How many processing cores can you convince yourself that you need? If the answer is a resolute “all of them,” then be prepared to spend a whole lot of money.
Intel’s current top-end Broadwell Extreme Edition (a.k.a. “Broadwell-E”) processor chip, the Core i7-6950X Extreme Edition$1,649.99 at Amazon, packs 10 cores and can handle up to 20 simultaneous operations via the company’s thread-doubling Hyper-Threading technology. The result is a stunning level of performance for tasks—and more specifically, software—that can take advantage of all those cores and threads. But that chip also costs a stomach-clenching $1,700, or thereabouts.
If you’re a well-paid media-crunching professional, or a research scientist churning through massive amounts of data in the pursuit of curing a debilitating disease, it could be well worth the money. But for the rest of us (gamers, performance tweakers/enthusiasts, and those who just want a doggone fast PC), there are better options that are nearly as fast—and in some ways, faster—available at a significantly lower price.
With eight cores (and 16 available threads), the Intel Core i7-6900K that we’re looking at here is one step down from Intel’s 10-core 6950X Extreme Edition chip that we reviewed back in May at its luanch time. But it’s cut from the same silicon as that chip, and it actually has a higher base clock of 3.2GHz (compared to the 3GHz-even of the Core i7-6950X). That means in tests that can’t take advantage of all available cores, the Core i7-6900K can actually be faster, as we’ll see later in testing. And it’s not far behind its deca-core counterpart in some of our demanding media-crunching tests, either.
That makes the eight-core Core i7-6900K a much better value for most users than the Extreme Edition 10-core option. But that idea of value is very, very relative, as the asking price for this part is still about $1,100. That’s a hefty savings of $600 for shaving off 20 percent of your cores and not losing a whole lot in the way of performance under most circumstances.
But unless you’re a professional who needs loads of cores and threads, or an enthusiast who will make use of the 40 PCI Express lanes built into this chip, you’re better off stepping further down Intel’s CPU ladder. The entry-level six-core Broadwell-E chip, the Core i7-6800K, sells for “just” $430, and the current top-end 6th-Generation Core/”Skylake” chip, the Core i7-6700K is cheaper still, yet powerful enough for most folks. The latter chip may have only four cores and eight threads, but it’s clocked much higher, with a base clock of 4GHz. Also, on occasion it goes on sale for as low as $300, and in some cases still manages to outperform the Core i7-6900K and even the $1,700 Core i7-6950X.
Broadwell-E Features & Chip Models
As we noted up top, the Core i7-6900K is one notch down from the 10-core Core i7-6950X Extreme Edition, but it still offers up an impressive eight physical cores and the ability to tackle up to 16 threads at once via Intel’s familiar Hyper-Threading technology. Hyper-Threading allows each actual core to work on two threads at the same time.
All of the new Broadwell-E chips will be backward-compatible with most X99-based Socket LGA 2011-v3 motherboards, provided that the motherboard maker offers up a BIOS update to support Intel’s latest chips. In fact, we used the same Asus X99 Deluxe motherboard for our testing that we did to test the previous-generation Core i7-5960X in 2014. A BIOS update for that board was available, quick, and easy. So upgraders who invested in the previous generation’s high-end platform should be able to save some money by using an existing motherboard. Note that earlier, non-“v3” Socket 2011 motherboards will not work. So if you’re rocking an “Ivy Bridge-E” or “Sandy Bridge-E” from a few years back, it’s new-motherboard time.
Here’s a look at the specs for the Core i7-6900K, as well as the trio of other Broadwell-E parts, direct from Intel.
Note that nearly all the new Broadwell-E chips will have an impressive 40 PCI Express 3.0 lanes leading directly to the CPU. That’s not an upgrade from the previous-generation “Haswell-E” platform, but having those lanes is arguably more important today, now that super-fast PCI Express/NVMe-equipped solid-state drives (SSDs) such as Samsung’s SSD 950 Pro$317.99 at Amazon are readily available and munch lanes, too.
For those who don’t need all those available lanes for storage and/or multiple graphics cards, the Core i7-6800K looks like it will be a nice alternative option on the lower end of the Broadwell-E platform, as something of a “bargain” chip, comparatively speaking, in the $400 range.
That chip is especially relevant seeing as the appeal of multiple-card Nvidia SLI and AMD CrossFireX support was somewhat in limbo when we wrote this. SLI and CrossFireX aren’t heavily supported for either the new DirectX 12 API or for virtual-reality setups (with either the HTC Vive or the Oculus Rift). And Nvidia is expressly not recommending anything more than two-way SLI for its GeForce GTX 1080, the company’s new top-end graphics card. So the Core i7-6800K may hold a lot of appeal with budget-minded enthusiasts, as its 28 lanes of PCI Express are still enough for a couple of high-end graphics cards and a fast PCI Express-based SSD (or two).
Turbo Boost Max Technology 3.0
Aside from the the new architecture carried over to the Broadwell-E family (borrowed from the 5th-Generation “Broadwell” chips we saw throughout 2015 in mainstream laptops, and to a much lesser extent, desktops), the primary new feature within the new lineup is something Intel calls “Turbo Boost Max Technology 3.0.” It’s available on all four of the new chips in this line.
Previously, Intel’s Turbo Boost technology allowed for individual CPU cores to ramp up to higher speeds under ideal thermal conditions when a given task wasn’t pegging all available cores. But which core was chosen to jump up to these higher speeds was arbitrary.
That’s fine if all cores are equal. But Intel says that some of its cores have the ability to ramp higher than others. So you might not always get the best performance that’s technically possible from your processor if the chip doesn’t know which core or cores have the highest possible frequency ceiling. That’s where Turbo Boost Max 3.0 comes in. The Broadwell-E chips, when combined with an Intel driver/utility, are able to automatically determine the best core for these types of tasks and shove it to the front of the line whenever a lightly threaded workload arises.
It’s an interesting idea that should let you squeeze some extra performance from your processor, and one that’s arguably more important in a chip like the Core i7-6900K, which has more than just a few cores to choose among. But of course, how much extra performance you might get depends on whether or not you happen to get a chip with a particularly overachieving core or two. So, just like overclockability, much of the potential benefit of Turbo Boost Max 3.0 is likely down to the luck of the draw—or a roll of the (silicon) die, as it were.
Intel, though, does claim that using Turbo Boost Max 3.0 allows these chips to achieve higher top clock speeds. And it somewhat confusingly sprinkles both spec numbers with and without Turbo Boost Max 3.0 in its own product pages and materials. For instance, the Core i7-6900K is listed in the Intel-provided chart above as having a top clock speed of 3.7GHz (without overclocking), and that number is also listed on this Intel ARK product page for the 6900-series chips. It also lists the Core i7-6950X has having a top speed of 3.5GHz. But if you click through to either one of the links to the actual specification pages for these chips, they’re both listed as having a “Max Turbo Frequency” of 4GHz.
An Intel rep told us the more conservative numbers are the top stock frequencies for Turbo Boost 2.0, so the 4GHz speed is likely the maximum possible with Turbo Boost Max 3.0. It’s confusing, though, that Intel lists two different sets of stock frequency ranges for these chips, and it muddies the waters for those evaluating the two. That’s because, on the one hand, Intel says there’s a 300MHz top-end clock difference between them, but then on the actual product pages, the company says the maximum frequency is the same: 4GHz.
We suspect part of the reason for Intel’s number-juggling is that, at least according to ZDNet, there is no Linux support (and none seemingly forthcoming any time soon) for the 3.0 version of Turbo Boost. So, at least for now, you’ll only be able hit the highest clock speeds here (again, without overclocking) by running these chips in Windows.
Speaking of Windows, Turbo Boost Max 3.0 is the kind of feature that, ideally, should be baked into the operating system itself. But at the moment, that’s not the case. Intel provided us with a piece of software (the company calls it a driver, but it also includes a user interface for tweaking features) that needs to be installed and running for Turbo Boost Max 3.0 to work. Here’s what the software looks like. It’s visually simple, and thankfully was completely stable while we used it.
The software lists all available cores, and it positions the fastest at the top. When running, the software should assign tasks that most benefit from fast clock speeds to the fastest core (or cores) automatically. You can also choose to pin specific programs to a core or cores, so they get assigned to them even if the program is running in the background while you’re doing something else.
This all sounds promising, and it may show serious benefit for some chips and in some specific cases. But the fact that you need to make sure the software is running to make Turbo Core Max 3.0 work makes the feature feel a bit cumbersome. We hope Microsoft and other operating-system developers will bake the feature into future OS updates to make the process invisible, or at least less hands-on.
Also, while we didn’t spend loads of time testing the feature and trying to find an ideal case where it make a significant difference, just as we saw with the Core-i7-6950X, we didn’t see any noticeable real-world benefit with Turbo Core 3.0. We tested the Core i7-6900K both with and without the software running, and performance remained either exactly the same, or so close that the difference could be put down to the standard 2 to 3 percent margin of error. With workloads that specifically push a single core, we might see more benefit from the technology. Or, we may have just gotten a test chip with cores that are all fairly evenly matched.
The Core i7-6900’s 3.2GHz base clock speed is 200MHz faster than that of the previous-generation top-end, eight-core Core i7-5960X Extreme Edition, as well the current-gen 10-core Core i7-6950X. So it can be seen as a sort of middle-ground chip between the 2014- and 2016-era flagship Intel enthusiast processors. Given that, we were eager to see how it would perform.
Our prediction? On fully threaded tests, the Core i7-6900K should outpace the last-generation Extreme Edition chip, while sticking at least fairly close to the new 10-core Core i7-6950X. But in areas where higher clock speeds are more beneficial, and where software hasn’t been optimized for massive core counts, Intel’s more mainstream chips—such as the Core i7-6700K and even the Core i5-6600K$243.99 at Amazon—might fare better. In a general sense, that’s pretty much what we saw.
To give a broad sense of how the new Extreme Edition processor stacks up to a wide sample of current processors, we benchmarked the chip against the above CPUs, as well as the low-end Core i3-6100$117.99 at Amazon, and a few AMD parts: the high-end, eight-core AMD FX-8370$220.75 at Amazon; the company’s current top CPU/GPU, the AMD A10-7890K (it includes impressive on-chip graphics); and the Athlon X4 880K$111.81 at Amazon, a chip that’s of most interest to budget-minded gamers, because of its four physical cores and low cost (under $100).
In Cinebench R15, an industry-standard benchmark test that taxes all available cores of a processor to measure raw CPU muscle, the Core i7-6900K stuck reasonably close to its 10-core counterpart.
Falling just 6 percent behind the Core i7-6950X, the eight-core Core i7-6900K impresses here, given the $600 price difference between those two chips. The difference between the current-gen and last-gen eight-core chips was much more significant, with the Core i7-5960X falling more than 20 percent behind the 6900K. The new eight-core Intel chip also turned in a score here that’s more than 2.5 times that of the closest AMD part. Of course, that AMD chip currently sells for about a fifth of the price of Intel’s part.
iTunes 10.6 Encoding Test
We then switched over to our venerable iTunes Encoding Test, using version 10.6 of iTunes. This test taxes only a single CPU core, as much legacy software does.
After its excellent showing our initial test, the Core i7-6900K didn’t look quite so hot here, though it wasn’t unexpected. It managed to easily best all the AMD chips and outpace the previous Extreme Edition part, as well as the much pricier Core i7-6950X. But all three of the lesser 4th-Generation/”Haswell”-based Core chips scored nearly as well or better on this timed test—even the sub-$150 Core i3-6100!
To be fair, single-threaded performance has never been a strong point for Intel’s high-core-count chips, and such software is growing increasingly less common. This isn’t any indicator of how the Core i7-6900K will perform on professional software designed for high-end hardware like this. But it’s still tough to see a chip this expensive get outrun on a common computing task by a part—specifically, the Core i7-6700K—that costs about a third the price.
These days, our older Handbrake test (run under version 0.9.8) now takes less than a minute to complete with high-end chips. (It involves the rendering of a 5-minute video, Pixar’s Dug’s Special Mission, to an iPhone-friendly format.) So, we’ve switched to a much more taxing (and time-consuming) 4K video-crunching test.
We’ve now switched to Handbrake version 0.9.9 and tasked the CPUs to convert a 12-minute-and-14-second 4K .MOV file (the 4K showcase short film Tears of Steel) into a 1080p MPEG-4 video…
Here once again, the Core i7-6900K looked quite good, returning to its spot just behind the Core i7-6950X, near the top of the CPU heap. In this timed test, the new eight-core CPU lagged just 16 seconds behind its 10-core counterpart, and about 18 percent ahead of the previous-generation top-end Core i7-5960X. It also managed to finish this test in less than half the time of the Core i5-6600K.
This is a better indication of why serious content creators might want to invest in a CPU like this. If you’re constantly editing hours of 4K video or stitching together footage from individual cameras for VR-friendly 360-degree video, a chip like the Core i7-6900K could save you hours of rendering time per day compared to a more mainstream chip—even the Core i7-6700K.
Next up is our Photoshop CS6 test, which taxes the processor to apply a series of complex filters to a large image.
Once again, the 6900K chip, while no slouch, wasn’t quite as impressive here as you might have expected. This test applies 11 filters, and the fact is that some filters (just like some stand-alone software programs) benefit more from multiple cores and threads than others. The eight-core Broadwell-E chip once again bested the previous (“Haswell-E”) Extreme Edition CPU here, and it even pulled ahead of its 10-core sibling slice of silicon, but it couldn’t outpace the Core i7 and Core i5 6000-series “Skylake” chips, which benefit from architecture that’s a generation ahead of the Broadwell-based Core i7-6900K.
POV Ray 3.7
Next, we ran the POV Ray benchmark using the “All CPUs” setting. This test challenges all available cores to render a complex photo-realistic image using ray tracing.
The Core i7-6900K climbed back near the top of the performance heap here. Like Handbrake, this test is a good indication of the benefits you’ll see with this chip running professional software in time-consuming workloads. If you do this kind of thing day-in, day-out, there would be a lot less waiting around for tasks to finish with the 6900K. And while the 10-core Extreme Edition part may let you finish a little faster, it doesn’t seem that the benefit warrants the $600 price hike, unless you absolutely have to finish your workloads in the shortest possible time.
Given the fairly generous overclocking we were able to achieve with the Core i7-6950X (bumping up to a top speed of 4.06GHz), we wondered what our Core i7-6900K would be able to hit, given its higher stock clock speeds but two fewer cores. We didn’t spend as much time tweaking as someone might who has invested the $1,100 asking price. But in our limited time overclocking, we settled on a stable clock speed of 4.2GHz. That’s not a whole lot more than the rated top Turbo Boost Max 3.0 speed of 4GHz for this chip. But every attempt to get above that resulted in either a lockup, or the sad-face Windows 10 blue screen.
At 4.2GHz, our Cinebench score jumped from 1,687 to 1,723 (a tiny bump of just over two percent), and we shaved 4 seconds off our iTunes Conversion Test time, getting down to 1 minute and 38 seconds. That’s not a huge improvement, but it does put the Core i7-6900K just 2 seconds behind the higher-clocked (but much lower priced) Core i7-6700K.
Our overclocked chip also finished our 4K Handbrake 0.9.9 conversion test 10 seconds sooner than at stock speeds, in 6 minutes and 34 seconds. It still can’t quite catch the 10-core Core i7-6950X, but that chip only landed 6 seconds ahead, which isn’t much of a win given the $600 price difference.
For media professionals and researchers who run time-consuming and fully threaded workloads day-in, day-out, the Core i7-6900K is arguably a much better value than its 10-core counterpart, which costs so many hundred dollars more. It won’t quite give you the highest performance possible from a single consumer-focused chip. But in our testing, at least, it gets close enough that the extra cores aren’t worth paying for, unless time is absolutely of the essence and your budget is wide open and a mile deep.
But for the vast majority of users, be they gamers, media prosumers, or those who just want a powerful processor that you won’t likely need to upgrade for several years to come, the Core i7-6900K is still a couple steps above the best-value chips in Intel’s Broadwell-E stack. Those who plan to connect lots of graphics cards and fast PCIe storage will want to look at the six-core Core i7-6850K, which has 40 PCIe 3.0 lanes baked in, costs not much more than half the price of the Core i7-6900K, and has the highest base clock speed of this lineup, at 3.6GHz. But those looking for the best bang for their Broadwell-E buck should strongly consider the Core i7-6800K. It also has six cores, and a higher base clock speed than the 6900K, at 3.4GHz, and currently sells for about $430.
That chip makes do with “only” 28 lanes of PCIe, but eight lanes is generally enough to handle current graphics-card bandwidth. So even if you drop in two cards (the maximum recommended for Nvidia’s current GeForce GTX 1080 and GTX 1070 monster cards), you’ll still have 12 lanes left over for fast storage, be that an internal drive like Samsung’s SSD 950 Pro, external storage and video via Thunderbolt 3, or both.