Slashdot Mirror
Locked Intel Skylake CPUs Can Be Overclocked After BIOS Update (techspot.com)
jjslash writes: For a few years now, Intel CPU overclocking has been limited to more expensive Core i5 and Core i7 'K' processors. Skylake launched this year with the rumor of strong non-K processor overclocking through an adjustable base clock, but that never eventuated... until now. In overclocking circles it was rumored that BCLK (base clock) overclocking might become a possibility in Skylake processors, but it would be up to motherboard manufacturers to circumvent Intel's restrictions. Asrock, Asus and a few other motherboard manufacturers are said to be issuing a BIOS update soon that will unlock base clock overclocking on Z170 motherboards. TechSpot has got an early look, overclocking a locked Core i3-6100 to 4.7GHz on air cooling.
Isn't that possible without overclocking anyway? Just being curious.
Or you can spend a few more dollars and just buy a CPU that won't burn up and fail from overclocking.
Old days: the processor should run at 200MHz. You can push 215MHz, but you need to modify the vcore. The processor might be unstable. You might need additional cooling. The gates might just not switch correctly at that speed (miller capacitance...) without a vcore high enough to blow the chip. It's stamped 200MHz for a reason.
Modern times: that's a 4.7GHz processor clocked at 3.8GHz. You buy it, you turn it up to 4.7GHz, don't mess with anything else, it runs 60C at full load under stock configuration. That processor came underclocked out of the box.
Support my political activism on Patreon.
I wonder if a rootkit could make the motherboard burn up the CPU? Poof! La computer muerte.
Depends on how the power delivery on the motherboard is designed, really. You'd have to be able to turn up the supply rails well beyond the margin the silicon is designed and tested for, and I don't think any competent manufacturer would design SMBus-accessible on-board switching regulators to do that; plus tune relevant PLLs for a higher output frequency, plus turn off the CPU cooling fan, plus disable all thermal management that would otherwise shut everything down. Much of this would have to be done at system power-on before everything gets locked down. You'd probably have to infect the BIOS itself, and aren't they all signed now to prevent such a thing from happening?
Are YOU using the TOOL, or is the TOOL using YOU? Think about it!
Or you could just not overclock. These CPUs — including the non-K versions — automatically overclock themselves anyhow.
The temperature sensors would sense this and shut down. A long time ago some review website posted a video of them removing the heatsink from a P4 with a game running. Not much happened, the game froze up and the CPU shut down. They did the same thing to an AMD K7, it let off a wisp of smoke and died.
Only the State obtains its revenue by coercion. - Murray Rothbard
Is eventuated the new hip business buzzword?
Skylake launched this year with the rumor of strong non-K processor overclocking
Within the Skylake generation, these CPUs might be relatively strong. But from the first AnandTech reviews in Q3 this year, I gathered Skylake itself wasn't all that special. 5-7% improvements compared to Broadwell, including a couple of regressions in certain circumstances.
And we're still waiting for the equivalent of the Haswell with Iris Pro, for high end laptops, IIRC.
8 of 13 people found this answer helpful. Did you?
I would overclock my celeron from 2.0ghz up to 3.0ghz and turn the fan speed to full. It worked fine except anything over 10% lost the audio.
The guys with the ATV vehicles go out and try to find places where they can 'mud.' They don't need to travel there, they go for the adventure of the trip.
Back when my main machine sported a Pentium 75 processor, I was ready for an upgrade. A guy at work (the QA manager, actually) jumped at the chance to buy my old Pentium 75 CPU. It wasn't because he needed the processing power for anything in particular, he just said it was 'a good processor to overclock.' It was a good deal for me because it paid for part of the Pentium-MMX 133 processor I put in it's place.
I guess I could mount some 25 Watt 1% resistors on an aluminum plate and dissipate 40 Watts of power out of them. Whoo. Whee!
That's soooo last millenium!!! Ever since multi core CPUs became common, accompanied by SMP OSs supported by multi-threaded apps, there is nothing that overclocking can achieve that can't be more easily, reliably and inexpensively achieved by more cores.
I actually burnt a motherboard and CPU a few weeks ago during an overclocking experiment gone bad, after using a little bit too much voltage.
There was smoke coming out of the VRMs on the motherboard. The machine did shut down, and did not catch fire, thankfully.
This was an AMD FX-8350 on a Gigabyte GA-990FXA-UD3 . The chip was confirmed dead when I tried it in another identical motherboard.
So yes, if a rootkit could actually change those BIOS settings, it could potentially cause this.
-- Julien Pierre http://www.madbrain.com/blog
Expecting overclockers to actually spend money... you must be new here.
It might only have to access it and that can be done through the OS now. It might be possible to piggyback on that, no?
My first AMD was a K6-2 350 that I had OCed to 500 Mhz (a little less) and it was stable but it wouldn't be stable past that unless I put it in a fridge while it was running. I had it sitting in a mini-fridge for a while for my own amusement. Anyhow, maybe they trip automatically now and don't allow it? I could have cooked that thing for sure if I could have just kept it on and doing *something* because it was damned hot when it hit about 550 Mhz. It wouldn't even get into the OS all the way before you could smell it and it would shut off. If I'd kept doing that, over and over, I'm sure I could have cooked it - or if I'd stayed in DOS or something, maybe? (I never tried.)
So, maybe? They've got apps to change BIOS settings from inside the OS. I haven't really done much OCing since those days. Computers just got "fast enough." I'm not even sure why I keep buying new hardware. I don't really seem to see much of a difference in the things that I do. I could probably do just fine on an old dual core system with a few gigs of RAM though my documentaries might not like to play as well. It's not like I'm taxing the CPU when I edit a PHP file and upload it. I even off-load my VMs and connect via VNC.
Hmm... Yeah, I guess I too am sort of returning to the days of the dumb terminal.
"So long and thanks for all the fish."
but that never eventuated
Your casualizing new usagenesses for words has me wondering if you ended up straight onto the Internet without actually attending through K-8th...
It might only have to access it and that can be done through the OS now. It might be possible to piggyback on that, no?
No, actually, not really. There are registers that, after a certain point of the bring-up process for the silicon, are locked out from access in production parts. In testing the boot process is halted at a certain point in order to 'unlock' parts and allow access to various registers. Once an OS is booted, it's no longer possible to access some of those registers.
Are YOU using the TOOL, or is the TOOL using YOU? Think about it!
There is potentially another way to kill either a CPU or mobo components too, if you do it right, and the OS doesn't watch hard for it, and it doesn't require changes to BIOS/UEFI. If you can trigger the right execution pattern to rapidly cycle between no load, full load on all cores, full load on 1 core and triggering turbo clocking if that's available, you can cause voltage regulators etc to overheat
Yes, actually, really.
Most recent example was a pair of monitoring utilities killing S2011-3 CPUs on Asus X99 boards by setting Vcore to 1.95v when both were running simultaneously.
Hey, that's exactly the combo I have. What settings did you fry your motherboard with? I've been thinking about going in there and trying to twiddle the settings to mitigate the massive vdroop problem and maybe OC a little more to get a few more FPS, and I don't want to let the smoke out. My board is a 4.0.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
That's programming a VR that lives on the SMBus, not accessing locked-down PLL registers to overclock the crap out of something, so no, actually, really, that's only one of the elements I was talking about.
Are YOU using the TOOL, or is the TOOL using YOU? Think about it!
Or you can spend a few more dollars and just buy a CPU that won't burn up and fail from overclocking.
If a processor could be guaranteed to work at a higher frequency, the manufacturer would just label it with a higher frequency and sell it like that.
It might only have to access it and that can be done through the OS now. It might be possible to piggyback on that, no?
No, actually, not really. There are registers that, after a certain point of the bring-up process for the silicon, are locked out from access in production parts. In testing the boot process is halted at a certain point in order to 'unlock' parts and allow access to various registers. Once an OS is booted, it's no longer possible to access some of those registers.
Yes, actually, really. What do you even mean by "Once the OS is booted". What about everything from the first instruction fetched after the BIOS/UEFI?
Little known fact: Intel K series processors don't support VT-d (Aka IOMMU) which allows routing your external device's DMA into a virtual address space. This is critical for security from external hardware (without it, your GPU that runs user code has full read write access to kernel memory among other things for example), and also vitalization. I assume this restriction is to prevent you from using overclocked desktop servers to replaces overpriced server processors. Its great to have a way around it though: the option to overclock your desktop and have improved security and PCIe pass-through to a VM is fantastic.
That said, I don't buy Intel anymore. AMD doesn't randomly fuse of useful security features and has more honest business practices. I know their processors are worse at what I need than Intel's, but I want to support their work (they do neat things like HSA), and would rather not fund Intel's business practices. As someone with cash to burn, I try to take the moral choice.
I've poked at a few of them and I think they required a reboot. They don't make any for Linux - that I know of and own, so I can't really double check that.
"So long and thanks for all the fish."
There are numerous utilities designed specifically to tweak CPU settings while Windows is running. Enthusiast motherboards allow just about every option from the BIOS to be ajusted live, which is meant to facilitate fine-tuning by eliminating some of the reboot-tweak-test cycles in the search for a stable overclock.
If some random asshole app were to reprogram those values, you could most certainly cook a processor by setting the voltage absurdly high and disabling the thermal protection (another option on most OC bards). Run some heavy FPU loops across all threads and the chip will be dead within seconds.
The difficulty of course would be to detect and "support" a wide enough array of popular boards to render this attack effective, as I can only assume the interfaces are vastly different, even within the same brand.
-Billco, Fnarg.com
Sure, and that's the very definition of binning, but they must also account for demand. If Intel's process improvements yield a higher ratio of top-binned chips than the market is willing to buy, those chips will be locked and sold as the faster-selling SKU. Better to sell the thing and still make a few bucks, than have it rot in a warehouse with a $1000 price tag.
-Billco, Fnarg.com
Theoretically, sure. But back in the real world, every manufacturer is supply constrained at the top end. There's no down-binning doing on anywhere.