Author: Sdougal & Xyala
HWBOT: Hi 0.0! Could you please introduce yourself…
0.0: My real name is Alex, nicknames vary from forum to forum. It seems each new forum I joined, previous names were already taken so it got to the point where I would just make up a new name on the spot rather than trying the old ones. Some may know me as CPL0, Dufus, µcode just to name a few. I live in the heart of Thailand, in the suburbs where weather is usually in the 30’s, so not the best weather for overclocking. ;)
HWBOT: When did you discovered overclocking and how or why did you start?
0.0: The first CPU I overclocked was a mobile P8400. I had already seen these overclocked by programming the master PLL usually using software called SetFSB, so I knew they could be overclocked. The laptop had plenty of thermal headroom but unfortunately the PLL was not programmable so I ended up having to hardware mod it. P8400 overclocked: http://www.overclockzone.com/forums/showthread.php/583730-Overclocking-laptop-Core-2-Duo-P8400-Mobile-CPU
There were the usual criticisms that laptops cannot be overclocked and would burst into flames etc. It’s still working fine today after 7 years and ~17000 hours of use, and the performance boost was welcome. Of course as with any overclocking one needs to be sensible about it and work within the parameters of the hardware. Besides hardware, firmware can have a lot to do with the devices’ capabilities.
I started delving into firmware during the early days of the laptop mentioned above after a standard manufacturer’s BIOS update went wrong and left the laptop bricked. The manufacturer’s solution through on-line support on a Friday was to have the mainboard replaced at a cost of USD300. Since I wouldn’t be able to send it in until Monday, I tried seeing if there were other alternatives over the weekend. Long story short, there was a crisis recovery option built into the Firmware not mentioned by the manufacturer, and luckily after much trial and error I got it to work, saving time and $300. Since then I have had an interest in firmware.
HWBOT: You’ve been brought to in the spotlight by a recent GFP achievement, can you tell us a bit more about it?
0.0: The MSI GE60 doesn’t come with support for BCLK adjustment, that was a feature cut with ME Firmware so that was the first thing I addressed. Having the BCLK run at 99.8MHz instead of a nice round 100MHz just doesn’t look proper and there is no center spread spectrum to adjust it. The next problem was power throttling. As the CPU relies on a slope value for current, it’s easy enough to trick the CPU into thinking it’s using less power than it really is and prevent power throttling. However that is not an ideal solution, I don’t really want to stress the battery. Power limiting in laptops is ultimately driven by the embedded controller (EC) firmware, so instead I modified that to give control back to the MSR while on AC power. Battery power left as is.
Now according to Intel data sheets, the i7-4700MQ is supposed to support coarse BCLK overclocking straps. There doesn’t seem to be any documentation on how to configure this mode if it’s not available as a BIOS setting, so I had to glean what I could and again, and use trial and error. I’m probably still missing something but one of the steps was to run without a microcode update in case it had been blocked. While I still didn’t have the system running at 125MHz BCLK it was at this stage the processor was running at 3.8GHz, 2 bins above the maximum limited unlock bin setting of 36. From there it was straight forward enough to use the processor unlocked, with an option of up to 80 bins or so (~8GHz).
Running the i7-4600M at 4.8GHz settings generates a lot of power (about 90W) and that means a lot of heat. For the heatpipe in the GE60 to be able to control that heat long enough to generate enough scores in the XTU bench it was best to start at core temperature at 20 degrees Celsius. Since ambients are high and normal idle temps are in the 40’s, I had to hold the laptop in front of an aircon for a minute or two to get it down to 20 degrees. Once there, there’s enough thermal latency to absorb that heat for a few seconds without hitting thermal limits. Power limiting was configured to allow 5 seconds of up to 110W then limit to 45W. This way the CPU does not hit thermal limits as power is limited after 5 seconds and no longer needs assistance from the aircon. Not sure why Intel configured the XTU bench this way, instead of over the whole run.
HWBOT: What were the major difficulties you overcame in order to get the microcode to work?
0.0: We’re not trying to get the microcode update to work, just the opposite. No microcode update = unlocked processor. Once we are at that stage it’s a case of programming the voltages and ratios. Again no documentation for this, just some hints and trial and error to overcome that. Once unlocked settings have been applied then a microcode update can be run.
Two of the things the microcode update helps with is displaying the correct VID voltage (it reads half without it) and being able to use VCCIN override. The i7-4600M at 4.8GHz ran a lot better on this hardware when VCCIN override was set to 2.2V. It would seem ratio overclocking is required in this situation and perhaps that’s why the desktop chips never got those limited unlocked bins the two previous generations enjoyed.
Maybe the door was left open by Intel on the mobile chips to test the waters. Having a locked desktop chip unlocked would be huge compared to its mobile counterpart. Perhaps Intel would be happiest if it wasn’t found at all however 2 years from launch would be pretty good news for them I would imagine. Or maybe they just like to throw in a few “Easter Eggs” now and then lol, idk. The microcode is a very powerful thing. If anyone outside Intel managed to figure it out, well there’s no telling what could be done.
HWBOT: Would you call this a “hack” ?
0.0: If the definition of a hack is to change something to enable operation outside of it’s design, then yes.
HWBOT: What did you learn from it and what is the next step?
0.0: Whether mobile or desktop the chips are basically cut from the same silicon; the mobile dual core i7-4600M has a TDP of 37W pulling power of around 90W at the 4.8GHz settings. So it’s interesting to see the relationship between mobile and desktop that way. The binning process may be a lot clearer under these conditions too. I would expect the higher binned chips such as i7-4800MQ and i7-4900MQ to perform even better unlocked than the i7-4700MQ. The guys with limited clocks on their i3’s did spectacularly to get those results and I had to pull a much higher core clock to have a chance beating them. I personally don’t think it’s taken anything away from those results other than having an unfair advantage with an unlocked processor. If I wasn’t learning something new I’d probably not pursue it so much, part of the fun.
HWBOT: Do you expect XOC being brought to some mobile CPUs thanks to this?
0.0: It would be nice to see, especially the MX chips that go for somewhere around USD1000 a piece. How much of that cost is supply and demand vs performance? The mobile scene does bring some challenges though, while it’s easy enough to pick up a desktop mainboard with plenty of grunt for overclocking it’s not so easy for mobile CPU’s. While the mobile CPU’s come without an IHS they have fewer power connections. RAM may be another factor to take into account, those i7-4600M results for example were being run on a 1.35V SODIMM.
HWBOT: What would be your advice and recommendations to others who would like to follow in your footsteps with this sort of “hack”?
0.0: Share information when possible, otherwise too much time is wasted in “reinventing the wheel”. Usually I find it best to not over supply info, but give enough to put people on the right track. Generally that way people tend to learn and understand it better. User ice.cold did an awesome job unlocking his Haswell mobile chip and found out some extra useful things during the process that I was not aware of. It was a real pleasure to converse with him.
HWBOT: In Thailand, what are your aspirations? Who do you look-up to within your community?
0.0: No aspirations really, while overclocking is interesting and fun I do like a quiet life with my family 😄 For the worldwide community I’d have to pick Kevin AKA Unclewebb, who has unselfishly spent possibly thousands of hours of his time providing free ThrottleStop software and support. It is used widely in the laptop community to help provide increased performance and possibilities for overclocking. While laptops seem to be it’s main users it can also can be used on desktops. He’s also very happy to receive feedback and suggestions.
Thank you for your time!