Introduction, Features Chipset
Most of the time, component makers such as Intel, Nvidia, Samsung, and AMD make predictable moves when it comes to product launches: a new product generation a year or so on from the last one, with a modest-to-moderate uptick in performance, and an upticked integer shoved somewhere in the product name.
Sometimes, a product comes along that’s a impressive enough to shake up the business, like AMD’s recent Ryzen CPU lineup. (We particularly like the six-core Ryzen 5 1600X.) Ryzen is giving Intel some long-absent serious competition. A similar phenomenon happened in 2016 with Nvidia’s GeForce GTX 10-series (“Pascal”) graphics cards, which, on the high end, still haven’t seen any real competition from AMD, more than a year on from their launch.
But sometimes a big company makes a move that just leaves us reviewers and editors scratching our heads. A prime example of the latter for 2017 just might be Intel’s decision to effectively launch two generations of silicon on its new Core X-Series enthusiast platform at the same time.
On the one hand, it’s not that surprising, given that both the 6th Generation (“Skylake”) and 7th Generation (“Kaby Lake”) architectures are much the same in terms of design and features. The later chips mostly just get slightly higher clocks, plus some hardware to allow playback of protected 4K content from the likes of Netflix and Amazon (more on that later).
The higher-end chips in the Core X-Series, unintuitively, use the older Skylake architecture; the lower-end ones, Kaby Lake. And the Skylake Core i9 chips certainly have potential to impress. The $1,000 Intel Core i9-7900X we tested, for example, sports 10 cores and 20 available processing threads, and thumps the company’s 2016-era 10-core chip, the Core i7-6950X Extreme EditionRyzen Threadripper enthusiast processors some serious competition up in serious-money-CPU Land. We love to visit that place, but we can’t afford to live there. (Ryzen Threadripper chips will top out, at least initially, at 16 cores and come in at $799 and $999 for the first two chips.)
But then…there are the Kaby Lake Core X-Series chips.
We’ve already taken a deep look at the four-core, eight-thread Core i7-7740X. And while it’s a fine performer, as far as fast-clocked four-core parts go, it doesn’t really break any new performance ground versus the Intel Core i7-7700KMSI X299 SLI Plus, rides right on that $250 line.) The thing is, also, the Core i7-7740X doesn’t pack the extra PCI Express (PCIe) lanes on the chip that the higher-end Core X-Series processors do. You’re stuck at 16 lanes with the Kaby Lake Core X chips, which is the same as you’ll find on “regular” Kaby Lake chips on Intel’s mainstream platform, like the Core i7-7700K, using socket LGA 1151.
Also, because Kaby Lake X CPUs have far fewer lanes than the higher-end, Skylake-based Core i7 and Core i9 CPUs (which have either 28 or 44 lanes, depending on the Skylake X chip in question), some motherboard features (such as extra M.2 slots, SATA ports, and graphics-card slots) can get switched off or run with reduced bandwidth if you drop in a Core i-7740X as opposed to, say, a Core i9-7900X. In other words, if you opt for a Kaby Lake X Core i7, you get the extra expense that comes with the X299 platform, but mainstream performance and a motherboard with hobbled features.
That brings us to the Core i5-7640X. This four-core Kaby Lake chip has one thing going for it compared to all the other chips on the Core X platform: its $249 price. That’s a quarter the price of the Core i9 we mentioned above, and about $100 less than the current going rate for its Kaby Lake Core X counterpart, the Core i7-7740X. This Core i5 is by far the least-expensive chip offered up for Intel’s enthusiast-grade “Extreme Edition” CPU platforms, going back at least several years.
But all the issues we already mentioned about the Core i7-7740X also apply to the Core i5-7640X. You get 16 lanes of PCIe and performance that’s reasonably improved compared to chips like the previous-generation Skylake Core i5-6600K. But performance should be about on par with current-generation Kaby Lake offerings in the same vein, such as the Core i5-7600K. Performance is similar to these existing Core i5 chips because the Core i5-7640X lacks the thread-doubling Hyper-Threading tech that’s part of all desktop Core i7 chips, like the previously mentioned Core i7-7740X.
Is there a case to be made for a mainstream Core i5-level CPU that works on a high-end platform like Intel’s Core X-Series? Perhaps…if you’re somewhat cash-strapped today but want to build a performance-hound PC that’s highly upgradable for when your component-budget dreams come true down the road. To answer the question in full, though, we need to investigate the Core X-Series platform in more detail, and take a look at what else is available these days when it comes to budget-friendly (though powerful) computing platforms.
But one thing is clear right away: The Core i5-7640X is a bit like a Honda Civic engine that you can drop into a supercar frame. To be sure, people have done stranger things, in both the computing and automotive worlds. But at best, this chip is going to appeal to a small subset of shoppers with very specific needs and desires. Then again, we’ve been saying the same thing about the company’s $1,000-plus Extreme Edition enthusiast chips for years. At least this chip is more affordable.
Meet the New Core X-Series
Topping out with the Core i9-7980XE—which, when it debuts, will be able to tackle up to 36 compute threads at once—the new Core X-Series of chips goes above and beyond what we’ve seen from any Extreme Edition processors to land in our test beds in the past decade.
The Core i9-7980XE, in fact, is the first CPU to bring more than a teraflop of compute performance to the consumer-computing table. That’s the kind of performance metric we’re used to seeing on graphics cards, not CPUs.
As we noted above, along with the highest-end chips, Intel will also bring a pair of “lesser” Core i7 and Core i5 processors—using the Kaby Lake architecture—to the new platform, using the same new LGA 2066 socket. Here’s a list of the company’s new chips, direct from Intel, which includes the five CPUs that have been released as of this July 2017 writing, plus the four Core i9 CPUs to come. Note that Intel is keeping many of the details of those upper four Core i9 chips under wraps for now, thus all of the dashes below. The company says these top-end Core i9 CPUs will arrive between August and October. Among the i9 chips, only the Core i9-7900X rolled out with the initial Core X-Series launch.
Now, if you’re looking at pricing, you’ll notice that the top end 18-core Core i9-7980XE commands a whopping $1,999. (At least, that is per 1,000 units, which is how Intel has long listed pricing for its highest-end CPUs, although retail pricing is usually about the same). Clearly, Intel isn’t worried that much about AMD at the top end of the stack. But maybe Team Blue should be a bit more concerned about Team Red, because AMD recently announced that its top-end 16-core, 32-thread Ryzen Threadripper 1950X will sell for $999, and have the ability to boost as high as 4GHz, while a 12-core, 24-thread Threadripper 1920X (with the same top clock speed of 4GHz) will sell for $799. At least on paper, that makes the highest-end of Intel’s new chips seem rather pricey before they have even been fully announced and put on sale.
But if you take a look at the 10-core Core i9-7900X, you’ll notice that at $999, it’s priced about $700 less than the previous-generation Core i7-6950X. So while prices are still fairly high, progress is clearly being made on the threads-per-dollar front. And Intel is offering many more options at the very high end than ever. So there’s no need to step up to (or even close to) the $2,000 range unless you absolutely need all the threads you can get. More choices are a good thing, and it’s likely we have them on the Intel side, at least in part, because of the pressure that AMD has applied so far this year.
On a related note, the eight-core, 12-thread Core i7-7820X, which will likely be the most direct competition to AMD’s top-end Ryzen 7 chip, is priced at $599. That’s $180 or so more than the current going rate for a Ryzen 7 1800X, but a whole lot more affordable than the $1,000 price of the previous-generation eight-core Core i7-6900K. Clearly, Intel is making an attempt to reclaim some of its core-per-dollar value over AMD, while maintaining some price premium over its desktop-chip competitor. It’s not waging all-out price war.
The other interesting difference with the Core X-Series launch is that, as we mentioned earlier, for the first time Intel brought two chip architectures (or at least two generations of silicon) to its enthusiast platform at the same time. The highest-end chips here are based on 6th Generation Skylake silicon, while those at the bottom of the above chart, the Core i7-7740X and the Core i5-7640X we’re looking at here, use 7th Generation “Kaby Lake” architecture, the same as you’ll find in Intel’s current top-end mainstream LGA 1151 CPU, the Core i7-7700K.
That’s important for a couple of reasons, but on the feature front, it means the Kaby Lake X chips (like the Core i5 we’re looking at here) will support 10-bit 4K HEVC encoding and decoding, as well as VP9 decoding. What does this mean, in non-geek-speak? You’ll need a chip from the Kaby Lake line (or later) if you want your system to support 4K streaming content from most major video outlets, or if you plan to, at some point, connect a 4K Blu-ray drive. A Skylake chip (even the 18-core beast) can’t handle that kind of content.
CPU PCI Express Lanes Chipset
As we alluded to earlier, the other major new wrinkle that arrives with the Core X-Series has to do with PCI Express lanes, which you can think of as the highway that moves the bandwidth-hungry data between the processor and things like graphics cards and fast NVMe solid-state drives. Intel’s enthusiast platform has long offered more lanes directly to the CPU than “lesser” chips such as the Core i7-7700K. For instance, the last-generation Core i7-6950X had 40 PCIe lanes built into its silicon, while the more mainstream-minded Core i7-7700K and its ilk have just 16.
But while the higher-end Skylake X chips, like the Core i9-7900X, get oodles of extra lanes, the Core i7-7740X and Core i5-7640X (the two current Kaby Lake processors in the Core X-Series) make you settle for the same 16 lanes as the similarly priced chips on Intel’s mainstream LGA 1151 platform.
Here’s a look at how some of the features break out between Skylake X and Kaby Lake X processors, compared to the features of Intel’s previous-generation high-end desktop platforms, on the X99 chipset…
Now that the Core X-Series platform spans two generations of silicon and ranges from a somewhat modest four-core/four-thread Core i5 chip to (eventually) a $2,000, 36-thread monster, the amount of PCIe lanes on the platform is a lot more complicated. For now, at least, as we noted up top, the two “entry level” Core i5 and i7 chips will offer 16 lanes, like their counterparts on the more mainstream platforms (like the Core i7-7700K). Stepping up the stack to the more midrange chips in this high-end platform, the Skylake-based Core i7-7800X and Core i7-7820X both offer 28 lanes of PCI Express bandwidth, and the Core i9-7900X offers 44 lanes. The extra four lanes versus the previous generation is there on the high-end chips, ostensibly, to make sure builders have the bandwidth to consider Intel’s new Optane Memory caching drives and eventually full-size Optane-based SSDs. Interestingly, Intel is still keeping the number of lanes available on its upcoming highest-end chips under wraps for now. But it’s a safe bet that they’ll have 44 or more.
If you’re wondering what you (or, say, someone who wins a PC-hardware lottery) might actually do with all those PCI Express lanes on the CPU, Core X introduces another feature that could be quite bandwidth-hungry: VROC, or Virtual RAID on CPU. The VROC technology lets you run multiple PCI Express/NVMe storage drives together in your choice of RAID flavor, directly through the CPU. Technically, the feature supports pairing up to 20 drives together in this fashion for a theoretical throughput of up to 128GB per second (and yes, that’s big-“B” gigabytes).
Before we dive into further details of VROC, note that Intel has yet to announce VROC details officially when we wrote this in late July 2017. Everything we learned about this feature came secondhand from motherboard makers at Computex 2017. So take this with a grain of skeptical silicon, and know that things may change when Intel reveals final, official specs and details about this feature.
Previously, on the consumer side of the CPU fence, you’d need a hardware RAID card to do something like a VROC setup. But as we’ll see, there are quite a few caveats here.
First off, you’ll need one of two Intel-provided hardware dongles plugged in to your X299-based mainboard to run drives in any VROC configuration other than RAID 0 (striped). And we were told those dongles would cost $100 or $200 respectively, depending on what type of RAID you planned on running. This is probably an attempt by Intel to keep enterprise customers from saving some money by using consumer boards and processors to run tasks that are designed for server-class boards and Xeon processors.
Second, you’ll need one of the Skylake X chips (like the Core i9-7900X) to use VROC. The Kaby Lake X chips (like the Core i5-7640X we’re looking at here) lack the PCI Express bandwidth to make such a setup possible.
The final VROC caveat is the big one. At least for now, for your VROC RAID drive to be bootable, you’ll need to run the array on Intel SSDs. Considering that it’s Samsung’s consumer drives (such as the Samsung SSD 960 Pro) that are generally the speediest, being forced to opt for Intel drives seems counter-intuitive if you’re going all this way (and spending so much money) to get the fastest possible storage speed. Hopefully, a future feature update will allow non-Intel drives to be bootable in a VROC setup. Within the context of this particular review, though, a wildly high-end VROC setup isn’t in the cards. Once you start plugging in more than a couple of PCI Express-bus x4 solid-state drives and a graphics card (or two), you’re going to start running out of lanes, and things will start disabling themselves on your motherboard. More on this issue soon.
Last on the feature front for Core X-Series chips is an update to the company’s Turbo Boost Max 3.0 technology. This technology debuted in the company’s previous-generation “Broadwell-E” chips, among them the top-end Core i7-6950X Extreme Edition. Now, instead of the chip being able to discern which one of its several cores is able to clock the highest and favoring it for certain tasks, Core X-Series chips can choose two cores that get the Turbo Boost Max treatment.
Like the Extended Frequency Range (XFR) feature found on AMD’s Ryzen chips (at least the models that end in an “X”), this is a good idea in theory. But it probably translates to a minimal appreciable advantage in overall performance, at best. Getting a couple of hundred extra megahertz in short-term bursts from one or two cores on a chip that has eight or 10 isn’t going to change how fast your system feels—unless maybe the “feeling” you’re talking about comes from seeing sky-high benchmark results. Regardless, it’s moot here; Turbo Boost Max 3.0 is enabled only on the higher-end Core X processors. You won’t find it on the Core i5-7640X that we’re looking at here, or on the Core i7-7740X. It’s another thing that Kaby Lake X makes you sacrifice versus Skylake X in this chip line.
Along with the new chips and the new LGA 2066 socket, Intel also will be offering up a new X299 chipset to support it all.
The chipset has up to 24 lanes of PCI Express built into the board. (Remember, this is in addition to the 16 lanes on the Core i7-7740X itself.) That’s a big jump from the eight lanes available on the previous-generation X99 chipset, and it means X299 boards will technically support more SATA ports, USB 3 ports, and fast M.2 storage natively. If you’re looking to connect all the things to your powerful PC, X299 should serve you well, provided you put one of the Skylake X chips in the board, which bring up to 44 extra lanes of bandwidth. But if you’re company-agnostic, keep in mind that that AMD’s Ryzen Threadripper chips will reportedly have an even loftier 64 lanes of PCI Express on the chips themselves. (And it won’t just be the highest-end Threadripper chips, but all of them.) So long as Threadripper’s performance can at least keep up, it looks like Intel will have strong competition in this high-end enthusiast space, as well.
The other wrinkle with X299 and Core X relates to memory. While the Skylake X chips support DDR4 memory running in quad-channel mode at speeds up to 2,666MHz (and higher when overclocked), the lower-end Kaby Lake X processors run in dual-channel mode (just like Kaby Lake chips do on Intel’s mainstream computing platform). That means X299 motherboards need to support both setups, and that you’ll have to configure your RAM differently depending on which subtype of chip (Kaby Lake X or Skylake X) you have installed.
It also means the amount of RAM a board will support will likely vary depending on which chip you have installed. For example, here are the RAM specifications for the Asus Prime X299-Deluxe motherboard we’re using as our testbed for Core X reviews…
As you can see, the DIMM support varies according to the Core X CPU core count. This setup also means, on a purely aesthetic basis, that you could end up with some strange RAM orientations, which could bother some builders with windowed cases and lighting. For example, on the above Asus board, when installing four DIMMs and a Kaby Lake X processor, all your RAM needs to be installed in the slots to the right of the CPU socket. That won’t be a huge problem for most people. But for veteran builders and aesthetes, it looks and feels strange to have all your memory sitting on one side of the CPU with a bank of empty slots on the other side of the socket.
In summary, the Core X-Series processors and the X299 chipset look to be powerful and well-equipped for all kinds of high-end hardware. But what exactly you can install in an X299 motherboard depends a whole lot on the exact chip you plan on putting in it. So, if you’re planning a build, you’ll want to spend some quality time with the manual of the motherboard you’re considering, studying block diagrams and specs to make sure the parts you want will all work with the CPU you’re planning on buying. And if you’re planning on including more than a modest amount of bandwidth-hungry devices such as PCIe SSDs and graphics cards, you’ll probably want (or likely need) to opt for a Skylake X chip, not one of the Kaby Lake varieties that we’re looking at here.
That’s a significant issue, to our minds, because you effectively get the same helping of PCI Express lanes whether you opt for an X299 motherboard and a Core i7-7740X or Core i5-7640X (16 lanes on the CPU and 24 lanes on the board), or a Z270 motherboard and a Core i7-7700K (16 lanes on the CPU and 24 on the board). That makes the initial Kaby Lake Core X chips tough to argue for unless (a) you want a platform you can grow into, adding more components and a better CPU down the line, or (b) performance is better on the Kaby Lake X chips compared to their Core counterparts on Intel’s mainstream LGA 1151 platform.
We’ve already determined that the latter isn’t really the case, at least with the Core i7-7740X. We’ll get a better look at how the Core X i5 chip fares when we get to the benchmark tests. We’ll do that shortly, but first let’s pull away from all of the backgrounder and take a closer look at the Core i5-7640X itself.
The Core i5-7640X: The Details
With the details about the platform and chipset out of the way, let’s take a closer look at the Core i5-7640X processor specifically.
This quad-core, four-thread chip’s 4GHz base clock is reasonably high, especially compared to the 3.3GHz base clock of the Core i9-7900X. But the Core i5’s maximum stock Turbo Boost is just 200MHz higher, at 4.2GHz. That’s not much of a bump as Turbo Boost upticks go, but it does match the top stock speed of Intel’s current high-end Core i5 in its mainstream lineup, the Core i5-7600K. We saw the same Turbo Boost clock speed matching between the Core i7-7700K and the Core X-Series Core i7-7740X. So it seems that Intel is going out of its way to make sure its Kaby Lake Core X chips deliver similar performance to its mainstream-computing-platform counterparts. The clock speed similarities certainly can’t be an accident.
And we saw in our review of the Core i7-7740X that it indeed very closely matched the performance of the Core i7-7700K. Given the above-mentioned clock-speed similarities, we expect the same to be true of the Core i5-7640X versus the Core i5-7600K, though we haven’t actually tested the latter chip. (We did test the Intel Core i5-6600K, its previous-gen equivalent.)
Intel also rates the Core i5-7640X at 112 watts TDP (thermal design power, a measurement of heat dissipation), the same as the Core i7-7640X. That’s lower than the 140-watt-rated Core i9-7900X, but that chip has 2.5 times the cores and five times the threads of the Core i5. The Core i5-7600K, which runs on the same Kaby Lake architecture, again has similar clock speeds and the same amount of cores, and it’s rated at 91 watts. We’re not sure where the extra 21 watts in Intel’s rating is coming from. If anything, the Core i5-7640X should have a lower TDP than the Core i5-7600K, as the Core X-Series chip lacks graphics silicon for gaming or just handling display output. With anything on the Core X-Series platform, you’ll have to provide a dedicated graphics card.