor list are outdated as most memory vendors are now using newer memory modules. with ram sicks they have a small jdec on them. most mb with good bios can read and set themself up to the correct voltage and speed of the jdec. cpu-z can read the jdec info for you and see what the speed and voltage needs to be. I would move your pc to another outlet and or check to see if you have an older hole that there a ground wire on your outlet. and that the person who build your pc did not use plastic standoffs or insulators on the mb standoff screws. the main ground of a pc is the power supply and the case. the problem is a lot of times people dont use all the standoffs they need or there a long tinn or a big heat sink that can ground the hot side of the mb to the case. as you said have an issue with the usb port check to see that the mb i/o shield legs are not shorting out to the mb. also take a look that there no dropped screws or pinched wires.
My specific ram is not supported, it is Patriot PGD38G1600ELK.
I am using 1600Mhz, 1.65V and 9-9-9-24-2T memory timings. These are manufacture specifications for 1600Mhz. The RAM is 1.65V rated, and 1.65V is what is listed in the XMP profile. Vtt (memory controller voltage) is listed as 1.25V in the XMP profile, but I have to manually set this. By default my mobo leaves it at 1.04V. I read somewhere that having a Vdram - Vtt voltage > 0.5V is bad. That is why I changed it to match the profile.
And to clarify I was not running my ram at 1.5V 1600MHz, I was running it at 1.65V 1600MHz (as set by XMP).
I will try the visual inspection as mentioned, will try to hunt down the ground loop, and do some stability testings at lower RAM speeds/voltages to see if I can get my PC stable.
I don't have another desktop, maybe I will try what is mentioned above and then try to trouble shoot the PSU. My gut tells me it might be a bad mobo do to the ground loop ESD events.
Will report back and thanks for all the help so far.
So it looks like setting my RAM back to 1.5V, 1.05V Vtt solved the stability issue. The strange occurrences in the bios have also disappeared.
Be warned about XMP. I found it increases my CPU voltage after asking "Do you want to increase CPU performance?", and caused a differential between Vdram and Vtt that likely fried my memory module. I suppose the ground loop is also possible.
I think my hardware is safe. And I will implement a fix I found on the internet regarding ground loop. (They did stupid wiring in the front media panel). I will return the RAM to my computer store and ask them to replace it with 1.5V 1600MHZ ram. Since I purchased the mobo/CPU/RAM as a bundle I will advise them of making sure they select hardware that meshes up better.