Review Index:

The Coffee Lake Story: Intel Core i7-8700K and Core i5-8400 Review

Author: Ryan Shrout
Subject: Processors
Manufacturer: Intel

SYSmark, WebXPRT, 7-zip

Testing Configuration

  • 7-zip Compression
  • Audacity MP3 Encode
  • Blender
  • Cinebench R15
  • Euler 3D
  • Geekbench
  • Handbrake
  • POV-Ray
  • SiSoft Sandra
  • SYSmark 2014 SE
  • WebXPRT
  • X264 Encode

The full testbed configuration is listed below.

Test System Setup
CPU Intel Core i7-8700K
Intel Core i5-8400
Intel Core i9-7980XE
Intel Core i9-7960X
AMD Ryzen Threadripper 1950X
AMD Ryzen Threadripper 1920X
Intel Core i9-7900X
AMD Ryzen 7 1800X
Intel Core i7-7700K
Intel Core i5-7600K
Intel Core i7-6700K
Intel Core i7-6950X
Intel Core i7-6900K
Intel Core i7-6800K
Motherboard ASUS STRIX Z370-E Gaming (Coffee Lake)
ASUS X399 Zenith Extreme (Threadripper)
ASUS Prime X299-Deluxe (Skylake-X)
ASUS Crosshair VI Hero (Ryzen)
ASUS Prime Z270-A (Kaby Lake, Skylake)
ASUS X99-Deluxe II (Broadwell-E)
Memory 32GB Corsair Vengeance DDR4-3200 (Running at DDR4-2400 in all configurations)
Storage Corsair Neutron XTi 480 SSD
Sound Card On-board
Graphics Card NVIDIA GeForce GTX 1080 8GB
Graphics Drivers NVIDIA 378.49
Power Supply Corsair RM1000x
Operating System Windows 10 Pro x64


SYSmark 2014 SE

SYSmark® 2014 SE (Second Edition) is an application-based benchmark that reflects usage patterns of business users in the areas of Office Productivity, Data/Financial Analysis and Media Creation. Joining these in SYSmark 2014 SE is a new Responsiveness scenario which models ‘pain points’ in the user experience when performing common activities. SYSmark 2014 SE features the most popular applications from each of their respective fields.

View Full Size

View Full Size

View Full Size

View Full Size

View Full Size

View Full Size

The 8700K has an outstanding showing in SYSmark 2014 SE with a score beat only by the 16-core and 18-core Skylake-X parts released last month. It comes in 11% faster than the 7700K and 28% faster than the Ryzen 7 1800X. Even the Core i5-8400 does well with a score of 1365, matching the Ryzen 7 and Threadripper parts. From a power consumption stand point, the 8700K and 8400 use less power than the 7700K as they are able to complete some of the workloads quicker.


WebXPRT 2015 uses scenarios created to mirror the tasks you do every day to compare the performance of almost any Web-enabled device. It contains six HTML5- and JavaScript-based workloads: Photo Enhancement, Organize Album, Stock Option Pricing, Local Notes, Sales Graphs, and Explore DNA Sequencing.

It runs these four tests seven times each:

  • Photo Enhancement: Measures the time to apply three effects (Sharpen, Emboss, and Glow) to two photos each, a set of six photos total.
  • Organize Album: Measures the time it takes to check for human faces in a set of five photos.
  • Stocks Option Pricing: Measures the time to calculate financial indicators of a stock based on historical data and display the result in a dashboard.
  • Local Notes: Measures the time to store notes securely in the browser's local storage and display recent entries.
  • Sales Graphs: Measures the time to calculate and display multiple views of sales data.
  • Explore DNA Sequencing: Measures the time it takes to filter eight DNA sequences for specific characteristics.

Each test uses different combinations of HTML5 Canvas 2D and Javascript, common elements in many Web pages, to gauge how well your device and browser work together in everyday Web browsing situations.

View Full Size

View Full Size

The Core i7-8700K is 9% faster than the Core i7-7700K and the Core i5-8400 comes surprisingly close to matching the Kaby Lake CPU as well.  

7-Zip Compression

View Full Size

The new 8700K generates the best single, double and quadruple threaded 7zip scores we have measured here at PC Perspective, quite a feat! Clearly the new Coffee Lake chips combining clock speed improvements with higher core counts is resulting in impressive results. Comparing the 6-core 8700K to the 8-core Ryzen 7 1800X, the Intel processor has the edge.

Video News

October 5, 2017 | 09:06 AM - Posted by Dark_wizzie

Thanks for the review, and I look forward to the high end watercooling article later. So you guys got a 7700k to 5.1 @ 1.375v @ up to 70C, while this 8700k also went to 5.1 but @ 1.2v and went to upper 80s.

No clock for clock testing? I would've thought it would be more applicable here than in Kaby Lake.

October 5, 2017 | 09:07 AM - Posted by Ryzen Shout (not verified)

I love the smell of coffee In The Morning

October 5, 2017 | 09:12 AM - Posted by Dark_wizzie

I love the smell of Ryzing Shrout in the morning. :^)

October 5, 2017 | 10:24 AM - Posted by Jimbo (not verified)

I love the smell of competition :)

October 5, 2017 | 09:23 AM - Posted by A (not verified)

How can you possibly not include the 1600x? Direct 6/12 core/thread comparison! And it would get better performance per $$ than R7.

October 5, 2017 | 10:57 AM - Posted by Ryan Shrout

Honestly, it was a graphing / data issue. We are working on an IPC follow up of 6c vs 6c!

October 16, 2017 | 01:29 AM - Posted by throAU (not verified)

I think you also need to include platform costs.

8700k, or any intel coffee lake cpu for that matter is useless without a board, and the only options right now are z370 based.

On the ryzen side, you can stick a 1600 or 1700 on a cheaper b350 board and still get overclocking ability if desired, or actually get a top end 8 core AMD part on that platform.

"$180" for the CPU cost on the 8400 is misleading - there's no cheap boards for it and won't be for some time.

Also, these parts are not shipping at anywhere near the RRP at the moment either.

October 5, 2017 | 09:43 AM - Posted by Amit (not verified)

Did you notice the mistake in the pricing that completely skewed your results?? $18 for the i5 - please fix this!!

October 5, 2017 | 10:57 AM - Posted by Ryan Shrout

Ha, that's a white font on a white background issue. It's set at $184. :)

October 7, 2017 | 09:39 AM - Posted by killferd

That was sneaky man...real sneaky.

October 5, 2017 | 10:11 AM - Posted by Brogen

It's nice to see the usual, but I'd like to see storage solutions of how the processors run also. Get an overall view on how things compare.

October 5, 2017 | 10:22 AM - Posted by Jimbo (not verified)

How could you not include the Ryzen 5? I really like PcPer, but when you don't include the obvious price/performance competitor you guys look biased at worst and incompetent at best.

That being said, the new Coffee Lake i5 looks like a great CPU for gamers and the price is right. I'll be recommending this to my buddy who is still on a 2700k and is very Ryzen-shy!

October 5, 2017 | 10:42 AM - Posted by JohnGR

I am a little confused here.

In the beginning of this page you talk about "current prices"

"Pricing is based on current pricing as of the writing of this review."

then I read this paragraph

"(Note that since the creation of this data, AMD Ryzen 7 1800X has been selling for $399 and the Ryzen 7 1700X has been selling for $299. This definitely changes the performance per dollar story towards AMD, and giving the 1700X in particular, a significant jump.) "

that makes me think that the above charts are based on MSRP prices(old AMD prices).

October 5, 2017 | 10:58 AM - Posted by Ryan Shrout

Writing and publishing are different. Also, there is a graph above that clearly shows what prices the numbers are based on.

October 16, 2017 | 01:30 AM - Posted by throAU (not verified)

Also... go price up a coffee lake CPU from somewhere that actually has them, then re-compare :D

October 5, 2017 | 10:55 AM - Posted by PricePerformanceMadPriceCutsIncoming (not verified)

Not that much gaming improvement and at higher resolutions things become more equal acroos all of Intel's eariler SKUs and even with AMD's Ryzen offerings. I looked at the TechPowerup/other sites figures also and Intel is ahead of AMD in gaming but Intel's is not far enough ahead of Intel's eariler SKUs for most other tasks and that Intel 6 core 12 thread K series SKU has to be overclocked to 5.0 to make up for its 2 cores less than the Ryzen 7 1700X/1800X on some multitheraded workloads that really like the extra cores. So how long will that 5.0 overclock be maintained for long periods of time relative to Ryzen 7 1700X/1800X 8 core/16 thread variants that appear to be doing fine clocked at 3.7GHz on some multithreaded workloads relative to the 8700k that is overclocked to 5.0GHz.

Intel will have to compete with its own eariler SKUs in addidion to AMD's Ryzen 8 core/16 thread SKUs that are really coming down in retail price lately and Ryzen's overclocks are getting better as the current die production matures and the overall binning improves in addition to the better memory overclocking that Ryzen really loves that is now enabled more with each AMD AGESA update.

So AMD is Can agressively respond with better Ryzen pricing and Intel has to compete with Itself(eariler K series SKUs) and AMD's Ryzen 7 serise SKUs that are improving over time with tweaks up until the Zen+ at 12nm products arrive in Q1/later 2018. AMD Ryzen has decimated Intel's low end product lines in the price/performance metrics and any overall downward pricing of the on the Ryzen 7 series SKUs fruther below the price of some Intel quad core variants will be that much worse on the lower end price/performance metric competition between AMD and Intel.

Intel's still has an Integreted graphics line that AMD is about to enter competition against in the mobile APU market at first from AMD's Raven Ridge Mobile SKUs. But AMD really will need to look at its 12nm Zen+ desktop APU options and maybe think about offering some Interposer based High end Desktop APU variants with Zen CPU cores, a larger Vega die, and HBM2 and I know that AMD has a workstation grade APU in the development phase for the workstation markets! And we all know that consumer Threadripper was derived directly from Epyc so maybe there will be the same sorts of consumer variant of any high power interposer based desktop APU made from that workstation grade APU that AMD has in devlopment.

October 5, 2017 | 03:12 PM - Posted by Photonboy

Well, there aren't many applications where the R7-1800X beats the i7-8700K when both are stock settings, so it's hard to justify Ryzen if you want processing power.

I think the highest I saw was the R7-1800X winning by 13%, so sure it probably works out you need to overclock the i7-8700K to about 5GHz to match the R7-1800X at its highest overclock for that application.

But, if you overclock to 4.8GHz then you'd be ahead or tied in all but a handful of the applications with only a very small boost to the R7-1800X (or R7-1700 if you can get a good overclock too).

8C/16T ZEN2 will obviously do better, and the i7-8700K is probably the best CPU this Intel platform will see, so investing in AM4 may make sense for some people...

I could see for example someone getting an R5-1600 for now, then maybe getting an 8-core ZEN2 CPU in the future depending on their budget and need.

(ZEN+ is a frequency boost on die shrink approx 2018, whereas ZEN2 is an architectural tweak as well approx 2019, so ZEN2 vs current Ryzen I would guess at 20% to 40% better performance, maybe even higher in some applications compared to NOW if we also see compiler optimizations)

*Basically I think Ryzen vs Coffee Lake depends completely on your budget and needs, though for a pure gaming machine my first choice would be the i5-8600K (6-core no hyperthreading, overclockable) which should be roughly $260USD.

I still have an i7-3770K @4.5GHz (and GTX1080), running games at 2560x1440 and don't feel a need to upgrade yet. I can run everything at mostly 60FPS with rare drops due to the CPU, and my money will go towards a good GSYNC monitor when prices drop a bit, not a platform upgrade.

I plan to keep this system five more years unless the motherboard or CPU dies.

By then it will be interesting to see if the few programs I might need a CPU with high core-count are optimized to benefit more due to a graphics card. OPENCL (not OPENGL) is an initiative that is starting to bear fruit, so for VIDEO EDITING in five years a good program may not need more than a few good CPU cores, but benefit mostly from a good graphics card. ADOBE has been pretty slow to adopt GPU optimizations, though other programs are doing much better with the GPU so hopefully Adobe does some major recoding.

October 5, 2017 | 04:50 PM - Posted by psuedonymous

"8C/16T ZEN2 will obviously do better, and the i7-8700K is probably the best CPU this Intel platform will see, so investing in AM4 may make sense for some people..."

Cannonlake has been rumoured to include an 8c16t processor as the i7 option, and to be on the x300 PCH platform. Whether Z370 boards will receive BIOS udpates to support that is unconfirmed but very likely (though we could end up in a P67 situation).

Intel has stuck to two-CPU-gens-per-socket since Socket H1 (Sandy Bridge), and with this being Socket H5 I'd expect two generations on it too. Not updating the physical keying was pretty silly though.

Whether Zen 2 will be able to bump clocks up is probably dependant on process optimisation. As we saw with Rx 4xx to Rx 5xx, adding extra headroom up top may come at the price of a disproportionate increase in power draw, or with Rx Fury to Rx Vega with an increase in die area spent on transistors to allow clocking up rather than additional functionality.

October 5, 2017 | 06:19 PM - Posted by ZenPlusHasSomePlusesAlso (not verified)

The Zen+ designs with a new stepping/tweaks will do some good also to increase AMD's first generation Zen micro-arch performance. If AMD on Zen+ can even get the memory controller working with faster memory then the Infinity Fabric latency will improve some more. And new tape-outs on a smaller process node mean more tweaking can be done than the average new stepping so do not discount the Zen+ improvments too quickly that will arrive before Zen2 comes to market.

No one even knows if the Infinity Fabric(IF) speeds are actually have to be coupled to the memory clocks or what IF to Memory Clock ratios that AMD may be able to provide on it's Zen+ designs, because the Infinity Fabric on Vega is not tied to any memory clock domains for that SKU. Zen+ will most definitely have clock speed improvments And AMD is able to fix any other problems with Zen's other functionality also that may be needing some tweaks.

Vega at 12nm may just get some nice clock speed bumps and some new variants with more ROPs(Vega has plenty of TMU capacity even more than the GTX 1080/1080Ti) and less compute than Vega 56. But even Vega 56 at 12nm may just do fine and Vega and DX12/Vulkan appears to be doning fine on Forza 7, other titles that are more optimized for the latest APIs that will replace DX11/OpenGL.

October 5, 2017 | 10:58 AM - Posted by Jim Lahey

I'm confused with the PCIe lanes. If I get the 8700k, I can run a 16x graphics card, but if I use just one 4x NVMe drive, I already max out all available PCIe lanes?

I feel like storage will be severely limited with the 8700k. Am I wrong?

October 5, 2017 | 11:23 AM - Posted by PricePerformanceMadPriceCutsIncoming (not verified)

Depending on what AMD does with its Threadripper 1920x pricing and getting the best binned Zeppelin dies utilized for that 8 core/16 thread X399 MB based product at maybe a lower price point and AMD making use of the X399's 64 PCIe lanes and 4 memory channels to pull folks more into the TR/X399 ecosystem. And that will help AMD win more sales before the Zen+ at 12nm competition arrives. If I where AMD I'd agressively lower the 1920X's pricing and use the X399 platform advantage to pull more customers into the TR/X399 ecosystem where updating without changing the motherboard will be a great option what with the TR/MCM being a SOC(notrhbridge/southbridge on the Zeppelin DIEs) anyways and and chipset improvments will come along with any newer Zeppelin dies.

AMD's 1920X binning proces can not only take advantage of the top 5% binned Zeppelin dies AMD can also choose the top perfirming CCX units on each Zeppelin die and get those clocks higher and memory overclcoking stable at higher memory clocks that the Infinity Fabric really loves and let the total TR/X399 platform standard features compete against Intel's more segemented product offerings. With AMD offering better standard features at no extra cost to win more sales against Intel on a overall platform to platfotm feature for feature basis.

October 5, 2017 | 11:27 AM - Posted by PricePerformanceMadPriceCutsIncoming (not verified)

Replace: 1920X
with: 1900X

damn brain not working!

But AMD could lower the 1920X pricing also.

October 5, 2017 | 11:56 AM - Posted by PCIeLane (not verified)

The motherboard chipset has many more PCIe lanes (z370 has 24 lanes) so you have to add MB lanes to CPU lanes, 24 + 16= 40

October 16, 2017 | 01:32 AM - Posted by throAU (not verified)

The chipset lanes are all running off a 4x from the CPU IIRC, so they have contention on them. If you run multiple devices at the same time on the chipset lanes they are only sharing 4x worth of bandwidth.

October 5, 2017 | 01:21 PM - Posted by psuedonymous

The drive connects via the PCH, which has 24 PCIe lanes to play with. If you;re purely transferring data between the CPU and SSD then you;re 'bottlenecked' via DMI 3.0 (effectively a PCIe 3.0 x4 link), but any DMA from the GPU or between other drives (also on the PCH) will avoid that link, and very few (if any) real world applications will ever come close to saturating the DMI link.

October 5, 2017 | 03:34 PM - Posted by Photonboy

The i7-8700K CPU has 16 lanes dedicated for graphics. (i.e. 1x16, or 2x8).

The CHIPSET is separate and supports up to 24 PCIe lanes.

The M.2 SSD's use the chipset lanes. So there are PLENTY of lanes available for a good graphics card, M.2 SSD's, SATA and other PCIe devices.

So to be clear, the CPU talks directly to the graphics card, whereas the M.2 SSD goes through the chipset first.

More info here (click the LEFT picture to see the block diagram for CPU and CHIPSET layout)

I thought that the M.2 SSD (PCIe) connected directly to the CPU, not the chipset but I can't find any evidence of that.

*EVEN IF THAT WAS TRUE it would not take away from the 16 lanes dedicated to graphics.

October 5, 2017 | 02:31 PM - Posted by eddman

@Ryan Shrout

What was the clock frequency under benchmarks? Did you get constant 4.4 GHz? Did you get 4.4 under cinebench MT too?

October 5, 2017 | 02:34 PM - Posted by Anonym (not verified)

Gold award for a CPU that has 0% IPC gain over previous generation and sucks more power than CPU with 2 extra cores and 4 extra threads. Guess Intel is shoving green up the rear side again.

October 5, 2017 | 02:36 PM - Posted by Dark_wizzie

You need to take a chill pill, bruh. It's just a CPU, the 8700k didn't kill your family or something.

October 5, 2017 | 04:35 PM - Posted by Allyn Malventano

Two additional cores at nearly the same price doesn't count for anything? Really?

October 5, 2017 | 10:05 PM - Posted by Anonym (not verified)

In a world where 1700x didnt exist it would have been revolutionary, but as things are right now more power draw than 8 core CPU, more expensive and just faster(or slower in multi threaded environment) than 1700x. its another crappy release from Intel. Lower end i3 and i5s are looking more attractive right now.

October 6, 2017 | 04:06 PM - Posted by Aparsh335i (not verified)


October 5, 2017 | 04:16 PM - Posted by Colin Pastuch (not verified)

Yay another generation of 0% IPC gains!

I find it pathetic that my 4.4ghz Sandy Bridge provides similar performance in 4k gaming as the 8700k. GPUs need to get a hell of a lot faster before I consider a CPU replacement.

October 5, 2017 | 05:09 PM - Posted by psuedonymous

Techreport tested the 2600k vs the 8700k. In most cases it was HALF the speed. Eurogamer did a test of an OCed 2500K last year, and it was lagging behind even a stock i5-6500. Even with OC, the gulf between the 2500k and 8700k is large.

October 5, 2017 | 05:38 PM - Posted by Colin Pastuch (not verified)

Tech report tested at 1080P which I haven't gamed on in over 6 years because it's shit. Try using a respectable display with at least 1440P and more importantly HDR support (TVs only sadly). At higher resolution the CPU isn't breaking a sweat because GPUs are so slow.

The biggest revolution in gaming is HDR displays but the computer monitor market failed miserably in that regard.

October 5, 2017 | 06:28 PM - Posted by Jeremy Hellstrom

That would be a GPU test, HDR isn't a CPU function by any stretch of the imagination.

October 7, 2017 | 04:23 AM - Posted by Tom Yum (not verified)

And that would be relevant if 57% of steam users used a gtx 1080 or higher amd therefore were in the position of being cpu limited. They don't, so it isn't relevant.

October 8, 2017 | 05:00 PM - Posted by not verified (not verified)

High end displays for gaming and content creation could make use of it, but there's so few games that support HDR right now it's hardly worth the investment.

October 7, 2017 | 03:53 PM - Posted by José Luís Ribeiro Lopes (not verified)

Ofc it has no gains at 4k performance, the limiting factor is the gpu not the cpu

October 5, 2017 | 04:52 PM - Posted by Anony_mous (not verified)

Where are the temp comparisons ?

October 6, 2017 | 02:56 PM - Posted by Allyn Malventano

Temperature comparisons are moot when you have power draw figures. Temperature varies widely based on cooler selection.

October 5, 2017 | 10:40 PM - Posted by agello24 (not verified)

im still good with my 1600.

October 6, 2017 | 12:55 AM - Posted by InTheHandsOfReviewersMostly (not verified)

Other than reviewers how much stocks of the Coffee lake K branded SKUs will actually be available and in what US and overseas markets and in large or small quantities. And what about the lower tier 300 series motherboards for any Coffee lake locked processor SKUs. Will there be a shortage of CPU/MB parts or will there be plenty of supplies to go around.

I'd like to see the Coffee Lake/300 series Motherboard roll out time line and does it streach across more than one calender quarter. Kaby lake is only around 10 months on the market and Ryzen has a lot to do with that short time span in existance because Intel has now had to pull Coffee lake in with some more cores. So the Z370 MBs are available but what about any B350, Q350, H370, Q370 chipset based MB SKUs coming this year.

October 6, 2017 | 10:26 PM - Posted by somebodys_kid (not verified)

For the Handbrake benchmark, what settings were used?

October 6, 2017 | 10:56 PM - Posted by SLOWION (not verified)

Coffee Lake is a solid release from Intel, I wouldn't mind getting my hands on an 8700K at some point.

Their HEDT release was a disaster but these products are much more sensible and appropriately priced.

October 7, 2017 | 10:34 AM - Posted by khanmein

I don't understand why don't want to use Handbrake 1.0.7?

October 8, 2017 | 01:33 PM - Posted by CNote

Id like to see 4k testing with the 1080 vs Vega 64 using these new 6 core vs ryzen 1600/X. See if there is any benefit to an all AMD setup.

October 10, 2017 | 05:56 PM - Posted by agello24 (not verified)

Fucked again. intel has gone and done it again. they made a new chip which requires a new board. smh, they where bound to get yall fan boys one way or another. finally they are about to move the industry from 4 core caps to maybe 6 core cap. however they screwed yall one more time. why make a chip with a base core of 2.8 and turbo the dam thing all the way up to 4.7?? the heat coming from those chips are insane!

October 13, 2017 | 02:03 AM - Posted by 1984 (not verified)

Intel released Coffee Lake to steal market share away from AMD. The last thing they wanted to do was sabotage their own product, Kaby Lake, and have customers move from KL to CL.

Intel could have released a version of Coffee Lake that ran on a Z270, lets call it i7-8700d for downgraded.

Just you wouldn't be able to overclock it and it would need to be paired with a high end motherboard that can provide the CPU with the power it needs, like the Asus Maximus IX Hero.

I think Intel couldn't be bothered with launching a product with such a complex requirement because thousands of people would end up buying it and finding their shitty Z270 motherboard weren't able to support CL especially since it would require a BIOS upgrade to make it work.

What I don't appreciate are the lies and deceit. I think the tech reviewers have done a disservice to consumers by allowing Intel to get away with this.

Intel claimed after much uproar that the 1151 pin configuration was changed but in reality it didn't change. They simply utilised reserved pins for the extra power draw on Coffee Lake.

They lie, and cheat customers who bought into their hardware and it seems like media/techn reviewers are happy to let em get away it.

October 16, 2017 | 08:27 AM - Posted by noony (not verified)

no testing against the 7400 either. for a comparison

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

This question is for testing whether you are a human visitor and to prevent automated spam submissions.