Intel Prescott Info
First of all I have to give Kyle props for being the first website I saw with any information posted on this Prescott voltage issue. It apparently is one of the new features Intel is using to keep these beasts cool. The processor senses when it's getting too hot and lowers the voltage to help cool itself off. The heat production of these processors is simply insane. If they were to keep the processor at 1.525v all of the time it would take way better cooling than any OEM would ever put in their systems. We all heard enough about Intel's struggles to go to the 90nm process and the delay the prescott and other products saw because of this. I don't believe they really fixed the issues they were having. I believe they used a bandaid and shipped it out anyway! Of course I can't prove this but it's just what it smells like from here.
Intel's response to Kyles article:
"The author has confused Dynamic VID with voltage variations due to the VR Load Line and current draw. They state: " We have been monitoring a Prescott 2.8E voltage under normal usage and have found the voltage to fluctuate greatly" which is normal and expected. The voltage must increase/decrease as the current draw decreases/increases. They were not monitoring the VID pins which would have held steady at their factory set level. "
The most useful document I found on Intels website was entitled Intel Pentium 4 90nm thermal and mechanical design guidelines these quote .gifs are from it unless otherwise noted.
Have you noticed when you check out the specs for these prescott CPU's why it says 1.25V-1.525V?
This is the most interesting thing I've found. While reading these white papers.
Prescott SHOULD see a boost in performance over the Northwood in things like Content Creation and business apps due to the added L2 cache, but it doesn't! So much for taking steps forward with a new processor release. Take a close look at all of the benchmarks and how the Prescott is alot slower than the Northwood in almost everything.
This diagram represents what I believe is a big reason why. Compare the normal clock to the resultant internal clock. This new internal clock is what happens when you have the thermal monitor enabled. I believe those flatlines are when the chip is cooling itself down. As the speed of the processor increases the heat will also. I am very interested in seeing how it affects performance to bypass this feature. I've emailed Wesley @ anandtech and Kyle at [H] about looking into this further. They're the benchmarking experts, not me.
I believe that with this technology enabled everybody who posted a review of this processor and tested in a cooler environment will have seen better results than anybody who was benchmarking in a hot room. When the CPU reaches a certain temperature it will halt it's calculations for a certain period of time. This period of time can be set by using ACPI MSR's in steps of 12.5% from 12.5% to 87.5%. The lowest setting is best for performance while the highest is best for staying cool and chasing bragging rights for highest clock speed. If this ACPI setting is set high it could halt it's operations for a considerable amount of time. This will hurt performance but makes you get insane overclocks. All of you people out there in a Mhz battle make sure you max this out ;)
Before I go to my next item I want to make sure you've read a couple things.
If it's so well designed then why does it need to do crap like this? Line 3 if you missed it.
From VRD paper:
Here is why it's so dangerous to change your voltage while running Dynamic VID. Basically if you try to overwrite the voltage setting they may not be able to control where the voltage stops when going either up or down. If it drops too low you'll crash and if it goes too high you'll fry.
Now, after this talk about how much this added "feature" sucks what can be done? Well, I can take things a step further and show you how to dump it! If we can disable the dynamic VID it will be safe to control the voltage again and you'll also have a better performing system.
How's this done?
It's done by setting a chipset register which is generally set in the BIOS. I would like to see some of the manufacturers step up to the plate and let me post a BETA BIOS for various boards that will let you disable this.
I will be posting further instructions as soon as I can. You'll need to use WPCredit to be able to modify these registers. Open WPCreditit ctl-D. If you see something that says ACPI P_BLK on that list then we're in good shape. We should be able to come up with something if we pool our thoughts on this one. Below I'm posting everything relevant on the 875 chipset. I haven't quite figured out the work around yet. I was hoping some of you could help me out here.
In review, I can tell you YES Intel is using Dynamic voltage. If you have proper cooling it should be safe to disable this feature. I will stand behind MSI when they said allowing voltage modifications with this dynamic voltage on is a really bad idea. I can tell you for sure that there is a way to disable this. Intel states in MANY places that dynamic voltage must be enabled, but I do not believe disabling it will cause any harm to your processor. What will happen is you will just have to keep an eye on what your voltage is set for. 1.5v may cause quite a bit of heat production. Any motherboard manufacturers out there, PLEASE get a hold of me to talk about maybe some kind of a very beta BIOS for some of us to play with.