Home / Component / APU / Intel Core i7-6700K ‘Skylake’ test results: 4–8% faster than Core i7-4790K

Intel Core i7-6700K ‘Skylake’ test results: 4–8% faster than Core i7-4790K

A web-site has published what it claims to be early benchmark results of Intel Corp.’s Core i7-6700K microprocessor based on the “Skylake” micro-architecture. If the results are accurate, then the new chips will be at least four to eight per cent faster than the current mainstream processors with unlocked multiplier. Still, it is too early to draw any conclusions.

CPU-Monkey on Tuesday published a page with alleged Intel Core i7-4790K (four cores with Hyper-Threading, 4GHs/4.4GHz clock-rate, 8MB LLC, dual-channel DDR3 memory controller) vs. Intel Core i7-6700K (four cores with Hyper-Threading, 4GHs/4.2GHz clock-rate, 8MB LLC, dual-channel DDR4/DDR3L memory controller) benchmark results. The web-site did not reveal any details about test bed configurations, when the performance tests were made and by whom. One of the possible sources of the benchmark results are Intel’s own documents for partners, another imaginable source is an Intel partner with access to the new hardware.

intel_core_i7_haswell_lga1150

Maxon’s Cinebench benchmarks based on the company’s Cinema 4D software that is used by studios and production houses worldwide for 3D content creation significantly stresses microprocessors and greatly benefits from multi-core central processing units. Cinebench R15’s test scene uses sharp and blurred reflections, area lights, shadows, procedural shaders, antialiasing, and much more, something that uses all available resources from microprocessors.

intel_core_haswell_vs_skylake_cinebench

If the test results are correct, then Intel’s “Skylake-S” micro-architecture is slightly faster in single-core mode than “Haswell”. In multi-core mode, the new-generation Core i7-6700K processor offers significantly better results than the current-gen Core i7-4790K due to higher internal parallelism on the micro-architectural level and better optimization of various blocks within the CPU.

PassMark CPU Mark is a set of tests that promises to replicate actual workloads and hence they take advantage of multiple processor cores. The benchmark measures performance in such operations as compression, encryption, execution of SSE instructions and so on. It also measures maximum performance integer and floating point units of CPUs. The result is presented in points, which means that it is impossible to determine strong and weak aspects of a microprocessor.

intel_core_haswell_vs_skylake_passmark

If the results are correct, then the new Core i7-6700K will be generally faster than the current Core i7-4790K in various client PC apps.

GeekBench 3 by Primate Labs is a cross-platform benchmark for central processing units that is designed to simulate real-world scenarios. GeekBench 3 takes advantage of multi-core microprocessors as well as of fast memory sub-system.

intel_core_haswell_vs_skylake_geekbench

In full accordance with previous benchmark results, the new Core i7-6700K “Skylake” processor beats its predecessor both in single-core and multi-core modes, which indicates that the novelty has a number of micro-architectural advantages. Higher performance benefit in multi-core test demonstrates that Intel continues to optimize efficiency of execution of multiple threads.

Intel did not comment on the news-story.

Discuss on our Facebook page, HERE.

KitGuru Says: As usual, early benchmark results should be taken with a huge grain of salt. Even if the current test results are correct, it remains to be seen in what type of applications Intel’s new chips can show their maximum performance advantages over predecessors. While mainstream benchmarks do indicate relative benefits, they may not demonstrate certain local breakthroughs, which may be important.

Become a Patron!

Check Also

Intel is reportedly launching Core Ultra 200 and Z890 next week

Intel has yet to announce the release date for its upcoming Core Ultra 200 desktop …

91 comments

  1. Mads-Ejnar Kehlet

    For me it’s not about speed anymore. It’s more about feature. Don’t get me wrong a fast CPU is awesome but right now they are the the bottleneck in a pc. So i like more the new feature in the cpu like DDR4 🙂

  2. ImpedingMadness

    Whilst I do agree with DDR4 and other features and I’d like to see the differences between this and old tech but still this is nothing major or headliners worthy for 4-8% faster.

    You can’t really feel that subtle improvement. Not some life changing experience like changing from old plate drive to SSD

  3. http://media.giphy.com/media/TbRXNJJJbgIkE/giphy.gif

    C’mon AMD give your FX series a kick in the pants. We need more competition.

  4. “The biggest performance jump since the Prescott to Conroe transition”

    “4-8%”

    The HYPE is strong with this one. I guess the Xeon family will benefit more, but there’s no point in waiting for Skylake for a desktop power user who needs to upgrade right now.

  5. Well nothing new on the performence evolution from Intel. So if AMD Zen architectuur is going to up to or better on performance per watt and energy efficientwe then Intel cpu architectuur in 2016, then Intel has to give during the next tic or toc a better perfonce upgrade, 15% performace and be as energy efficient to keep customers onboard otherwise they gone loss market share to AMD. IMHO about time to give better competition to Intel.

  6. But there are people like me who would want to upgrade from an AMD CPU (FX-8350 in my case) to an Intel i7. I have been waiting for 2 years to switch and Intel Skylake would be my opportunity. Even if its just a small improvement over the i7 4790K, it is still a big leap from FX-8350.

  7. looks like there is no incentive to upgrade to skylake. Or it could be we have already reached the peak of desktop pc market.

  8. ddr4 doesnt have any advantage over ddr3 in desktop computers since we do not have any memory bottleneck. It has more advantage on mobile devices and servers due to lower power consumption.

  9. well zen cud b an even better upgrade ..

  10. intel is w8ing for amd to catch up

  11. How does it compare to broadwell? If it’s a bit faster, than I will just wait for 2 more months with upgrade.

  12. Barbara Quiles

    ∎∎∎∎∎∎✈✈✈✈✈Take Easy with kitguru < my buddy's step-mother makes $74 hourly on the computer . She has been without a job for 7 months but last month her paycheck was $14216 just working on the computer for a few hours.

    official website ===—->-> SEE MORE DETAIL

  13. Presumably Broadwell was literally JUST an optical shrink, and would have had no advantage. Indeed, with Intel’s 14nmFF woes, I suspect the clocks would have decreased and it would have ended up being slower. Hence why Broadwell has been all but cancelled … I doubt they’ll ship more than 5% of the originally intended volume outside of tablets and mobile.

  14. Q2 2016 is far away

  15. i dnt think zen will launch before Q3 2016..

  16. so to put it in perspective the newest i7-6700k-skylake ‘tock’ architecture compared to the i7-4790k Haswell Processor ‘tock’ is ONLY 4-8% faster and just about inline with the refreshed Haswell …. did people forget the Broadwell Tick shrink and is that only giving 2-4% speed increase if any ,against the current i7-4790k Haswell Processor ‘tock’

  17. Etienne Boutet boucher

    Except the DDR4 features i don’t see great feature enough to make me change my old I5 3570k, except maybe the new intel 1xx serie chipset with skylake…

  18. Brian Diffenderfer

    your wallet will feel it

  19. It was posted about a month, few weeks ago and I believe those were confirmed that are fake.

  20. You don’t have to wait until Skylake, especially if it is barely any faster that what is out now.

  21. Kristijan Vragović

    that’s my suspicion also, but it could be that AMD introduces it sooner, as they postponed skybridge

  22. Kristijan Vragović

    You are right in a way. But Intel has major advantage over AMD, and doesn’t have to do anything but wait and when the time comes strike AMD, which could end up an easy task. AMD will need a huge jump in a IPC, in my opinion more than a 40 percent. If Skylake has something around 10 percent over Haswell than 50 percent jump for AMD would match Skylake, but to surpass Intel AMD need a bit more than 50.

  23. Didn’t Intel say this was to be the biggest change since Nephalem and be so special etc etc.
    8% increase sounds like just another iteration of Sandy, ivy, Haswell, Broadwell.

    Where is this amazing thing Intel?

  24. Patrick Proctor

    Broadwell has between a 3 and 5% IPC gain over Haswell, and most of it is in floating point math where mul/div were shaved from 5 cycles to 3.

    For Skylake vs. Haswell, we have 1.08/(4.2/4.4) = 1.1314 or 13.14% IPC gain. That’s about right for a decent gain. It’s no SandyBridge jump, but it’s pretty fair.

    Zen will not overtake Intel. Intel’s only making sure AMD survives through 2019 when all their debt comes due. At that point Intel will make a quantum leap and leave AMD behind, with enough money to keep going but not enough to challenge Intel. Perf/Watt will come with the Xeon line and AVX 512.

  25. Looks like I can sit on my i7 3370K at just 4.2 Ghz until the sun burns out and we die.

    CPU’s are so extremely underwhelming now, 4 generations later and they have some 20% more performance in total over the years.
    DX12 will just help out on the CPU side, so even LESS of an incentive.
    DDR4, sure, maybe, but GPU’s take care of that in the future with HBM and large VRAM.

  26. Patrick Proctor

    It’s not meant to be an SOC. Internal Ethernet is pointless. SIMD throughput will increase where needed: server chips.

    HMC and HBM are both too expensive for the consumer segment right now and offer no gains other than for iGPU. Not to mention HMC production is already too little for the number of Knight’s Landing packages Intel intends to sell in 2016.

  27. Patrick Proctor

    Having NVMe native should be reason enough for most, and the extra PCIe lanes for PCIe/M.2 storage. The iGPU will also be great for OpenCL 2.x developers who don’t need a dGPU to test algorithms.

  28. Zen will not overtake Intel. Intel’s only making sure AMD survives
    through 2019 when all their debt comes due. At that point Intel will
    make a quantum leap and leave AMD behind, with enough money to keep
    going but not enough to challenge Intel. Perf/Watt will come with the
    Xeon line and AVX 512. – Laughed out loud at that.

  29. Patrick Proctor

    Why? There’s no way AMD is going overtake the best chipmaker in the world financially strapped as it is while building on inferior nodes (Samsung 14nm is no denser than their 20nm). And from a business perspective, Intel wants AMD alive, but only just. It would much rather keep half-dead competition alive than usher in new, better equipped competition like Samsung, IBM, Nvidia, or Apple.

  30. Wooo

  31. That’s terrible! Guess I can keep rocking LGA1366 until the heatdeath of the Universe, so that’s a plus

  32. That’s……… a pathetic gain.

  33. as far as we know amd .. they will only delay 😛
    intel will also debut skylake-e n cannonlake at that time … Q3 2016 is going to very very interesting 🙂

  34. Kristijan Vragović

    Hope not. Atleast for delays. 😀

  35. “For Skylake vs. Haswell, we have 1.08/(4.2/4.4) = 1.1314 or 13.14% IPC gain”

    Oh good mathematics. But in real world performance, the difference is just around
    4–8% faster than Core i7-4790K which is way below all people expectation the way Intel been hyping for years. What else desktop version of Skylake CPUs won’t support AVX 512 which is another disappointment.

  36. Mads-Ejnar Kehlet

    True normal pc user will not feel anything from DDR4 over DDR3, but DDR4 is still a new feature/tech 🙂

  37. Mads-Ejnar Kehlet

    SSD today is really getting cheap. If you don’t have one you really need to put a SSD on top of you list what to get next. 🙂

  38. yea lets w8 n see how things turn out .. 🙂

  39. I’m waiting for TDP reports to show up. Intel, please make air coolers relevant again in gaming systems.

  40. Any ideas why the U variants will not be utilising DDR4 (or LPDDR4), rather than sticking with LPDDR3? I had been lead to believe that part of the performance bottleneck, particularly with graphics, was due to the speed of the memory being used with current Broadwell chips.
    Also, with the supposed benefits of lower power consumption, DDR4 would surely make sense for devices using the U variants…

  41. I don’t really see the advantages here. Maybe some the users of this website/forum (enthusiasts, and professionals) who can afford to take advantage of this technology would benefit from it, but most people won’t. The biggest perceptual shift comes with switching from hard disks to solid-state, but beyond that a higher throughput SSD does little for overall perception of the system’s speed.

    Hell, I’m still running an SSD on SATAII (LGA1366) and it’s damn fast. Subjectively, I can tell no difference between that and a newer Sandy-based system with a SATAIII SSD my SO uses. Maybe Windows loads a little faster? I also however have more stuff loading on startup, so the difference the SSD itself makes is statistically negligible from a perceptual point.

  42. Patrick Proctor

    The LPDDR4 standard has just been ratified, and the expense is still too much for now.

  43. Patrick Proctor

    They said an equally big IPC increase as from Conroe to Nehalem, and at 13.14% (1.08/(4.2/4.4)), it is. Remember this is comparing the fresh, first parts to the refined, refreshed parts of Haswell. When the Skylake 6770K or its equivalent lands, likely at 4.5 GHz boost, I’m sure we’ll all be pleased by the results.

  44. Patrick Proctor

    I’m just saying we need to be careful here. Intel said an IPC increase as big as Conroe to Nehalem, which is true. Now, this is also the starting parts, not the refined refreshes along the lines the 4790K was. If Skylake gets a refresh and a 6770K SKU with a 4.5 GHz boost and even better overclocking relative to the 6700K, enthusiasts will end up being plenty happy.

    No software in the consumer world even takes advantage of AVX 256. It’s a pointless feature to include at this point, though it’s desperately needed in the server/HPC world.

  45. Skylake is also about hdmi 2.0 and running 4k displays at 60hz.

  46. “If Skylake gets a refresh and a 6770K SKU with a 4.5 GHz boost…..”

    Lol… you count IPC by higher clock speed. Bafun……

  47. بني حماد

    I think I will ditch my LGA775, I’m sick of waiting.

  48. funny thing is this 4-8% increase skips a generation broadwell. if broadwell is only 2-3% over haswell, means skylake is 2-3% over broadwell. thats just sad, sad thing for intel to do, almost skip a generation of CPU to make it look like they are improving over 5% which they’re not.

  49. But they’re including it on other processors in the Skylake range, so the ‘cost’ of having it as option is going to be nothing on the U and Y as well, especially given that it is potentially the processor range that would benefit from it the most.

  50. Patrick Proctor

    Um, the LPDDR4 in the Xeon range is all ECC stuff, and it costs 30% more per GB than the LPDDR3 ECC RAM of previous generations.

  51. Patrick Proctor

    No, I count PERFORMANCE of the same generation by clock speed scaling. That’s entirely valid. With the 13.14% IPC gain over Haswell, even a 0.25% clock boost would yield visible differences, and I’m sure the scores we’re seeing are not comprehensive across all of Intel’s instructions. When the cycle counts for Skylake are released in the next edition of Intel’s x86 manual, then we can say for certain.

  52. Mr.Intel Advocate, Where did you get the 13.14% IPC gain over Haswell ?? The best possible scenario ??

    What is available information at the moment is a mere increase in performance of just 4–8% faster than Core i7-4790K that’s simply not noticeable . You need a Stopwatch to differentiate that 4-8% performance gain over haswell i7 4790K.

    Intel failed to deliver what’s been hyped over these years regarding Skylake.

  53. Patrick Proctor

    8% performance boost over the 4790K at ~5% lower clocks. I gave the direct math.

    1.08/(4.2/4.4)

  54. 8% is pretty low IPC improvement what Intel been hyping for years regarding Skylake. Intel again cheated it customers by promising and hyping big but delivering pathetic improvement if any.

  55. Patrick Proctor

    8% is the performance improvement. 13.14% is the IPC improvement.

  56. hahaa… that’s not the performance shown here. Only 4-8% improvement in performance compared to i7 4790k.

    Oh boy….. again Intel advocate failed here

  57. Patrick Proctor

    No, Intel didn’t cheat or lie to anyone. The IPC improvement is there, but whether or not a limited range of benchmarks reflect it all is another matter. That said, Intel’s focused on the HPC space, and everyone should know that by now. People upgrading from Kentsfield or Sandy Bridge will be very happy. And with DX 12 making iGPU useful, it’s a good investment.

  58. Question–does anyone think that the socket for this new chip, the LGA1151, could be around for a while?

    I ask because I’m looking to upgrade, and it looks like the 1150 is coming to an end after a couple years of use. And if I hold off on the upgrade and wait for these to come out, I might be able to upgrade the CPU a couple times before I have to start all over again.

    Thanks in advance.

  59. “The IPC improvement is there, but whether or not a limited range of benchmarks reflect it all is another matter. ”

    Yes its not reflecting on normal day to day applications people use. That’s what I said Intel cheated its customers by hyping too much and then underdeliver.

  60. Patrick Proctor

    Intel did neither. It is as big a change as Conroe to Nehalem. This is indisputably true. Intel can’t cheat customers it hasn’t started selling to yet either. The PC space is done and complete. AMD can have the last few drops of blood if it can catch up. Intel’s got more important clients.

  61. “It is as big a change as Conroe to Nehalem. This is indisputably true.”

    You are trying escape from the truth by smoke screen trick.
    Where is that “big change” in performance ?? 4-8% is big ?? Intel cheated its customers by hyping too much and then underdeliver.

  62. Patrick Proctor

    It hasn’t underdelivered or overhyping. You’re just not the customer it cares about most. Wait until we see some virtual machine benchmarks, transcoding, and more. We have a very limited set of benchmarks we’re currently basing this 4-8% figure on.

  63. Then why you are barking about unreleased skylake CPU performance ?? You should also wait till full set of benchmark arrives. Your Intel marketing won’t work.

  64. Patrick Proctor

    I’m not praising Intel at all. I’m saying they’re being truthful.

  65. ” I’m saying they’re being truthful.”

    No Intel cheated all there customers expectations.

  66. Patrick Proctor

    How? It hasn’t even started selling them yet. There’s no way to do that.

  67. From the available benchmark 4-8% performance improvement is pretty weak and failed upto Intel hype and expectations of customers.

  68. Patrick Proctor

    Not at all. It’s not a comprehensive set of benchmarks, and you can actually see the instruction cycle counts in the newest revision of the Intel x86 manual. The IPC increase is 13.5%. That’s as much as the difference between Conroe to Nehalem. Now, that difference is not uniform across all instructions, and some workloads will benefit more than others. There’s no cheating, and no lies. And, I was right: lots of virtualization improvements ahead of W10 which uses a fair amount of it.

    http://www.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-instruction-set-reference-manual-325383.pdf

  69. “…..and you can actually see the instruction cycle counts in the newest revision of the Intel x86 manual. The IPC increase is 13.5%. That’s as much as the difference between Conroe to Nehalem. ”

    That’s all the theory that we adjusted this, we added this and that etc. Actual performance matters to consumers. From the available benchmarks skylake performance improvement is pathetic of just 4-8% that’s unnoticeable. That’s a clear consumer fraud of hyping too much and underdeliver.

  70. Patrick Proctor

    It’s not theory. Without those numbers you have no performance.

    Again, no there isn’t, because we haven’t seen even 1/3 of the possible workloads of the consumer space. There’s plenty more to come, and when the release actually happens and we get actual reviews, I’m going to come back here, tell you to admit you’re full of crap, and you’ll stamp your feet and deflect like a typical troll.

    Intel hyped nothing for the consumer space and said Skylake would be disruptive to HPC only, promising a change as big as Conroe to Nehalem. Not once did Intel mention the consumer space, and nor has it any reason to, or even pay attention to the consumer space which is drying up in desktop and stagnating in mobile. You’re making a mountain out of a mole hill. Intel hyped nothing and hasn’t even delivered yet. We also have no confirmation this benchmark set wasn’t spoofed, something that’s been proven pretty easy to do. And last but not least, again, this benchmark set is far from comprehensive of typical consumer workloads.

  71. Talking too much without any facts points to Buttthurt.

    Benchmarks that releases early before launch of a hardware, usually were cherry picked to show the best possible performance gain. So if this 4-8% is the best the rest of the application performance usually won’t be any better. If any improvement is there then its in power efficiency due to 14nm node.

  72. Patrick Proctor

    I have more facts supporting me than you do. These benches are unsubstantiated, and it’s too early for engineering sample benches. Those usually come out 30 days before the launch of the product which is in late August.

  73. “I have more facts supporting me than you do. ”

    Where is that facts ??

  74. Patrick Proctor

    Other than hard evidence that Intel’s average IPC gain is as high as they said it was, even if those gains aren’t in these specific workloads or focused on consumer software? Although, consumer software needs to go SIMD and multithreaded anyway. SISD is a waste, but it’s where we’re stuck because most software is built to support the minimum of Windows XP: Pentium 3.

  75. Yes Skylake is a waste CPU for desktop

  76. Patrick Proctor

    Not remotely true. There’s plenty more workloads to cover before you can conclude that, and the benches we have aren’t validated.

  77. “,and the benches we have aren’t validated.”

    Who is We ??

  78. Patrick Proctor

    The benchmarks you and I are currently looking at are independently confirmed/validated. Seriously, are you just trolling, or do you miss a lot of obvious stuff?

  79. Oh….. first you said “the benches we have aren’t validated” and now you are saying
    “Benches are independently confirmed/validated.” hahahaaa……

    Your brain is even more malfunctioning. You are a clown both in looks and talks. Get a life.

  80. Patrick Proctor

    Every once in a while Dragon messes up. I meant “aren’t,” not “are.” Edited and fixed. And would it kill you to carry yourself with an ounce of decency and maturity? I feel like I’m communicating with a thirteen year old with a chip on his shoulder.

  81. aaaahaahahahahoooohahahaaaaaa……Buttthurt

    >>>>>>Deal with it.

  82. Patrick Proctor

    Your parents must be real of you.

  83. Your parents must be real of you.

  84. Patrick Proctor

    Your parents must be real of you.

  85. the worst thing is amd is behind the sad thing .no competition at all why would they increase anything when they are far way ahead of amd? this is capitalism man open your eyes.

  86. if they dont increase it, then 99.9% of enthusiast won’t purchase, and tbh a lot of people out there get info from these enthusiast. they’ll prob just hear us talking that its 1% increase over previous gen but cost 100$ more LOL.

    in other words intel not dum so they had to increase performance somewhat to keep ti going.

  87. WRONG,the only thing wasn’t improved with intel is gaming, not all
    “enthusiast” are gamers,intel had reasonable performance increase in their products for other “enthusiast” but gamers.
    as we have very advanced video cards the systems are 99% bottlenecked by GPU cpu doesn’t get in line that much in this case on pro gaming systems so they can do nothing to improve gaming much.
    they always can sell their latest products to new users , but you are right bringing new lineup would gain them more,it’s just we see they are not in such a hurry.

  88. Nvidia_ALWAYS THE BEST

    I want atleast 20%+ improvement over my 4790k, then only I’ll upgrade….

  89. Only 4-8% what a disappointment i will stay with my Core i7 4770k

  90. Andre Luiz Barbieri

    Passmark result for 6700k is wrong… These are the correct values: http://www.cpubenchmark.net/high_end_cpus.html and http://www.cpubenchmark.net/singleThread.html (4790k is the King!)

  91. Why did not include singlecore PassMark? https://www.cpubenchmark.net/singleThread.html