Do you use a Rampage Formula (a real one, not a flashed Maximus) with an Intel 45nm CPU? If so, this thread applies to you! Do not vote in the poll if you do not have a Rampage Formula and a 45nm CPU!
Ok, I'm sure you all know what Vdroop is. You know, that nasty little thing that causes your CPU's voltage to significantly drop when under load? I'm sure you all also know of the various motherboard modifications that people have devised to counter the effects of Vdroop, most notably the pencil mod. Some motherboards even have an option in their BIOS to disable Vdroop, the Rampage Formula is one such motherboard. However, it has a little known problem with its method of disabling Vdroop:
If you use it with a 45nm CPU, your system will be fully prime stable, but it will crash and BSOD under absolutely no load.
AnandTech was the first source to draw somewhat widespread attention to this issue. A quote from their writeup on the subject:
Here is a link to the whole writeup:Originally Posted by AnandTech
http://www.anandtech.com/cpuchipsets...spx?i=3184&p=5
When I first read the article, I disregarded it, and used LLC on my Rampage Formula with a Q9450. Things were great, fully prime stable, everything was going well. That is of course, everything but the random crashes and BSODs while the system was under absolutely zero load. It took me a while to figure out what was causing the crashes, as I was hoping it was not LLC, because it let me use a significantly lower voltage.
If you read deeply into that AnandTech writeup, they make it sound like Vdroop control is not possible to utilize on 45nm CPUs and remain stable due to technical reasons (even though their justifications are way off base). I was starting to believe that it actually might be true, until I found this:
http://www.xtremesystems.org/forums/...05#post3021605
Apparently, the LLC feature on the Asus P5E works perfectly with 45nm CPUs, but not on the Rampage Formula. Which is just as I thought: It is possible. I did some closer reading of the AnandTech article, and it says that the instability is caused by a sort of "reverse Vdroop", or in other words, the idle voltage is lower than the load voltage.
However, I have done some testing and it appears that the delta between idle and load is much smaller with LLC enabled than without, going by the CPU-z voltage readings (inaccurate, I know). With LLC enabled, I set voltage to ~1.36750 in the BIOS which resulted in an idle reading in CPU-z of 1.352v, and a load reading of 1.360v. With LLC disabled, I had to increase the BIOS voltage to 1.40625v to remain stable. The idle voltage reading in CPU-z was 1.400v, and the load is 1.352v, the same as the idle voltage with LLC. So, with apparently the same stable load voltage without LLC, the CPU is unstable with LLC at idle. I don't know why this is, but just to make sure, I set the BIOS voltage to 1.38v with LLC and the crashing/BSODs remained at idle. But to illustrate, the voltage delta with LLC is .008v (hardly anything), and .048v without (a lot), as well as the deltas being reversed.
Clearly, there is something flawed with the implementation of LLC on the Rampage Formula, and AnandTech is wrong about it not being compatible with 45nm CPU, regardless of their fancy looking graphs. Because now the fact stands that LLC works perfectly on 45nm CPUs with the Asus P5E, which is supposed to be an inferior motherboard.
We need to find a way to jointly contact Asus regarding this issue and have them work on it in a BIOS release. I have already sent two tickets regarding this issue and have not received any response. If many people do so, maybe they will do something.
Anyway, those with the Rampage Formula and a 45nm CPU, please test with LLC on your overclock; does the system crash when idle and remain stable under load? Please vote in the poll.
Bookmarks