o keep in mind. I've been trying to rework this thread, and with the formatting problems persisting, I've ported over to a new thre

dennisburke

Just an update on how that 3.6GHz is going. Been running 3.6GHz with 1600MHz for almost two days and all my previous problems seemed to have disappeared. Not really sure what the problem was, but I'm starting to develope a theory. I ended up sticking with a VCore of 1.250V and a VTT of 1.150V and since yesterday I can restart, cold boot, and warm boot. I really don't believe the problem with hanging on restart has anything to do with voltages. At one point I decided to pull my computer out of it's cubby hole and took the side cover off to look at the error code and to my surprise it was FF D4. For some reason I thought I was immune to D4, because people refered to it as a loop, and in my mind I imagined the computer continuously trying to restart.
 
Anyway, I was getting notified in the BIOS that I had a Watchdog Timeout and some here have reported it as a memory timing problem, and it probably is, but not because of faulty memory. The bios is probably as good a place to fix the problem, but I'm starting to think it could also be a Windows 7 problem, and maybe even the federal governments fault. I think it all ties into the ACPI,  the fed calls the tune, and Intel, Microsoft, and the bios are all invited to the dance. In short I think D4 is a timing and switching issue...but I'm usually wrong.
 
So anyway, I updated my guide and included the 3.6GHz LinX test. Below is something I clipped out of a recent i5-xxx review from Anandtech, and a short run of Prime to test the 3.6 out.
 
 

 
 

cult to find a clear definition of what the PLL voltage actually does. Apparantly it is set to 1.80V by default.  

dennisburke

raja1 

Running 200 BCLK with a VTT south of 1.20V is very low. This will guive rise to erratic behaviour during the post process at times. The D4-FF loop is to do with read/write levelling primarily, which is reliant on optimum levels of VTT and VDIMM. Error generation on the edge of stability is random and not always repeatable in exactly the same way. Further, if things are that close to instability, thermal factors come into play which is why cold boot issues manifest at times.

One other thing, I see people advocating Intel maximum voltage thresholds with overclocked bus frequencies, remember sticking to Intel's stated VID applies to stock operatring frequencies only. Increase the frequency and all of those max VID tables go out of the window. Play safe if you want to, but don't think you're within Intel's guideleines if you breach the stock operating frequencies of your processor. All of the figures shown in the Intel white papers are related to power - and this is derived from operating frequency. 

Out of all of the specifications, two are probably wise to observe regardless of operating frequency, the first is the VTT to VDIMM ratio (Intel may have changed things under the hood to increase resilience - this threshold is to do with a maximum drain to source DC delta), you want to keep VDIMM and VTT within 0.40V at all times (the absolute maximum delta will vary from processor to processor and will also be somewhat affected by cooling).  
The other rail that can damage a processor rather quickly if over-volted too high is PLL voltage. With PLL on 24/7 OC's, if you find you've reached a bus speed that can only be stabilized by increase voltage to CPU PLL, you'd be wise to peg the bus speed back to a point where PLL does not require a ramp in voltage. Of course, the risks may not be severe in either instance, but some things are more susceptible to damage than others and if you are playing the 24/7 game and have fear of longevity you'll want to keep certain things within bounds. The main thing is to use common sense.

Wow, this is a lot of food for thought. I'm certainly not lacking in common sense, but when the rules are not clearly defined for breaching the stock operating frequencies of my i5, I'm trying my best to learn how to read between the lines. I remember when I first discovered I needed to overclock my i5 to use the 1600MHz memory I purchased at the same time, it was/is a great learning experience, although challanging at times for someone learning about overclocking. The reason I created the guide is to help those green to overclocking and to share my experiences. I want this guide to be done right, so I will edit to make it so. If you don't mind, I would like to ask a couple of questions so that I might better understand what you addressed above.
 
 
___"Running 200 BCLK with a VTT south of 1.20V is very low."
 
So I guess what your saying here is that by increasing the frequency of the core clock will increase the frequency output of the memory controller, and as such the VTT needs to be increased. So in essence, running a 3.6GHz overclock using (x18) x 200MHz will require more VTT than (x20) x 180MHz, irregardless of VCore...correct? 
 
___As you have probably noticed there has been considerable discussion and frustration with the D4--FF loop in this forum. What you have to say makes sense to me, but many here believe the problem lies in the BIOS. Could the BIOS in some way be a factor?
 
___I understand what you have to say about the VID, which has me wondering about voltage and thermals. If a system is stable, and the thermals are in the normal range, then the voltages used will have little to no impact on the life of a processor...just a theory...right?
 
___"...VTT to VDIMM ratio...maximum drain to source DC delta..."   [Delta_"...variation of a variable..."_Oxford]
 
This one has had me and some others scratching our heads. In my BIOS I set the VDIMM to what the memory manufacture specifies. I'm assuming it is what powers the DIMM slots and once set becomes a fixed value. VTT powers the memory controller and once set also becomes a fixed value. So what your referring to is some type of voltage regulator that factors in both VDIMM and VTT?
 
When my system is running at default, my memory is running at 1333MHz @ 1.50V and my VTT is running at 1.05V.
Is it safe for me to assume my delta is in the range of 0.45V? I have read overclocking guides for the i7 on the X58 that state 0.50V is the range. The normal overclock I run 24/7 is 3.2GHz, my VDIMM is 1.62V and my VTT is 1.10V, that's a 0.52V difference. Stabilty and thermals are great.
 
I'm using some pretty plain jane memory: 1600MHz @ 1.65V @ 9-9-9-24. Corsair says my memory is rated at 1.65V, but does that actually mean it should be run at 1.65V @ 1600MHz. If I were to run my memory at 1.55V @ 1600MHz would that be a problem?
 
___PLL voltage
 
It was difficult to find a clear definition of what the PLL voltage actually does. Apparantly it is set to 1.80V by default.
 
I clipped this out of a bit-tech overclocking review of the i5-570
 

 
I clipped this out of Intel's Datasheet