All aboard the Skylake: How Intel stopped worrying and learned to love overclocking Nope, it’s not the next instalment of the James Bond franchise, but Intel’s eagerly awaited successor to the Broadwell platform. Skylake, like Broadwell, is built on a 14nm process but this time we get yet another new socket, as the new CPUs use the LGA1151 Socket. Intel Skylake unlocked Core i5 and i7 CPUs Ready for a CPU …
From memory with ASUS BIOS's, they have basic signature checking of updates, but it can be disabled / fooled.
So, it'd be interesting to see how easy this is to MITM. Could be quite simple if they're not using some form of cert's or other useful auth. That could lead to some pretty large scale compromises.
The DMI 3.0 interface between the CPU and the Z170 chipset has a bandwidth about the same as a PCIe 3.0 x4 link (just under 4GB/sec). If the high speed ports from the chipset are heavily used then they will be starved for bandwidth. This CPU and chipset combination is not suitable for more than a 2 way SLI or Crossfire configuration.
This isn't comparing like with like. These are supposedly 'high end' chips, but it's the initial high end, not workstation class.
People are comparing Haswell-E against Skylake. A direct comparison is Skylake vs Haswell.
In that case they are very similar - the processor lanes are the same. The chipset lanes are much better in Skylake, because Haswell used DMI 2!
If compared against X99 (Haswell-E), the processor lanes are better, because it's high end enthusiast/workstation. The chipset lanes are again linked by DMI 2.
I'm of the opinion that the work Intel is doing to promote power saving simply has the side-effect of allowing more overclocking. That and the fact that fewer customers than before will spend more money on a chip purely for a clock speed increase, which means there's less financial incentive for Intel to control the clock post-sale.
I believe Intel's concern about overclocking was over unscrupulous PC makers buying cheap CPUs, overclocking them and then selling them as higher specced models. This would not only hurt Intel's bottom line but could also annoy customers who have paid extra for a chip that Intel wouldn't replace if it failed due to overclocking.
That said, that was their argument about 15 years ago when there were still smaller companies building desktop computers. These days most people want laptops and the few desktops made all tend to come from Dell/HP/Lenovo. It's only the enthusiasts who really still build the big gaming rigs so I guess that's why Intel are relaxing things a bit. That said, they were doing Extreme Edition unlocked chips about 12 years ago so I think overclocking goes in and out of fashion at Intel.
Chip manufacturing Has always been a mad business the failure rates etc, but the real reason for Intel's hostility to O'Cing is not as you think, for instance take the Intel PII 200/333/450 where all actually the same chip just the lower speed versions where clock locked at the factory , it was just cheaper to make only one version and lock the speeds rather tool up for a range , the chip runs with damaged caches became the Celeron range.
Also People would O'C chips in the past without Water cooling or all the modern O'C designed kit which just didn't exist back then so people would be tampering with Stock kit not really designed for operating out with it's intended limits.
Yeah it is a bit different now , but another seller for the Chip manufacturers is Clocked chips never last as long .
When I bought my first gen i7-920 I never bothered OC'ing it. I took it out of the cupboard a few weeks back and bought a cheapo water cooler system and it's running quite happily on a smidge under 4Ghz and is now the happy owner of some kick-arse DDR3 OC'd memory - the whole thing is now my ESXi host with a VM container that represents the old OS that used to inhabit it.
Nothing to it - there are loads of guides out there to prevent you going too far wrong. Don't go mad (unless that's your thing) and you can get a lot more use out of older kit with OC'ing :D
...I've only ever had one PC setup where it overclocked by default (an old AMD Athlon setup, unlocked CPU and IIRC an Abit mobo) - TBH the board didn't really understand the CPU, which should have been 100MHz FSB x18, and set it to x20 without asking, which is about +11%. It never batted an eyelid for about 5yrs or so before I needed to save space and retired it in favour of laptops.
On the build before last I finally did the overclocking route just for the hell of it. [That and I could afford smoking it at that time.] There are whole sites to the topic. I searched on a few using my CPU and motherboard [i5-3570K, Asus Z77 Fatal1ty] and the first result was a step by step guide by a guy who does this with all the "enthusiast" gear and had no dings. Worth the effort. That CPU clocks along at 4.8 GHz, has 2400 MHz memory, and is my workstation, thus Quadro+Tesla. The icing on the cake is the OCZ Revo Drive.
I'm not a gamer. Heck aside from marathon sessions of Solitaire while waiting on one or more, usually much more, computers, I don't game at all. No multimedia. Nothing special except virtualisation wherever you look. [NB: Want to see an Amiga on steroids? Yeah, this is my emulation machine.]
So go forth safe in the knowledge that "The Truth is Out There."
I mostly build my own CAD workstations, and this is my history. All but one were air-cooled and stable, and all were overclocked on day 1 :
First PC - Pentium 1 100mhz / O'C-ed to 120
Second PC - Celeron 350mhz / O'C-ed to 450
Third: P3 850 - O'C-ed to 1 Ghz
Fourth: Intel Q6600 2.4Ghz / O'C-ed to 3.0Ghz (water cooling)
Current: Intel I7 3960 ??Ghz / O'C-ed to 4.4Ghz
Question is, what are the new chips good for?
This board and chipset really sounds like a beast.
I did have one system I overclocked, to very good effect, without CPU speedup at all. I had a K62-450 (450mhz), which I found under stress test of rebuilding firefox was not even stable at 475mhz. But, I could set it from stock 4.5x100mhz to 4x112mhz, and it was rock solid. The 12% RAM (and PCI, it was based off main system clock on this system) speedup must have knocked off a wait state or something because the benchmark speedup was closer to 25%.
I was really excited when Intel finally announced its extremely late to the party 14nm Broadwell chips earlier this year. Shortly afterwords I was thinking WTF... where are they? Broadwell came and went faster than a virgin foreigner in a Thai brothel. Sadly Broadwell, overall, was no better than the excellent Haswell 22nm chips. The long overdue reduction from 22nm to 14nm die process must have been problematic as Intel only released slower versions.
What's even worse are these shite Skylake CPUs since the 4% IPS increase can't make up for their slower than Haswell clock speeds. This is what happens when there's no competition. Sadly, underdog AMD simply hasn't had the financial resources to provide any real competition to Intel for over a decade and Intel is taking advantage of its monopoly position in the ancient X86 arena. To get the full story go check out Anandtech.com at: http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation. Their conclusion is don't even bother upgrading unless you're still running a Sandy Bridge rig or are dependent on Intel's ramped up built-in graphics capabilities.
AMD... PLEASE WAKE UP before it's too late.
Not bothered with it myself normally 10% is neither here nor there for me.
What I am intrigued with is - is there any intelligent feedback in the setups? Presumably now Intel are on board with overclocking they can rapidly work out why it fails and provide software tests to allow every last drop of umph out of a chip possible by testing the bits most likely to fail at varying temperatures etc.
I could be asking my grandmother if she knows how to suck eggs but every overclocker I've met just ramps it up until it fails and then backs off. And then wonders about subtle little things that crop up from time to time.
A custom overclocker testing program that call home would be good for Intel in the long run.
Feel free to downvote but I still maintain that LGA 1151 is a halfway house for moderate setups.
True enthusiast boards are LGA 2011 exclusively.
Just look:
PCI lanes on offer: 1151 - 16, 2011 - 40!
SATA 3 (6 Gbps): 1151 - 6, 2011 - 10!
Cores: 1151 - Max 4, 2011 - Max 8
L3 Cache: 1151 - Max 8MB, 2011 Max 20MB
So, unless I misread something, the LGA 1151 is what I'd call sporty city car and not exactly grand tourer. In other words, the 6700K is hardly a flagship, more of a capable middling.
My 2 cents:
386-16 o/c to 386-20 (with faster ram)
486-sx33 to 486-sx40
A few more in-between - all Intel
Q6600 to 3.0ghz+
i5 2500k to 4.3ghz
All on-air o/c from day one and totally stable.
My current i5 still does everything I want to do, and more, and it will be awhile before I upgrade it.
There was a time when an upgrade meant just a faster CPU, possibly more ram and a regular display card purchase.. As those days have gone and it now requires new CPU+MB+RAM and the inevitable display card, like many of my fellow gamers, my upgrades just don't happen as often - overclocking helps me hold on to the old stuff a lot longer.
That's a nonsense figure. I doubt there are 1.2 billion PC owners, let alone gamers.
That would be about 1 in 6 people being a PC gamer - which might be feasable (just) in rich first world countries, but is hardly a likely statistic when you consider the relative poverty which much of the world lives in.
Are they getting a payout from component manufacturers from forcing users to upgrade everything else when they buy a new Intel CPU, or is there a legitimate technical reason they keep switching sockets, that couldn't be achieved by any other means?
It's literally the only reason I switched to AMD. Okay, also from a bang for buck point as well, but that's becoming a bit of a bigger chasm these days with AMD's performance dropping enormously, whilst still sucking down all the power.
Most of my overclocking tended to be finding chips which were "Underclocked" versions sold cheaply at a lower FSB setting.
Dual Celeron 266 -> Dual 400Mhz (running NT with SMP HAL on a BP6)
Celeron (300A) 300MHz -> 450MHz
Celeron 700E 700MHz -> 933MHz
The xeons were a nice find, I got a Dell 490 and put in a pair of E5350 (2.67GHz) then clocked them to 3.33GHz, eight cores, and that was five or six years ago.
The harper xeons were also pretty good, as the 1333Mhz units are just 1600MHz and the L5408 (2.13GHz) can do 3.2GHz with nothing more than some tape and silver paint, I've got 3.6GHz from a pair of E5450's (stock coolers) and 4GHz from a pair of X5470's (on water)
For a budget system you can get a cheap E5420 (obviously a 771 chip), hack it to 775 and overclock on a cheap 1600 775 mobo and you have four cores @3GHz or try for 3.6GHz with a E5450 and these things are as cheap as chips (well, they would be wouldn't they?).
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
