Earlier this year when we were getting giddy for Ryzen, rumors began surfacing about how Intel might respond. At the time, we knew Ryzen would bring 6 and 8-core CPUs to the mainstream and it already looked as though Intel's quad-cores would be unable to compete. Granted, the 7600K and 7700K still look strong when gaming, but their time in the sun seems limited.
Along with preparing a series of Skylake-X processors, Intel's counter to Ryzen includes a Kaby Lake-X lineup consisting of the Core i5-7640X, which is basically a renamed 7600K, and the Core i7-7740X, a 7700K in disguise.
Leading up to their arrival, it was suggested that the 7640X might feature Hyper-Threading support, essentially making it a Core i7, while the 7740X would perhaps tout 6-cores. Those upgrades didn't seem particularly likely to me, and it also seemed improbable that Intel would release existing mainstream CPUs on their high-end desktop platform. After all, what would that achieve and how would this in any way be a counter to Ryzen?
It seemed far more probable that Intel would repackage its 7600K and 7700K Kaby Lake chips as Kaby-Lake X parts and call it a day, and that's precisely what the company has done.
On paper, you'll see that the Kaby Lake-X parts come with a mild factory overclock and sport a higher TDP while pricing remains the same. However, you can expect to spend more on a Kaby Lake-X build thanks to its pricier X299 platform.
It's hard to imagine why anyone would spend twice as much on an X299 board for these quad-core CPUs when they can get all the same features and capabilities in a Z270 board for half the price. Not only that, but half the features won't even work on the X299 boards when using a Kaby Lake-X CPU as they lack the required amount of PCIe lanes and only support dual-channel memory.
Setting that mess aside for now, let's move on to see whether the new Core i5-7640X and Core i7-7740X perform any faster than the familiar i5-7600K and i7-7700K...
Synthetic & Application Benchmarks
As mentioned, the Kaby Lake-X CPUs still only feature a dual-channel memory controller, unlike the Skylake-X parts such as the 7800X which boast quad-channel memory support. Unsurprisingly, when running DDR4-3200, we saw the same 31GB/s of memory bandwidth for the Kaby Lake-X parts as we did with the 7700K and 7600K.
The numbers in Cinebench R15 weren't unexpected. Despite their slight frequency advantage, the 7740X and 7640X are both only able to match the scores of their LGA1151 equivalents. At this point we largely know what we need to know, but I ran a heap of benchmarks anyway so we might as well check them out -- there are a few interesting numbers to note.
Take 7-Zip's figures for example: the 7740X managed to pull ahead of the 7700K yet the same can't be said for the 7640X versus the 7600K.
This time the 7740X was only able to match the 7700K in our Excel test while the 7640X falls behind the 7600K.
Moving to the PCMark 10 Essentials benchmark, the older mainstream Kaby Lake parts provide the best results here, narrowly outscoring newer Kaby Lake-X models -- a rather disappointing result for Intel's new high performance quad-core processors.
The 7700K and 7600K were again able to outscore the 7740X and 7640X, this time by a quite convincing margin in the productivity tests.
Something odd is happening here. The 7740X and 7640X were significantly slower for the photo editing test and I'm not sure why. These results don't really make sense but I can assure you that I double checked them.
The last PCMark 10 test simulated editing performance and here again the older Kaby Lake CPUs outpaced the 7740X and 7640X, albeit by slim margins.
Rendering and Encoding Benchmarks
The Corona benchmark also favors the 7700K and 7600K by a slim margin.
We saw the same trend in our Blender test, in which the 7700K and 7600K were clearly faster. I ran this test three times on each CPU after a full system reset and each time the LGA1151 CPUs came out on top.
This can also be seen in HandBrake where the 7700K and 7740X were much the same though the 7600K did pull ahead of the 7640X.
The last application we looked at was Premiere Pro CC and again the mainstream Kaby Lake parts delivered the best results.
We saw how the 7740X and 7640X struggled in the applications tests and here we see much the same in our first gaming benchmark using Battlefield 1. The 7700K was 5% faster than the 7740X when looking at the minimum frame rate and that's a decent margin, much the same was seen when comparing the Core i5 CPUs.
Again this time when testing with Mafia III we see that the mainstream Core i7 and i5 CPUs provide the best results.
Hitman's performance was similar between platforms and the Kaby Lake-X and Kaby Lake CPUs really traded blows here.
The same is true for Ashes of the Singularity as performance was much the same on either platform.
Conclusion: What Just Happened?
If you thought Kaby Lake-X looked pretty ordinary so far, hold on to your thermal paste. For no added performance (in fact, less performance more often than not), the Kaby Lake-X CPUs consumed between 15 and 20% more power under load in our Excel test. Perhaps worse still, X299 platform can be seen using almost 20% more power at idle.
We have even more bad news for would-be Kaby Lake-X buyers. Above we see that during extended periods of load the 7740X and 7640X consumed over 40% more power! WTF?
What Just Happened?
Going into this review, we suspected Kaby Lake-X wouldn't be anything too impressive, mostly because these quad-core CPUs have no place on a high-end desktop platform that they can't fully drive. The fact that they provide weaker performance and greater power consumption than the original Kaby Lake processors only adds insult to injury.
These performance issues could be blamed on the X299 board we used and motherboard manufacturers sure appear to be somewhat of a scapegoat for this sloppy release. That said, the Asrock X299 Taichi appeared flawless for me when testing the higher-end 7800X, 7820X and 7900X CPUs.
After recording such odd results I looked around to see what other reviewers found during their testing. Mixed results are the norm with Kaby Lake-X being slower for the most part across a range of hardware. However, the confusing power consumption figures I recorded haven't been seen by everyone.
For instance, ProClockers found the 7740X to consume 15% less power than the 7700K when comparing total system consumption and they used the Asus Prime X299-Deluxe motherboard for those wondering. Meanwhile, LAN OC found the 7700K and 7740X to consume the same amount of power under load, though at idle the 7740X was 16% hungrier, and they used the Asus Prime X299 Deluxe as well, so those two reviews alone are quite conflicted.
Whether we blame the motherboard makers, the rushed platform in general or perhaps even the different testing methods used by us reviewers, we all seem to agree that Kaby Lake-X is overall slower than Kaby Lake, and that's obviously not good news.
The Core i7-7740X and Core i5-7640X make absolutely no sense to me. I can't think of one good reason why they exist or what Intel's plan might have been here. To pull a quote from my Intel Core X vs. AMD Threadripper article published at the start of Computex:
Perhaps even more puzzling is the $240 Core i5-7740X which is another quad-core part, though it lacks Hyper-Threading support, so just 4 threads are on offer here. For $20 more than the Ryzen 5 1600, Intel is giving us a traditional Core i5. I won't bother reading out the specs, I'm pretty sure this is going to be another Core i3-7350K situation where I simply tell you guys, don't buy it.
Here we are five weeks later and I'm confident in suggesting that you shouldn't buy a Kaby Lake-X CPU. Glad to have that off my chest.
Pros: Kaby Lake-X offers no apparent advantages over Kaby Lake and actually performs worse in every category.
Cons: Slightly slower than Kaby Lake yet consumes considerably more power and costs more courtesy of the X299 motherboard.