Intel's 7th generation processor has arrived; after the several leaks, being referred by some as NDA violations,  we can deliver the raw data Madshrimps has collected over the past weeks. Kaby Lake again does not adhere to the Intel Tick-Tock principle. Many would have expected a die shrink (Tock) because Skylake was manufactured on the same 14nm technology as the  unsuccessful desktop Broadwell generation. Nevertheless Intel has refined or reworked the 14nm technology and refers to it as 14nm+. Numerous transistor enhancements allow for higher clocks at the same voltage or otherwise requiring less voltage for similar clocks. We at Madshrimps prefer more speed for the same voltage. More is better isn't it?

 

 

On the left we have the Core i7 Skylake 6700K, on the right the new kid on the block the Core i7-7700K aka Kaby Lake. A small difference in heatspreader design and on the backside also minor changes noticeable.

 

 

Quickly comparing the two high end models of the 5th, 6th and 7th generation clarifies much:

 

 

A serious clock boost for the Core i7-7700K versus the previous Skylake model not only in CPU core speed, but also sporting a 300MHz higher Turbo Boost. The Core i5-7600K gets a 300MHz higher core speed and a Turbo Boost of a mere 100MHz higher versus the Skylake Core i5-6600K. No Changes in TDP, nor Cache over the 6th generation lineup announced.

Intel though has included a beefier Integrated Graphics, the HD 630, yet still not up to scratch with the performance of recent  AMD APUs or even the Iris Pro Graphics from the Broadwell series, but gaming on the IGP  is not the focus of Intel's latest release, however it supports VP9 decoding and even 4K HEVC 10-bit decode/encoding.  According to Intel the new Kaby Lake is capable of decoding 4K video at blistering speeds (60 fps, with a max bit rate of 120MBbps). The integrated GPU does all the calculations now, so in fact the Kaby Lake CPU is idling. This results in the aforementioned performance and for the Notebook users improved battery life.

 

Our review test bed for Intel's 7th generation processor has been the brand new Gigabyte Z270X Gaming7 from the Aorus series. The Aorus lineup was limited before to just some high performance Gaming laptops and peripherals. Gigabyte is adding some high quality motherboards to their Aorus Gaming brand.

The initial Bios was F2, which had some fail safe boot issues and a weird mouse lag in the BIOS; F4g solved those issues and the latest F4k bios even allows better compatibility with 4000Mhz memory modules. Without going into further technical details, this mobo was a breeze to work with and performed as it should. The BIOS layout and features have been simplified, making it easier to navigate through as most of them abundant technical settings have been left out.

A very solid board to say the least, Intel and Killer Network NICs onboard and of course the Intel Thunderbolt couldn't be left out, featuring nVIDIA SLI support and the mandatory RGB heaven galore features :)

 

 

 

We paired it with: 

• Intel Core i7-7700K.
• Gigabyte Aorus Z270X Gaming7
• Cooled by Be Quiet Silent Loop 240.
• 16GB G.Skill TridentZ 3600C17 Dual channel memory.
• MSI Lightning R290X graphics card.
• 2TB Western Digital Caviar Green HDD.
• Seasonic Snow Silent 760W power supply.
• Streacom BTC Open Benchtable.

 

First up the stock impressions, without a doubt the brand new Kaby Lake series provide a decent performance increase over its predecessor. The Core i7-7700K running out of the box 200MHz higher clock speeds, versus the 4Ghz of the i7-6700K. Logically, the Turbo frequencies also were raised to 4.5GHz for the Core i7-7700K versus the maximum 4.2GHz Turbo of the Skylake i7-6700K.

Take note that we used the 2133MHz memory divider instead of the 2400MHz one (now officially supported by Intel)

In the single threaded SuperPi 32M, the new Core i7-7700K delivers the fastest quad core result on the chart. The Devils Canyon Core i7-4790K comes close as its Turbo is just 100MHz shy of the new Kaby Lake i7.

 

 

 

On the older Maxon Cinebench R10 the brand new 7700K quad core is battling it out with the high end hexacore brethren of the Haswell-E series; pretty remarkable as these have 2 extra cores to crunch with, though the newer versions again shows a strong performance versus any quad core on the market, but this time the higher core specced CPUs are logically pulling away in this multi-threaded benchmark. Without a doubt the Core i7-7700K is Intel's fastest mainstream processor to date.

 

 

 

 

In our 2D test suite, mainly thanks to the increased clocks, the new Core i7-7700K blows the other mainstream competition away. The available computing power is breathtaking and is still getting boosted year after year. Let us see how this fairs in several games or synthetic benchmarks.

In Ashes of Singularity the Kaby Lake flagship tops the charts again in 1080P  or 4K mode. Gaming has become less and less dependent on raw CPU power over the years and the bottleneck at current gaming resolutions is far more GPU driven than before.

 

 

 

 

 

It still surprises me why Intel's PR machine still says the high end X99 platform is the way to go for maximum Gaming performance. Maybe if you stream, download, record and game at the same time, they could have a marketing point. However for casual gamers that run, also the usual Team Chat, a quad core solution is more than plenty. Investing in a higher specced GPU is way more beneficial to achieve higher Frames Per Second. Investing in the most extreme processor on the market, really no need for a casual gamer.

 

 

 

Benchers will applaud the extra MHz capabilities of the new Kaby Lake generation. As most synthetic benchmarks still scale with added processor oomph under the hood. But as soon the GPU becomes the bottleneck we are observing similarities as in the above game tests. At Ultra resolutions and maximum detail preset not even the fastest processor will contribute a lot to produce them extra Frames Per Second.

 

 

 

 

 

Time to add a second MSI Lightning R290X to see how well the Z270X and the Kaby Lake can handle the extra graphics power. Even though the R290X isn't the top of the bill anymore, these cards in Crossfire mode still pack some serious punch. In our two game benchmarks the 7th generation just out edges the Skylake setup, though a 5 FPS gain is minimal at plus 200 FPS isn't it? Once we up the resolution at 4K both our Radeon cards are being maxed out and there's hardly any difference anymore. Again it enforces our claim that there's no real need for Hexacore nor Octacore processors in a daily gaming setup.

 

 

 

 

 

 

 

To find out if there are any performance improvements due to architectural changes, we limit the clock speeds of our test suite processors to a mere 4GHz. In fact for our Core i7-7700K we had to down clock it to make the  Instructions Per Clock comparison (IPC) happen.

Let us start off with a single threaded benchmark, being again our beloved SuperPi 32M. A benchmark relying heavily on bandwidth combined with CPU speed. With all our processors clocked at the same speed it will become more clear if the new Kaby Lake CPU generation has a gain over its predecessors.

 

 

Mmm okay we are seeing no specific performance improvement here. So is Intel pulling off another Devils Canyon on us? Time to verify our initial finding with the other benchmarks in the test suite.

 

 

Well hate to bust the bubble, Cinebench Release 15 shows a similar score as the Sky lake generation. No IPC improvement to be spotted here, so performance wise Kaby Lake is a solid clock boost over the previous generation, but don't look for further architectural improvements besides the much improved encoding/decoding capabilities of the HD 630 iGPU.

 

 

 

And it continues, no matter the benchmark or game we throw at the brand new Core i7-7700K, clock per clock it performs the same as the previous Sky Lake generation.

 

 

 

 

 

 

 

 

 

 

Time to see how well the new Kaby Lake generation responds to increased bandwidth. Since we didn't see any gains in Instructions Per Clock we assume it will scale in the same fashion memory wise as the Skylake series. Take note that the Integrated Memory Controller of the Kaby Lake series is capable of handling +4000Mhz easily.

To conduct our tests we are varying the memory speed from 2133MHz up to 3733MHz, the max our Gigabyte Z270X Gaming7 allowed us to do with BIOS F4g. Newer BIOSes or other boards might push our G.Skill B-Die kit close to the 4000mark, though that is not the concept behind this test, nor what this Gigabyte Aorus motherboard was designed for. The idea is to spot if a high end tweakable memory kit paired with the i7-7700K brings any significant performance advantages.

 

 

We maintained the memory timings at C15-15-15-35 2T between the below speeds:

• 2133MHz
• 2666MHz
• 3000MHz
• 3200MHz
• 3600MHz
• 3733MHz

Why keep the timings fixed you might ask? Otherwise it will be comparing apples with oranges.. Memory performance is always a mixture of the used memory ICs, their rated speed and timings. If we compare the popular Hynix MFR IC based 3200C15 memory with a B-Die based 3600C16 kit, there won't be much of a performance difference, even though that 3600 kit certainly looks faster on paper. By just adjusting the memory speed we exclude the timings variable; secondly by using the same memory ICs we also rule out any performance hits there.

 

 

The AIDA64 bandwidth tests shows the solid scaling when selecting a higher memory divider, especially going from 2133 to 3000MHz results in a big boost. From speeds beyond 3600 the performance increase becomes less and less prominent. We can spot similar behavior on the below benchmarks: take for example the Cinebench R15 benchmark below, which in general loves to be backed up by the memory and a high performance subsystem, so scales with any added bandwidth you can throw at it. Note the scaling has lost in most benchmarks its momentum with system memory speeds once over 3600MHz.

 

 

 

But not every application needs tons of bandwidth to perform. We are throwing the Blender benchmark into the benchmark suite, this to prepare for another important upcoming CPU release, so we get a feel for the way it renders the scene. A very minor performance improvement is observed, though nothing earth shattering. The same accounts for the two game benchmarks. Memory speed is not that important to gain any noticeable Frames Per Second in your favorite games. If you are a gamer look elsewhere, a decent main stream memory kit will be just fine, no need to cash out!

 

 

Another variable we can play with in the Bios is the CPU cache speed. Is it responsible for the bridging between the bandwidth demands of the CPU cores and the installed memory on the motherboard, especially with X99 we have observed solid performance improvements in combination with fast memory when the CPU cache speed was overclocked above the stock value. With Kaby Lake Intel already boosted the Cache speed from 4000MHz to 4200MHz versus the former Sky Lake offering.

 

 

We tested with our i7-7700K CPU running the cores at 4800MHz, the 16GB Dual Channel memory runs at 3000MHz and the below different Cache speed scenarios were tested:

• 4000MHz (underclocked to Sky Lake speeds)
• 4200MHz (stock speed)
• 4400MHz
• 4600MHz
• 4800MHz (even though read out in CPU-Z as 4600MHz we observed a performance boost)

 

 

The AIDA64 test shows only a mild improvement when increasing the Cache speed. Cinebench R15 and the Fritz Chess benchmark however show some noticeable scaling.

 

 

 

 

Both the rendering and game tests only show a tiny performance boost, nothing really significant to recommend running the cache speed at higher levels. The Intel stock value of 4200MHz is pretty spot on for most applications. Only tweakers and overclockers will try to go for a 1:1 ratio between the CPU speed and Cache to achieve maximum performance.

 

 

The integrated GPU solutions of the Intel brand haven't been too well performing in games. Only the Iris Pro iGPUs found in the high end Broadwell processors (e.g. i7-5775C) are truly a match for the AMD APUs. Even though the new HD 630 has some raw encoding horsepower under the hood we were unable to conduct proper tests to confirm this, so we have to come back on that matter as the CPU sample had to go back to Gigabyte. For now we can only highlight the gaming performance.

Like mentioned before AMD still has the upper hand here providing decent game performance at 1920x 1080 resolution, no matter the detail preset. Between the last Intel generations the performance improvements from the Intel HD 4600 (i7-4770K) were huge. Stepping up from Intel HD530 (Skylake) to Intel HD 630 is less phenomenal. Let us take a closer look:

 

 

Stock runs:

 

 

If we compare the HD 530 and the HD 630 purely on gaming performance, in Bioshock  we see hardly notice any improvement. Only the Iris Pro iGPU and the recent AMD APUs can run this game on Medium detail preset at 1080P. In Tomb Raider there's a small performance gain observed, Anyway the game looks pretty outdated in a Low detail configuration.

 

 

In the graphics test of Futuremark's Firestrike we notice the same tiny boost over the HD 530, though nothing Earth-shattering compared to the difference we got with the HD 4600 series and the HD 530 when reviewing Skylake.

 

 

iGPU Memory scaling:

 

The above stock runs were performed with the system memory running at 2133MHz. Since these iGPUs use the same system memory, faster memory should provide a boost in performance. DDR4 3000Mhz is pretty common nowadays and DDR4 3600MHz is already a bit more niche but might become a standard soon.

 

 

Both games benefit nicely when using faster system memory, especially at the low preset there's more headroom for lag free gaming, even Firestrike puts out a way higher score. Nevertheless if one intends to game on these iGPUs, lowering the resolution is the only viable path to take.

 

 

 

Memory allocation:

 

Since we couldn't overclock the iGPU on our Gigabyte  Z270 Gaming 7, the last option for us to test was the amount of allocated memory to the iGPU. To cut a long story short we went from 96MB to 256MB and saw hardly any changes in output. It might have a substantial effect when decoding, but for gaming purposes this BIOS feature seems pretty pointless.

 

 

 

 

What we have seen from the previous pages is that there is not that much new with this 7th generation. Higher clock speeds together with the encoding power seem to be the biggest assets for Kaby Lake. But hold your horses, tweakers and overclockers will cheer for Kaby Lake. Remember Sandy Bridge when people were running their processors at that magical 5GHz speed on air or water? No extreme cooling was needed to get very strong performance back in those days! Now with each CPU generation Intel released we saw those daily overclocks being reduced generation on generation. Okay the new releases were more efficient clock per clock, though the sweetness of those OC'ed Sandy Bridges still left most with a hunger for more speed.

 

(click to enlarge)

 

Well sadly we didn't reach 5GHz on our test sample, but 4.9Ghz on all cores was pretty easy with a slight bump in voltage. Out of the box our CPU was set to run at its default speeds of 4.2GHz and 4.5GHz Turbo clocks at 1.35Vcore.  By adding +0.05Vcore we reached 4.9GHz on ALL cores, which is kind of awesome to increase the clocks all round without really the need of massive voltage adjustments.

An important setting to play with is the AVX offset value, to gain higher clocks it is advisable to leave this usually a few multipliers below the core ratio setting (negative offset). So when the AVX instruction is not used you have the full processor speed available, once AVX is used it will drop all cores to the set AVX Offset value.

 

 

Here are some comparisons between stock and our OC'ed setup running at 4.9Ghz:

 

 

 

For 2D applications that can use all the power you throw at them, they will love the headroom these Kaby Lake processors have under the hood. It is up to you to unleash the beast! Gamers will have plenty of power with the out of the box clocks.

 

 

So how come we weren't able to reach 5GHz and beyond? Well the answer is pretty simple: HEAT. Even at stock the Kaby Lake was crunching at around 74°C, this while being cooled by the Be Quiet Silent loop 240. Not the best performing AIO on the market, but a very decent product after all with an excellent cooling versus noise ratio. When we raised the clocks to 4.9GHz we saw a drastic performance increase but also the core temperatures were going bonkers. In Realtemp we logged 95°C on some cores.

We effectively ran 5GHz but the multithreaded scores got lower as the CPU activated the throttling mechanism to protect itself from overheating. To really drop the temperatures, delidding the Kaby Lake processor is a necessity. On the internet temperature drops of 20 to even 30°C have been reported, which is massive. Sadly the processor sample provided had to go back in its original state, so delidding was not an option. Next week we will provide more data from CES where we will have the chance to play with the new 7th generation on liquid nitrogen.

Our chaps at Eteknix already launched their delidding article and the temperature gain over the stock Intel thermal paste is around 20°C, allowing their CPU to hum at a whopping 5GHz and according their conclusion never even breaching 80°C.

 

 

(pictures courtesy of Eteknix.com)

 

 

Early LN2 results have shattered the rankings at the overclockers' favorite website, HWBot.org. On liquid Nitrogen American overclocker "Splave" aka Allen Golibersuch has set new World Records in most 2D benchmarks. Powered by his trusty ASRock Z170 Formula he even surpassed 7GHz with just 2 cores running. In full multithreaded mode he was able to rock hard at 6.8Ghz, something which only a handful gem Skylakes could achieve. There will be more to come on the CES trade show for sure.

 

 

 

Well ladies and gentlemen there you have it. Intel's latest 7th generation aka Kaby Lake looks at first glance as a small evolution in processing power. Indeed we were a bit intrigued by the fact that there were no logical Instructions Per Clock gains to be detected. Was that needed? Not really, but one expects from a new generation that something has to be better in regards of its predecessors. In return Intel has upped the clocks by a fair amount.

The Redmond Giant has deviated from their own heralded Tick - Tock mechanism before and that will be why many talk about a Devils Canyon revisited. Well it might look like it, but it is not exactly true. Devils Canyon was great on paper as it was Intel's first 4Ghz desktop processor, but a decent Haswell CPU could clock as high if not higher than the DC counterpart. On more exotic cooling Devils Canyon was blown away by its predecessor, even though the marketing promised the inverse. Intel has learned from the past and is not making the same mistake here.

Yes they did a similar frequency increase with Kaby Lake, but this generation has got plenty of unlocked power under the hood: it is up to you to unleash the real beast. Kaby Lake is logically a tiny bit more energy efficient at out of the box speeds, But this has a far bigger effect in the mobile segment than on the desktop platform.

Only thing one has to worry about is how to keep these crunching monsters cooled when aiming for that magical 5GHz frequency for daily usage. Delidding is probably the option to just achieve that, though there are rumours that Intel is able to improve on short term the quality of the silicon for even higher daily clocks.

Second good thing is that we don't need to upgrade our motherboards to the latest Z270 chipset, this processor generation runs fine on the Z170 platform (after a BIOS update), a chipset which has matured well over the last year. If you need to have all the functions Kaby Lake will allow the new chipset is needed.

 

 

Besides the increased clocks and superb overclock headroom, the en/decoding strength of the new HD 630 is probably the biggest asset of the new Kaby Lake generation. Supporting HDCP 2.2, Playready 3.0 from Microsoft and able to encode or decode VP9, even in 4K 10-bit HEVC. Question is will many desktop users use this platform to do just that? Maybe again more for the mobile segment, though it is nice to have those features under the hood. Let the HD 630 handle those 4K streams, it will be a delight!

So what's the deal here, some might say not really a big deal. Well it is all about refinements made to an already amazing CPU generation. I see no real need to start upgrading to Kaby Lake if you own a modern powerful CPU like the Haswell series, as always it will depend on your specific needs and the system you are currently gaming/working on, though this CPU generation is a perfect blend of some nifty new features integrated inside the HD 630 iGPU, an improved Memory Controller and the combination of them sick clock speeds from Sandy Bridge with the performance benefits of the Sky Lake generation. What's not to like?

 

 

Though it is hard to give a specific award, consider the Kaby Lake as a refresh of its predecessor with some extra refinements. The increase in clock speeds and available overclock headroom easily make up for the lack of IPC improvement. We can vouch that it truly performs and it makes Intel claim the title of the fastest mainstream desktop processor once again.

 

 

 

 

We wish to thank Gina and Bernice from Gigabyte for the Z270X Gaming 7 AORUS review sample. Kuddos ladies, we owe you (again) Hopefully we can update the current article soon with some nice pictures and hardcore footage from CES Las Vegas and update the database with other Kaby Lake models.

 

 

 

Thanks to Nils from Seasonic for the Snow Silent PSU and the Tones crew for the years of support!

 

Technik PR for the Be Quiet Silent Loop 240 series