Intel Skylake Core i7-6700K IPC & Overclocking Review

Today we finally get to share with you our Intel Skylake experiences. As we like to, we are going to focus on Instructions Per Clock / IPC and overclocking this new CPU architecture. We hope to give our readers a definitive answer to whether or not it is time to make the jump to a new desktop PC platform.

continued...

Please keep in mind as you go through our data that all our CPUs and RAM in our top group are clocked identically. 4.5GHz CPU clock and 1866MHz memory clock. Memory timings are spelled out on our test system page, and I would give those a look because as we scale our memory clocks, our timing lantencies do in fact increase greatly and you will see this impact some of our testing. In the group of data in the middle we show Skylake performance while scaling memory from 2133MHz to 3600MHz while keeping the CPU clock at 4.5GHz. Finally in the bottom line of data we show Skylake overclocked to 4.8GHz with 3600MHz memory.

Also it is very worthy of pointing out that 2133MHz is the "stock" or "default" memory clock on a Skylake / Z170 chipset platform. We are showing retarded 1866MHz DDR4 speeds in order to have an "apples to apples" comparison to our previous DDR3 platforms and we were simply curious about this and we think that having it in our charts paints a truer picture of the Skylake processors.

It is highly unlikely that anyone will be using DDR4 at these lower speeds. According to PC Hound, as of writing this the most popularly purchased DDR4 memory kit is rated at 2400MHz. The two most popularly purchased memory kits behind that are rated at 3000MHz.


Multimedia Testing

Outside of gaming and encoding, there are few applications on the desktop that will push our systems to the limits, this especially becomes apparent when we start talking about multi-core processors that are now the norm. Some multi-thread aware encoding and content creation applications are starting to reach into all available threads and truly utilize the processing power of these multi-core CPUs.

The benchmarks below all represent very real world situations just like you would run into at home while encoding video from your camcorder - or while using a picture or video editing program - or while encoding music for your MP3 player- or encoding a DVD for saving it to your hard drive or mobile device to allow you easier access to the content. We have included WinRAR, a very popular compression program used by many when sending or sharing files.

We have also included the synthetic Cinebench 11.5 and POV-Ray benchmarks as well that should give you an idea about how 3D production programs will perform when rendering scenes.

We have simply timed our various tests on the different systems and supplied you with the amount of time it took for the system to fully build the file. SSD Scratch drives and SSD OS drives were used properly as well to make an attempt to not bump into any IO bottlenecks.


Cinebench 11.5

Article Image

Cinebench gives us a much clearer look at what Skylake can do for us on the "desktop," at least for those of us using it to render graphics with. We see a solid 11% increase of Skylake over Haswell, a 20% increase over Ivy Bridge, and a 25% increase over Sandy Bridge.

In this benchmark, we see that Skylake memory bandwidth has very little impact on its overall score scaling.

POV-Ray v3.7 RC3 64-bit Benchmark v 2.00

Article Image

POV-Ray is a ray tracing for creating high quality graphics. We are using the benchmark included in the software and using its multicore ability. Again we see Skylake stretch its legs. Skylake rewards us with a 16% decrease in render time compared to Haswell, a 21% decrease compared to Ivy Bridge, and a 27% decrease compared to Sandy Bridge.

Like in Cinebench, we see no gains to be had by scaling the memory bandwidth in this specific instance, in fact again, we see relaxed memory latencies peek up again.

Handbrake 0.9.9.5470 64-bit

Article Image

Here we are using a 30 minute movie ripped from a DVD and encoded with X.264 utilizing the "Android" preset in Handbrake. Handbrake takes advantage of threading on different levels depending on how the encode is set up.

Article Image

In this instance we are using Handbrake to encode a 2.5 minute movie trailer ripped from a Blu-ray Disk to a 1080P MKV file again using the X.264 codec.

The performance summaries for both our DVD and Blu-ray Handbrake encodes are virtually identical. Skylake encode times are decreased 10% compared to Haswell, decreased 19% compared to Ivy Bridge, and decreased 24% compared to Sandy Bridge.

LAME v3.99

Article Image

LAME is a widely used single threaded application that is used here to convert a 1 hour long .WAV ripped from a CD to a 128-bit MP3 file. Again, nothing too exciting to see here, but here are the stats. 4% decrease over Haswell, 8% decrease over Ivy Bridge, and 10% compared to Sandy Bridge.

WinRAR v4.2 64-bit

Article Image

This application, is dependent on both clock speed and memory bandwidth. Values were repeatable. All of these very similar times are blindingly fast for a zipping up a 560MB folder full of misc. files of all types and sizes.

We have always known WinRAR to be extremely CPU clock and memory bandwidth sensitive and we get to see one of the first cases where memory bandwidth actually makes a difference in CPU performance scaling.