RAM settings in BIOS to get 6 sticks stable (and scary burning smell)

avianrand

Distinguished
Apr 18, 2010
155
0
18,680
See my member config or the very bottom of this post for my setup.

I've only running with 3 sticks of RAM because I could not get all 6 to run stable with Prime95 for more than an hour or so. I'm very new to any BIOS tweaking and was going to wait to do any of that until I had a chance to learn. I found out after I started setting things up (per bilbat) that 3 sticks is usually stable out of the box but 6 usually isn't and it requires some tweaking. So I was having trouble (bilbat tried to help before, see note below) and figured I'd call on tech support at Gigabyte. They have been very helpful when I call them so I figured I'd try with this issue. It seemed like it should be a simple matter for them to be able to tell me how to get 6 sticks running stable on their board that is designed to do this. It was NOT a good idea.

The specs on my RAM are DDR1333, 7-7-7-20, 1.65v

Here are my BIOS settings for the CPU and RAM:

BIOS%20MIT%20Page%201.JPG


BIOS%20MIT%20Page%202.JPG


BIOS%20MIT%20Page%203%20Marked%20Up.JPG

(this is currently set to auto but I'm showing what I tried below)

And here are my CPU Tweaker settings:

CPU%20Tweaker%20Readings.jpg


The tech person recommended setting the DRAM voltage from auto to 1.64 to come closer to the specs on the RAM. I was a bit uneasy about this as I am not clear about which voltage is which for RAM. But the tech seemed very confident about this so that's what I did.

I rebooted and things seemed to be running fine. So I opened up the case, popped in the other 3 sticks and proceeded to run Memtext86+. Within seconds, a foul burning smell was coming from my case :fou: . I shut down very quickly and opened it back up in a bit of a panic of course. The RAM was very hot but I could touch it. Mind you it took a few minutes between the time I turned the machine off and the time I got to the RAM so it had had a chance to cool a bit. The NB heatsink seemed no hotter than normal and same with the CPU pipes near the CPU. There are no scorch marks on anything but with the heat spreader covering the RAM chips, I can't see them but what I can see seems fine.

Anyway, I'm sure it's possible to get this running on 6 sticks but not with my limited experience at this (I'm a software guy by trade and a lot of this is new to me). Right now I'm more worried about damage that may have been done although things seem to be running fine. I took out the extra 3 and set the DRAM back to auto. I ran Prime95 (with RAM set at 6144MB) for a few hours tonight and everything seems back to normal. I'm going to run Memtest86+ again tonight for a bit. I had run it before with the regular BIOS settings on each set of RAM (3 sticks at a time only) for several hours each and no issues. I'm sure I'll have the same results again, but I am concerned that there could have been damage but it's not obvious. Would it be? Would I get errors right away if there was unapparent damage?

Also, if you take a look at my screenshots, what would you suggest I do to get all 6 sticks running stable? Keep in mind, this is very new to me. One of the things that is unclear is the fact that some of the wording in CPU Tweaker is not the same as in the BIOS so if you make any suggestions, keep that in mind. What's clear to you might not be clear to me. :)

bilbat tried to help me here http://www.tomshardware.com/forum/forum2.php?config=tomshardwareus.inc&cat=30&post=273675&page=1&p=1&sondage=0&owntopic=1&trash=0&trash_post=0&print=0&numreponse=0&quote_only=0&new=0&nojs=0 but the thread kind of covered too many issues and I didn't provide enough screenshots. I want to start fresh.

Motherboard = GIGABYTE GA-X58A-UD5 (BIOS v 5)
CPU = (1) Intel Core i7-930 Bloomfield 2.8GHz LGA 1366
RAM = (12GB total) CORSAIR XMS3 6GB (3 x 2GB) DDR3 1333 (PC3 10666) 7-7-7-20 Model TR3X6G1333C7 G
Cooler = Noctua/XION NH-U12P SE2 (2 fans)
Video = (1) EVGA 512-P3-N871-AR GeForce 9800 GTX+
Sound = HT | OMEGA CLARO Plus+
Case = LIAN LI PC-P80 or PC-A77F
PSU = CORSAIR CMPSU-HX1000
Storage = (2) Intel X25-M 80GB SATA II MLC SSD RAID 0 (on the Intel controller)
Storage = (4) yet to be determineed HDD\'s (probably going to use WD7502ABYS) RAID 10 (on the Intel controller)
Storage = (1) Western Digigal WD1001FALS 1TB HDD for my DVDs
Storage = (1) Western Digigal WD5001AALS 500GB HDD partitioned in 3 parts for clones of the OS drive in case of emergencies (this is in addition to clones and backups that are on spare drives outside the box)
Optical = (2) LITE-ON DVD Writer - Bulk - Black SATA Model iHAS224-06
OS = Windows 7 Ultimate 64 bit
 

bilbat

Splendid
Dog just got me up from a knock-out med-induced sleep - will take a bit of time, but if nothing's physically damaged, I think we can do this starting in the AM - several issues: for i7, & fast RAM, Vddr needs to be within a half volt of Vcore, which is set, often, by the default, too low to work; Vddr may need to go up a hair for two sticks per channel, and QPI will likely have to be 'bumped' considerably... [:fixitbil:5][:fixitbil:5]

Info you posted is excellent, & should give me all we need...
 

bilbat

Splendid
Still don't have an idea about 'hot smell' (eeekk - :??: ) and looks to be exactly what I always recommend for 1156/1366 RAM, not too godawful high a clock speed, but really low latency! - but see a couple things in setup...

first thing is on the first BIOS screen, it shows 'Profile QPI' voltage around where it needs to be at 1.25V (for one DIMM/channel, for two, I'd go 1.3 - 1.35), but, on second BIOS page, shows it actually set to 1.175 - check & reset...

second thing, shows uncore at x20, which is absolute minimum for x10 memory multiplier; for two DIMMs/channel, bump up to x21...

third thing - RAM timings look good - it's 'normal' for Tweaker/MemSet to show current tRAS one higher than actually set - don't know why? But, tRFC must usually be 'bumped' for two DIMMs/channel as well - I'd take it to 82, & 'walk it down', two at a time, after stable (read section in 'sticky' labeled BIOS - Saving CMOS - will save you a ton of time & effort 'tweaking' this stuff!

BTW - timing defs:

BIG FOUR! normally shown: CAS-tRCD-tRP-tRAS
CAS (tCL) Timing: CAS stands for Column Address Strobe
tRCD Timing: RAS to CAS Delay
tRP Timing: Row Precharge Time
tRAS Timing: Min RAS Active Time [normally tRAS = tCL + tRCD + tRP (+/- 1)]
+++++++++++++++++++++++++++++++++++++++++++++
sub-timings, in 'Tweaker's SPD order...
Command Rate: amount of time (cycles) between chip select and command acceptance - lower (1T), faster; slower (2T), more stable
tRC Timing: Row Cycle Time
tRFC Timing: (Refresh To Act Delay) Row Refresh Cycle Timing
tRRD Timing: (Act to Act Delay) Row to Row Delay or RAS to RAS Delay
tWR Timing: (Write to Precharge Delay - Write Recovery Time)
tWTR Timing: (Rank Write To Read Delay)
tRTP Timing: (Read To Precharge Delay)
tREF Timing: Time before a '1' bit needs to be refreshed so it does not lose its charge and 'vanish' (in micro-seconds [µsec])

Will need at least 1.65 Vddr (you've got the right place!), and I'd check it with something like HWMonitor, to see what your setting is actually giving you on six DIMMs, might also need a (tiny) 'bump'...

Good luck! Post back with any questions - I'm conscious for a bit yet - dunno if you've seen my expl, but I'm a raving manic-depressive - normally insomniac, and insomnia causes mania, which causes more insomnia, which causes more mania - and pretty soon I'm pacing aroung the ceiling with a tin-foil hat on [:fixitbil:9] [:bilbat:9]
 

avianrand

Distinguished
Apr 18, 2010
155
0
18,680
Cool. Thanks. Lots of info, and before I make any changes I have questions. :) After the burning incident I'm a little nervous about this. I memetested all the RAM separately (3 sticks at a time) for 8 hours each or so and got no errors. I also ran Prime95 for a few hours on just one set and things were fine but I'm not entirely confident there was no damage because at one point today I had all 6 DIMMs in place (without changes to the BIOS) and ran Prime. The first time, it ran for about 2 minutes when one of the workers shut down and went red. The other 7 were still running. The 2nd time it just shut down after about 15 minutes (not the whole computer, just Prime95). I'm hoping it's just because I hadn't made any BIOS changes to accommodate 6 DIMMs but I had run Prime95 a few days ago on all 6 and it ran fine for about an hour. I'm nervous that one of the controllers in the CPU could have been damaged but maybe I just don't understand this enough yet and it's really okay.

It seems like there are several ways to refer to some of these settings and I'm not always clear as to what's being referred to. It also seems (as you've pointed out with the QPI voltage) that some of them appear in more than one place with different values. So that's confusing me as well.

So here are my questions.

You said to change Profile QPI to 1.3 or 1.24. It has 2 different readings as you said, 1.175 and 1.25. Which is correct? Here are some screenshots (also, is this where to change it? - as you can see from the 2nd screenshot, the numbers go pink at 1.355 so I'm nervous about that):

BIOS%20setting%20QPI%201.JPG


BIOS%20setting%20QPI%202.JPG


Is this where I need to change the unCore (3 screenshots showing sequence of where to change this)?

BIOS%20setting%20unCore%201.JPG


BIOS%20setting%20unCore%202.JPG


BIOS%20setting%20unCore%203.JPG


The next 5 screenshots show where I think I need to change the tRFC (but I'm not clear on this because it appears that one way you can only change channel A but the other way you can change A, B and C):

BIOS%20changing%20tRFC%201.JPG


BIOS%20changing%20tRFC%202.JPG


BIOS%20changing%20tRFC%203.JPG


BIOS%20changing%20tRFC%204.JPG


BIOS%20changing%20tRFC%205.JPG


If I choose the "quick" way, where it only allows changes to "A", is this actually changing all 3 to be the same?

In my other thread where you were first helping me with this you mentioned changing EIST and C1E here. Should I do that again? (I'm guessing the screenshot below isn't showing cuz maybe there's a 10 screenshot limit per post?)

BIOS%20EIST%20from%20before.JPG


You also said something before about changing the Command Rate from 1 to 2 (in hardware monitor it says 1N but you said "T" and there's no "T" or "N" in the BIOS so I wasn't sure where to set that but command rate was in one of the screenshots above).

If the profile DDR says 1.65, why does it show 1.5 at the bottom of that BIOS screen where it indicates "auto" (this is where the gigabyte tech person had me change it to 1.65 and then I got the burning smell because I'm guessing the RAM was overvolted at that point.

In CPU Tweaker, I noticed the timing says "PC3-10600 (666MHz) @ 9,9,9,24 - 1.5v. Why does it show 9,9,9,24 instead of 7,7,7,20 (or 21)?

Oh, and here are my current settings with all 6 sticks installed:

CPUID%20and%20CPU%20Tweaker%20Settings%20with%2012GB%20in%20place.jpg


One other thing, with BIOS set to optimized defaults (pretty much screenshot above) except setting RAM to XPM, if I run Prime95 like this:

Prime95%20Test%20Settings.jpg

(or any high RAM value from 8000 to 12000 or so)
... Prime kills a worker or 2 right off the bat, almost no delay:

Prime95%20Worker%20Stopped.jpg




EDIT:

Don't know if this helps but with all of the above settings (the ones I show for having 6 DIMMs installed), so far Memtest86+ has run without errors for 4 passes (6 hours). I'm going to continue to run it for a while longer, maybe until Monday morning.
 

bilbat

Splendid
OK - how's the testing going? I think I'm up to par as I'm gonna get - spent half the day reading at TweakTown, Extreme, and OCer's; pretty much found out what there is to find out, I think... First - think the 'smell' issue can be attributed to the variance between high Vddr and too low a QPI/Vtt - seems to be the most common 'cooker' of memory controllers. Second, still have some 'mechanical' issues - exactly what adjusts what in the actual BIOS, but the couple things still up in the air, we can 'learn as we go'... 'Nother thing - seems like there are 'issues' involving XMP with all these boards - and I think I have a procedure to 'sidestep' the majority of them!

I think you may have slightly misunderstood my last post - I did not mean to say we have to adjust the 'Profile QPI'; we have to boost the actual QPI/Vtt on the voltages page. I believe what is happening is there is a 'mechanism' in the BIOS that 'looks at' the XMP's 'suggested' DDR voltage, and then, based on the sSpec ID returned by the processor, 'calculates' what the 'safe', higher QPI/Vtt should be - but I don't believe it can set it! Bad BIOS! [:isamuelson:8] I'm pretty sure this is the situation, as the XMP does not (in fact, can not) contain an actual register for this value - it's got to be gotten from the processor type and stepping info... I have determined that relatively astronomical QPI's are completely safe, and often specified by DDR makers themselves for fast modules. It's the 'split' between high Vddr and Vtt that'll getcha! Also, bumping the Vddr is a common practice for getting two modules per channel to work - I've come across people running Vddr's as high as 1.8 Volts!! :ouch:

One last suggestion before we tackle this - what do you think of taking the CPU directly to (a pretty 'tame') 3.5GHz? The 'maths' work out right there, and if everything works out nice and stable, you can then probably just 'walk it up' to around 3.65 t0 3.73 just by increasing the Bclk a count or two at a time, 'till she fails prime - than 'back down' two!
 

avianrand

Distinguished
Apr 18, 2010
155
0
18,680
12 full passes of memetest86+ (and nearly done with pass 13) and going on 23 hours with no errors. Does that mean the RAM controller in the CPU is okay? It occurred to me yesterday that the Prime95 errors I was getting when running it on a lot of RAM could mean that the cooking smell I got the other day could have damaged the CPU. When you said "seems to be the most common 'cooker' of memory controllers" that got me concerned too that I could be right. Is there any way to know for sure? The thing is that this is going to be my primary work computer once it's all set. I can't afford instability. If a gaming computer screws up now and again it's not a big deal, but I can't afford to have any of my data messed up due to RAM issues. That has me a bit worried. Of course I'd rather not but If I have to replace the CPU I'll do that. They're on sale at Microcenter for $200 still. Quite a deal. I thought to take the one I have out of the motherboard so I can check for scorching and also check the pins carefully. Is it possible the board could have been damaged? That's a bigger concern as it was in the low $300 range. I can deal with replacing the CPU but not the board.

(side note: I cannot read sites like overclockers - white text on nearly black background actually hurts my eyes; why they do that is beyond me; it looks cool at first but it's not cool at all :pfff: )
 

avianrand

Distinguished
Apr 18, 2010
155
0
18,680
Okay, you're going to be mad at me. I thought I had the right version of prime but there's no way to tell from its properties or the name. I may have been testing using the 32 bit version. That might be why some workers were not happy when I would run a custom test with a lot of RAM. I am running the 64bit version (2511) and at least for a 1/2 hour or so with all the blend settings but running custom with 11000MB RAM, it's running okay so far. (BIOS with optimized defaults but running RAM with XMP on and CPU on "turbo" setting)

I'm still worried that I may have damaged something the other day. Since memtest86+ ran fine for 23 hours and if prime (the right one) runs okay for a dozen hours or so, should I not worry that something could have been harmed?
 

avianrand

Distinguished
Apr 18, 2010
155
0
18,680
Prime95 with settings shown below has run fine for about 5 1/2 hours

Prime95%20Test%20Runnnig.jpg


I'm guessing at this point that the BIOS on this board handles 6 DIMMs just fine with most of the optimized settings and that the reason I was having trouble with Prime95 originally was due to using the wrong version. Ugh.

I'm still concerned that the bad info i got from Gigabyte tech support telling me to set the voltage as high as I did could have caused some damage somewhere but it's just not showing up. Maybe took away some of the useful life of something. The RAM seems physically fine. I took a flashlight and a magnifying glass so I could see under the heat spreaders and all looks good. As far as the CPU goes, seems like with all the protections built into the newer CPUs for shutting down if there's a heat problem that that should be okay. Just my guess. I also thought the burning smell could have been the PSU but without opening it up i looked inside with a flashlight and that looks okay too and the fact that they also have a lot of built in protection makes me think it could not have overheated. Still wondering what got so hot that it smelled like something was on fire. It was pretty bad.
 

bilbat

Splendid
The memory controller actually has its own thermal controller sub-system, so I think, too, you're good to go... If you'd like to do that 'bump', heres the reason why: boosting the Bclk to 167 would give you 1336 DDR3, with an eight multiplier, but would give you a 21 x 167 3.5GHz CPU - bet you could 'see' the difference immediately!
 

avianrand

Distinguished
Apr 18, 2010
155
0
18,680
Yeah, that'd be cool. I definitely want to give it a shot but not today. I'm going to let Prime run over night just to be safe with current settings. Just want to be sure hardware's all good. It's 8 hours so far. Tomorrow morning will make it 24.

Would anything else have to be changed other than changing bclk from 133 to 167?
 

bilbat

Splendid
That's prtty much it - change a couple multipliers - memory timing will remain the same, so no timings need to be altered; I still would like to 'close the gap' a bit, between your QPI/Vtt voltage, and your DDR voltage - safer, long term, that way...
 

avianrand

Distinguished
Apr 18, 2010
155
0
18,680
Cool. Looking forward to it. Thanks.

And by the way, burning smell solved! It turns out that a couple of weeks ago someone in the house had dropped a piece of onion into one of the stove burners :ouch: . The kitchen is very close to my office. At the exact same time that I had started up memtest to test the voltage changes I'd made, someone else turned on the burner to cook something and the onion in the burner gave off a very foul smell :sweat: . The timing couldn't have been more perfect. I know this because it happened again just a few minutes ago and the smell is so strong and so foul you'd think wiring or some other electrical parts were burning :fou: . It really does smell like a circuit gone bad. I'm very happy that it wasn't one of my new parts! :bounce:
 

bilbat

Splendid
Murphy's damned law, in action!! Brings to mind two stories:

Used to work for a firm that rebuilt (and computerized - that's why they had me!) machine tools - so the whole place was essentially one big, huge machine shop! Once in a while, somebody'd be cutting phenolic, which was the stuff, at the time, that all circuit-board 'substrates' were made of... All us 'electronics guys' would have to 'go on strike'; return to our office and cubicles, 'cause we couldn't bear the smell of 'burning circuit board' 'till the 'air had cleared'!

Working in the same place, I had done a good sized machine that had a 'twin bay' electrical cabinet - inside one door was the computer rack, and all its I/O cards; inside the other, a bank of rack-mounted DC drives, all sharing the rack's (like 100V at 80 or 100 A!) PSU... So, 'start-up procedure' was: open both doors, one guy in front of each door, guy in front of the end door with his hand on the disconnect (more or less the 'main' 440 power switch), ready to 'kill it' at a yell. I'm workin' with an electrician who smoked cigars, and have him at the switch. He takes a big lung full of smoke, 'drops the hammer', and simultaneously blows the lungful of smoke into the cabinet, and out my side!! Damn near had a heart attack right there - must've been thirty-thousand worth of drives in that rack!!