How quickly things change. Just nine months ago, Intel introduced the Core i7-7700K, its seventh in a line of four-core, eight-thread enthusiast processors that stretches back to 2010. We contented ourselves with about a 5% performance increase on average from that chip compared to the Core i7-6700K before it. A proven formula got a little bit better, and that was that.
Shortly after that quiet bit of incremental progress, AMD upended the status quo like an errant speck of dust in a 193-nm stepper with its Zen architecture and Ryzen CPUs. Builders could suddenly get more cores and threads for their dollar than ever before. Every one of those chips was overclockable, and they didn’t suffer from nipping and tucking to satisfy product-segmentation whims. Most importantly, AMD spoke directly to the hearts and minds of the brightest spot of today’s PC market: enthusiasts and gamers. Nobody less than CEO Lisa Su talked up the company’s focus on high-performance products for that audience, and AMD backed up that talk with a full range of desktop chips spanning prices from about $100 to $1000 and everywhere in between.
In the face of this onslaught, the i7-7700K easily maintained its dominance in single-threaded performance and high-refresh-rate gaming, but it was suddenly matched in just about every other measure of performance by chips selling for about $100 less. Whatever Intel’s plans may have been at that time, it was clear that the four-core, eight-thread formula wasn’t going to be enough to hold the line any longer for mainstream PCs.
The Core i7-8700K
That brings us to Intel’s eighth generation of Core desktop processors, code-named Coffee Lake, that launch this morning. These CPUs don’t combat the Ryzen threat with a major microarchitecture change or even a process shrink. Instead, the company has continued to improve its 14-nm process for another generation of CPUs powered by the Skylake microarchitecture. Although the exact nature of those improvements is classified, they are at least partially responsible for letting the company put more cores in its mainstream CPUs at every price bracket, and those higher core counts are how Intel hopes to extract the performance leap that the eighth-generation Core name suggests.
Make no mistake: this is a major change for Intel’s entire product stack. Coffee Lake Core i7s ditch the four-core, eight-thread configuration that’s been the top-end Core i7 formula for the better part of a decade. Instead, those chips now have six cores and 12 threads at their disposal. Core i5 chips now have six cores and six threads, and Core i3s have gone from dual-core designs with Hyper-Threading to native four-core parts without. Given Intel’s clock-speed and architectural advantages over today’s Ryzen CPUs, this full-court press seems poised to slow AMD’s momentum in regaining desktop CPU market share.
The six-core Coffee Lake-S die. Source: Intel
Outside of those very high-level changes, however Intel is basically treating these chips as black boxes, aside from the fancy die shot you see above. Die size, transistor count, and the exact nature of the process improvements Intel has been able to extract from its 14-nm fabs remain mysteries.
We do know that the 14nm++ process, as it’s called, can handle a 23% to 24% higher drive current (a measure directly associated with how fast a transistor can operate) with only minor increases in leakage current, at worst, thanks to a mantra of constant improvement over the life cycle of a given process node. The company disclosed as much at its Technology and Manufacturing Day earlier this year.
In total, the company claims the improvements in 14nm++ can allow for transistor performance increases of up to 26% at the same power level compared to the process’ status in 2015. Perhaps more astoundingly, the company can get the same level of performance on 14nm++ as one of its 2015 transistors for 52% less power. These developments are most likely a large part of what allowed the company to perform feats like putting four cores in a 15W TDP and six cores in a 95W TDP.
Outside of what it’s presented in its manufacturing day events, however, we don’t have any specific inkling of how those process improvements are put to use in Coffee Lake. In the finest tradition of Andy Grove, the company now seems unwilling to let even the slightest hint of competitively useful information slip its clean-room walls. Given AMD’s resurgence, this paranoia is perhaps justifiable. Tipping off your competitors about how, exactly, your fabs are improving is perhaps not the wisest course of action.
Just like the birds and the bees in the age of the Internet, however, curious folk can find enough things out on their own to be dangerous. Someone will certainly figure out the die size of Coffee Lake-S by simply popping the IHS off a chip and using some calipers—not exactly nation-state levels of industrial espionage. Transistor count and process improvements can’t be gauged nearly as easily, but firms like TechInsights will surely take a guess (or probably better, for those willing and able to pay). So long as that’s all that happens, however, Intel seems to be happy to let the rumors swirl.
This secrecy is kind of a shame, because the company’s process technology at 14 nanometers is clearly one of its greater competitive advantages. It’d be neat to hear what has allowed the company to boost the performance of its transistors quite so much on the same node, but in such a cutthroat industry as fabrication, Intel has no reason to tip its hand to competitors. For now, though, let’s examine the implementations of Coffee Lake that are launching today.
Sizing up the lineup
Turbo Boost 2.0
|Core i7-8700K||3.7 GHz||4.7 GHz||6/12||95W||1.573
|Core i7-8700||3.2 GHz||4.6 GHz||65W||$303|
|Core i5-8600K||3.6 GHz||4.3 GHz||6/6||95W||9MB||$257|
|Core i5-8400||2.8 GHz||4 GHz||65W||$182|
|Core i3-8350K||4 GHz||N/A||4/4||91W||1.05
|Core i3-8100||3.6 GHz||65W||$117|
As it’s implemented in the six chips launching today, Coffee Lake really is just a six-core Skylake die at an architectural level. The increase to 12MB of L3 cache on Coffee Lake Core i7s is the natural consequence of adding two more cores with 2MB of L3 each. Coffee Lake Core i5s and Core i3s get 1.5 MB of L3 per core to play with. Outside of those broad changes, Coffee Lake is, as we’ve noted, largely a product of process refinement and small tweaks, like the boost to natively-supported DDR4-2666 RAM on some parts.
A block diagram of the Skylake client core. Source: Intel
This is by no means a bad thing. Despite launching in 2015, Skylake is still a world-class architecture, and Intel probably wants to get as much out of it as it can before firing another valuable microarchitecture or process-shrink arrow from its quiver. Doesn’t help that the company seems to be struggling with its 10-nm process, either, but that’s another topic for another time.
Although price points largely haven’t changed in the move to Coffee Lake, buyers are getting more for their money at each rung on Intel’s product ladder than ever before. Core i3s now offer four full cores to toy with, and the Core i3-8350K is basically a Core i5-7600K for almost $100 less. That’s quite the rough break for AMD’s four-core, eight-thread Ryzen 5s and four-core, four-thread Ryzen 3s, as the i5-7600K had enough oomph to match the $190 Ryzen 5 1500X in our productivity tasks while beating it handily in our gaming benchmarks. That’s before we consider the i5-7600K’s overclocking prowess, as well.
|Model||Cores||Threads||Base clock||Boost clock||Max XFR
|Ryzen 5 1600X||6||12||3.6 GHz||4.0 GHz||100 MHz||16MB||95W||$249|
|Ryzen 5 1600||3.2 GHz||3.6 GHz||50 MHz||65W||$219|
|Ryzen 5 1500X||4||8||3.5 GHz||3.7 GHz||200 MHz||$189|
|Ryzen 5 1400||3.2 GHz||3.4 GHz||50 MHz||8MB||$169|
Up the stack, things look slightly better for the red team. The six-core, six-thread i5-8400 has a Turbo clock of 4 GHz and a base clock of 2.8 GHz in a 65W power envelope. It goes for $20 less than the Ryzen 5 1600, an unlocked, six-core, 12-thread chip for $215 at retail. AMD’s effort has a much larger L3 cache, twice the L2 per core, and six threads on the i5-8400. Its stock cooler will likely be much nicer than Intel’s boxed heatsink, too, a nice perk for budget builders. I’d expect the i5-8400 to have better performance in lightly-threaded workloads, but the Ryzen 5 1600 could pull ahead in multithreaded tasks, and it could stretch any lead it has even further with some overclocking.
The real part of interest to many enthusiasts may be the $257 Core i5-8600K, whose price naturally pits it against AMD’s hottest six-core part, the Ryzen 5 1600X. With six unlocked cores and the considerable overclocking potential of Intel’s 14-nm process to play with, the i5-8600K seems an easy list-topper for any midrange system. Even before we look at overclocking, though, the i5-8600K is besting the Ryzen 5 1600X’s Turbo speeds with XFR in the picture. The 1600X’s SMT support might give it a slight edge in some heavily-threaded workloads at stock speeds, but I imagine overclocking the i5-8600K could easily bridge that gap.
The Core i5-8400
We do have a Core i5-8400 in our labs that arrived alongside the i7-8700K, but in the interest of time, I chose to focus my efforts on the high-end chip first. Intel hasn’t sent us an i5-8600K to play with so far, so we may have to source one of our own to see how it handles as part of our i5-8400 review. Neither the i5-8600K nor the 1600X come with a cooler, so buyers are essentially being asked to pick their poison with regard to overclocking or not. As we’ll soon see, Coffee Lake’s overclocking potential could easily tip the scales in Intel’s favor, but I still expect this battle to be a close one, dollar for dollar.
Coffee Lake Core i7s are the most interesting of this bunch, because they mark the first time so many cores and threads have ever been available on Intel’s mainstream desktop platform. In fact, these chips have specifications that used to be reserved for the blue team’s high-end desktop CPUs, and I have every reason to expect their performance will belong in that tier.
For $303, the i7-8700’s base clock matches the Turbo Boost speed of the Broadwell-E Core i7-6800K, and its Turbo speeds go all the way up to a very impressive 4.6 GHz. No, you’re not getting unlocked multipliers, quad-channel memory, or a ton of PCIe lanes to play with from the i7-8700, but I still think many builders will be elated with this kind of performance from a mainstream CPU in a 65W power envelope.
Finally, the $359 Core i7-8700K brings six unlocked cores and 12 threads to the table. In the transition from i7-7700K to i7-8700K at the top of the lineup, Intel found 200 MHz more Turbo Boost frequency for lightly-threaded workloads even as it dropped two more cores into roughly the same power envelope. Compare these numbers to Intel’s own $379.99 Core i7-7800X and its six Skylake Server cores, and it’s clear the gap between the company’s mainstream and high-end desktops is narrower than ever.
The i7-7800X is kind of the ugly duckling of the X299 platform. It doesn’t have Turbo Boost Max 3.0 support like other Skylake-X chips, so it only clocks up to 4 GHz in light workloads. It has less shared L3 cache than the i7-8700K (but four times the private L2). It’s limited to DDR4-2400 RAM at stock speeds, albeit four channels of it as opposed to two. The i7-7800X is also built with Intel’s server-class mesh interconnect, which sometimes does not respond well to tasks like high-refresh-rate gaming. All together, it seems that only those with a need for AVX-512 support, gobs of memory bandwidth, or 28 PCIe lanes would need to consider the 7800X for $20 more. The i7-8700K is a very potent chip, and we can say that even before we look over to the Ryzen 7 side of the aisle.
Although some will point to the lower base clocks on Intel’s eighth-generation desktop CPUs as the way it held TDPs to only minor increases over seventh-gen chips, I’m not sure those base clocks will be entirely representative of real-world performance. The Core i7-8700K can sustain all-core Turbo speeds of 4.3 GHz in non-AVX workloads without a hitch, at least under the 280-mm liquid cooler we use on our test bench. That’s down just 100 MHz per core from what we observed from the Core i7-7700K. Admittedly, we’re looking at a 95W chip instead of 65W models, but still. Put a good cooler on these chips, and I expect the base clock could be a rare sight.
Z370 increments by 100
New Intel CPUs generally mean new motherboards every couple generations or so, and Coffee Lake is no different. These CPUs will still drop into an LGA 1151 socket, but don’t be fooled—it’s not the same as the LGA 1151 we know from Z170 and Z270 boards. Intel’s official statements on this platform change vaguely cite the need for improved power delivery to support six cores, but we found that position confusing given the already-overbuilt power delivery circuitry on many enthusiast Z270 boards.
Thankfully, the enterprising David Schor of WikiChip has ferreted out exactly what’s changed in the Z370 version, and the answer appears to be more power-delivery and ground pins taken at least in part from a group that were previously reserved. Note to Intel: “we changed the pinout” was all you had to say without giving anything more away.
Even with the different pinout and lack of support for older CPUs, Z370 motherboards still use the LGA 1151 name. I suppose this is technically correct, but continuity implies compatibility, and I expect some builders will be surprised when they learn that Z370 boards can’t be used with physcially-compatible older CPUs. Intel says it’ll avoid confusion among buyers by putting “Supports eighth-generation Intel processors” on motherboard boxes, and early packaging renders suggest “Requires Intel 300 series-based chipset motherboard” will be on at least one side of the box in somewhat-prominent text.
We still don’t think that does nearly enough to avoid befuddlement, however. Buyers will still see “LGA 1151” on both CPU boxes and motherboard boxes, and Z270 supported both sixth- and seventh-gen Intel processors without complaint, so some will inevitably assume one can mix and match all of these things freely. Call the socket LGA 1151 v2, call it LGA 1151+, we don’t care—just call it something different. Newegg seems to have settled on the term “LGA 1151 (300 Series),” and that’s a good enough start.
You can put a sixth- or seventh-gen LGA 1151 CPU into a Z370 motherboard, and it will drop in and latch just fine. I tried as much with a Core i5-7600K, and it didn’t harm anything even after I tried to power it up. Z370 boards will not even POST with these older chips installed, however, and I wasn’t brave enough to risk anything by putting our i7-8700K in a Z270 board to find out what would happen.
Really: just cross the 7 and 2 out and write in an 8 and a 3
Beyond the electrical changes in the socket, the Z370 PCH itself offers the exact same resources as Z270 before it. Motherboard makers can tap 24 flex I/O lanes for use as PCIe 3.0 slots, SATA ports, M.2 slots, and more. Up to 10 USB 3.0 ports and up to 14 USB 2.0 ports can also be called upon for peripheral connectivity. Six SATA ports round out the package. Intel advertises Thunderbolt and Optane support from Z370, as well, just as it did with Z270.
The lack of new features in Z370 is honestly a bit underwhelming. Z270 was already the basis for some fine motherboards, and I never found myself wanting much of anything from them for peripheral I/O. Still, given that the life of a system is often five years or more these days, it’s annoying that Z370 itself offers no new features, connectivity, or quality-of-life improvements compared to its outgoing counterpart.
AMD’s X370 and B350 chipsets each offer two native USB 3.1 Gen 2 controllers for motherboard makers to tap, and Ryzen 3, 5, and 7 CPUs can devote four lanes of CPU-connected PCIe to M.2 slots in addition to the 16 available for graphics cards or other expansion cards. That dedicated x4 link for storage devices is especially nice given the lane-sharing circus that can ensue if builders try to use every port available from modern Intel boards.
Touring the Z370 platform with Gigabyte’s Z370 Aorus Gaming 7 motherboard
Perhaps with the extremely modest changes in Z370 in mind, motherboard makers are offering their highest-end Z370 boards for relatively sedate prices. I got Gigabyte’s Z370 Aorus Gaming 7 in the TR labs ahead of the arrival of our Coffee Lake chips, and I performed all of my Core i7-8700K testing with it for both stock and overclocked numbers. This board is the highest-end model in Gigabyte’s Z370 lineup so far, and it’s served me admirably over the past few days. Despite being the fanciest board of the Aorus bunch, this mobo only carries a $250 price tag.
For the money, buyers get a top-end board almost as fully-featured as any from Gigabyte’s under-$300 Z270 lineup. The most prominent feature of the Gaming 7 is RGB LEDs, though, and there are lots of ’em. All three primary PCIe slots are illuminated with multicolor goodness, and the blinkenlights also nestle into the power circuitry, I/O shroud, and even the VRM and chipset heatsinks themselves. A lighting strip on the right edge of the board completes the visual statement. These various lighting zones should be individually tweakable through Gigabyte’s RGB Fusion Windows software, although a handy firmware function allows for basic setup if your tastes run more to the monochrome.
The Gaming 7 is also the most extreme exponent of Gigabyte’s Z370 styling trends. Instead of the plain white and black shrouding that we saw on the company’s Z270 boards, the Z370 Gaming 7 has a sort of mecha- or cyberpunk-inspired style that manifests as elaborate metal and metal-look accents all over the I/O shroud and heatsinks. Although I quite liked Gigabyte’s simple and clean design language in the Z270 generation, this look is distinctive without evoking medieval torture devices or occult rituals. It should stand out in any windowed case.
This board isn’t all about flash, though. To supply the necessary juice to Coffee Lake CPUs, Gigabyte taps a 10-phase power design incorporating VRM and PWM control circuitry from Intersil. This VRM array is paired with what Gigabyte calls “server-level chokes,” and the units so employed do appear similar to some of the higher-end components we’ve seen on the company’s X99 boards in the past. While I can’t comment in depth on the specific components involved, this setup does appear more than ready to supply the juice the i7-8700K needs to reach its maximum potential.
Like many high-end Z270 boards, the Z370 Gaming 7 offers three PCIe x16 slots and three PCIe x1 slots, all meeting the PCIe 3.0 standard. Two of these are powered by the CPU’s 16 lanes of connectivity. The topmost x16 slot gets 16 lanes from the CPU with one graphics card installed. Deploy a second card in the middle x16 slot, and the Gaming 7 splits those lanes into a pair of x8 channels. The third x16 slot gets four PCIe lanes from the Z370 chipset, and each PCIe x1 slot gets a lane from the PCH, as well.
For PCIe storage devices, the Gaming 7 has a whopping three M.2 slots, the topmost of which sits above the first PCIe x16 slot for better thermal resilience against the graphics card that will presumably sit directly below. The first slot is also protected by a handsome M.2 heatsink with a pre-applied thermal pad.
Since Z370 is simply an evolution of Z270, loading up the board with expansion cards and storage devices could result in some resource-sharing conflicts. The first two PCIe x1 slots get a single lane from the chipset at all times, but the third shares its bandwidth with SATA port 0. Plug an expansion card into that slot, and SATA port 0 goes dark, and vice versa. Install a PCIe or SATA storage device in M2M_32G, the first M.2 slot, and SATA ports 4 and 5 turn off. M2A_32G, the middle M.2 slot, gets four PCIe lanes for M.2 devices at all times, but installing a SATA device in it will disable port 0, as well. Finally, M2P_32G (the bottom M.2 slot of the bunch) shares four lanes of PCIe with the bottom-most PCIe x16 slot. Install a device in one slot or the other, and its unused counterpart will go dark.
Those resource conflicts are a bit frustrating on a motherboard this expensive when we consider that the board only has six SATA ports to work with. Use an NVMe SSD in the first slot as your system’s boot device, and you immediately lose two of these ports to lane-sharing. It’s a bit bemusing that Gigabyte didn’t flip the allocation of lanes for M2M_32G and M2A_32G when it was laying out the board, considering that the middle slot doesn’t ever conflict with SATA devices with an NVMe SSD installed. You can move the M.2 heatsink to this second slot, at least, but then you’re subjecting your M.2 device to the jet blast of the PC’s graphics card. This is a decidedly sub-optimal arrangement for the storage-hungry.
On the audio front, Gigabyte employs its usual arrangement of Nichicon and Wima capacitors in the board’s analog audio path. The codec that feeds this array is Realtek’s now-ubiquitous S1220 codec, complemented by an ESS Sabre 9018Q2C DAC. This chain is claimed to be good for an analog SNR of 121 dB, or on par with the claimed specs for many high-end onboard setups. We’ll delve into this more when I review the Gaming 7 in depth, but my listening experiences on this board were as pleasant as any high-end motherboard’s of late. Nothing quite touches the Z270X-Gaming 8’s Creative ZxRi setup, though.
The Z270X-Gaming 7 offers plenty of possibilities for peripheral I/O. All of the back panel’s USB ports are of the 3.0 standard at a minimum. The leftmost yellow ports offer Gigabyte’s DAC-Up voltage-control feature, which purports to provide more juice to power-hungry devices on long cable runs if it’s needed. Although it’s unlikely they’ll be used on such a high-end board, Gigabyte offers an HDMI 1.4 jack and a DisplayPort 1.2 connector for Coffee Lake’s integrated graphics processor, as well. The lack of a separate converter chip for HDMI 2.0 means the HDMI port only supports 4096×2160 at a maximum, and only then at 30 Hz. Those looking for tolerable IGP output probably want to use the DisplayPort, which can handle 4096×2304 displays at 60 Hz.
The two blue USB 3.0 ports to the right of the gold-plated display outs draw their connectivity from the Z370 PCH. The USB 3.1 Gen 2 Type-C connector and the red Type-A port both draw connectivity from ASMedia’s latest ASM3142 USB 3.1 controller. Gigabyte backs this chip with two lanes of PCIe 3.0 from the chipset for a potential 16 GT/s of bandwidth, a reserve that might come in handy when transferring lots of bits over both ports at once. Above the Type-C port, we get a Killer Gigabit Ethernet jack powered by the company’s E2500 controller. If you still have a thing against Killer for some reason, Gigabyte accommodates with an Intel controller behind the second Gigabit Ethernet jack. The final USB 3.0 port also comes from the Z370 PCH.
Perhaps because of the power-draw potential of a Coffee Lake chip, Gigabyte includes a tiny fan for active VRM cooling duties under the I/O shroud. This tiny fan joins similar cooling approaches from Asus on its X399 boards, and I’m not really a fan. Tiny fans like this have the potential to completely spoil the noise characteristics of a system, and I feel like more effective heatsinks with greater surface area and less decorative bric-a-brac would be a more effective choice. That said, I never heard the fan spin up in my testing, so it’s likely one would have to push the Gaming 7’s power delivery subsystem to true extremes to get it to run.
Like Z270 boards, the Gaming 7 offers four DIMM slots with support for up to 64GB of RAM. Coffee Lake ups the base DDR4 speed to 2666 MT/s for JEDEC RAM, but the Gaming 7 should offer support for much higher speeds through XMP and manual tweaking. Gigabyte’s QVL offers options at instane speeds up to 4166 MT/s. I got my G.Skill DDR4-3600 sticks running on the board without a hitch simply by flipping on the XMP profile in the firmware.
I have more to talk about with the Gaming 7 in my full review, but Gigabyte’s early firmware seems well-baked, and the company slowly continues to address some of my biggest nitpicks with the interface. I had no issues with this board’s performance while gathering numbers for our review, and overclocking the i7-8700K manually was swift and effective on the Gaming 7. If you want a top-end Z370 board to toy with, this one is a fine choice already.
Some quick overclocking explorations
As is usually the case with every new generation of its chips, Intel is exposing some new knobs for overclockers both casual and extreme to toy with on Coffee Lake and Z370. The most prominent of these is per-core overclocking, a feature reserved for recent Broadwell-E and Skylake-X CPUs at the very least. Overclockers will also be able to tune a Coffee Lake chip’s memory timings without a reboot. Extreme OCers will enjoy memory multipliers for speeds up to 8400 MT/s and better phase-locked loop (PLL) control.
We’re not putting some of those more advanced overclocking features to work today, but it seemed criminal not to overclock our particular i7-8700K like a regular enthusiast might, given Intel’s claimed process improvements. Because I like to live dangerously, I swung for the fences and tried for 5 GHz on all cores from the get-go. With a 280-mm cooler, it turned out a Blender-stable 5-GHz overclock was easy enough to hit, but I could never quite get Prime95 Small FFTs enough voltage to be stable at the same speed before we ran into thermal limits.
Eventually, I compromised and set a 5-GHz all-core speed with a -2 AVX offset, good for 4.8 GHz on all cores under AVX workloads. Incredibly, that configuration was happy even under Prime95 loads with an observed 1.284-1.296V using dynamic Vcore on our Aorus motherboard, so it wasn’t difficult to cool at all— 80° C or so was the order of the day with a Corsair H115i on top. Those willing to delid their chips and apply more exotic thermal compounds would seem to have plenty of voltage headroom left to pursue even more extreme overclocks, assuming our particular chip isn’t an exceptional cherry-picked example.
On the same note, we’re finally giving one of AMD’s Ryzen 7 CPUs a formal overclocked review. I picked our Ryzen 7 1700 to serve as the guinea pig for this article, and I was able to achieve 4 GHz on all cores with a pretty aggressive 1.41V with it. Thanks to AMD’s use of solder under Ryzen CPUs’ heat spreaders, though, thermals were never an issue under the AMD-supplied EK Predator 240-mm all-in-one I have on hand for Socket AM4 motherboards. My concerns about safe voltages and the limits of our given chip were much more relevant road blocks to higher speeds. Not bad for an eight-core, sixteen-thread CPU selling for $299.
I didn’t stop turning up the clocks there, either. Intel’s Sandy Bridge Core i7-2600K is rightly regarded as a legendary overclocker, so I pushed up the sliders on our particular chip and reached 4.6 GHz on all cores. With both stock-clocked and overclocked Sandy Bridge results in our stable, builders who have chosen to sit out Intel’s incremental march of improvements over the past few years can see what they’re missing.
Now that we’ve poked, prodded, tweaked, and tuned our test subjects, it’s time to put up some numbers. Let’s hop to it.
Our testing methods
As always, we did our best to deliver clean benchmarking numbers. We ran each benchmark at least three times and took the median of those results. Our test systems were configured as follows:
|AMD Ryzen 7 1700||AMD Ryzen 7 1700X||AMD Ryzen 1800X|
|CPU cooler||EK Predator 240-mm liquid cooler|
|Motherboard||Gigabyte Aorus GA-AX370-Gaming 5|
|Memory type||G.Skill Trident Z DDR4-3600 (rated) SDRAM|
|Memory speed||3200 MT/s (actual)|
|Memory timings||15-15-15-35 1T|
|System drive||Intel 750 Series 400GB NVMe SSD|
|Intel Core i7-2600K||Intel Core i7-3770K||Intel Core i7-4790K||Intel Core i7-7700K|
|CPU cooler||Corsair H110i 280-mm liquid cooler|
|Motherboard||Asus P8Z77V-Pro||Asus Z97-A/USB 3.1||Asus ROG Strix Z270E Gaming|
|Chipset||Intel Z77 Express||Intel Z97||Intel Z270|
|Memory type||Corsair Vengeance Pro Series DDR3-1866 (rated)||G.Skill Trident Z DDR4-3600 (rated) SDRAM|
|Memory speed||DDR3-1866 (actual)||3600 MT/s (actual)|
|Memory timings||9-10-9-27||16-16-16-36 2T (DDR4-3600)|
|System drive||Corsair Neutron XT 480GB SSD||Samsung 850 Evo 512GB|
|Intel Core i7-8700K|
|CPU cooler||Corsair H115i 280-mm liquid cooler|
|Motherboard||Gigabyte GA-Z370X-Gaming 7|
|Memory type||G.Skill Trident Z DDR4-3600 (rated) SDRAM|
|Memory speed||3600 MT/s (actual)|
|Memory timings||16-16-16-36 2T|
|System drive||Samsung 960 Pro 500GB|
They all shared the following common elements:
|Storage||2x Corsair Neutron XT 480GB SSD
1x HyperX 480GB SSD
|Discrete graphics||Nvidia GeForce GTX 1080 Ti Founders Edition|
|Graphics driver version||GeForce 385.69|
|OS||Windows 10 Pro with Creators Update|
|Power supply||Seasonic Prime Platinum 1000W|
Our thanks to Intel and AMD for all of the CPUs we used in our testing. Our thanks to AMD, Intel, Gigabyte, Corsair, Cooler Master, and G.Skill for helping us to outfit our test rigs with some of the finest hardware available, as well.
Some additional notes on our testing methods:
- You’ll note that Intel’s Core i7-6700K is missing from our results. We figured that the i7-7700K is a close enough substitute given its minor performance improvements over the Skylake chip—about 5%, on average. We used the time saved this way to test more overclocked CPUs and to explore use cases like streaming in more depth. We hope this is an understandable decision.
- Unless otherwise noted, we ran our gaming tests at 1920×1080 at a refresh rate of 165 Hz. V-sync was disabled in the driver control panel.
- For our Intel test system, we used the Balanced power plan, as we have for many years. Our AMD test bed was configured to use the Ryzen Balanced power plan that ships with AMD’s chipset drivers.
- All motherboards were tested using the most recent firmware available from the board vendor, including pre-release versions provided exclusively to the press where necessary.
- All available Windows updates were installed on each test system before testing commenced. The most recent version of each software application available from each vendor was used in our testing, as well.
Our testing methods are generally publicly available and reproducible. If you have questions, feel free to post a comment on this article or join us in the forums.
Hitman‘s DirectX 12 renderer can stress every part of a system, so we cranked the game’s graphics settings at 1920×1080 and got to testing.
Get used to this sight. With a GTX 1080 Ti and an i7-8700K working in tandem, we get the highest average frame rates and lowest 99th-percentile frame times out there in Hitman. Yes, this is a big gap, but it’s not unprecedented. Our recent review of the Core i9-7980XE showed similar gains for the Skylake architecture compared to Zen.
These “time spent beyond X” graphs are meant to show “badness,” those instances where animation may be less than fluid—or at least less than perfect. The formulas behind these graphs add up the amount of time our graphics card spends beyond certain frame-time thresholds, each with an important implication for gaming smoothness. Recall that our graphics-card tests all consist of one-minute test runs and that 1000 ms equals one second to fully appreciate this data.
The 50-ms threshold is the most notable one, since it corresponds to a 20-FPS average. We figure if you’re not rendering any faster than 20 FPS, even for a moment, then the user is likely to perceive a slowdown. 33 ms correlates to 30 FPS, or a 30-Hz refresh rate. Go lower than that with vsync on, and you’re into the bad voodoo of quantization slowdowns. 16.7 ms correlates to 60 FPS, that golden mark that we’d like to achieve (or surpass) for each and every frame.
To best demonstrate the performance of these systems with a powerful graphics card like the GTX 1080 Ti, it’s useful to look at our three strictest graphs. 8.3 ms corresponds to 120 FPS, the lower end of what we’d consider a high-refresh-rate monitor. We’ve recently begun including an even more demanding 6.94-ms mark that corresponds to the 144-Hz maximum rate typical of today’s high-refresh-rate gaming displays. Finally, we’ve added a 5-ms graph to see whether any of our chips can sustain 200 FPS or better for any length of time.
Our 16.7-ms graph shows that none of these CPUs hold up the graphics card for any appreciable length of time past that mark. What’s really incredible is that the Kaby Lake and Coffee Lake chips spend less than one-fourth the time of the next-best chip hampering the GTX 1080 Ti’s efforts to stay above 120 FPS, and they continue that exceptional performance even when we consider the 6.94-ms mark. One has to stretch all the way out to 5 ms before these chips start accumulating substantial amounts of time in our graphs. That’s hair-raising high-refresh-rate gaming at peak visual quality in Hitman, and nothing else on the market can come close.
Although Crysis 3 is nearly four years old now, its lavishly detailed environments and demanding physics engine can still stress every part of a system. To put each of our CPUs to the test, we took a one-minute run through the grassy area at the beginning of the “Welcome to the Jungle” level with settings cranked at 1920×1080.
The section of Crysis 3 we use for testing loves lots of CPU cores, and that shows in our results. The overclocked Ryzen 7 1700 rockets to third place thanks to its eight cores and 16 threads. Having slightly fewer CPU cores running at breakneck speed seems to be the perfect recipe to get the best out of this section of gameplay, though, and the i7-8700K produces a considerable leap in average FPS above that of the Ryzen 7 1700. Stock for stock, the 8700K is delivering double the frames, on average, compared to the i7-2600K, and it delivers 99% of its frames in less than half the time the i7-2600K needs to feed the beast by the same measure.
Our time-spent-beyond-X graphs show just how stupefyingly quick the i7-8700K-and-GTX-1080-Ti combo are in this title. Even at stock speeds, the 8700K spends just a little over a second on tough frames that log time past 6.94 ms. You want 144 FPS virtually all the time? You have it. The overclocked Ryzen 7 1700 is delivering a fine experience, to be sure, but wth about four seconds logged past 6.94 ms to its name, the experience on the AMD CPU might be just slightly less glassy. Our average-FPS graph shows the i7-8700K can provide higher frame rates to go with its smoothness, though, so it wins out.
Grand Theft Auto V
Grand Theft Auto V can still put the hurt on CPUs as well as graphics cards, so we ran through our usual test run with all of the game’s settings turned all the way up (save for MSAA and extended distance scaling) at 1920×1080. Unlike most of the games we’ve tested so far, GTA V favors a single thread or two heavily, and there’s no way around it with Vulkan or DirectX 12.
Let’s face it: Grand Theft Auto V is getting up there in years. Its original code base dates back to 2013, and its PC release is a little over two years old at this point. Unlike Crysis 3‘s forward-looking friendliness to many-core chips, GTA V still needs an extremely fast single thread to extract maximum performance at our test settings. That might explain why the i7-7700K takes the top stock-clocked spot in this benchmark.
Put the spurs in, however, and the i7-8700K finds an incredible second wind by delivering a 15% higher average frame rate than even the i7-7700K. The overclocked Ryzen 7 1700 can’t keep up with those frame rates, even if its 99th-percentile frame time is still quite respectable in this company.
Starting our time-spent-beyond analysis at the 8.3-ms mark, the Kaby Lake and Coffee Lake chips spend barely a couple of tenths of a second holding up the GTX 1080 Ti here. Again, if you want 120 FPS or better virtually all of the time, Intel’s latest cores deliver. The overclocked Ryzen 7 1700 would otherwise lead the pack, but it logs over three seconds of time past 8.3 ms where Kaby and Coffee don’t.
What’s really illuminating is how these chips behave past the 6.94-ms mark. The Core i7-7700K’s lower core count seemingly lets it log less time spent beyond this mark against the stock-clocked i7-8700K, but the extra-caffeinated Coffee spends less than a second on frames that take 6.94 ms or more to produce over the course of our entire one-minute test run. That’s the kind of performance where I’m tempted to start throwing words like “perfect” around, even though that’s technically not the case.
With 11 seconds spent beyond 6.94 ms, the Ryzen 7 1700 in overclocked form is the next-best thing going in our group, but it’s a distant, distant fourth place. Coffee and Kaby are the clear champions of high refresh rates and smooth gameplay all at once in GTA V.
Watch Dogs 2
Watch Dogs 2 can seemingly occupy every thread one can throw at it, so it’s a perfect CPU test. We turned up the eye candy and walked through the forested paths around the game’s Coit Tower landmark to get our chips sweating.
Watch Dogs 2 will take every core and thread one can throw at it with our test settings, but as we saw with Crysis 3, fast seems to beat mass. The i7-8700K’s killer combo of fast and numerous cores lets it open a wide lead over every other chip in our stable.
At the 8.3-ms mark, the i7-8700K spends less than half the time holding up the GTX 1080 Ti than even the i7-7700K does. Weirdly, overclocking the chip doesn’t seem to help it much at all, suggesting another bottleneck lies elsewhere. Still, if you’re looking for some of the highest frame rates and the smoothest possible gameplay from Watch Dogs 2, the i7-8700K is the chip to get.
Deus Ex: Mankind Divided (DX11)
With its rich and geometrically complex environments, Deus Ex: Mankind Divided can prove a challenge for any CPU at high enough refresh rates. We applied our preferred recipe of in-game settings to put the squeeze on the CPU and got to it.
I warned you to get used to this, didn’t I? Our Coffee Lake and Kaby Lake chips take a wide lead over the rest of the pack in DXMD. Just like with Hitman, these results aren’t unprecedented, either. We saw similar leaps in performance over AMD CPUs while running a similar test setup on Intel’s Skylake-X chips.
There’s not really much more to say here: the Kaby Lake and Coffee Lake chips spend far less time holding up the GTX 1080 Ti than the rest of the pack, whether you consider 8.3 ms, 6.94 ms, or 5 ms your piece of pie. Their gaming performance in this high-refresh-rate scenario is simply without equal.
That brings us to the end of our pure gaming tests for these CPUs. I don’t have much to say that hasn’t already been said, except wow. I’m a fiend for high-refresh-rate gaming, and the i7-8700K is simply the best pairing available with the fastest consumer graphics card around on that mission, and the race isn’t even close most of the time. It’s just flat-out fun watching single-thread performance bottlenecks evaporate in the face of a 5 GHz Coffee Lake chip. As Coco Chanel once said, “luxury lies not in the richness of things, but in the absence of vulgarity.” If you consider poor 99th-percentile frame times and fuzzy frame-time plots as vulgar as I do, then the i7-8700K embodies that statement perfectly. Just, y’know, that it’s not priced like couture. Or something.
Streaming performance with Deus Ex: Mankind Divided and OBS
One of the uses that Intel and AMD have hyped the most for their highest-end desktop processors this year is single-PC gaming and streaming. The most avid Twitch streamers, as we understand it, have tended to set up dedicated PCs for video ingestion and processing to avoid affecting game performance, but the advent of these many-core CPUs may have opened up a world where it might be more convenient to run one’s stream off a single PC. Intel calls this kind of thing “megatasking,” and it claims the i7-8700K is quite good at it. Let’s find out.
Although one might wonder why people are still making a hullabaloo about CPU encoding performance when hardware-accelerated game streaming is available from both major GPU software packages, the fact of the matter seems to be that the most demanding professionals still choose to use software encoding. The reason for this is that Twitch and other streaming services have restrictive bit rates for streamed content. GPU-accelerated services like GeForce Share (née Shadowplay) and Radeon ReLive make it easy to stream without affecting gaming performance that much, but they might not offer the highest-quality viewing experience to fans within the bounds of those bit rates. For achieving the best results possible, the name of the game is still software encoding with x264.
Let’s start off by noting that gaming at 1920×1080 and streaming that same gameplay at 60 FPS will be difficult, if not impossible, with a CPU-bound title like Deus Ex on four cores and eight threads. The gameplay experience and the stream alike look like crap, even on our otherwise excellent Core i7-7700K. You can safely forget trying on older quad-core chips. One could perhaps get away with looser encoder settings and lower frame rates on such chips, and you might not see the same problems with games that aren’t primarily CPU-bound, but this is a benchmark, dangnabit. We’re trying to find limits, not shy from them.
Even with that in mind, the Core i7-8700K couldn’t quite keep up with DXMD running at 1920×1080. I found that easing the load on the CPU by upping the game resolution to 2560×1440 and downscaling the capture to 1920×1080 at 60 FPS was the optimal solution. In simpler terms, we’re giving the CPU some breathing room by shifting more of the load to the graphics card. I maintained this approach across all of the chips I deemed streaming-capable.
If you really want to do no-holds-barred 1920×1080 gaming and stream it from the same machine, it seems you really want a Ryzen 7 1800X at the very least. A Core i7-7820X or a Core i9-7900X are even better yet, if Hitman is any guide. No, this is not a cheap way to do things, but if you’re dead-set on single-PC streaming performance, you will eventually have to pay the piper for the equivalent of two systems’ worth of processing power.
The GTX 1080 Ti can still play DXMD in an enviably fluid fashion at 2560×1440, so it’s not much of a sacrifice to play at the higher resolution and broadcast at the lower one. To get my final setup, I played with OBS’ various x264 settings to achieve what looked (to my eye) like the best visual quality possible without unduly bogging down any of my test rigs. To my video-snob tastes, that was the “faster” x264 profile. For the curious, though, it’s interesting to note that looser encoder settings didn’t make 1920×1080 gaming and 60-FPS streaming possible on the i7-8700K, either. That suggests streaming framerate is a much bigger initial obstacle to clear when tuning one’s setup than the choice between “veryfast” or “faster,” for example.
We’re still not considering x264 encoder performance in isolation for this review. While metrics like dropped frames are certainly important to the viewer experience, we don’t have the methods to effectively process or present that data yet. We did monitor stream quality during our testing and ensured that our particular encoder settings weren’t producing choppy or otherwise ugly stream delivery, and we figure that what looks consistently good to the eye is fine in light of the fact that there are possibly dozens of network hops between us and a final viewer. We might consider these metrics in future articles, but for now, we’re mostly worried about the gameplay experience these CPUs deliver to the streamer.
There are a lot of moving parts in the graphs above. We’ve presented frame-time plots, average frame rates, and 99th-percentile frame times for both streaming and non-streaming gameplay, so take some time to flip through the various graphs above to get a full sense of the performance picture.
For all that, the results are stark. The i7-8700K suffers only a minor performance hit with OBS running, while the Ryzen 7 CPUs lose about 20% of their performance potential (as measured by average FPS). DXMD is apparently GPU-bound at these settings, as evidenced by the nearly identical performance across the board for these parts without OBS engaged, so it seems flipping on a stream is enough to expose a major bottleneck on the Ryzen chips. Once again, this result is not unprecedented: the Ryzen 7 1800X’s average frame rate plunged 30% in our Hitman streaming tests during our Core i9-7980XE review.
x264 is one of the more prominent applications with AVX support, so it’s possible that Coffee Lake’s wider AVX pipes may be giving it a leg up here, among other things. Whatever the cause, it’s clear that the i7-8700K can deliver a smoother and more fluid gaming experience at our test settings than even a Ryzen 7 1700 can with all cores ticking away at 4 GHz. The large performance drop we observed with the Ryzen 7 parts might not be as severe with less CPU-bound games, and admittedly, that describes more of today’s titles than not. Still, it’s not like Intel is charging huge amounts more money for the i7-8700K for the pleasure of this experience, CPU-bound games or not.
One thing our results don’t capture is that at stock speeds, the Ryzen 7 1700 is not quite capable of an entirely smooth stream. I noticed a number of frame drops while watching my test stream as I played, so I held onto my log file and found that the stock 1700 was dropping about seven percent of the frames destined for Twitch. Blame the chip’s constrictive TDP and low resulting base clock, I guess. If you want to stream CPU-bound games with the 1700, you definitely want to take a trip into the BIOS and turn up some multipliers.
Our time-spent-beyond-X graphs basically confirm what we’ve already discussed: the i7-8700K is delivering a much smoother gaming experience than the Ryzen 7 parts while streaming, no matter what threshold you choose to examine. Let’s see if our productivity results show as wide a gulf between these competitors.
Compiling code with GCC
Our resident code monkey, Bruno Ferreira, helped us put together this code-compiling test. Qtbench records the time needed to compile the Qt SDK using the GCC compiler. The number of jobs dispatched by the Qtbench script is configurable, and we set the number of threads to match the hardware thread count for each CPU.
Compiling likes lots of threads, but there’s a whiff of Amdahl’s Law to the proceedings, and the i7-8700K’s many cores and high clocks give it a minor edge on even the overclocked Ryzen 7 1700.
File compression with 7-zip
In our 7-Zip testing, the i7-8700K takes an early lead in the compression portion of the test. In the arguably more common decompression operation, though, it falls behind the entire Ryzen lineup at stock speeds. Even an overclock is only enough to bring the i7-8700K even with the Ryzen 7s running at stock speeds. The overclocked 1700 is the undisputed champion here, for once.
Disk encryption with Veracrypt
In the accelerated AES portion of our benchmark, the i7-8700K is just behind the Ryzen family at the top of the chart at stock speeds, and it ends up in about the same place in our non-accelerated Twofish test. Overclocking the i7-8700K to 5 GHz takes it to the top of the chart in AES work, and it only trails the overclocked Ryzen 7 1700 in the non-accelerated portion of the benchmark. Pretty impressive considering the core-count imbalance here.
The evergreen Cinebench benchmark is powered by Maxon’s Cinema 4D rendering engine. It’s multithreaded and comes with a 64-bit executable. The test runs with a single thread and then with as many threads as possible.
The multithreaded portion of Cinebench really lets the Ryzens shine. The i7-8700K can only match the Ryzen 7 1700 at stock speeds, and it can’t catch that chip at 4 GHz even in its overclocked form.
Blender is a widely-used, open-source 3D modeling and rendering application. The app can take advantage of AVX2 instructions on compatible CPUs. We chose the “bmw27” test file from Blender’s selection of benchmark scenes to put our CPUs through their paces.
With the latest version of Blender, it’s a close race between the Ryzen 7s and the i7-8700K at stock speeds. Overclocking the i7-8700K lets it chase the OCed Ryzen 7 1700 at the top of the chart, though.
Here’s a new benchmark for our test suite. Corona, as its developers put it, is a “high-performance (un)biased photorealistic renderer, available for Autodesk 3ds Max and as a standalone CLI application, and in development for Maxon Cinema 4D.”
The company has made a standalone benchmark with its rendering engine inside, so it was a no-brainer to give it a spin on these CPUs. The benchmark reports results in millions of rays cast per second, and we’ve converted that figure to megarays for readability.
Corona seems to like Intel CPUs, but that’s not enough to let the i7-8700K overcome its cores-and-threads disadvantage versus Ryzen chips. It merely pulls even with the Ryzen 7 1800X at stock speeds, and narrowly edges out the overclocked Ryzen 7 1700.
Handbrake is a popular video-transcoding app that recently hit version 1.0.7. To see how it performs on these chips, we’re switching things up from some of our past reviews. Here, we converted a roughly two-minute 4K source file from an iPhone 6S into a 1920×1080, 30 FPS MKV using the HEVC algorithm implemented in the x265 open-source encoder. We otherwise left the preset at its default settings.
x265 should tap AVX instructions where it can, so the i7-8700K takes the top stock-for-stock spot. It also eclipses the overclocked Ryzen 7 1700 when we pull out all the stops.
CFD with STARS Euler3D
Euler3D tackles the difficult problem of simulating fluid dynamics. It tends to be very memory-bandwidth intensive. You can read more about it right here. We configured Euler3D to use every thread available from each of our CPUs.
Before we discuss our results, it should be noted that the publicly-available Euler3D benchmark is compiled using Intel’s Fortran tools, a decision that its originators discuss in depth on the project page. Code produced this way may not perform at its best on Ryzen CPUs as a result, but this binary is apparently representative of the software that would be available in the field. A more neutral compiler might make for a better benchmark, but it may also not be representative of real-world results with real-world software, and we are generally concerned with real-world performance.
Given the i7-8700K’s speed, number of threads, and its memory bandwidth, it’s no surprise it takes the top spot here.
Digital audio workstation performance
One of the neatest additions to our test suite of late is the duo of DAWBench project files: DSP 2017 and VI 2017. The DSP benchmark tests the raw number of VST plugins a system can handle, while the complex VI project simulates a virtual instrument and sampling workload.
We used the latest version of the Reaper DAW for Windows as the platform for our tests. To simulate a demanding workload, we tested each CPU with a 24-bit depth and 96-KHz sampling rate, and at two ASIO buffer depths: a punishing 64 and a slightly-less-punishing 128. In response to popular demand, we’re also testing the same buffer depths at a sampling rate of 48 KHz. We added VSTs or notes of polyphony to each session until we started hearing popping or other audio artifacts. We used Focusrite’s Scarlett 2i2 audio interface and the latest version of the company’s own ASIO driver for monitoring purposes.
A very special thanks is in order here for Native Instruments, who kindly provided us with the Kontakt licenses necessary to run the DAWBench VI project file. We greatly appreciate NI’s support—this benchmark would not have been possible without the help of the folks there. Be sure to check out their many fine digital audio products.
At 96 KHz and a buffer depth of 64 for DAWBench VI, we are most likely testing per-core throughput more than anything. That means the i7-8700K shoots to the top of the chart at a buffer depth of 64, and its lead widens when we relax the buffer to 128 samples. Six Coffee Lake cores are no joke.
The DAWBench DSP test is a much more even playing field for our test subjects at 96 KHz. Still, the i7-8700K grabs the stock-clocked lead at 64 samples, and the overclocked Ryzen 7 1700 is only enough to match it. The overclocked i7-8700K is the leader by a wide margin. At 128 samples, the i7-8700K still maintains its lead at stock and not-stock speeds.
DAWBench VI loves Intel CPUs, and lowering the sampling rate just lets the Core i7-8700K widen its lead at both buffer depths. The superb latency characteristics of Intel’s client core seem to make a big difference in this benchmark.
DAWBench DSP tends to be a great equalizer, but the i7-8700K bucks the trend by taking a wide lead at both buffer depths at 48 KHz, both stock and overclocked.
Given its chart-topping performance across the board, the Core i7-8700K seems to be the best thing going in affordable CPUs for digital audio workstation duty. There’s really no contest.
Power consumption and efficiency
We can get a rough idea of how efficient these chips are by monitoring system power draw in Blender. Our observations have shown that Blender consumes about the same amount of wattage at every stage of the bmw27 benchmark, so it’s an ideal guinea pig for this kind of calculation. First, let’s revisit the amount of time it takes for each of these chips to render our Blender “bmw27” test scene:
Because Blender is a steady-state workload, we can take a reading off our Watts Up power meter and multiply it by the time required to complete the render for each chip to come up with an estimated task energy in kilojoules. We can then plot that figure against the time each chip needs to complete its task to get a simple visualization of effciency.
Unfortunately, we’ll only be considering Intel versus Intel for the task energy portion of these tests at press time. Recent firmware updates to our X370 motherboard of choice have vastly increased both idle and load power consumption from our Ryzen 7 CPUs for no discernible reason, and we don’t feel comfortable reporting power-consumption numbers for what appears to be a firmware issue on the AMD side. This power management issue didn’t affect clocks or performance in any of our other benchmarks, but it does make this comparison impossible. We’re looking into this issue and we’ll update the review with new numbers if we can.
Under load, the i7-8700K system’s instantaneous power consumption is up from generation to generation, as one would expect from adding two more cores on a similar process. The increase isn’t substantial, though. Overclocking the chip does cause our testbed to draw far more power from the wall, but that’s to be expected, and about 60W more isn’t that bad.
Plot our estimation of task energy against time to completion, and the i7-8700K is impressively efficient at stock speeds. Its higher system power draw is offset by the fact that it gets the job done much faster than every other chip here, so its overall power consumption is lower than any four-core, eight-thread part. That’s an excellent advance for Intel.
It’s time to condense all of our test results into our famous value scatter plots. We use a geometric mean of all of our real-world results to ensure that no one test has an undue impact on the overall index. First up, let’s look at gaming performance.
The i7-8700K doesn’t leave me much to say. In both gaming performance potential (as measured by average FPS) and in delivered smoothness (as measured by its 99th-percentile FPS figures), literally nothing else on the market can touch this chip for high-refresh-rate experiences. Simple as that.
Not much to say here, either. The Core i7-8700K’s world-beating single-threaded performance, Ryzen 7-matching multithreaded performance, and exceptional DAWBench results put it well over the top for the best overall productivity CPU of this bunch. Overclocking only sweetens the deal.
All told, the Core i7-8700K is Intel’s most groundbreaking enthusiast CPU in years. Its productivity performance and power efficiency are outstanding. It offers enviable new heights for high-refresh-rate gaming experiences, especially on our critical 99th-percentile FPS metric of delivered smoothness. Our particular CPU had plenty of overclocking potential, too. Whatever wizardry Intel’s manufacturing team is working on the company’s 14-nm process is paying off handsomely, even if we are getting a third round of the same Skylake core we’ve known since 2015.
Although Intel took its sweet time in delivering the coup de grace, the i7-8700K dramatically puts the final nail in the coffin of 2011’s Core i7-2600K. Clock for clock, the i7-8700K delivers 2.5 times the performance of Sandy Bridge’s finest four-core avatar in our productivity index. Overclocking the i7-8700K delivers nearly three times the performance. ‘Nuff said. Coffee Lake’s gaming prowess is head, shoulders, knees, and toes above the 2600K’s for high-refresh-rate experiences, too.
Even though it’s down two cores and four threads on AMD’s Ryzen 7 family, the i7-8700K spells trouble for those chips at their current prices. The Coffee Lake chip is a fair bit faster than AMD’s eight-core parts in our productivity index. For high-refresh-rate gaming, there’s no contest. Those playing at 2560×1440 and 4K might find less to get excited about in Coffee Lake, but I think gamers of all stripes will prefer the i7-8700K’s as-yet-unmatched combo of world-beating single-threaded performance and 12 threads for more demanding work. Ryzen 7 CPUs can’t quite cover both bases yet, and you’ll feel the i7-8700K’s advantages everywhere, not just in gaming.
All that performance was available before we turned up the multipliers for an easy overclock to 5 GHz on our particular CPU, too. Ryzen CPUs just can’t clock that high under typical conditions today. That’s not to say Ryzen is completely left in the dust in the overclocking stakes, of course. I pushed my $300 Ryzen 7 1700 to 4 GHz across all its cores, and that boost let the chip claw back some of the gap with the $360 i7-8700K in productivity tasks at stock speeds. Overclocking the i7-8700K opens the gap right back up, though, and if you’re going to the effort of tweaking to begin with, the rewards of the i7-8700K seem much more tantalizing to me.
As if all that wasn’t rough enough, the i7-8700K might snuff one of AMD’s brightest selling points for its Ryzen 7 parts: single-PC streaming power. Whether at stock speeds or overclocked, the i7-8700K delivered a much more fluid client-side gaming experience than the Ryzen 7 chips for a given stream quality (at least in our largely CPU-bound testing, which won’t always be typical). Streamers who considered Ryzen 7 CPUs a default pick over Intel’s quad-cores now have a compelling alternative.
No matter what team you root for, though, AMD deserves enthusiasts’ gratitude for finally offering Intel some stiff competition in x86 CPUs again. Without that development, it’s not clear how long we would have had to wait until Intel delivered higher core and thread counts in its mainstream socket. AMD seems to have an aggressive CPU roadmap of its own planned for the near future, so we can only hope this renewed competition will remain vigorous for some time to come.
For now, the Core i7-8700K is the most well-rounded CPU available. With virtually no weaknesses, it’s among the easiest TR Editor’s Choice picks I’ve ever made. If you’ve been sitting on an upgrade from Sandy or Ivy for the past few years, grab your hammer and piggy bank and wait for our next System Guide. It’s time.
Like this review? Help us continue to perform the dozens upon dozens of hours of hands-on testing that go into articles like this one by becoming a TR subscriber today. Subscribers get exclusive benefits as part of their membership, and you can contribute any amount you’d like to support us. Best of all, we were doing crowdfunding before it was cool, so your entire contribution (less transaction fees) helps us continue our work. Thanks in advance for your support.