1366 x58 Xeon Enthusiast overclocks club

[Napoleon]

Sorry to double post, but just ordered an X5670, hope it works! A plus would be to match/exceed my current 4.2ghz

 

[Bill1024]

Yeah, the x5650 is 22, the 5660 is 23, the 5670 is 24 and so on.

 

[primetime]

And to some degree it depends on the bios your using...the normal stock one for mine is one notch lower than what bill posted iirc....of course cross flashing can help fix this when needed. If you run into blck limitations every little bit helps

 

[Napoleon]

According to this Meet The Xeon 5600 Family - Intel Xeon 5600-Series: Can Your PC Use 24 Processors? the x5670 at 2.93/3.33 equates to a 22x/25x(turbo) multipliers based on blck of 133mhz. Who know what it'll do on my board though, as log as it turn on ill be pretty happy. If it's at 24 I'll prob never hit 200 blck but at least high total OC and tight timings

 

[JEKYLL]

Just did my final cpu upgrade on my Asrock X58 Extreme..Went from 960 to the W3690 xeon.. This board has been a tank for about 6 years, so I thought one last heart transplant would be nice..Seems to be pumping along pretty well..Wish I could find a Asrock X58 Extreme6 for a good price..

 

[Rajveer]

What OC did you stabilize at? How has your experience been?

I'm running an i7 950 @ 4.2GHZ on an X58A-UD3R rev2 and am thinking about buying an X5670 on Ebay, would you recommend it?

It seems all these Xeon OC's are based on FSB, is the memory multplier locked or can you still keep it at 8x 200 = 1600mhz? I'd like to try for a 4.4ghz xeon OC on my current hardware

I replied to your PM before I saw this post, so I'll post my PM reply here too for completeness:

I stabilised my X5670 at 4.2GHz, only because I wanted to keep my case fans as quiet as possible and didn't want to have to increase them. It was a fairly easy overclock and I could have easily hit 4.4-4.5GHz, but didn't like the noise of the case fans.

I went from an i7 920 @ 4GHz to my X5670, which I ran stock until a few weeks ago. I used to play Battlefield 4 on 64-bit servers with everything maxed out with my GTX 980ti and it was a beast (only going below 70FPS on a handful of occasions for split seconds). I only overclocked recently as I needed a higher core-clock for the Oculus Rift, and that was only stuttering ever so slightly in specific scenes - it's perfect in all games now.

It's definitely worth the upgrade IMO! I run my overclock with all C-states, EIST and Turbo Boost on, and use dynamic voltage so it's cool when idle/low use but fast when I need it.

Regarding QPI: I run mine at a lower ratio now closer to it's original clock, however the next step of my overclock is to increase that as much as I can. I just...totally forgot, until you mentioned it.

 

[windianrecords]

Mind if I join the club? Just bought a X5660 off eBay to replace my i7-920, have it at 3.63 before temps get too high (80C) on my current Noctua. Going to install an AIO before I move it any higher, but thanks everyone for keeping interest/information in these chips alive!

 

[Rodall]

Just got my X5670 today and started overclocking it. So far I've been able to get it up to 4680 (195 x 24 at 1.35 vt) Temps are maxing out around 65C, but it keeps BSOD or freezing up. May have to dial it back a bit.

Lots of great info here about overclocking Xeons.

 

[primetime]

Just got my X5670 today and started overclocking it. So far I've been able to get it up to 4680 (195 x 24 at 1.35 vt) Temps are maxing out around 65C, but it keeps BSOD or freezing up. May have to dial it back a bit.

Lots of great info here about overclocking Xeons.

drop it back to 191blck and 22 multi and enjoy easy 4.2 bullet proof stable. drop the blck to 183 if you want to use 24 multi

 

[Rodall]

drop it back to 191blck and 22 multi and enjoy easy 4.2 bullet proof stable. drop the blck to 183 if you want to use 24 multi

Thanks for the tip. I'm running at 200 x 22 at the moment, and so far so good

Time to stress it and see if she holds up.

 

[Napoleon]

Got my X5670, running as well. I was able to boot into BIOS/Windows at 22x200, but it was unstable and would restart. I messed with voltages a bit and backed it down to 22x191~4.2ghz for now (gotta go to work). I'm using the same voltages as my i7-950@21x200~4.2ghz right now

I've never had issues with blck before, but I guess the memory controller on these is different.

Validation link: CPU-Z VALIDATOR

Voltages:

Vcore: 1.35625 (reads as 1.344 in HWmonitor)
CPU PLL: 1.8
QPI/Vtt: 1.315
VDIMM: 1.6 (reads as 1.584 in HWmonitor)
QPI PLL: 1.2
IOH Core: 1.2
ICH I/O: 1.5

Ram multi: 8x
Uncore multi: 12x

All power saving features off, Line load calibration: Level 2

Observed a 10% physics score increase in 3dmark, didn't do any before testing to compare anything else.

Thoughts? If I can't get more blck I'm debating trying out 'Turbo'

EDIT: not completely stable, spontaneously restarted (no blue screen) after about 20-30 mins in Prime95. Ideas on what this means?

I bumped the QPI/Vtt to 1.335 an dVDIMM to 1.62. Running Prime again, not sure why, thinking some sort of RAM instability, although I was solid at DDR1600 8-8-8-24-2 for months (at 1.65vdimm though)

Should I just set the vcore and hit 'AUTO' on the rest (or AUTO on all?), set it to 200 bclk and then check via Easytuner6 as to what the mobo sets the voltages do for 'stability'?

 

[LigTasm]

1.35V is a shitload for that clock speed. Can you do 20x200 at 1.25V?

 

[primetime]

1.35V is a shitload for that clock speed. Can you do 20x200 at 1.25V?

that may be but overheating is not his current problem....he needs to get it stable. He should use turbo and go for a lower blck on that board

 

[LigTasm]

that may be but overheating is not his current problem....he needs to get it stable. He should use turbo and go for a lower blck on that board

Honestly if he just swapped out the i7 and went with it I would start fresh and work up from the bottom again. What the Xeon likes might be much different than a 45nm bloomfield.

 

[primetime]

Honestly if he just swapped out the i7 and went with it I would start fresh and work up from the bottom again. What the Xeon likes might be much different than a 45nm bloomfield.

yea of course.....it a hole lot of trial an error all over again

 

[Napoleon]

yea of course.....it a hole lot of trial an error all over again

Ok. I'll turn on turbo and turn down the BLCK and vcore. I'm not sure where to start with the other voltages, I may either leave them or set to AUTO

With turbo on, should I turn on any of the power saving features? I'm not sure how clock scaling works with the TURBO function (or when TURBO turns on)

It looks like there is a +2 multiplier for 3 to 6 cores and +3 multiplier for 1 or 2 cores.

I'm not sure how far back to drop the blck to start, perhaps 180?

22x180=3960
24x180=4320 (3-6 cores)
24x180=4500 (1-2 cores)

 

[primetime]

Ok. I'll turn on turbo and turn down the BLCK and vcore. I'm not sure where to start with the other voltages, I may either leave them or set to AUTO

With turbo on, should I turn on any of the power saving features? I'm not sure how clock scaling works with the TURBO function (or when TURBO turns on)

It looks like there is a +2 multiplier for 3 to 6 cores and +3 multiplier for 1 or 2 cores.

I'm not sure how far back to drop the blck to start, perhaps 180?

22x180=3960
24x180=4320 (3-6 cores)
24x180=4500 (1-2 cores)

all i know is blck limitations is what typically make the overclocks more difficult. I also seem to remember people with certain board when switching to hexa cores all of a sudden having trouble reaching blcks they easily hit with quads....If your one of those then use the turbo by all means. The way you described turbo seems correct to me, but it varies slightly with different bios and boards so it will be different for some. 180 blck sounds like an ok starting point to me.....if its not stable back up so more

 

[Napoleon]

all i know is blck limitations is what typically make the overclocks more difficult. I also seem to remember people with certain board when switching to hexa cores all of a sudden having trouble reaching blcks they easily hit with quads....If your one of those then use the turbo by all means. The way you described turbo seems correct to me, but it varies slightly with different bios and boards so it will be different for some. 180 blck sounds like an ok starting point to me.....if its not stable back up so more

200 FSB/blck has been the sacred cow number I have shot for for YEARS; too bad I might have to abandon it...

 

[Rajveer]

200 blck doesn't mean anything, don't aim for blck, aim for frequency. Let your CPU scale the frequency from x12 - x24 using it's power-saving features (so it's not hot all the time and is quieter when idle), rather than locking at a certain multiplier. 175 Blck will give you the 2100MHz -> 4200MHz range.

 

[Napoleon]

I just finished 8 hours prime blend at 175x24 (turbo enabled, multi set to 24 in bios)

Before I tried 22x200 - spontaneous restart in windows
I also tried 22x191 - would run 3dmark, restart after 20 mins running prime blend

Vcore: 1.20625 + 0.15 (reads as 1.344 in HWmonitor)
CPU PLL: 1.88
QPI/Vtt: 1.335
VDIMM: 1.66 (reads as 1.648 in HWmonitor)
QPI PLL: 1.2
IOH Core: 1.2
ICH I/O: 1.5

CPU Clock drive - 700mv
PCI Expres Clock Drive - 700mv

Ram multi: 8x
Uncore multi: 12x
QPI Link - 36x

All power saving features off, Line load calibration: Level 2

I used to use AUTO for most voltage settings, like I did on my [email protected] with no issues; only recently did I start actually setting the voltages, the day before I swapped in my X5670; I'm not even sure it was stable before.

I may go back and only adjust Vcore/Vdimm and leave the rest on Auto to see if it will go higher, I am not sure what my 'weakest link' is right now. Perhaps my QPI link freaks out when it is above 6.4ghz? I would like to use less voltage or get a higher OC at this voltage

I never exceeded 69C on any core after 8 hours, normally it runs about 10C cooler, my water just heated up (normally I use only for 1-5 hours total, not full load). At least I can rule out thermal issues

HWmonitor: hwmonitor_x5670_42_8hrprime.jpg

One thing that is concerning is the large variance of my voltage lines; does HWmonitor typically read badly or can I realiably blame this on my PSU?

• 3.3V - 3.136-3.248
• 5V - 4.758-4.919
• 12v - 12.288-12.416

 

[Rodall]

I just finished 8 hours prime blend at 175x24 (turbo enabled, multi set to 24 in bios)


HWmonitor: hwmonitor_x5670_42_8hrprime.jpg

One thing that is concerning is the large variance of my voltage lines; does HWmonitor typically read badly or can I realiably blame this on my PSU?

• 3.3V - 3.136-3.248
• 5V - 4.758-4.919
• 12v - 12.288-12.416

I recently went from a Corsair TX 850w to a EVGA 750P2 and noticed my voltages show practically no variances now.

 

[primetime]

This is the setting i use, but it seems like every board and cpu is a bit different in what they like

cpu voltage 1.35
pll voltage 1.9
qpi/dram 1.35
dram 1.65
all spread spec disabled
dram speed 1531
qpi link 6893
overclock xmp
dram time all default
i use highest multi for me 22
speedstep disabled
c-states all on
blck 191

 

[Eric1285]

Here's my first update since putting in my new X5670 a few days ago...

Pertinent specs:

Intel X5670 - bought used off of eBay for $70. It's a "B" chip. Didn't think to write down the rest of the details before I popped it in, but I'll probably be pulling it out to lap the IHS in a few days.

12GB G.Skill Ripjaws X (3 x 4GB) (rated 1.5v, DDR1600)

Asus Rampage III Gene

Anyhow, I got chip put in a few days ago and immediately started playing around with it to try and find its OC potential. I set vCore just below 1.35v, VTT at 1.3v, and bumped up both the IOH and ICH voltages by 2 notches each. I have the RAM set to run at just 1400 and kept the UCLCK between 1.5-2x the ram speed. Was able to boot to Windows at 190 x 24 for 4.5ghz, but Prime95 would stop or crash after a few minutes. I messed around a bit more trying to see how stable I could get 4.5ghz but I never managed to keep Prim95 running for much more than 30 minutes.

A few things I noticed during the initial testing:

• Vdroop on my board is horrible, and enabling CPU LLC has almost no effect. Current theory is that CPU LLC may only work nicely in offset voltage mode, so I'll be playing with that at some point. Right now I have to set vCore around 1.36 in the BIOS, which gives me ~1.35v (per HWmonitor) at idle and drops to 1.296v at full load.
  
• Temps seemed fine with an Arctic Liquid Freezer 240. Idle core temps are around the low-mid 30's and full load will hit 63C. Difference between cores is usually no more than 4-5C.

I was encouraged by the fact that I could boot into Windows at 4.5Ghz, so I figured a stable 4.2Ghz would be easy...well, maybe. I had hoped to be able to just knock the BCLCK down, pull the vCore back a bit, and call it a day but it wasn't meant to be. Even with the same vCore (~1.36v BIOS, 1.35v idle) Prime95 would still stop or crash after ~30 minutes, so I started to mess around with the other BIOS settings.

One of the things I read in this thread (and over at OCN as well) was that the CPU PLL voltage could affect stability. Most people mentioned bumping this up towards 1.9v, but there are a few folks who said they found greater stability by reducing CPU PLL voltage. I tried both, as my board gives me a range of 1.2v - 2.0v. I first tried taking the voltage up in increments of .025 or .05 volts. Going above 1.8v didn't do anything for me so I started decreasing the voltage in .05v increments. I found slightly better stability around the 1.6-1.7v range and even better stability around 1.25-1.45v. I left it at 1.25v as that seemed to give me consistently longer runs of Prime95 before stopping.

At this point I was still reading through this thread and browsing various X58 overclocking guides. I don't remember where I noticed it, but there was a post or guide saying that both CPU and PCIE Spread Spectrum should be disabled. Well, I must have missed this in my first pass as I still had both of them enabled. Some quick research confirmed that the overwhelming consensus is to turn these off (unless you have weird EM issues) so I did. With no other changes, I rebooted and found that I could still boot into Windows, but it would lock up within the first 2 minutes (even just idling). So I went back to CPU PLL voltage and started testing that again. Voltages around 1.7-1.8v would let me boot Windows and give me a few minutes before ultimately freezing again. I figured the spread spectrum settings must have been significant, so I decided to take most of the voltages back to stock and first figure out the best setting for the CPU PLL voltage. Taking the ICH, IOH, and DRAM voltages back to stock seemed to help. I could get Prime95 to run for 15-20 minutes now with the CPU PLL at 1.8v. So I started decreasing CPU PLL again and ended up back at 1.25v with Prime95 stable for ~30 minutes...so pretty much back to the level of stability I was at before discovering the spread spectrum settings.

Next thing I did was read up on the 3 types of Prime95 torture tests. I had been using blend, but decided to play around with the small & large FFT tests as well. I started with the small FFT tests and to my surprise the system is stable for hours on end. I tried running it overnight last night but I think Win 7 installed updates and restarted so I don't know how long it lasted. I let it run for 2 hours this morning before stopping it to try the large FFT test. This one yields pretty much the same results as the Blend test - stops after ~30 minutes. I played around with the ICH and IOH voltages to see if I could increase stability by lowering or raising them, but didn't see any notable improvements. At this point I was a bit stumped, since the small FFT results suggested that the CPU was stable and the errors in the large FFT and blend tests hint at a memory issue. I thought this was strange since the ram was operating below spec (1400 vs. rated 1600). I tried increasing the DRAM voltage all the way up to 1.65v but those resulted in windows freezing within 1-2 minutes. So I went the other direction and lowered it to 1.45v. Surprise, surprise! I was able to run the large FFT test for over an hour without it stopping. I've switched to the blend test now and it's been running for almost 2 hours now!

I've still got some work to do, but I wanted to share my findings. It sounds easy enough to bump all the voltages to their (safe) limits and work your way down, but there are so many variables in play and every CPU (and motherboard) is different...it's a lot of things to try. Who would have thought that lowering DRAM voltage would bring about such an increase in stability? That, and all the various settings I tried for the CPU PLL voltage (which seems to only garner a passing mention in most posts) make me think I might be playing around with this chip for a while. I'm not going for anything crazy - 4.2ghz stable would be great, although I'll quite happily take 4.4ghz =) I'm also still keeping my eye out for a cheap X5675 to try, just so I can test that 23x multiplier.

Sorry for the long post! Will report back when I have more.

 

[primetime]

i get the hole try everything approach your doing but 4.2 should be very easy to do, but you appear to be focused on 4.4-4.5ghz.....that where it gets difficult . Those rampages don't seem to need the the higher pll voltages from what i remember.....but it does help these p6t models....

cpu voltage 1.35
pll voltage default
qpi/dram 1.35
dram 1.65 or what ever yours is rated for
all spread spec disabled
dram speed 1531
qpi link 6893
overclock xmp
dram time all default
multi 22
speedstep disabled usually
c-states all on (doesn't hurt a thing)
blck 191

I would be amazed if this wasn't bullet proof stable right of the get go for you. higher clocks between 4.2-4.6 should just be a matter of slightly higher vcore but getting 4.4 or higher stable just isn't meant to be for every cpu....it either is or isn't really (most are not)
4.2 or below is pretty much guaranteed

actually its almost always easier to do with less voltage at lower blcks....since your board can lock in the 24 multi unlike mine go ahead and use it with a lower blck 175 to get 4.2 with even less voltage

 

[LigTasm]

Do any of you guys run mixed RAM setups? I'm waiting on a new kit of memory to come in for my board but I figured I would just play with what I have on hand and see how it runs, and what I have now is 2x4GB, 2x2GB, and 3x1GB. So I went ahead and put both 4GB on A1/B1, both 2GB on A2/B2 and two of the 1GB on A3/B3 for 14GB total. It seems to be running fine in triple-channel mode since the amounts are balanced, and I haven't noted any stability issues at 1600mhz.

 

[Napoleon]

I recently went from a Corsair TX 850w to a EVGA 750P2 and noticed my voltages show practically no variances now.

Have my played with your OC/voltaged since this upgrade? Have your temps changed? Stability changed?

 

[Eric1285]

i get the hole try everything approach your doing but 4.2 should be very easy to do, but you appear to be focused on 4.4-4.5ghz.....that where it gets difficult . Those rampages don't seem to need the the higher pll voltages from what i remember.....but it does help these p6t models....

cpu voltage 1.35
pll voltage default
qpi/dram 1.35
dram 1.65 or what ever yours is rated for
all spread spec disabled
dram speed 1531
qpi link 6893
overclock xmp
dram time all default
multi 22
speedstep disabled usually
c-states all on (doesn't hurt a thing)
blck 191

I would be amazed if this wasn't bullet proof stable right of the get go for you. higher clocks between 4.2-4.6 should just be a matter of slightly higher vcore but getting 4.4 or higher stable just isn't meant to be for every cpu....it either is or isn't really (most are not)
4.2 or below is pretty much guaranteed

actually its almost always easier to do with less voltage at lower blcks....since your board can lock in the 24 multi unlike mine go ahead and use it with a lower blck 175 to get 4.2 with even less voltage

Well, ran into another twist last night after my post. I decided to uninstall the Asus AISuite utility - I occasionally used this in the past to control fans and check temps. After rebooting I found that my vCore was much more stable. It looks like LLC now works in Auto mode (was previously showing huge vdroop). Unfortunately, after adjusting the vCore in the BIOS to get back to where I was before (~1.3v under load) I lost the improved stability I had found yesterday. Messing with the DRAM voltage and CPU PLL (the things I was playing with yesterday) didn't change much, so now I'm going to take a few shots in the dark to see if I can luck into a stable collection of settings. I'm also going to turn off the QPI LLC since it results in a huge jump under load...from 1.25v to 1.35v.

 

[Napoleon]

I tried auto without much success. While I can boot into windows at 4.3/4.4, it doesn't want to be stable.

Vcore: 1.20625 + 0.15 (reads as 1.344 in HWmonitor)
CPU PLL: 1.86
QPI/Vtt: 1.335
VDIMM: 1.66 (reads as 1.648 in HWmonitor)
QPI PLL: 1.16
IOH Core: 1.16
ICH I/O: 1.5

CPU Clock drive - 700mv
PCI Expres Clock Drive - 700mv

Ram multi: 10x
Uncore multi: 15x
QPI Link - 36x

Seems I'm stable at 4.2ghz, I'm happy where I'm at. I will try again when I get a new PSU/memory.

My CPU mark score is 10,769 which I'm pretty happy with, based on their comparison just a tad under core i7-6700k@4ghz - CPU marks core 11004 PassMark Software - Display Baseline ID# 618181

My FPS in CPU dominated games like Starcraft2 and Heroes of the Storm has improved vs the [email protected]. I think this will be a good setup when I get a new GPU

 

[primetime]

I tried auto without much success. While I can boot into windows at 4.3/4.4, it doesn't want to be stable.

Vcore: 1.20625 + 0.15 (reads as 1.344 in HWmonitor)
CPU PLL: 1.86
QPI/Vtt: 1.335
VDIMM: 1.66 (reads as 1.648 in HWmonitor)
QPI PLL: 1.16
IOH Core: 1.16
ICH I/O: 1.5

CPU Clock drive - 700mv
PCI Expres Clock Drive - 700mv

Ram multi: 10x
Uncore multi: 15x
QPI Link - 36x

Seems I'm stable at 4.2ghz, I'm happy where I'm at. I will try again when I get a new PSU/memory.

My CPU mark score is 10,769 which I'm pretty happy with, based on their comparison just a tad under core i7-6700k@4ghz - CPU marks core 11004 PassMark Software - Display Baseline ID# 618181

My FPS in CPU dominated games like Starcraft2 and Heroes of the Storm has improved vs the [email protected]. I think this will be a good setup when I get a new GPU

What blck and multi did you settle on? Same as your previous quad?

 

[LigTasm]

Well I got my X5670 in, a quick and dirty 200x21 with 1.250V got me up and running. Interesting IMC on this chip, it will not do 1600mhz DDR3 if the FSB is 133 but it does it just fine at 150/166/191/200. The i7-920 had no issues with 1600mhz at 133 FSB on this Rampage III Gene.

And of course now I realize I have Windows installed under IDE mode.

 

[Napoleon]

What blck and multi did you settle on? Same as your previous quad?

24x multi with "Turbo" on, 175 BCLK, I tried 24x180 but it wouldn't get stable at voltage levels I'm comfortable with. I decided to bump up the ram and it's doing fine so far at 8-8-8-24-2 1750mhz DDR3

Well I got my X5670 in, a quick and dirty 200x21 with 1.250V got me up and running. Interesting IMC on this chip, it will not do 1600mhz DDR3 if the FSB is 133 but it does it just fine at 150/166/191/200. The i7-920 had no issues with 1600mhz at 133 FSB on this Rampage III Gene.

And of course now I realize I have Windows installed under IDE mode.

nice! I'm quite jealous of your 200 blck and voltage. Looks like both your chip and board are gems! My i7-950 could handle the higher BCLK but not this X5670. If I would have known better (6 years ago), I would have spent an extra $50-$100 on my MOBO for a higher tier model; I recall researching the Gigabyte X58A-UD3R Rev.2 and thinking it was a solid board

 

[Rodall]

Have my played with your OC/voltaged since this upgrade? Have your temps changed? Stability changed?

I installed the new ps just before I got my X5670 so I don't know if there's any difference between OCing with the old ps and the new one. Like I said the voltages are more consistent now though, and I think that will help with OCing.

 

[Rodall]

Well I got my X5670 in, a quick and dirty 200x21 with 1.250V got me up and running. Interesting IMC on this chip, it will not do 1600mhz DDR3 if the FSB is 133 but it does it just fine at 150/166/191/200. The i7-920 had no issues with 1600mhz at 133 FSB on this Rampage III Gene.

And of course now I realize I have Windows installed under IDE mode.

You running win 10?

 

[LigTasm]

You running win 10?

No, windows 7 right now. The problem is even once I do the AHCI switch with regedit, it will only boot if I manually select the drive each time and Windows now sees the hard drives as USB hot swappable. Its a fresh install so no biggie to re-do it.

 

[primetime]

fuck win 7 and all those pesky windows updates...clean win 10 install takes all of 5 min

 

[Bill1024]

Do any of you guys run mixed RAM setups? I'm waiting on a new kit of memory to come in for my board but I figured I would just play with what I have on hand and see how it runs, and what I have now is 2x4GB, 2x2GB, and 3x1GB. So I went ahead and put both 4GB on A1/B1, both 2GB on A2/B2 and two of the 1GB on A3/B3 for 14GB total. It seems to be running fine in triple-channel mode since the amounts are balanced, and I haven't noted any stability issues at 1600mhz.

You can mix ram size. What happens is: The MB will take the smallest stick of ram and every other stick will run with it in triple channel.
The rest of the memory will be in single channel mode.

Say you have two 2gb and one 4gb, you will have 8gb total. 6gb will be in triple channel and the other 2gb will be in single channel.
Say you have two 4gb and one 2gb you will have 10gb total. 6gb will be in triple channel and the other 4gb will be in single channel.
The triple channel will be used first

 

[LigTasm]

You can mix ram size. What happens is: The MB will take the smallest stick of ram and every other stick will run with it in triple channel.
The rest of the memory will be in single channel mode.

Say you have two 2gb and one 4gb, you will have 8gb total. 6gb will be in triple channel and the other 2gb will be in single channel.
Say you have two 4gb and one 2gb you will have 10gb total. 6gb will be in triple channel and the other 4gb will be in single channel.
The triple channel will be used first

I know it can do it and it will run but stability was my concern. The X5670 will not run those 14GB stable like that while the i7-920 will. Seems like my i7 had a much stronger IMC.

 

[Rodall]

fuck win 7 and all those pesky windows updates...clean win 10 install takes all of 5 min

Win 10 is fine if you don't mind MS having access to all your data.

 

[{NG}Fidel]

I know it can do it and it will run but stability was my concern. The X5670 will not run those 14GB stable like that while the i7-920 will. Seems like my i7 had a much stronger IMC.

This is the same experience i had.

 

[MongGrel]

I know it can do it and it will run but stability was my concern. The X5670 will not run those 14GB stable like that while the i7-920 will. Seems like my i7 had a much stronger IMC.

I still just like using three of these with my P6T7 and a X5680 @ 4.5

G.SKILL Ripjaws Series 12GB (3 x 4GB) 240-Pin DDR3 SDRAM DDR3 1333 (PC3 10666) Desktop Memory Model F3-10666CL7T-12GBRH - Newegg.com

They scale well and are low latency.

Win 10 is fine if you don't mind MS having access to all your data.

No free I-Pads ?

Kidding a bit.

Lots of good tips over here too, Burpo is pretty much a X58/Xeon Expert I'd say.

Xeon L5639 Overclocking on X58 - Page 106 - AnandTech Forums