Cold boot issue, setup help with F3-12800CL7T2-12GBPI

[GeddyT]

System is in the sig.

This is my second attempt with this kit, as I just got my RMA replacements. The first kit exhibited the exact symptoms of the cold boot issue that's all over this forum (unbeknown to me at the time). On cold boot only 10GB of 12GB would be reported by Windows. Interestingly, though, Memtest 86+ in every case detected all 12GB and tested fine over multiple cycles. Problem was only in Windows. If the computer was then rebooted and timings changed in BIOS (sped up, slowed down, didn't matter), system would boot up and Windows would see all 12GB. I had no idea of the hot/cold nature of the problem until I read it in this forum, then started to test for it and found it to be the case. Until then, I just assumed it wasn't detecting the memory at all. Instead, restarting and changing BIOS settings would get things detected. (I've read other reports that restarting several times in succession would achieve the same result, but I always had to change settings to get it to detect.)

Tried using X.M.P. profile, and problem persisted. Tried using manual settings found elsewhere on this forum, and problem persisted. Tried using X.M.P. profile, but de-tuned to 1333mhz--no luck. In all cases Memtest would detect no problems, Windows would drop 2GB. Sent this kit in for RMA.

New kit came yesterday afternoon. Only had a short while to try it out, but this kit seems to be even WORSE than the last! Loaded X.M.P. profile in BIOS and it called for 1.9V VDIMM! Last kit's X.M.P. profile called for only 1.5V, so this didn't seem even close to right, and made me nervous considering I'm only air cooled. Nevertheless, I tried it; all settings to auto, X.M.P. profile enabled. System would not post. Restarted the machine, this time the system made it through post, but cut out at some point with a message that "overclock was unsuccessful," or something like that, and that I needed to load BIOS and change settings.

Entered BIOS and changed some more settings. Lowered VDIMM to 1.6V, raised QPI/DRAM to 1.35V manually. System made it through post, Windows loaded. Unfortunately, every time I tried to access any program, gibberish errors popped up. Clearly something was still amiss. Restarted the system, this time Windows worked fine. Except Windows was reporting only 8GB of RAM: I'm going backwards here...

This is as far as I've been able to get before having to go to work. Since obviously I'm going to have to try manual settings with this kit to come even close to getting it to work, can someone please suggest a list of settings that might make this kit behave in this system? I would appreciate it.

On a side note, is this just standard nowadays? I used to be a hobbyist system builder that built about a rig per month. By the end of that time (about four years ago), hardware, drivers, and Windows had gotten to the point that I could throw just about any hardware into a box, install Windows, and things were going to work right out of the box 99% of the time. No need to get too busy with BIOS settings, no need to hunt down drivers and flash BIOS to the latest version; things just went together and worked. Is the industry regressing, or am I just unlucky with this build?

And kudos to G.Skill for their excellent RMA policy and fast service. Apparently since I was last building a lot of systems Newegg decided to start shafting their customers and take the opposite approach, which is too bad...

[GSKILL TECH]

Sorry to hear you are having extensive problems. Check out this link for possible causes: http://www.gskill.us/forum/showthread.php?t=5136

Make sure you have the latest BIOS for the motherboard. Should not have such a problem with this combination. Keep me posted on the results and I will be able to further assist you.

Thank you
GSKILL TECH

[GeddyT]

My motherboard is a fairly new model, so there are only two BIOS revisions. In fact, there is only one official BIOS, the v0108. From what I've read, they briefly released a v0204 BIOS, but then pulled it shortly afterward and it's still not available on ASUS' site. For everything I did with the first kit I had, I was using the v0108 BIOS. I have tried both with the new kit and am currently on the v0204.

Today was, once again, frustrating:

1.) Reset BIOS to default settings.
2.) Reinstall Windows. There is now a clean, stable, Windows install.
*Note: at this point, still showing only 8GB, even at 1066mhz and relaxed timings!
3.) Install and run CPUZ. Shows all 6 DIMMs, all 12GB.
4.) Run Memtest86+ off of bootable disc through one complete cycle. Zero errors, all 12GB detected.
5.) Load memory X.M.P. profile. Don't like that 1.9V DRAM voltage, so lower to 1.6. Raise QPI/DRAM to 1.35V. Raise CPU Voltage to 1.2V. Set timings to manual. Manually input every timing setting that I've read into on these various forums (believe me, I've been over every word of that link you suggested and a dozen others like it! ).
6.) Save and reboot. CPUZ sees all 12GB. Windows shows 8GB.
7.) Power down, remove memory from 1,3,5 slots. Power back up. CPUZ shows 6GB. Windows shows 4GB.
8.) Power down and start shuffling DIMMs. First two I swapped out didn't improve anything from 4GB. Third one I swapped brought it up to the correct 6GB.
9.) Powered down and started adding sticks back in one at a time, checking in-between, making sure to avoid reinserting sticks into slots that didn't work before. Each time I added one, memory came up 2GB as it should until all 12GB were recognized by Windows.
10.) Back to BIOS. Raised DRAM Frequency from 1066 to 1600. Rebooted to Windows.
12.) Windows sees all 12GB. CPUZ sees all 12GB at 1600mhz.
13.) Do a happy dance, go to bed.

Six hours later I got out of bed and fired the computer up to check. As I feared, back down to 8GB.

Turns out that nothing I had done had mattered. The reality of it was that in the process of changing settings and shuffling memory sticks around in the sockets the computer restarted enough times to start hot and detect memory. To test this theory, I started restarting the computer repeatedly, checking Windows for memory amount after every restart. Five restarts later all 12GB were there on display.

To further test things, I continued to restart the computer. Two more restarts and all 12GB stuck around. Powered down and walked the dog for an hour and came back. Powered up and still all 12GB. I can guarantee one thing, though: when I get home from work tomorrow morning and power the computer up Windows will only recognize 8GB of memory...

Here is where my BIOS settings ended up:

Ai Overclock Tuner Manual
CPU Ratio Setting 21.0
SpeedStep Enabled
Xtreme Phase Full Power Mode Auto
BCLK Frequency 133
PCIE Frequency 100
DRAM Frequency DDR3-1600MHZ
UCLK Frequency 3200MHZ
QPI Link Data Rate 4800MT/s

CPU Voltage Control Manual
CPU Voltage Auto
CPU PLL Voltage Auto
QPI/DRAM Core Voltage 1.35000
IOH Voltage Auto
IOH PCIE Voltage Auto
ICH Voltage Auto
ICH PCIE Voltage Auto
DRAM Bus Voltage 1.60
DRAM DATA REF Voltage on CHA Auto
DRAM CTRL REF Voltage on CHA Auto
DRAM DATA REF Voltage on CHB Auto
DRAM CTRL REF Voltage on CHB Auto
DRAM DATA REF Voltage on CHC Auto
DRAM CTRL REF Voltage on CHC Auto

Load-Line Calibration Auto
CPU Differential Amplitude Auto
CPU Clock Skew Auto
CPU Spread Spectrum Disabled
IOH Clock Skew Auto
PCIE Spread Spectrum Disabled

DRAM Timing Control
1st Information : 7-8-7-24-5-88-12-8-24-0
DRAM CAS# Latency: 7 DRAM Clock
DRAM RAS# to CAS# Delay 8 DRAM Clock
DRAM RAS# PRE Time 7 DRAM Clock
DRAM RAS# ACT Time 24 DRAM Clock
DRAM RAS# to RAS# Delay Auto
DRAM REF Cycle Time 88 DRAM Clock
DRAM WRITE Recovery Time Auto
DRAM READ to PRE Time Auto
DRAM FOUR ACT WIN Time Auto
DRAM Back-To-Back CAS# Delay Auto
2nd Information : 2N-57-58-58
DRAM Timing Mode 2N
DRAM Round Trip Latency on CHA Auto
DRAM Round Trip Latency on CHB Auto
DRAM Round Trip Latency on CHC Auto
3rd Information : 8-8-18-8-8-8-7-6-4-7-7-4
DRAM WRITE to READ Delay(DD) Auto
DRAM WRITE to READ Delay(DR) Auto
DRAM WRITE to READ Delay(SR) Auto
DRAM READ to WRITE Delay(DD) Auto
DRAM READ to WRITE Delay(DR) Auto
DRAM READ to WRITE Delay(SR) Auto
DRAM READ to READ Delay(DD) Auto
DRAM READ to READ Delay(DR) Auto
DRAM READ to READ Delay(SR) Auto
DRAM WRITE to WRITE Delay(DD) Auto
DRAM WRITE to WRITE Delay(DR) Auto
DRAM WRITE to WRITE Delay(SR) Auto

CPU Ratio Setting 21.0
C1E Support Enabled
Hardware Prefetcher Enabled
Adjacent Cache Line Prefetch Enabled
Intel Virtualization Tech Enabled
CPU TM Function Enabled
Execute Disable Bit Enabled
Intel HT Technology Enabled
Active Processor Cores All
A20M Disabled


Also, I've tried thinking about whether or not it may be my other hardware that is causing the problem. I know some others have had problems with bent pins, shorted CPU coolers, etc. Whether or not my CPU or motherboard are damaged is impossible for me to know, as I don't have a warehouse full of parts to test them out with. I did remove the CPU and inspect the pins and couldn't see any that were bent or damaged. Visually, I see no problem with the CPU, either. Although I suppose it's possible that my CPU cooler is shorting out and skewing the results for this memory kit that I'm currently trying, when I was testing the last kit I was using the stock cooler, which only attaches via plastic to the board. Since the problem has persisted with two different coolers, I'm ruling that out for now. Tomorrow I will be going to my brother's place to borrow his OCZ RAM and testing it in my rig. He says it's been running stable right out of the box in his from day one, so this may help me pinpoint a hardware issue other than the memory kit.

Also, sadly, I'm exactly one day past RMA period for both the board and chip, anyway, so I'm stuck with them.

So are there any of those BIOS settings that I could change to help the situation out?

And why does this kit's X.M.P. profile call for 1.9volts!? From everything I've read it's incredibly unusual to have to go past 1.66V even for the 2000mhz kits.

[GSKILL TECH]

Very odd, it should not. I would suggest manually setting everything. QPI/DRAM Voltage may need more than 1.35V, so that is somethng you will want to test for the lowest stable voltage.

Thank you
GSKILL TECH

[GeddyT]

Newegg customer service was pretty cool when I contacted them. I've got an RMA refund approved for board and CPU, and I will try this before I bang my head against a wall with BIOS settings anymore.

If one must go above 1.35V on QPI/DRAM to get memory to be recognized, this says to me that something's wrong with the hardware, not just that settings are off. To me, if memory doesn't work until you've fiddled with the BIOS for four hours, then basically that memory doesn't work. So I'll rule out motherboard and CPU issues before jumping to conclusions, but if problems persist with new hardware, I think I may just go another way when it comes to memory.

So... now that I'm fortunate enough to know AHEAD of time about this widespread memory dropping issue on X58 systems before buying my replacement motherboard, what is the X58 board that people are having the least amount of problems with? Are they all the same and it's just a lottery? Or is there one particular brand/board that is seemingly bulletproof?

[GSKILL TECH]

It is not necessarily a problem. Operating performance memory always requires manual settings and voltage adjustments. What is occuring is the memory controller, embedded on the CPU, is not receiving the proper voltage amount to operate the target frequency. The only "better" motherboards are those that automatically supply more voltage, such as motherboards specifically made for overclocking. The ability to automatically over volt when necessary, makes it much easier for the beginning user unfamiliar with voltage changes.

Thank you
GSKILL TECH

[GeddyT]

Wow, I have to say this is the first time I've been disappointed with the support side of G.Skill. Up to this point G.Skill has been very fast with responding to me and handling RMA, but it seems the service I'm getting is degrading. I'm wondering if you even read my whole post now (sorry, I realize it's long, but I figured in a situation like this details are important to eliminate wasting anybody's time troubleshooting).

You write: "Operating performance memory always requires manual settings and voltage adjustments."
I beg to differ. One of the things that convinced me to buy this memory kit is a reviewer on Newegg that said the memory worked right out of the box with the X.M.P. profile. Obviously making memory work doesn't "always" require hours of BIOS tweaking, according to this happy customer. Regardless, were this statement true it would be very dishonest of G.Skill. To me this is the equivalent of saying, "If you want to buy a performance sports car, you have to manually rotate the tires yourself to find the right combination so that it will drive straight." That's garbage. You advertise a spec. NOWHERE do you advertise that in order to get it to run at this spec you may have to tap your head while rubbing your belly and doing a rain dance while praying to the gods--all the while risking the rest of your hardware in the process of trial and error. Were I to know this is the case, I would have spent my money elsewhere.

But this is a moot point. When you write write, "What is occurring is the memory controller, embedded on the CPU, is not receiving the proper voltage amount to operate the target frequency," you're clearly missing the fact that this memory will not function properly at ANY frequency. Even at BIOS defaults of 1066 and extremely relaxed timing, BIOS does not detect all memory. Even after holding the handy "MemOK" button on my motherboard to load failsafe timings, BIOS does not detect all 12GB of memory. Forget hitting the "target frequency," this stuff can't even hit the frequency that the cheap memory easily does.

So, you say "It is not necessarily a problem." Let's clarify a bit:
Here is the product page at the store I purchased from. You'll notice the only thing mentioned in the specs about voltage is that the memory is specced to run at 1.5V. This is confirmed on the similar spec page on your website. To someone like me, who believes that listed products specs should be trusted, this tells me that if the memory doesn't work at the specified timings at the specified voltage it IS "necessarily a problem." I bought 1.5V 1600mhz memory, not a month-long hassle.
And let me ask you if this would be a "problem": Were I to load the X.M.P. profile and either not know better or not check the numbers, then boot up the system... when my CPU is fried by the 1.9V that the X.M.P. profile specifies, is this a "problem"? And will G.Skill replace it?
JEDEC standard for Core i7 is a max of 1.65V on VDIMM. I've tested up to that, along with up to 1.45V on the QPI/DRAM (which is also in the red). I notice that your responses are never specific, and basically sound like, "Keep upping the voltage!" It's as if they're intentionally vague in order to avoid responsibility.
Well, is G.Skill prepared to put its money where its mouth is? I would like you to specify here EXACTLY what voltages you'd like me to try up to. I want to see that in writing. I also want to know that G.Skill will reimburse me for any damaged hardware that results from testing memory up to these voltages that exceed Intel spec.

Or, alternately, just honor my RMA request--SECOND RMA request--for this kit so I can be done with it. The request is in the RMA department's inbox at this very moment.

I'm a pretty patient guy, and feel like I've given G.Skill a huge benefit of the doubt so far in this ordeal (a five week old $1700 computer that I have yet to be able to use), I would hope for reciprocity on G.Skill's part as opposed to these kind of cop-out, "not our fault" responses.

[GSKILL TECH]

Let me clear some things up as I think you misunderstood me.

Setting this up shouldn't take hours, agreed, but sometimes things happen such as bad hardware/settings/incompatibility that can cause these types of problems.

Obviously, there may be something wrong with the XMP on the memory or BIOS since it is setting DRAM Voltage to 1.90V. If you can post a screen shot of the SPD tab in CPU-z, it would give me a better idea.

If you are having issues operating at any frequency, test them individually, run memtest. So test each module to see if one may be causing the problems. If all test out fine, then you may have a bad piece of hardware such as the CPU or motherboard.

If you want to follow Intel spec, you should have purchased DDR3-1333 CL9.

To get one thing straight. No one is copping out. I was just giving a general statement in response to yours. I'm not going anywhere until your system is up and running to spec. Even afterwards if you want to overclock I'll be here to help. I'm in this just as much as you are, we are simply discussing and troubleshooting to find possible culprits.

My job is to help you figure out whether or not your hardware will work together, and if it does, how. There is no point in saying it's not our fault, and we will never say that here at G.Skill. Most other companies will, but not here. Just ask around and many will tell you the same. What I meant was that there are more possible reasons; simply because memory is plugged in and they do not work does not mean they are defective.

Request to have the memory package tested before shipping, this will insure that the kit you received is working flawlessly.

Sorry to hear your expensive system isn't working yet, but work with me and it will.

Thank you
GSKILL TECH

[GeddyT]

Alright, sorry if I got a bit snappy, I'm just very frustrated. I've got at least ten hours into this now and I'm a bit worried that in the end I'm going to have to eat hardware cost in some form or another.

Here's the situation: I've got a brand new motherboard and CPU arriving in the mail tomorrow. Newegg made an exception to allow me to RMA after the deadline, so I'm quite sure this is my last shot with them before things get REALLY expensive.

So, would you recommend I put this memory kit into that brand new hardware and try it out? I hope you understand why I might be a little hesitant. Here's a better question: If the memory is faulty, is there a risk of damage to other components in the system? I'm willing to give this kit a shot and try it on the new board, but not if it's risky.

More questions, just out of curiosity:

What is the difference between running Memtest86+ with individual sticks as opposed to the whole kit? If Memtest reports no errors when ALL of the memory is in place, why would it report errors with just one DIMM? I haven't tried the individual DIMM test for this reason. BIOS and Windows always drop 4GB when cold. Inexplicably, CPU-z and Memtest86+ always show everything just hunky dory and report no errors. I can't get a CPU-z screenshot because the computer's apart at the moment, but the SPD tab looked exactly like it should except for the 1.9V in the VDIMM spot under the X.M.P. column. All other timings were correct to memory spec. It's not a BIOS problem, either, as the first kit's X.M.P. profile correctly called for 1.5V. (And detected 10GB as opposed to this kit's 8GB, but that's another story...)

I've posted pretty complete BIOS settings above, but haven't really gotten any specific feedback. That's what got me a little frustrated. I get it that the idea is that I need to raise voltages, but to how high? When does it get dangerous? And why, then, is the product sold as 1.5V memory? And were there any other issues with my BIOS settings? And why did the memory have the same issue even when running much lower speeds and relaxed timings?

Again, I realize I got a little hot-headed in my last post and I'm not trying to be confrontational here, just looking for some answers so that I might get this figured out. First I need to know if it's safe to plug this memory into my new hardware and, if so, answers to these questions might get me started on trying to make things work.

And, in response to, "If you want to follow Intel spec, you should have purchased DDR3-1333 CL9":

At this point if I could get all 12GB of memory working at DDR3-1333 CL9, I'd do a freakin' happy dance and leave you guys alone forever! I've given up on speed a long time ago in this process and just want all my RAM. I didn't buy this kit because I wanted the fastest speed on the planet, I bought it because it was recommended on another forum as a great value, and I didn't see the point in paying more for slower memory. That and my board claims to support 1600mhz memory natively.

[GSKILL TECH]

The memory will not damage the new hardware. Although, I would suggest manual settings as you stated that XMP shows 1.90V for DRAM Voltage. What you can also do is test them one at a time. If one simply doesn't boot up, then you know it is defective. Especially if you still have the first kit, I would suggest this so you do not continue using defective modules assuming they are good. It just confuses you further.

Memtest does not tell you which module is defective. Testing them individually also eliminates any possible memory controlller/CPU instability. But if they all tested fine together, then no need to retest individually.

You will definitely want to try a new kit if XMP on the memory is 1.90V.

As far as voltage changes, the memory will always be fine at 1.50V, the problem is finding the QPI/DRAM Voltage (memory controller voltage). This voltage should not need to exceed 1.45V for DDR3-1600. It is safe as people operate DDR3-2000 all the time.

If the memory is having issues at lower speeds and timings, then seems like it may just be a bad matching set. So if you replace it, the new tested set should have no problems at all.

It won't be worth the time to try this set in the new motherboard. Once you get the replacement set, everything should fire up and be ready to go.

Thank you
GSKILL TECH

[GeddyT]

Cool, that answers all of my questions. Appreciate it.
Yeah, Memtest detects all memory and runs indefinitely with no errors with the full kit in place. The issue (that everyone with this problem seems to have) only seems to affect BIOS and Windows memory detection.

Good to know that 1.45 QPI/DRAM is the safe limit. Although I never tested that high, I did test all the way up to 1.4 with no success, even at very relaxed speed and timing.

G.Skill customer service continues to be great. I've got another RMA number and a promise to test the kit before shipping to guarantee that they'll work as promised.

Thanks for your help and sticking with it.