It seems like only yesterday we were reviewing the Skylake CPU, but it turns out that was nearly 17 months ago! Back then, Skylake was the “tock” of Intel’s “tick-tock” release strategy, but recent times has seen them add a phase to this, almost a tick, tock, and refine. Since we have seen the tock, we now expect even further refinement. Queue the 7th generation of Intel CPUs, dubbed, Kaby Lake (pronounced KAY-BEE). With Kaby Lake, you are looking at “Intel’s most advanced 14nm+ process technology and delivers new levels of performance and security.” Intel has made some refinements which for us overclocking enthusiasts should yield higher overclocks. It will also save power for the lower power/laptop CPUs in the Y, U, and H, series CPUs, and for all of us, better built in security, and a better 4K streaming experience.
But we are here to talk about the S series, in particular, the flagship Intel i7-7700K Kaby Lake we have to review. Though IPC improvements may be sparse at best, my hope is to squeeze a couple hundred MHz more out of the top end. The higher base clocks may be appealing for those looking to get a new platform. I think its time to stop the chit chat and get moving on those details.
Specifications and Features
Looking at the specifications table below, the i7-7700K is, as expected, a quad core with Hyper Threading (HT) for a total of eight threads. The base clock comes in at 4.2 GHz (vs 4.0 GHz for the 6700K) and boosts a single core to 4.5 GHz, while all cores boost to 4.4 GHz. Another 200 MHz improvement over Skylake. The new CPU sports a refined 14nm+ process while still using the Tri-Gate 3D transistors. Power use/TDP is down to 91W, a savings of almost 5% over the 95W 6700K. The cooling medium/thermal paste is the same found in Skylake. De-lids of the 7700K are still possible according to the web (yes someone already performed a de-lid). This should not be necessary in most cases, short of the Benching Team really pushing things to the limit.
Memory on this CPU/platform hasn’t changed in that it supports a total of 64GB with the base specification of DDR4 2400 or DDR3L 1600, both in a dual channel configuration. It does not support ECC memory.
On the PCI Express (PCIe) side of the house, Kaby Lake offers a total of 20 lanes out of the CPU now which will allow a bit more flexibility for multiple cards, PCIe based NVMe SSDs, and other PCIe based devices.
|Intel i7-7700K Specifications|
|# of Cores||4|
|# of Threads||8|
|Clock Speed||4.2 GHz|
|Max Turbo Frequency (2.0)||4.5 GHz (single core), 4.4 GHz all cores|
|Instruction Set Extensions||SSE 4.1/4.2, AVX 2.0|
|Lithography||14 nm Tri-Gate 3D Transistors|
|Thermal Solution Spec||PCA 2015D|
|Integrated Graphics||Iris Plus HD630 – 24 Execution Units up to 1150 Mhz|
|Max Memory Size||64 GB|
|Memory Types||DDR4 2400 / DDR3L 1600|
|# of Memory Channels||2|
|ECC Memory Support||No|
|PCI Express Revision||3.0|
|PCI Express Configurations||1×16 or 2×8 or 1×8+2×4|
|Max # of PCI Express Lanes||20 Lanes|
|Intel Data/Platform Protection Technology|
|Intel Device Protection w/Bootguard||Yes|
|Trusted Execution Technology||No|
|Execute Disable Bit||Yes|
The large image below is a list of the S-SKU lineup (remainder in thumbnails below it). In it we see the 7700K is the top with its 4C/8T configuration, followed by its little brother, the i7 7600K. There are a slew of locked chips to satisfy any performance point.
Take a look at the last thumbnail, specifically the i3 line. There is the fancy new i3-7350K we heard about. This little gem is an unlocked 2C/4T chip! It should be a very easy way to have a four thread budget overclocking machine. And for the benching team, perhaps a beast on single/dual threaded benchmarks if it screams.
Below is a list of Key Features identified from Intel:
- Amazing Platform Performance and Responsiveness
Intel’s advanced 14nm+ process technology is the foundation for the entire 7th Gen Intel Core processor family, delivering performance improvements over previous-generation processors. For the enthusiast, choosing unlocked notebook and desktop processors provides the opportunity to tune performance to its fullest potential in order to get the most out of games and virtual reality experiences, with performance headroom left over for live broadcasting.
- 3D Graphics Performance alone is 6.6X better than a 5-year-old desktop PC
High performance 7th Gen Intel Core H-series and S-series processors deliver amazing VR experiences through incredible CPU performance providing balanced specs that AAA games demand in order to maximize play.Intel HD graphics in 7th Gen Intel Core platforms enhance gameplay at 720p.
- Extended Battery Life
Intel continues to drive power efficiency improvements at the processor and platform level to increase battery life and enable even smaller and thinner4 devices. Dedicated hardware acceleration reduces power consumption dramatically. With the Y-series processor family, 2 in 1s are reimagined, enabling thin and fanless designs for ultra-mobility. On U-series processors, enhanced productivity and creativity are possible in increasingly slim form factors without sacrificing battery life –in fact new 7th Gen Intel® Core™ U-series processors now provide up to 10 hours of battery life for the whole workday. Modern Standby ensures laptops resume with data ready in a flash.
- I/O Support
Thunderbolt™ 3 – Doubling the bandwidth of the previous generation, Thunderbolt 3 transfers data 8x faster than USB 3.0.6 Plus, for the first time, you can run data, video, and power over a single cable. Notebooks with the versatile Thunderbolt 3 technology, the USB-C that does it all, provide an incredible I/O experience. A single cable conveniently supports up to 40 Gbps transfer speeds, two 4K 60Hz displays, system charging up to 100W, external graphics, and Thunderbolt networking to bolster productivity.
Gen 3 PCIe Support – 7th Gen Intel Core processors offer Gen 3 PCIe support for higher data transfer rates of 8 GT/s versus 5 GT/s with PCIe Gen 2. The latest Intel® Rapid Storage Technology supports NVMe PCIe x4 solid state drives and is capable of utilizing Gen 3 PCIe Speeds. DT systems will offer 15% more I/O capability3 over the previous generation.
- Security (vPro processors)
7th Gen Intel® Core™ vPro™ processors include a variety of hardened security features built into the silicon. Intel® Authenticate provides a robust multifactor solution that is protected in hardware, reducing users’ exposure to identity theft attacks such as phishing and screen scraping. With fingerprint, Bluetooth proximity, protected PIN, location, and facial recognition, there are plenty of options for policy customization on PCs.
Online shopping is more secure with Intel® Software Guard Extensions (Intel® SGX) and Intel® Online Connect, enabling service providers to allow personal biometric credentials to be used to make purchases online, minimizing worry about identity theft or credential loss.
- Windows 10 Support
Intel has partnered with Microsoft* to optimize Windows* 10 experiences on 7th Gen Intel Core processor-powered devices. Intel’s platform innovations combined with Windows* 10 deliver benefits for both consumer and business users. Using Windows* Hello8 removes the hassle of remembering and typing passwords. Other features such as Windows* Ink11 provide a more natural way to interact with your device, making it as easy to use as pencil and paper.
In the slideshow below are some excerpts of the Intel press deck for Kaby Lake. In it we see the entire 7th generation processors range from an exceptionally low 4.5W all the way up to our i7-7700K Kaby Lake at 91W. It goes over the different series of the 7th gen and their applications such as 2 in 1’s and compute sticks, to light laptops, performance laptops, all the way to desktop performance, and value to all-in-one computers. The next slide shows the Intel HD Graphics 620 compared with the Iris Plus Graphics 650 being around 65% faster (according to Futuremark’s 3DMark Skydiver benchmark graphics score). That is a significant improvement if you need it over some of the lower SKUs.
In order to bring 4K content to the mainstream, Intel has added new media engines in their HEVC 10-bit and VP9. These newer codecs offer higher compression rates without having huge bandwidth requirements.
The next slide talks about Intel’s new Optane memory. This new technology is said to serve as go between from your HDD to the Optane memory module improving boot times and application response. These come in two form factors, the first being M.2, and the second being more like a memory DIMM (which should be faster, but mainly on the server side it seems?). From the explanations I’ve seen, it looks to be more of an integrated cache for the HDD with less volatility than previous caching technologies. We can expect to see M.2 form factor Optane modules in 1H 2017. Below is a video from Intel which should help explain 3D XPoint Memory and Optane technology.
The last slide covers improved overclocking capabilities. The 7th gen CPUs are now B Clock Aware and have a voltage/frequency curve. Intel stated this “improves stability and reliability when overclocking as it knows where the BCLK is. It will make choices on voltages to achieve higher overclocks and better stability.” This sounds to me like better overclocking on auto or with motherboard software as the chip is more self aware. You can also utilize the AVX offset ratio for more stable and higher overclocks on non-AVX2 frequencies.
The picture below shows a high-level breakdown of the Intel Z270 chipset, the performance/flagship chipset of the 7th generation processors. It is the only chipset which gives us the ability to overclock. Also included in this lineup are the other chipsets, Q250/Q270 (corporate), B250 and H250 (consumer and corporate). I have also included a Z170 chipset diagram below this one (thumbnail) for comparison.
One of the main differences between the Z170 and Z270 platforms are the increase in PCIe lanes, up to 24 total, and native support for DDR4 2400. Each board is going to be a bit different in how it breaks things down when multiple PCIe lanes are in use, but you get the general idea of where things are coming from and what it supports. Native USB3.0 is also included while USB3.1 Gen 2 will still be provided by a third party chip (ASMedia I would imagine on most). We can expect to see the full gamut of boards from the add in board partners… and reviews from us as the boards are released.
Below is our gratuitous picture of the CPU’s circuit map! Looks pretty similar to the 6700K circuit map.
Below are some sample images from Intel of the product packaging for the 7th generation CPUs. Like past generations, any unlocked SKU does not come with a cooler.
Next up are pictures of the i7-7700K sample we have. It’s the same size PCB as Skylake, including the thickness, but has a slightly different Integrated Heat Spreader (IHS) design when compared with the 6700K next to it. Flipping the CPUs over, we are hard pressed to find many differences here either. I think one of the caps on the bottom were in a different orientation! Otherwise, there isn’t really much to see here except for a new CPU!
The data we have gathered will give us a great idea of its performance both at stock (no turbo), and matching clockspeeds to see IPC performance increases between Skylake and Kaby Lake. I have included Broadwell results with the i7 5775C, as well as throwing in a hex-core Haswell-E CPU in the 5820K, round out the comparisons.
|Motherboard||ASRock X99 OC Formula||ASUS Maximus VII Formula||ASUS Maximus VIII Extreme|
|Memory||Kingston HyperX 4x4GB DDR4 3000||G.SKILL TridentX 2X8 GB 2400 MHz||G.SKILL Trident Z 2X8 GB DDR4 3200 MHz @ 3000 MHz|
|HDD||Samsung 950 Pro 500 GB||Samsung 840 EVO 500 GB||512GB OCZ RD400|
|Power Supply||Seasonic Platinum 1kW||Corsair HX1050||Seasonic Platinum 1 kW|
|Video Card||Gigabyte GTX 980 Ti Extreme||EVGA GTX 780 Ti Classified||MSI GTX 1070 Quick Silver|
|Cooling||EK LTX CPU Block, Swiftech MCP655 Vario (set at 3),|
Swiftech MCR320 + PA 120.2, 5x Yate Loon High @ 1K RPM
|EK-Supreme LTX Water Block|
360 mm Radiator
|EK LTX CPU Block, Swiftech MCP655 Vario (set at 3),|
Swiftech MCR320 + PA 120.2, 5x Yate Loon High @ 1K RPM
|OS||Windows 10 x64||Windows 10 x64||Windows 10 x64|
And the test system:
|CPU||Intel i7-7700K @ stock (4.2Ghz) and 4.5 GHz all cores (motherboard default).|
|CPU Cooler||Custom Loop with EK LTX CPU Block, Swiftech MCP655 Vario (set at 3),|
Swiftech MCR320 + PA 120.2, 5x Yate Loon High @ 1K RPM
|Motherboard||MSI Z270 XPower Titanium Gaming|
|RAM||G.SKILL Trident Z 2x8GB DDR4-3866 18-19-19-39 2N|
|Graphics Card||MSI GTX 1070 Quick Silver|
|Hard Drive||512GB Toshiba OCZ RD400|
|Power Supply||Seasonic Platinum-1000|
|Operating System||Windows 10 Pro x64 (fully updated)|
Both MSI and G.SKILL were kind enough to send a board and memory for the Kaby Lake review. MSI delivered a shiny new flagship motherboard in the Z270 XPower Gaming Titanium, while G.SKILL sent over some smoking fast 2x8GB DDR4 3866 CL 18-19-19-39 2N kit (F4-3866C18D-16GTZ). Be sure to check out the links above for details on those products and how they fared when we pushed on things a bit. We have some screenshots below of the hardware and the test system.
And the test system. See? I told you the MSI GTX 1070 Quick Silver I reviewed would look good on one of these boards! Wish I had the all black or white/grey G.Skill!
All benchmarks were run with the motherboard being set to optimized defaults (outside of some memory settings which had to be configured manually). When “stock” is mentioned along with the clockspeed, it does not reflect the boost clocks, only the base clocks. I tested this way as it seems motherboards are different in how they work out of the box. The MSI I have sets all cores at the max turbo (4.5 GHz) for the CPU while other boards follow the Intel Turbo specification. So, it is more of a “run what you brung” type of testing for stock speeds.
After testing, we then shifted to comparing the 6700K and 7700K at the same clockspeeds and memory settings. This testing will flesh out the difference in Instructions Per Clock (IPC) between the two.
- AIDA64 Engineer Memory, CPU, and FPU Tests
- Cinebench R11.5 and R15
- x265 1080p Benchmark (HWBOT)
- POV Ray
- Intel XTU
- SuperPi 1M/32M
- WPrime 32M/1024MB
All CPU tests were run at their default settings unless otherwise noted.
All game tests were run at 1920×1080 and 2560×1440. Please see our testing procedures for details on in-game settings.
- 3DMark Fire Strike Extreme
- Rise of the Tomb Raider
- Crysis 3
- Dirt: Rally
- Ashes of the Singularity
Our first results are using AIDA64 Engineer using the CPU, FPU, and Memory tests. The results speak for themselves here as there isn’t a discernible difference between the quad core CPUs outside of its different stock clock speeds and margin of error. This will be a theme moving through the stock testing.
|AIDA64 CPU Benchmarks – Raw Data|
|i7 5775C @ 3.3 GHz||45880||23655||326.8||16365||3918|
|i7 6700K @ 4.0 GHz||48776||25823||358.2||18248||4594|
|i7 5820K @ 3.3 GHz||59395||30319||434.1||22974||5179|
|i7-7700K @ 4.2 GHz||51215||28584||381.8||19117||4805|
|AIDA64 FPU Benchmarks – Raw Data|
|i7 5775C @ 3.3 GHz||5873||24513||13184||4625|
|i7 6700K @ 4.0 GHz||7532||34048||18322||4830|
|i7 5820K @ 3.3 GHz||6674||39957||21475||4974|
|i7-7700K @ 4.2 GHz||7850||35630||19174||5053|
|AIDA64 Memory Benchmarks – Raw Data|
|i7 5775C @ 3.3 GHz||37589||37376||42403||53.1|
|i7 6700K @ 4.0 GHz||43542||44661||39900||43.4|
|i7 5820K @ 3.3 GHz||51596||46939||58553||59.9|
|i7-7700K @ 4.2 GHz||50469||50123||46718||42.2|
Next we will move on to something a bit more tangible/productivity based with compression, rendering, and encoding benchmarks. When left to their base clocks only, we can see just how much the cores matter in many of these tests when you see the gaps between the Broadwell CPU and Haswell-E. As far as the 6700K and 7700K there isn’t much difference outside of a few percent here. This is what was expected, due to the different clock speeds.
|Cinebench, 7zip, POVRay and x265 Benchmarks – Raw Data|
|CPU||7Zip||CB R11.5||CB R15||POVRay||x265 (HWBOT)|
|i7 5775C @ 3.3 GHz||23517||8.39||774||1560.85||23.79|
|i7 6700K @ 4.0 GHz||25812||9.79||893||1900.37||21.98|
|i7 5820K @ 3.3 GHz||30617||11.00||1012||2082.87||22.42|
|i7-7700K @ 4.2 GHz||26492||10.5||974||1957.19||32.93|
Keeping the momentum going, we move into our benchmarking applications in Intel XTU, WPrime, and SuperPi. In these tests, we didn’t really see much of a difference here either. The 7700K is faster throughout when compared to the 6700K (outside of the XTU result). Take a guess at which application is single threaded…
|Intel XTU, SuperPi, and wPrime Benchmarks – Raw Data|
|CPU||Intel XTU||wPrime 1024M||wPrime 32M||SuperPi 32M||SuperPi 1M|
|i7 5775C @ 3.3 GHz||996||183.221||5.899||520.370||10.374|
|i7 6700K @ 4.0GHz||1296||158.924||5.111||471.972||9.141|
|i7 5820K @ 3.3 GHz||1351||142.087||4.763||541.953||10.883|
|i7-7700K @ 4.2 GHz||1268||145.188||4.797||441.233||8.407|
As far as the games go, we tested both at 1080p and 2560×1440 resolutions to see if we can see a difference between them. I’m betting no… and won that bet. All results here between these games at our settings showed maybe 1 FPS difference which can be attributed to run variability. No improvements here as expected.
As for the synthetic benchmark, 3DMark Fire Strike Extreme, no difference here at the same clocks and memory speeds.
Head to Head Results
In our head to head results, we put both my 6700K and 7700K against each other at 4.5 GHz and walked out of the room. When we returned, both were still standing. The results speak for themselves I think and are just what we expected… nary a difference outside of margin of error. The real good stuff, in this chip for our crowd is with having potentially higher overclocks over the 6700K.
Below is a quick screenshot of this CPU stable at 5 GHz. As you can see I threw most everything I had at it, just south of P95. I ran the AIDA stress test for nearly 11 hours (trust me, it was an accident, I fell asleep). I landed at 1.41V (confirmed with a multi-meter).
To reach the XMP memory speeds of 3866, the MSI XPower Gaming Titanium is putting through 1.45V to the System Agent and 1.4V to the I/O. This seems like a ton to me, using Skylake as a knowledge base, but after asking MSI about it, they have confirmed this is intentional and is to support weak IMC’s and fast ram as I have. That said, I was able to lower the SA and IO voltages to 1.25 and 1.20V respectively for this clockspeed. Below that, the board would have problems booting and figuring itself out.
One thing you will notice are the temperatures. They are pretty hot here in my loop peaking at 88 °C (room temp was 25 °C – readings in the graphs were normalized to 22 °C). So, we have more overclocking headroom, but we are in a temperature conundrum. Time to get some LN2 on this thing soon! In the meantime, while gaming, temperatures only peaked around 70 °C so we were just fine with normal use at this clockspeed. I’d ride a CPU like this daily!
So I decided to push on things a bit and see where I could get from a benchmarking perspective on this CPU. I dropped HT to save a few degrees, raised the voltage to 1.45, and end up at 5.1 GHz. Not a huge increase. 5.2 GHz “stability” for the benchmarks below, jumped to 1.5V+. Not many systems could keep that chip cool enough to run daily without de-lidding and a little prayer.
The memory was solid at 4K and automatically jumped to CL19 (from CL18). I didn’t have enough time to really crank on the timings or push the speeds. For reference, I needed to run 1.3V System Agent (SA) voltage, and 1.3V for IO voltage to reach and runs those clocks. For those using the XMP profile on the Z270 XPower Gaming Titanium, make sure you look at the SA/IO voltages. At least with my IMC and sticks (as always, your mileage may vary), I was able to lower the voltages dramatically from the 1.4V it sets, even when running DDR4 4000 MHz. It may save you a couple of degrees too.
Power Consumption and Temperatures
In the graph below we tested power use of the system across several situations from idle, to Prime 95 Small FFT (with FMA3/AVX) to playing Crysis 3. The highest power draw from the wall I read was 271W while playing Crysis 3. Remember folks, this is with using the GTX 1070, a very fast GPU coming it at ~150W. If I overclocked the GPU, we may see 300W there while gaming.
Below are the results I gathered for the temperatures. Stock temperatures peaked at 65C running at 1.2V in this loop while AIDA64 it hit 61C on the hottest core. I stepped up to 5Ghz 1.4V and things got a bit more toasty. we really hit what many would be considered the limit on these high stress applications like P95 (uses FMA3/AVX), and the AIDA 64 FPU tests temp wise hitting 90C. I still had some headroom while running the regular AIDA 64 test, my preferred stability test. Gaming is several C less than AIDA so we have the reserve still if needed. This particular sample needs a de-lid to push further since there is voltage headroom. I wish they would bring back the solder TIM in the 2600K. It was said some of these can hit 5 GHz on air, however, I’m not sure my sample would qualify for such a feat considering its temperatures under a pretty beefy water cooling loop.
Where to start with this one folks? On one hand, we see what I feel is the face of complacency on the part of Intel with not having any real IPC gains in this refined 14nm+ CPU. On the other hand, why would they with no real competition in this segment (come on Zen). For the enthusiast, there is more to love here with the higher out of box clockspeeds as well as the better overclocking top end than Skylake.
But not every one is like us. Many use the iGPU for playing their games or watching their media. With the update to the Iris Plus graphics, and using the VB9 and HVEC 10-bit codecs, it will enable you seamlessly stream 4K video from your most popular providers (think Netflix, etc). Security was improved at a hardware level with Intel’s Authenticate providing a multi-factor solution protected in hardware which should minimize exposure to identify theft attacks. Using their Software Guard Extensions (Intel SGX) service providers can now use personal biometric credentials for secure online purchases. The new chipset brought with it an increase in PCIe lanes as many devices are now coming out which use them, such as the PCIe NVME based SSD’s and Intel’s own Optane memory which will use the M.2 form factor.
Let’s not forget the Optane memory either. We should see some drives come out (16GB flavor) in the 1H 2017. I am not sure how much better this will be for those that have an SSD, I have a feeling we will really see the benefits of this technology when the DIMMs come out… and hopefully they trickle down to the consumer level (if I am understanding things correctly). As you saw from that video, its many times faster in that form than even SSDs. We will see what kind of impact this has on the market or if it goes the way of the cache drives of today.
One thing we haven’t mentioned so far, but you may have seen it listed, is price. The 7700K will come in at $305 MSRP. This is $45 lower than the 6700K which is great news for all assuming we have availability of the CPU and prices do not inflate. The price of $305 is attractive as it stands with Skylake still hovering around the $350 mark in widely available locations. I have a funny feeling we will see Kaby come in a bit higher than that and Skylake could see a price reduction. Both are very solid platforms.
In the end, Intel has improved upon the venerable Skylake CPU and brought us Kaby Lake making it more secure, with more I/O support including additional PCIe lanes, an upgraded iGPU, and a higher top end for overclocking for the unlocked processors. From our perspective here, I don’t really see the point in upgrading to Kaby Lake from Haswell on forward unless you are looking for more connectivity or better overclocking, which many are. In my opinion, anything from Ivy Bridge on down, there is enough here to justify going right to it when buying new, especially if the price stays close to that $305 MSRP.
The question then becomes if the clockspeed increases, new/improved media capabilities, and power savings (for laptops) are going to get you into this platform. I wish Intel didn’t make it so tough. On one hand you have a couple generations of processors back to back that perform admirably, but on the other, that shows we are not seeing the bigger differences in performance anymore with the elongation of the latest architecture. Intel really does have a faster, by clockspeed, CPU in Kaby Lake over Skylake. My sample overclocked better than my above average 6700K did at the same voltage and we have more available PCIe lanes on the platform to boot. If you are looking for the latest and greatest on the mainstream platform, it has arrived in the form of Kaby Lake, and for us overclockers, the i7-7700K.
Joe Shields (Earthdog)