MSI Tech Support and the X58 QPI Mystery

[joker1973]

I sent MSI Tech another lengthy group of overly Technical Questions and it took them a week to come up with this,


"We are sorry to say due to the hardware design limitation, BIOS and any software can not detect the QPI voltage.
But we have test all the function at our lab, so it's safe for your CPU."

So who is going to be the first to test their theory?   By the way they haven't opened the doors on what their testing procedures consisted of, so I asked for details and we hopefully will get an answer in about a week.

 

[jakethesnake011]

Still no luck  Man I am really looking for something to at least give me some type of reading, I hope to get my Thermalright and new fans this weekend so I want to push this i7 up to 4.0 area from the 3.5 I have now and i am worried about that QPI and the relation it has to the other voltages,  I just want to save time and not guess at which voltage to use, anywhere in 1.2 to 1.35 range.  Joker you tole me before to keep QPI about 0.01V higher than my Vcore voltage, but what about when the Vcore gets past 1.35V.  Say I need to go to 1.40 on Vcore to get stable at 4.2ghz, should I keep the QPI at the 1.35?  It is not a 24/7 OC just a benchie to show myself the chip can get to 4.0 (4.2 even better  ).  SOme one tole me to keep QPI 0.5 V close to the RAM voltage, which makes sense, but want to confirm. 
Also testing.... hmmm, Wouldn't MSI be acting like we are now with the QPI reading if they "tested" the board, similar to what we are doing now with overclocking, I am sure they overclocked some i7 in their tests, but the QPI never occurred to be a variable?

 

[joker1973]

I mean that i have been setting the QPI at .16V and the Vcore at .15V offset. This is for the bios settings.  I have read that the max for Qpi is 1.35V .  I have not needed to go over that yet.  The most I have needed is .19V  .19+1.1= 1.29V.  You need to keep the QPI within .5 of the Vdimm .  I haven't had to push my ram at all to get 1600 (1.55V) so I haven't tested anything like I have read about like 1.9Vdimm and 1.4VTT (QPI).  At least on my Eclipse with my 920 I have found that a Vcore setting of .15V and QPI of .16V, Vdimm at 1.55V.  Bclk 190.  Rock hard stable 3.8GHz- 4.0GHz w/turbo.  Without turbo and BCLK 200 I get it stable with Vcore .18, QPI .19. Vdimm 1.60V, but not over night prime cause I was afraid of heat and don't really see a diff other that bragging rights.  Just remember what i said about CPU/QPI and 0.01V diff  had to do with the settings in MSI's bios design, not real time voltage.  The whole +- ADJ set-up design.  Hope this makes sense.

 

[joker1973]

Batch# 3837A732

 

[manuxtreme]

Have the same A batch as well and similarly enough got my QPI = +0.15 and VCore = +0.14 and BCLK = 190Mhz, stable with turbo on. I set my PLL = 1.8V

Also found that reducing the IOH and ICH voltages to a minimum does help reduce temps and has no adverse effect on stability. I have my ICH = 1.5V and IOH = 1.0V

Have my Corsair RAM at 1.5V running at 1520Mhz as well, mind you these rams are suppose to only run at 1333Mhz. Surprisingly stable with no errors

 

[HU16E]

Batch # 3839A635 under construction (I7 920/X58 Pro build for relative). Will update later with results. Sure hope that 1.12 is right! (CPU Temp. 35c, IOH 66c idle).