Hi I have posted elsewhere on this forum, I will put link on the end of this post.
I have been trying to OC my sys, and having trouble making 3.82GHz stable. My heat is getting on the high side too, around 80C max. at Bclock 191 (air cooling).
I have seen other ppls get to 4GHz (on air), and I can't even seem to get 3.8. I am posting my BIOS template below W/current settings, they cause P95 to err in just a few minutes max. I can't quite get my mem to run at full speed specs, but close, I have gotten it to run fine at 2:10 with Bclock 159 (mem 1590MHz) quite stable, with the settings shown. Reduction of the latencies at this speed causes errors or lockups - but I did not try them individually (yet) just all three at once.

In the (recent) past, I have ran Memtest86+ v4.0 all night and no errors. Even with higher clocks. Just a day ago I ran Prime95 (large) for hours, and LinX for 43 passes in 9:40 at Bclock 171, mem 2:8, and no errors. Max temp during this time was 76C on RealTemp for core1, other cores were a couple lower. The other settings were as shown below in my BIOS template.

My ultimate overclocking strategy:

I am desiring to OC to the max, then back off the Bclock (and therefore the memory speed as well) 10% to keep the system stable over years time as dust builds up as well (I will clean sometimes). And I also need to keep the core temps low, as I don't have AC and it normally gets into the 90's inside here. So ultimately I am looking to get my final OC below 80C, preferably 75 or less.

Limitations:

I don't want to make any changes that cause my core temps to go above 85C at any time. I don't want to turn on Virtualization, nor anything else that lengthens my boot time. I prefer VDroop enabled. EDIT: Not willing to reduce RAM to gain clock. And not spending any more $ on this rig.

Read below my BIOS template to see what I have already tried.


* * * EDIT 3/3/10 * * *:  This is now an obsolete BIOS template, see Post #38 on the next page to see the update.


Mother Board ( EVGA X58 SLI E758 )                                                       
Drivers ( ? )
Bios ( SZ2X )                                                                                                     
CPU (920 D0 E-Leet says "stepping: 5" )
CPU Cooler ( Zalman CNPS9900LED (air cooled), AS5 )
Memory (Corsair HX3X12G1600C9 SRAM set of 6 (2GB each stick) DDR3 1600 triple-channel memory 9-9-9-24 1.6V )
PSU ( Corsair HX1000W )
GPU ( Dual ASUS Radeon HD 5850 in crossfire )
Drivers ( 8.660.0.0, "Driver packaging version 8.66-090918a-088905E-Asus" )
Operating System (Windows 7 64b Home premium )


Frequency Control
CPU Clock Ratio ( 20X )                                                                                         
CPU Host Frequency (Mhz) ( 191 )                                                                      
MCH Strap ( auto )
CPU Uncore Frequency (Mhz) ( auto )                                                
CPU Clock Skew ( 0 ps )
Spread Spectrum ( Disabled )
PCIE Frequency (Mhz) ( 100 )

Memory Feature        
Memory Speed ( Standard )
Memory Control Setting ( Enabled )
Memory Frequency ( 1528Mhz / 2:8 )                                                                
Channel Interleave Setting ( 6 Way )
Rank Interleave Setting ( 4 Way )
Memory Low Gap ( Auto ) EDIT: 2G
tCL Setting ( 11 ) * * * EDIT * * *: Whoops it was actually 10, sorry
tRCD Setting ( 11 ) * * * EDIT * * *: Whoops it was actually 10, sorry
tRP Setting ( 11 ) * * * EDIT * * *: Whoops it was actually 10, sorry
tRAS Setting ( 31 ) * * * EDIT * * *: Whoops it was actually 30, sorry
tRFC Setting ( 88 )
Command Rate ( 2T )


Voltage Control
EVGA VDroop Control ( Without VDroop )
CPU VCore ( 1.275V )                                                                                                
CPU VTT Voltage ( +150 / 1.25V - E-Leet says 1.35V monitored, but 1.25V set )
CPU PLL VCore ( Auto / 1.8V )
DIMM Voltage ( 1.65 )
DIMM DQ Vref ( +0mV )
QPI PLL VCore ( 1.2V )
IOH VCore ( 1.2V )
IOH/ICH I/O Voltage ( Auto / 1.5V )
ICH VCore ( Auto / 1.1V )
PWM Frequency ( 800 )


CPU Feature
Intel SpeedStep ( Disabled )
Turbo Mode Function ( Enabled )
CxE Function ( Disabled )
Execute Disable Bit (Enabled )
Virtualization Technology ( Disabled )
Intel HT Technology ( Enabled )
Active Processor Cores ( All )
QPI Control Settings ( Enabled )
QPI Link Fast Mode ( Enabled )
QPI Frequency Selection ( 4.8GT/s )
OC Recovery ( Enabled )

PnP/PCI     PCI Express
Maximum Payload Size ( 256 )
 
Power Management Setup
ACPI Suspend Type ( Default / S1 (pos) )   
HPET Support ( Enabled )     
HPET Mode ( 64-bit ) 



What I have already tried:

I have tried adjusting Vcore up & down between 1.250 and 1.3125, and at Vcore 1.30 the core1 temp got up to 85 in just a few minutes.
I have tried lowering CPU PLL Vcore in 0.05 steps, and at 1.70 P95 app completely crashed.
I have tried adjusting QPI PLL Vcore between default 1.10 and 1.275V in 0.025V increments - I also noted that this did not seem to make any difference in core temps.
I tried changing the memory to 2:6, no difference. This proves it is not a memory-related issue.

The instability:

Prime95 workers error out after just a few minutes or less. I have seen about three different types of errors, sometimes "Fatal error", often "Illegal SMOUT." And when I changed to some of the extreme settings listed above, sometimes Windows said that Prime95 has stopped working and must close.


Any ideas please?


My previous troubles thread:
First SATA failures, now dead replacement, months of anguish

You need better CPU cooling it looks like to me.  If at only 1.3v your getting 85c, then your going to have major difficulties trying to get higher clocks stable under load.

I'd like to just get this speed stable first!
Prove to me I can go faster before I spend money on something that won't help me.

The Zalman is not going to cut it.  Part of your instability may very well be the cooler.  I was using the V10 cooler (massive massive cooler) before my Megahalems and could only get the CPU to 4.1 stable.  I was pushing 85C as well.  I then put in the Megahalems and was able to get it to 4.3 with no issues and under 85C with a higher Vcore. 

You also have to consider you may have received a not so great overclocking chip.  Most D0s overclock very well. 

Also have you tried running only 6 gigs of ram?  You may want to see if that helps with stability as well.

Dude. 83c is equal to 180 degrees. That would have fried my Q6600 in no time. Thankfully the i7 is stronger. If you buy the Megahalems and added  2  130cfm 120mm fans (push/pull), it would drop your temps at least 10 degrees. So instead of hitting 85c you will be more around 72c - 75c.

In your template your ram is rated 1600mhz 9-9-9-24 and you have it running 1528 11-11-11-31. Don't loosen the timings so much, That can cause instability as well, because now your ram is highly underclocked but still using the spec voltage, and we all know that too high Vdimm can also be unstable. When I lower my ram below spec, I can then lower the Voltage below spec.

You can leave all the IOH, ICH/IOH, and ICH voltages on Auto. Since you are hitting near 85c you are not going to be able to raise the vcore anymore. I would back down the overclock to around 3.8ghz until you get yourself a better cooler. Take a look at my guide if you haven't already. http://forums.evga.com/tm.aspx?m=81651

Also start using LinX instead of Prime95 when doing the first tests. Use prime95 later after you have passed 20 or so tests with linX.

Something like this would be a nice overclock with low temps, and your memory will be at spec.

Frequency Control
CPU Clock Ratio ( 19X )
CPU Host Frequency (Mhz) ( 201 )
MCH Strap ( Auto )
CPU Uncore Frequency (Mhz) ( 3216Mhz { 20X } )
CPU Clock Skew ( 0 ps )
Spread Spectrum ( Disabled )
PCIE Frequency (Mhz) ( 100 )


Memory Feature        
Memory Speed ( Standard )
Memory Control Setting ( Enabled )
Memory Frequency ( 1067Mhz / 2:8 )
Channel Interleave Setting ( 6 Way )
Rank Interleave Setting ( 4 Way )
Memory Low Gap ( Auto )
tCL Setting ( 9 )
tRCD Setting ( 9 )
tRP Setting ( 9 )
tRAS Setting ( 24 )
tRFC Setting ( Auto )
Command Rate ( Auto )


Voltage Control
EVGA VDroop Control ( Without VDroop )
CPU VCore ( 1.25 )
CPU VTT Voltage ( +150 )
CPU PLL VCore (1.8)
DIMM Voltage ( 1.60 )
QPI PLL VCore ( 1.2 )
IOH VCore ( auto )
IOH/ICH I/O Voltage ( auto )
ICH VCore ( auto )

CPU Feature
Intel SpeedStep ( Disabled )
Turbo Mode Function ( Disabled )
CxE Function ( Disabled )
Execute Disable Bit ( Disabled )
Virtualization Technology ( Disabled )
Intel HT Technology ( Enabled )
Active Processor Cores ( All )
QPI Control Settings ( Enabled )
QPI Link Fast Mode ( Enabled )
QPI Frequency Selection ( 4.8gt )
OC Recovery ( Enabled )   

you may of needed to have it set to 16 or 20 with/depending the timings / freq. you were at in your BIOS. Get your RAM running 100% spec at 1.6v even if it involves lowering your clock to do so (as you said you were willing to do in your PM). RAM stable first, rest of OC second.

Then find your highest clock you can get with the RAM at spec, and save until you can get your CPU temps lower. The difference between 3.6 at the lowest or 3.8 if your cooler can handle it will not be noticeable to you most likely unless you are a benchmark freak.

You can play with the tRFC as well if auto doesn't want to be stable. Play around between 80 and 100. 12 GB of RAM is a difficult beast for stability (mainly 4.0 and over tho) so we should be able to find a happy medium somewhere with the clock you can do with your current cooler.

Trying to get stock RAM settings to work, trying tRFC 100 right now, 74 nor 88 worked. LinX err.

I just had an anomaly occur about three hours ago: The icons on the desktop and on the taskbar for RealTemp and LinX screwed up. Now they are the default Windows programs Icons. At first I thought it was just a result of OC, and one of these times I reboot it would be OK. But after a half dozen reboots it did not fix. Then I went into RealTemp properties, and chose "change icon" and it was there, correct and everything. So I pressed change file and navigated to the file, and chose the icon again, pressed OK twice, and it did not fix. I tried Rclick on desktop and "refresh." And when I tried the change icon button on LinX, it said there are no icons in that file (the LinX 64b exe). It used to have such a nice "check mark" icon. Now it's just gone, and no other file in that directory seems to have it. What the????


And just now my NZXT "Sentry LX Aluminum" fan control quit working right (I use it for fan speed and temp readouts only, it does not control my fans). Every second or so one fan speed comes on, and winks out, and everything else is blank, except for a few LED segments that intermittently work. Damn I have bad luck. I have not troubleshot this yet.


And nope, tRFC 100 doesn't work either. Can't get stock RAM timings to work at Bclock 189, LinX err. I had VDIMM set to 1.60V. But it works at 10-10-10-88-2T.



Update EDIT: After power cycle my fan controller was fine.

You're trying to run 1.65V RAM at 1.6V...

You're using a CNPS ...

Relying on E-LEET for voltages..

So many things wrong with this picture.

XXChronicXX

You should be able to get it stable if you raise the vdimm a little. did you try 1.63 or 1.65?

I was just doing what was suggested in Post #9, my OP setup was 1.65.

XXChronicXX

Also try 9-10-10-24 or 9-10-9-24

Yeah I will try that when I get everything else tuned. For now I want to be sure that my RAM timings are not screwing with my OC results. But I do appreciate your contributions here. So, later with that.


MJinZ: That's OK, nobody is making you try to help me. Inoxim-NorCal and XXChronicXX are helping me. Thanx anyway.

Inoxim-NorCal

Get your RAM running 100% spec at 1.6v even if it involves lowering your clock to do so (as you said you were willing to do in your PM). RAM stable first, rest of OC second.

Then find your highest clock you can get with the RAM at spec, and save until you can get your CPU temps lower.

XXChronicXX

Oh, I thought you already had everything else tuned. You shouldn't be messing with the ram until you know that everything else is correct. Thats a big no no.

Can I pretend to be soo confused?

Anyway, I am now running LinX to verify my setting PCIe to 101 has not caused a prob, mem timings 10-10-10-88-2T 1.60V. So far 4 good passes. OK 5 good passes, I'm now trying lowering CPU Vcore and IOH.

OK now I have proven stable W/CPU Vcore down to 1.2625V and IOH down to 1.125V (@ Bclock 189). But drop either of these V even one notch and LinX errors. So I am going to go up one notch with both, and try QPI PLL increase to allow me to increase uncore multiplier. So my CPU Vcore will be 1.26875V and IOH will be 1.15V.

OK I have tested and found out a lot more of my limitations and a couple of possibilities I am going toward.

1. At 189 Bclock I cannot go down below 1.125V IOH Vcore.

2. By reducing Bclock to 187 I can run uncore 18x.

3. At 187 and 18x I have reduced IOH Vcore to 1.10V (default), and CPU VTT to 1.125V ("+25mV") stable, at least with LinX and short runs W/P95 (requires more testing to be sure). RealTemp peak temps at these settings were down to 79,77,75.

Dropping VTT resulted in at least a 10% increase in LinX speed! (Exactly how much is questionable as I was running AV and browsing during testing.) I now see >700 speeds on LinX, two days ago and previous they were all high nine-hundreds.

Oh, I just had a white-screen lockup, the type that I get when my uncore is too high. I'm bringing Bclock down to 185 and testing again.

Then I will try bringing CPU PLL Vcore down some, target = 1.65V.

Which do you think is better: Bclock 187 & uncore 17x; or Bclock 185 & uncore 18x?

yea no need to up the uncore higher than x16 if you are using 2:8. Its not going to give you any better performance in games or apps.