Feedback

Our Core i7-8086K Experiences: Problematic at Best

Author:
Subject: Processors
Manufacturer: Intel

A Weekend of Misadventure

Last Friday marked the release of the Intel Core i7-8086K to consumers through retail channels like Amazon, Newegg, and Microcenter. Announced just earlier that week at Computex, the i7-8086K is essentially an i7-8770K, running slightly higher clock speeds, and is meant as a limited edition item to commemorate the 40th anniversary of Intel’s 8086 processor, which marked the beginning of the x86 microarchitecture.

Eager to test this new CPU, I picked one up from our local Microcenter on Friday evening, and plugged it into our Coffee Lake CPU testbed, powered by a Gigabyte Z370 Aorus Gaming 7 motherboard (updated to the latest BIOS), let my first pass of automated CPU benchmarks run, and went off with the rest of my evening.

View Full Size

Saturday, when I came back to look at the results, they seemed mediocre at best, with the i7-8086K trading blows with the i7-8700K. While the extra 300MHz of clock speed seemed like it would provide more of a benefit than I was seeing, it wasn’t entirely unexpected that performance might not be spectacularly higher than the i7-8700K so I continued to run through the rest of our standard CPU benchmarking suite, as well as our CPU gaming benchmarks.

Finally looking at all of the data together, it appeared there was no change from the i7-8700K to the i7-8086K leading me to do some more digging.

Equipped with Intel’s Extreme Tuning Utility, I began to measure the clock speeds during several benchmarks.

View Full Size

Much to my surprise, even on purely single-threaded workload, such as Cinebench R15 in Single mode, the processor wasn’t getting close to its 5.0GHz Single Core Turbo Boost frequency, in fact, I never saw it get above 4.5GHz. We corroborated these issues with another piece of CPU monitoring software, HWInfo64.

As you can see in the screenshot from XTU, the processor was sitting at a cool 48C while this was going on, and no other alerts such as the motherboard power delivery or current limit throttling were an issue during our testing.

Moving to another motherboard, the ASUS Strix Z370-H Gaming, again on the latest UEFI release, we saw the same behavior.

So far, we have been unable to get this processor to operate at the advertised 5.0GHz Turbo Boost frequency, on a multitude of different hardware and software setups.

However, if we manually overclock the processor, we can get an all-core frequency of 5.1GHz, although with a temperature around 85C.

At this point, we are left puzzled and disappointed by the launch of the i7-8086K. This is the same hardware and software setup we used for all of our CPU benchmarking for the recent Ryzen 7 2700X review, with no issues. We even tried a fresh, fully updated Windows install on a separate SSD, to help eliminate any potential for weird software issues.

Jeff at The Tech Report used the same Gigabyte Z370 Aorus Gaming 7 motherboard as us, and while he didn’t see great performance overall, you can see explicit scaling in pure single-threaded workloads like Cinebench in his review.

As far as the ASUS motherboard we also tried is concerned, the i7-8086K is listed on ASUS’ CPU compatibility list for UEFI 1301 (which we are running), so it seems there should be no issue.

View Full Size

This morning, the i7-8086K we ordered on Amazon showed up, and did the exact same thing, in both test setups. 

To be fair, based on the reviews that we have seen pop up thus far, including The Tech Report, the resulting performance if things were configured correctly doesn’t appear to be worth the extra cost.

View Full Size

What was meant to be a celebration of Intel’s 40 years of the X86 architecture seems more like a rushed release than a fully baked product. Remember, we bought this processor directly from a retail outlet with no intervention from Intel. Without the proper BIOS-level support, and potentially a more widespread issue affecting normal consumers building machines with i7-8086K.


June 11, 2018 | 03:18 PM - Posted by Brother Michigan (not verified)

This could be a silly question, but do you see any 8700Ks hitting the proper turbo frequencies in these setups? I see where you said these rigs performed normally in the 2700X review, but I wouldn't expect that you guys go out and verify turbo frequencies every time you put a system together.

June 11, 2018 | 06:47 PM - Posted by Forstral (not verified)

So does that mean I wasted my money, do I need to just return it to newegg once it arrives?

June 17, 2018 | 01:11 PM - Posted by PaulH (not verified)

I bought two, (8086K) because I cracked my z370 motherboard doing install, the original chip may be fine. I think I bent the pins on board and chip. New board and chip, it let me run windows with cheap Intel cooling fan on 5 cores at 5.0GHz using Intel XTU, sixth core is at 4.9.

All adaptive defaults, changed nothing.

I saw 32 degrees celcius mostly on all cores, the 8700K could only do 4.7GHz sometimes 4.8GHz on five cores.

I'm not trying anything else today, after doing cables for power on motherboard. (MSI z370 Gaming Pro Carbon AC).

I could not run, start, or shutdown W10 before at 5.0GHz, it crashed in less than sixty seconds.

June 17, 2018 | 01:16 PM - Posted by PaulH (not verified)

I also got from 4.3 to 4.7 CACHE on CPU, INCREASE, and memory OC to 3600MHz, "stable" what was 3400MHz it goes to 4000MHz, but, not with a high OC of CPU.

June 18, 2018 | 08:00 AM - Posted by PaulH (not verified)

This morning I put all six cores at 5.0GHz, Intel XTU, Intel air cooling with heat pipes $25 dollars (funny). It ran fine, W10, and a backup which might be a better test than most benchmarks.
It's what you need to do, fast, and I had no Thermal throttling.
So I think there is allot of value in it, you only have one shot at delidding, you can ruin the chip and the motherboard easy. Why not buy a chip that does it for you just with air cooling, really safe, and you cool the room like data center do. They had (now Oracle) servers and super computers with 128 cores on one board, they blast fans at them, and cool the data center. Maybe 66 degrees Fahrenheit. Water cooling was used on older mainframes the fortune 100 (75%) still run, not the newer models. I like the 8086K, they say there will be a 28 core 5.0GHz chip in Q4 also (Intel). That might be a mortgage payment on a luxury home in NJ.

June 18, 2018 | 08:03 AM - Posted by PaulH (not verified)

The average speed was 4.99GHz with XTU OC on six cores to 5.0GHz. The only point was, "I wanted 5.0GHz". I was tired of waiting for it to be quite honest.

June 18, 2018 | 08:06 AM - Posted by PaulH (not verified)

That was wrong, 4.91GHz average, maximum 4.99GHz.

June 19, 2018 | 12:04 PM - Posted by PaulH (not verified)

Today, with 8086K I tried 4000MHz memory and 5.0GHz on six cores, it crashed during PassMark physics test.

I did run each of there tests on 5.1GHz on six cores, with 3600MHz memory. The score was 8843, my best, but 9000 alludes me.

The average on 8086K was 5.01GHz and the maximum was 5.10GHz.

I had the memory back at 3600MHz, just a little better than 3400MHz I could run with 8700K at 47x5, 46x1 GHz on it (the multiplier).

Maybe, its just because I have the A/C cooler there was a heatwave that felt like 101 degrees. Still pretty cool that most of what I tried worked very well without any throttling.

July 11, 2018 | 09:27 PM - Posted by Cortez (not verified)

IBM z13 mainframes are still water cooled and isn't that old. I was on the deployment team for an agency upgrading a z196.

August 12, 2018 | 09:44 AM - Posted by PH (not verified)

Well, Sun Microsystems bought Cray Computers supercomputers, 1997, they then created 10K with 128cores, in 1998. By 2008, Oracle acquired Sun Microsystems, made there money back in one year, and created EXADATA super computers connected 7-8 servers with no derogation of speed with high speed links. Both, not one drop of water. I have not touched a mainframe since 1995, many people, have not, no jobs. We automated it to lights out operations to save our own jobs, then they outsourced and downsized IT with SAP off the shelf software, the best practices, are the defaults.

So, I don't feel bad, not many people work with mainframes, the ones we ran, in the fortune 100, no water cooling. No water cooling on UNIX super computers either. They control humidity and temperature in the room, under the floor tiles is very cold, they cut a form out, the fans of the computers push the cold air were it is required.

Last time someone said water tower, was in the 1980's, NYC and Texas data centers, it was a older mainframes even then. The new ones that fortune 100 company bought did not require any of that.

August 12, 2018 | 10:14 AM - Posted by PH (not verified)

https://www.computerworld.com/article/2515283/high-performance-computing...

The last water cooled mainframe was ES9000, 1995, they announced a new one 2010, z196.

The point, was, why would you want one, they had big water tanks on top of a hundred story build in NYC, the mainframes then could not fit in a freight elevator, they closed down 42nd street with a crane to put it in through a large window.

It had to cost a small fortune, the idea, supercomputer, just like mainframes, many high end servers connected together, parallel processing, layers, like client server. Minimum requirement was "fit in a freight elevator". It took Sun Microsystems and Oracle likely more than ten years to do that, we talked about it in 2002, 2008 is was real. I thought it was 2011 they brought it to market (Oracle). Anyway, I know 75% of fortune 100 still runs mainframes, but, why would you want to. Likely huge work loads, and its much more secured, encrypting everything, and always private networks only.

June 11, 2018 | 09:47 PM - Posted by KingKookaluke (not verified)

I thought as much. Just a Marketing ploy,with a different lid.

June 12, 2018 | 04:40 AM - Posted by Spunjji

This is even lazier than Lenovo's limited-edition Thinkpad. Amazing!

June 12, 2018 | 06:30 AM - Posted by Meaker (not verified)

It's not really about stock performance, this is a binned part for a bit extra.

June 12, 2018 | 07:26 AM - Posted by psuedonymous

Yep, it's basically "Silicon Lottery's pre-binned parts, but official". If you weren't already considering a pre-binned 8700k (or for some reason wanted to buy up a pile of 8700Ks to bin yourself) then there's not any real reason to go for an 8086K.

June 22, 2018 | 01:50 AM - Posted by jihadjoe

The problem with that is Intel scooped out the best Coffee Lake S parts for the 8086k. Notice that siliconlottery used to have a 5.3GHz bin for 8700ks that disappeared, and now they're offering binned 8086ks we finally see the reappearance of that 5.3GHz bin.

I think the mere presence of the 8086k very nearly guarantees that we won't find very good 8700ks for a while.

June 12, 2018 | 07:55 AM - Posted by Don't panic! (not verified)

8086k differs from 8700k only by a 300Mhz higher 1-core Turbo frequency, multicore frequencies are exactly the same (inc 2 core). In real world this translates no real gains in performance.

I guess it could be a better binned model, so it could prove useful for extreme overclockers or collectors.

June 12, 2018 | 11:32 AM - Posted by Rocky123 (not verified)

" While the extra 300MHz of clock speed seemed like it would provide more of a benefit than I was seeing, it wasn’t entirely unexpected that performance might not be spectacularly higher than the i7-8700K so I continued to run through the rest of our standard CPU benchmarking suite, as well as our CPU gaming benchmarks."

That extra 300MHz clock speed is only on 1 core when going beyond more than 1 core the rest run exactly the same as the 8700K Anandtech has a pic of this showing it. That's how Intel dupes everyone by saying oh look we just released a 5.0GHz CPU but forgetting to say hey guys that is only on a single core. Kind of useless since most work loads use more than a single core these days.With that said all people see is 5.0GHz and assume that is what the CPU will always be running at when in fact it will almost never be running at that in normal work loads well unless you force it on in the UEFI bios with MCE turned on which then takes the CPU out of stock configuration.

June 12, 2018 | 11:45 AM - Posted by Rocky123 (not verified)

I wanted to add that in my own testing on my i7 2600K I have played with the turbo settings to see if that was a better way to do things for OC. What I found was when I staggered the cores lets say 5.1GHz 5.0GHz 4.9GHz 4.8GHz just for testing of coarse.

What I found was it never really hit 5.1GHz even in Cinebench yes it would bounce up to that once and a while but it never changed my single core score to higher since it sat at 5.0GHz for single core tests. Then I tried 5.1GHz 5.1GHz 5.0GHz 4.9GHz
Now it hit 5.1GHz in single core test and my score went up.

IN the winter months I just run 5.1GHz all cores without the turbo setup at all as it is turned off. IN the summer months I use the turbo with 5.1Ghz 5.1Ghz 5.1Ghz 5.0GHz which keeps the temps right where I want them at around 68c max under full load on all 8 threads.

If I do 5.1Ghz on all cores in the summer it gets into the 70c range well 74c to be exact which for me is out of my comfort zone. I need better cooling as I am on air right now was thinking on the EVGA AIO 280mm water cooler unless there is something better in that price range.

In 98% of the games whether it is 5.1Ghz or 5.0GHz the temps range from 43c to 56c & 64c max.

June 12, 2018 | 01:45 PM - Posted by ipkh

The question is, at default settings, how much power is the CPU using. Is there a noticeable difference between the 8700k and 8086k in power draw at equivalent frequencies. Similarly I'd like to know what the power draw looks like at 5.0Ghz across all cores between the 8700k and 8086k. Intels bullshit official tdp numbers for frequencies the CPU isn't programmed to use is extremely disingenuous and makes it very hard to compare cpus.

June 17, 2018 | 07:11 PM - Posted by quest4glory

So far, so good with mine. Paired with an Asus Maximus X Hero WiFi, G.SKILL Trident X RGB @ 3466 and a Corsair H115i Pro RGB. It’s obviously the RGB advantage.

Currently running at 5.1 GHz on all cores at 1.275V, easing my way up to 5.3 if possible.

Yes, I know what are the temps etc etc. etc.

I’m only saying I’m happy with mine as a binned 8700K and future collectors item status maybe 10 years down the line.

June 18, 2018 | 05:22 PM - Posted by elites2012

take it back and get a refund. let the misled intel fans buy it. some are still stuck on a single core world in a multitasking environment.

June 19, 2018 | 12:08 PM - Posted by PaulH (not verified)

I think that's terrible advice, my 8086K is much better than 8700K, "night and day", look at the tests I was able to run up to 5.1GHz on all cores, no delidding, water cooling, just Intel heat-pipes and fan to 3100RPM's.

June 19, 2018 | 05:48 PM - Posted by Anonymous12345 (not verified)

I have an i7-8700 non K running on an Asus Z370H motherboard running Windows 10 and I have never seen the turbo ratio go to the X 46 single core ratio. It never goes past X 45. I have HwiNFO64 running all the time and regularly check. I'm using the latest BIOS. Maybe there is something deeper going on here, or maybe one single core will never get pushed to its max multiplier under Windows 10 daily use with several programs and apps installed.

June 23, 2018 | 03:07 AM - Posted by PaulH (not verified)

I surprised at how random the results can be, its the exact chip, I did not want to believe. My 8700K was terrible.

I changed air cooler to NH-D15, the 8086K in Z370 MSI Gaming Pro Carbon AC went to 4000MHz Memory, 5.0GHz in BIOS all six cores. With INTEL XTU it went to 5.2GHz all cores (six), I OC'd TITAN XP +250 +550, ran 960 PRO 512 GB RAID0 in write back cache, no buffer flushing. "9246" in Passmark, I've never seen anything like that before.

The NH-D15 has two fans, huge, ugly, air cooling, don't care, I put the room temperature (A/C) at 66. "No" Thermal throttling, "none". Very cool. One trick to that score, just for 2D score, you earn it. You have to lower resolution. Here is one of the scores: https://www.passmark.com/baselines/V9/display.php?id=104075219644

I was figuring out why scores are so random, not with same settings, 5.2GHz, 2D, 3D +250 +600 if you can get it on Titan XP, Memory 4000MHz, "9246".

Room Temperature with air cooling matters allot I've learned. Likely, any type of cooling.

June 23, 2018 | 03:41 AM - Posted by PaulH (not verified)

My 8700K was crap, I tried 48 and 47 multiplier in XTU six cores it crashed, it seem random, the cooler "was" MSI Freezer, it did not like high memory.

As I was typing this the system rebooted, I put it back to 5.0GHz and 4000MHz memory.

That is enough I believe.

For benchmark, this NH-D15 cooler, with room (A/C) at 66 degrees allowed to run on 5.2GHz six cores (all) from Intel XTU, 4000MHz memory, Titan XP OC +250 +600 (or something close to it) with Afterburner.
Then I turned RAID0 to "WRITE BACK" cache, no buffer flushing. Ran everything, my version, of "flooring it".

The 2D, lowers all your scores with PassMark. The importance, to me, is questionable.

My point, I like the 8086K, NH-D15 cooler, Passmark, not that much. It is a tougher benchmark, I don't always agree with there calculations of score, not reporting it as is.

June 23, 2018 | 07:15 AM - Posted by PaulH (not verified)

I would install Intel XTU to at least monitor speed and temperature, for what its worth, I did not change any power or volt settings. At least now, 5/23/2018, it does it by itself.
You can or can not go faster, my volts says 1.365V maximum, minimum 1.065V, average 1.355V for 5.2GHz six cores. The temperature 29, 77, 32 minimum, maximum, average.

I read allot of articles on overclocking, it itself, is not that interesting, the "results" are.

I have been fighting with this build for 3 months, I just happened to notice the 8086K Anniversary announcement. I did crack my motherboard, and kill a 8700K and 8086K chip (likely anyway on the later). So I burned $692 dollars, I was not going to use the 8700K again anyway.

June 23, 2018 | 03:48 AM - Posted by PaulH (not verified)

I did get 47x and 48x multiplier on 8700K on five core, 46 and 47 on sixth in XTU, its not the same, but it improves your performance with a safety net, with, allot of heat it goes to lower multiplier.

Yes, not having to do any of that is much better, 1st, time at that level for me, "Oh, Happy Days?" Ahh. I do like the scores. It does seem faster at 5.0GHz all six cores and 4000MHz memory.

Maybe, just in my head, I am hyperactive, always been that way, I sense it. Even like the sound of the word 5.0GHz. Just do.

June 23, 2018 | 06:20 AM - Posted by PaulH (not verified)

Well, I went round in circles, had to re-install INTEL XTU, Extreme Tuning Utility, it was 6.4.1.23 I swore I was running 6.3.1.24. It does not let you overclock past the BIOS setting I have (MSI z370 Gaming Pro Carbon AC) of 5.0GHz on all six cores of 8086K, also true of memory at 4000MHz for 8086K. The Passmark score went to down to best case "9094", still using 5.0GHz on all cores, no thermal throttling, "maybe" there doing me a favor? I was just making CPU speed faster in XTU, odd.

So maybe, the thread is correct on problematic, I just had terrible problems with 8700K, much worse, things worked, than failed, in regard to performance. This 8086K is working out better still, room is at 66 degrees with the new NH-D15 cooler. Yes, I'm cold.

I did "also" learn the hard way PCI Express lanes go to 8 by 8, with two 16 PCIE cards, I tried a hardware RAID card, and the NVDIA Titan XP, the hit to graphics was 2000 points (1400). The RAID card, Highpoint, ran to 8GB a second, three 960 Pro solid state drives. I needed two active 16 PCI Express lanes. So I pulled it, 5.2GHz really have the scores a kick (9246 vs 9096 now). It gets harder to improve your scores at that level. In day to day activity, not going to matter. I find all of that interesting. A combinations of events you organize to succeed. I have to wait for a newer release of XTU exists. Perhaps, email them (INTEL).

June 23, 2018 | 07:00 AM - Posted by PaulH (not verified)

I got it "Maximum Core Frequency" defaulted to 5.0GHz, it was 83x multiplier in the last Intel XTU version I used.

I got a higher score "9249" Passmark Benchmark, I ran CPU, Memory, and Disk one test at a time at 5.2GHz all six cores on 8086K. Then 2D and 3D graphics at once.

If you keep highest score, and run short tests it makes it easier to do many steps with the right gear, it did take me forever.

It's not exactly stable at 5.2GHz and 4000MHz, I can randomly, fail, who knows, humidity, temperature, likely both.

The case and cooling system is not exactly like a data center that blasts A/C up through a vent in the data center floor tiles. The high end systems did run fans, pushing that cold air to components. I was in a data center for three decades, so it is seemed normal, like just another day in the week.

I don't think it can go any faster with a CPU and motherboard with two active 15 PCIE lanes, it was mentioned for x299 motherboards, not Z370.

June 23, 2018 | 07:02 AM - Posted by PaulH (not verified)

16 PCI E lanes, two of them, my bad.

June 24, 2018 | 03:54 AM - Posted by PaulH (not verified)

"9282" 5.2GHz six cores, 4000MHz memory, cache "write Back", OC GPU +250, +600. 66 to 65 degrees room temperature? (A/C running at) with NH-D15 full speed (CPU cooler).

I set it to 5.0GHz all cores on 8086K (6) and keep 4000MHz memory after benchmark. That matches the BIOS OC settings.
I tried the OC many times over the course of days and its very consistent. "Finally".

The part, that's not funny, I can't start my computer if the room temperature is hot. The reason not to keep it at maximum speeds. I had to reset CMOS once, you power off, unplug, drain power (power on unplugged), and touch JBAT1 with a flat head screw driver. How, embarrassing.
I ran out of things to test and Overclock.

I can't plug in cooler to CPU pin, the NH-D15 is really sharp, at DC and full speed it does not matter, using some volts I would not use in smart mode fan control. I just wanted it cold. I had to use system fan 4 pin. I literally had to put cardboard on radiator.

It runs at 31 degrees after each benchmark, maximum 77, average 37, minimum 28 degrees Celsius.

It's pretty much, CPU, DISK, much faster, then memory, and 2D, and 3D. I read two NVDIA Titan XP running 8 by 8 PCIE lanes is slower. The hardware RAID card, would work great in 16 lanes, in 8, I got 8 GB/s. It lowered 3D to 14000 with 8 lanes. Too much heat, and not enough lanes for both. I tried 16 by 4, it ran specification of (960 Pro SSD 512GB).

I'm done, "I quit".

June 30, 2018 | 05:29 AM - Posted by PaulH (not verified)

I finally got to "9430" Passmark score, benchmark from Passmark, its limited by the lowest score, 2D, in my case. It's not a lower resolution, it lower scale, 300% 4K, 150% HD, and then with "adjust the appearance and performance of Windows settings". I seem to have to manipulate RAID disk volume from cache "write back" to "write through" to get highest disk score for 960 Pro 512GB RAID0 volume.

It does really reward changes including OC, its supposed to be 101 degrees again, not Over clocking anything, past 5.0GHz and 4000MHz memory. It seems random, more like the temperatures, its not that warm, a package temperatures are 33 Celsius. ON 5.0GHz, at 64 degrees in the room, even 5.2GHz ran at 29 degrees Celsius.

I was thinking about 970 PRO 1TB drives at $499, yes, just for the benchmark. "Maybe" the 512GB model is to close to 960 Pro 512B SSD specification. I

August 12, 2018 | 09:54 AM - Posted by PH (not verified)

In several sessions, I get to 10,131 with PASSMARK, with the 8086K chip, all tests 5.2GHz, 5.3 all tests not Physics CPU test, sometimes threaded memory test, but, not in a consistent manner,
3D stops, at 5.3GHz, 2D, Memory, and 970 Pro 1TB disks, run at 5.4GHz. They don't accept the high score, they took 9,500 this month. It's was a very good score for competing even with Xeon Processors, I can not afford. It's a very "good" score, almost, the best for any configuration, even a server.

It's not delidded, maybe, I will send it out and have it done, not, seeing the point, I do want to have two (2) 16 lane slots active, not seeing anything on market.

July 14, 2018 | 05:17 PM - Posted by William D (not verified)

Considering the Aorus motherboards had no direct support for the 8086k until July 5 2018 I can see why you got this result. Perhaps you should have checked into this before putting out a bs article?

July 14, 2018 | 05:18 PM - Posted by William D (not verified)

Considering the Aorus motherboards had no direct support for the 8086k until July 5 2018 (I imagine others not much before) I can see why you got this result. Perhaps you should have checked into this before putting out a bs article?

August 24, 2018 | 07:14 AM - Posted by Boris (not verified)

Hi guys!
Will i7-8086K processor work with this memory (4000mhz, Intel XMP 2.0 ready) and motherboard?
https://gskill.com/en/product/f4-4000c18q-32gtzsw
https://www.asrock.com/MB/Intel/Z370%20Extreme4/index.asp#Memory

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

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