Short Bytes: Intel's Devil's Canyonby Jarred Walton on August 4, 2014 11:50 AM EST
- Posted in
- Devil's Canyon
- Short Bytes
In early July, Intel released two new processors that go by the codename "Devil's Canyon" (DC). You can read our full review, but if you just want the executive summary it's pretty simple. Take the core processor design of Haswell (4th Generation Core Series) and modify the packaging to improve a few areas that will mostly be of interest to overclocking enthusiasts. Specifically, Intel has used a new Thermal Interface Material (NGPTIM) and they've added some capacitors to improve voltage stability/delivery. What does this mean to the layman? For most users, it means a drop in CPU core temperatures of around 10C under load, which can help with noise, cooling, and overclocking as well as improved performance.
There are only two DC CPUs at present, the i5-4690K and the i7-4790K. These are both K-series CPUs, so they're unlocked and target the overclocking enthusiast for a slight price premium, and they supersede the earlier i5-4670K and i7-4770K Haswell parts at similar prices ($5-$10 more). The only noteworthy specification changes: the TDP (Thermal Design Power) has been bumped from 84W to 88W, base/turbo clocks have been increased, and the DC CPUs are the first K-series parts to support VT-d and TSX-NI (mostly useful for workstation/server type environments). In the case of the i5-4690K, the jump in clock speeds is only 100MHz (base and turbo clocks), but the i7-4790K boasts a far more impressive 500MHz increase over the i7-4770K. The new CPUs should work with all 9-series and 8-series chipset motherboards, but 8-series boards will require a BIOS/firmware update so you'll want to verify compatibility before making a purchase.
What this boils down to is how much you want/need to increase CPU performance; you can see the quick summary of performance above. Without overclocking, the i5-4690K is functionally equivalent to the i5-4690 while costing $15 more, so it's almost solely targeted at overclockers. The i7-4790K on the other hand has the highest stock clocks of any CPU Intel has ever released; it's clocked 400Mhz higher than the i7-4790 while costing $25 more. That extra 400MHz translates into a 10-11% improvement in CPU clocks (9% overall performance increase on average), which is quite good all things considered – i7-4770K was only about 7% faster than i7-3770K on average in our testing. Gaming performance, if you're wondering, is largely bottlenecked by GPUs and so any of the CPUs listed above will result in very similar frame rates.
For those interested in overclocking, our internal testing suggests that the new sweet spot for maximum clocks on air/water is going to be around 4.7GHz, which is slightly improved from Haswell where there was more variability. Those chasing extreme overclocks who have already investigated other options for improving overclocking performance (e.g. delidding the CPU package to replace the TIM and integrated heatspreader) will likely find little to no difference between Haswell and DC. Unless you're after the top overclocking spot, DC isn't going to do much for enthusiasts who already own a fast CPU.
Users that aren't looking at overclocking will have an easier choice to make. If you're buying a new system, DC is of course worth considering, and the i7-4970K is the fastest consumer CPU to date. If you're thinking of upgrading, at the top of the product stack the i7-4790K still offers a 9% increase in average CPU performance over i7-4770K and 17% over i7-3770K, which is certainly enough to warrant consideration. But if you're running heavily threaded workloads, you're going to be better off waiting for Haswell-E and X99, which should bring 6-core and 8-core processor's to Intel's Extreme platform next month.
Post Your CommentPlease log in or sign up to comment.
View All Comments
Alexey291 - Tuesday, August 5, 2014 - linkIn half a year or so (when everyone on this forgot that this cpu even existed) there might be a review on it.
Or more likely yet - never.
Instead they will review some 4 year old phones and look at how colour accuracy has improved from 2010 to 2012... Or something :/
JarredWalton - Tuesday, August 5, 2014 - linkUmm... we already have a full review of Devil's Canyon; it's linked in the first paragraph.
homerdog - Monday, August 4, 2014 - linkDevil's Canyon CPUs also support VT-d and TSX, whereas the 4670K and 4770K do not. I'm just putting that out there.
JarredWalton - Monday, August 4, 2014 - linkThanks! I was actually in the process of updating the text to note this when I saw your comment. :-)
Kevin G - Monday, August 4, 2014 - link"The i7-4790K on the other hand has the highest stock clocks of any CPU Intel has ever released..."
Not entirely accurate. The Xeon X5698 has a stock speed of 4.4 Ghz. The only catch is that the X5698 was an OEM only chip due to the niche it was targeting.
JarredWalton - Monday, August 4, 2014 - linkTo be fair, it's still tied for the highest clock and is thus "the highest". Plus, dual-core Nehalem is a bit long in the tooth now.
Kevin G - Monday, August 4, 2014 - linkExcept 4.4 Ghz was that Xeon's base speed, not turbo.
JarredWalton - Monday, August 4, 2014 - linkYes, but there was no turbo on the X5698 AFAICT. It was a weird chip, regardless, and the fact that Intel doesn't even list it in ARK makes it even more of an oddity. So it was a six-core chip with all but two cores disabled, with the clock speed goosed to 4.4GHz and a codename of Everest, and it was largely ignored. Basically it was the last gasp of the "clock speed is king" era IMO. Now we're back up to 4.4GHz, but at roughly half the power draw as Everest.
TiGr1982 - Monday, August 4, 2014 - linkDon't be so picky - I guess,
97% of desktop customers are unaware of such an oddity from the past as X5698 :)
Only server guys and geeks may remember now about this SKU, I think :)
brucek2 - Friday, August 8, 2014 - linkOK, I'm stumped. I've been trying to puzzle out the nature of the "Short Bytes" series. There's only this + one more that comes up when I click on the tag for it so not a big sample size to extrapolate from so far. Where do these fit in vs. say reading just the conclusion of the original review? And why have it come so far after the original review? Maybe they're intended to provide updates once there's been more field experience to draw from (ie like the likely 4.7 max o/c provided here?)
I'm guessing there's an answer and its bothering me I can't figure it out...