Originally Posted by OCmember
In my personal experience they have. A guide I came across recently described the two like this:
"QPI pll- The pitcher's pitch! When you are reaching (within .5GHz) the physical limit of the QPI clock for your chip this is usually a culprit for stability. At a max of 1.5V (DON'T GO HIGHER) some stubborn chips and boards can reach high BCLK OCs. The IOH Core Voltage needs to be = to the QPI pll voltage for best results. The two voltages should be the same at stock .....
" Cited here.
I personally would not go above 1.25v with either of them.
I had time to have a look into this today and as far as I can see my Rampage II Gene BIOS only has the IOH Core Voltage option and doesn't have the QPI PLL Voltage option. If there's a way to change QPI PLL Voltage on RIIG please let me know.
I thought I'd have a go with IOH Core Voltage anyway even though they are supposed to be set the same and I can't find the other one.
So I first found the limit of the QPI at auto IOH core voltage (1.11341) by reducing my stable 200 BCLK and increasing the QPI multi by one in the following steps:
BCLK QPI RESULT
170 7498 Memtest starts
185 8159 Fails to boot
177 7808 Fails to boot
174 7674 Memtest starts
176 7763 Weird slowness
175 7718 Memtest starts
So QPI of 7718 (175 BCLK with 2nd QPI multi) gets as far as starting memtest but QPI of 7808 (177 BCLK with 2nd QPI multi) fails to boot every time (I tried a few times).
I then tried increasing the IOH core voltage:
BCLK QPI IOH Result
177 7808 1.20 Fails to boot
177 7808 1.25 Fails to boot
So, increasing IOH core voltage didn't seem to help get from booting QPI of 7718 to QPI of 7808.
I tried a few other things too but didn't find anything that could make the system boot with 7808 QPI. Now I'm writing this I realize I didn't try higher VTT. I'll give that a go at some point.
After that it occurred to me that my previous max BCLK had a QPI quite close to the QPI limit I found for the first time today so I thought I'd have another go at max BCLK with QPI slow mode which I hadn't used much before...
Trying for a new max BCLK using slow mode:
210 Memtest starts
220 Memtest starts
230 Memtest starts
240 Fails to boot
235 Fails to boot
233 Fails to boot
232 Memtest starts
So I managed to find a new max BCLK of 232 with slow mode which is good enough to boot.
Now I'm thinking that my previous 5GHz+ attempts may have been QPI limited too so I have another go at 5GHz+:
1 Core 1 Thread Slow Mode:
BCLK VCore CPU Multi CPU Freq Result
232 1.7V 22 5104 Memtest starts
232 1.7V 23 5336 Memtest starts Windows fails
232 1.7V 24 5568 Fails to boot
232 1.7V 20 4640 Gets to windows desktop
232 1.7V 22 5104 Gets to windows desktop
232 1.75V 23 5336 Gets to windows desktop CPU-Z fails to load
232 1.775V 23 5336 http://valid.x86.fr/8kw135
Now I'm not sure if that was my +2GHz OC or if I need to get +2GHz over the max turbo frequency so I continue...
232 1.775V 24 5568 Memtest fails
232 1.8V 24 5568 Memtest fails
232 1.825V 24 5568 Windows fails
232 1.9V 24 5568 Fails to boot
I try my stable overclock... Windows fails.
I try BIOS defaults... Windows fails.
I power off and on again... still more faults in Windows.
I power off the machine and take the metro to the computer shop but it's the last day before the spring festival here and it's impossible to get another Xeon because all the delivery people are on their way back to the countryside to see their families so I return thinking I'm going to have to install my i7 920 and suffer that for a couple of weeks.
On the way back I realize I'm using 1.375VTT and the 1.9V is >0.5V above. I wonder if this is what killed the CPU or if I simply melted the one core I was using.
But I try the machine when I get back and the Xeon is magically working again and will still do 5GHz+
I guess it was just waiting for the idiot user to go away