I think my memory is dying

[zero08]

I have an EVGA E760 Classified motherboard and 6GB of GSkill Model F3-16000CL9T-6GBTD and I've been running this memory now for about a year at a little over spec speed of 2033MHz with an overclock of 4GHz on my Intel Core i7 920 without any problems until recently. Over the past month or so I've started getting a lot of random BSOD shut downs and I couldn't figure out what was causing the problem until I cut my memory speed down to 1600MHz. With my memory speed at 1600MHz I don't get any BSOD, but the minute I crack it back up to what I used to run it at I start getting BSOD either during Windows load or shortly after Windows loads. I haven't changed anything like BIOS update or anything else that would all of the sudden effect what I had stable and run for almost a year now. The only change I have made was upgrading from Windows Vista Ultimate 64-bit to Windows 7 Ultimate 64-bit. However, even before that recent change I was getting BSOD in Vista until I dropped the speed. This is very weird since I've run this memory for about a year without any problems. I did use MEMTest86+ to test the memory and that has very weird outcomes, as well. One run over 5+ hours showed no errors then another run at 3+ hours in showed all kinds of errors. Not sure what is going on, but like I said it's run like a champ for 11 months at spec until recently. I've always kept my memory at 1.65v with a CPU VTT Voltage at +200 it was the only way to even get it stable at 2033MHz with a 4.0GHz overclock. I ran this machine when I first overclocked it for 24+ hours in Prime95 and it passed without a hitch. I'm starting to think that my memory is dying, so any help here would be appreciated. I know you ask for a full BIOS setting deal when it comes to matters like this, however after 11 months of stable run I doubt it has anything to do with a BIOS setting. I do run a 1200 WATT Thermaltake ToughPower PSU so I'm doubting it is a power problem. I've also tried raising my voltages higher no effect. Doesn't matter what I do as soon as my memory is set to anything near 2000MHz auto fail.

[Schwanke78]

Bro that is way out of spec, I can see 1650 or so but 433+mhz is allot. Also running them at the max voltage what they are rated for does not help either, I would also have to say they are dying or your CPU and Motherboard may be biting the dust. Keep in mind the memory controller is on the CPU now, vs. the old FSB way of doing things! Over-clocking is fun but anytime you run something out of spec there is that risk of malfunctions.

[wevsspot]

Bro that is way out of spec, I can see 1650 or so but 433+mhz is allot.

His memory is the Trident series and it is rated at DDR3 2000 speeds with timings of 9-9-9-24 2N and 1.60v - although most have to run that particular memory at 1.64 - 1.66 DRAM voltage to get above DDR3 2000 speeds.

To the OP - due to the fact that we have to run fairly high QPI/DRAM (CPU/VTT) voltages in order to get this memory kit to DDR3 2000 speeds, it's possible that the on-die memory controller on the cpu is getting weak. I don't think it's the memory, but it could be.

I see that you've been running at +200mV CPU/VTT voltage since you started, do you happen to know what the actual CPU/VTT voltage is? I'm only guessing that default CPU/VTT is 1.10v? Does that sound right?

So if default CPU/VTT is 1.10v and we add .2 that puts your CPU/VTT at 1.3v - and if you've been running at DDR3 2000 for almost a year at that voltage you've been blessed, because most setups require quite a bit more than that. You might try a small bump in CPU/VTT up to 1.35v or so and see if that helps. Remember, this recommendation is based entirely on my best guess of what your BIOS default CPU/VTT voltage is.

[Schwanke78]

Good catch man I didn’t even realize he was running the Trident's, so yeah he is only running a slight overclock very slight! I Should have caught this one, but I was thinking to myself damn 433+mhz over spec that's crazy!

[wevsspot]

gskill

Since you are having Memtest errors I'll bet that if you contact G.Skill RMA department they will authorize a return and replace your faulty memory modules for you. Especially considering that they are only 11 months old. No harm in asking.

[zero08]

wevsspot - thanks for the reply, but I don't think it's a matter of needing more voltage not after 11 months solid no fail performance. I'm seriously thinking the memory has started to fail at that speed after 11 months of use running at 2033MHz. Everything I've checked out even the BSOD logs points to memory failure and with the memory being so spotty during memtest86+ then it kind of makes it clear. I guess the only thing to do here is to test each stick one by one until I find the stick that is causing the failure.

[zero08]

Since you are having Memtest errors I'll bet that if you contact G.Skill RMA department they will authorize a return and replace your faulty memory modules for you. Especially considering that they are only 11 months old. No harm in asking.

I could do that, but since this memory is in my primary computer I'd still have to purchase new memory because I can't afford the downtime. I like the memory and it's held out great and even in benchmarking this machine flew in high digits, and this memory maxed out the rating in Windows 7 but like everything of today I think its time has come and gone :P

[zero08]

Bro that is way out of spec, I can see 1650 or so but 433+mhz is allot. Also running them at the max voltage what they are rated for does not help either, I would also have to say they are dying or your CPU and Motherboard may be biting the dust. Keep in mind the memory controller is on the CPU now, vs. the old FSB way of doing things! Over-clocking is fun but anytime you run something out of spec there is that risk of malfunctions.

Thanks for the reply, but actually it's not out of spec at all these are 2000MHz rated chips and they ran solid for 11 months. The memory voltage is set to spec at 1.65v which is the required voltage it's not like I'm running my DIMM voltage at 1.85 that would be out of spec. My CPU VTT Voltage is only +100 over stock which too isn't out of spec if you want to run stable and since most people pump it to +300 and that is also what G-Skill recommends you do when running these chips at 2000MHz I think I've been doing fine. I don't think the motherboard or CPU have any problems since they both run fine my CPU passes all test with flying colors and since the memory seems to be failing random times through memtest86+ I would tend to believe the memory is failing which isn't uncommon on these memory modules if you look around. 1600MHz is recommended if you are using 12GB plus, but since I'm only running triple channel 6GB 2000MHz+ is supposed to be a breeze for these chips. I've been temped to replace these with a different brand of 2000MHz and see what happens because if a different brand fails to run like these used to then it might be something else though based on information I believe these chips are the problem. I'm sure they where tested to run at 2000MHz which is why they are sold as 2000MHz DDR3, but where they tested on how long they could run at that spec? The CPU is way out of spec for sure 1.34GHz increase over stock spec, but its holding strong and I've never had to crank the heck out of my voltages for this over clock to be stable. My CPU voltage alone is only running 1.275v at a 4.066GHz overclock which is pretty bad **** if you ask me so it's not like I've over volted my rig on any level.

[Schwanke78]

Yeah I jumped the gun; thinking you was running 1600Mhz Ram @ 2033Mhz I was saying no way! Don?t know what to tell you man, what's you temps like on your CPU and Ram? 4.066GHz is pretty sweet been to 4.2Mhz myself but saw no point in running it 24/7 @ that speed or really any gain for that matter. Also 1.275v is pretty much in standard for a i7-920 or similar chips, I take it your running a 920 or 930. Personally I like to undervolt my stuff even my Sager Laptop is undervolted, just like to max the life out of my hardware. Anyways far as your hardware issues not sure what to tell you, something is starting to falter there?s no doubt about that.

[wevsspot]

I guess the only thing to do here is to test each stick one by one until I find the stick that is causing the failure.

I think that makes the most sense from a trouble shooting standpoint.

[zero08]

Yeah I jumped the gun; thinking you was running 1600Mhz Ram @ 2033Mhz I was saying no way! Don?t know what to tell you man, what's you temps like on your CPU and Ram? 4.066GHz is pretty sweet been to 4.2Mhz myself but saw no point in running it 24/7 @ that speed or really any gain for that matter. Also 1.275v is pretty much in standard for a i7-920 or similar chips, I take it your running a 920 or 930. Personally I like to undervolt my stuff even my Sager Laptop is undervolted, just like to max the life out of my hardware. Anyways far as your hardware issues not sure what to tell you, something is starting to falter there?s no doubt about that.

I run the 920 and the voltage is just crazy because I'm not that far over stock voltage, but with a huge over clock. Really nice when you see most other people have to push voltage to 1.3 or 1.4 to get a 4.0GHz stable. My temps are great and since I'm not that far over voltage and I use liquid cooling my cpu stays around 31 -34 degrees Celsius idle and about 42 - 50 degrees under max load. I monitor my core temps, as well and those are always in the mid 40s idle or upper 50s if I'm gaming. The hottest my core temps ever got was when I ran Prime95 they hit around 75 core temp which is great at the speed I'm running because that is still 25 degrees from thermal junction. I've pushed 4.2GHz for benchmarking and trust me that was awesome especially when I benchmarked against a few Mac Pros using dual quad Zeon processors and smoked them with a single core i7, but I pretty much keep it at 4GHz on a normal bases with a base clock of 203 and a multiplier of 20. It does suck that the memory has started to fail, but in all honesty I haven't noticed a performance loss from coming down from 2033MHz to 1686MHz my system is still super fast. The only thing that troubles me is I like the hardware I buy to run at the spec it was designed for and when it fails even after 11 months just sucks. This memory wasn't the most expensive ram, but still I think it should last longer then a year in life.

[zero08]

I think that makes the most sense from a trouble shooting standpoint.

I agree I just have to find the time to run it. The first couple times I ran memtest86+ was with all three sticks. This time I'll have to do single sticks and nobody ever looks forward to that task

[Schwanke78]

Yeah liquid works wonders my i7 build was the first time for water-cooling. Your CPU looks dead on but what about your RAM temps, I have a probe on mine since the Koolance RPK1000 gives you three and I run around 32c on load and that is without a RAM fan mind you. I know that some of the other brand RAM cooks at 2000Mhz, so I was just curious. I have also verified mine with a no-contact infrared thermometer. But as already said running one stick at a time should help narrow it down, but yes it?s a pain and probably three hours you don?t want to spend on doing it. Either way let us know how it goes and if you find a remedy.

[wevsspot]

Zero,

Even if only one stick is bad, GSkill will want the entire kit back to do a warranty exchange for you.

Maybe testing each of the three sticks individually will make you feel better knowing exactly which one or two are causing the problems - but like I said, they will still want the entire kit back.

And it really sucks that you have to buy a new kit just to keep you machine from being down. On the flip side, once your warranty replacement arrives you'll have a whole kit that you can sell to help offset the cost of the replacement kit you're buying.

[zero08]

Yeah liquid works wonders my i7 build was the first time for water-cooling. Your CPU looks dead on but what about your RAM temps, I have a probe on mine since the Koolance RPK1000 gives you three and I run around 32c on load and that is without a RAM fan mind you. I know that some of the other brand RAM cooks at 2000Mhz, so I was just curious. I have also verified mine with a no-contact infrared thermometer. But as already said running one stick at a time should help narrow it down, but yes it’s a pain and probably three hours you don’t want to spend on doing it. Either way let us know how it goes and if you find a remedy.

Hey Schwanke, not sure what my ram temps are I do know that they don't get hot to the touch, but meh you never know. I use the Thermaltake Armor+ because I went with the Tri-SLI plus a physx card setup so the case gets good ventilation and airflow. I've often thought about purchasing a ram cooler, but the heatsinks on the Trident is so large I didn't want to spend the $40 bucks for a decent cooler only to have it not fit. I know that running liquid cooling does effect the ram cooling because the traditional fan / heatsink combo helps move air around in that area a lot better then liquid cooling. I'm going to run them in singles and test them later on this week I can't do it now since I design websites I have a few clients on deadlines I have to get finished. I'll post the results forsure once I find out what is going on.