Intel's 8th Generation Core Family - Coffee Lake (LGA 1151, 6C/12T)

Where do you expect Core i7-8700K's Turbo to land?

  • 3.8/3.9 GHz

    Votes: 0 0.0%
  • 4.0/4.1 GHz

    Votes: 3 23.1%
  • 4.2/4.3 GHz

    Votes: 6 46.2%
  • 4.4/4.5 GHz

    Votes: 3 23.1%
  • 4.6/4.7 GHz

    Votes: 1 7.7%

  • Total voters
    13
  • Poll closed .
Picked up my 8700K on Sunday along with a gigabyte Gaming 7. It's barely stable at 4.9Ghz, and it's hot as fuck with a Mugen 5 on it. 85-90c under Realbench stresstests. Think I might RMA this back to microcenter when they get new stock.

Also... My Mugen 5 is kind of... Sagging? It's really odd. https://imgur.com/a/K2Ji3 Never seen a heatsink do this, but it's on pretty solid.

Now I just need to find someone with a delid tool in NJ. I'd hate to pay $50 and buy my own for a one-off delid.
 
Rumor has it that CFL U is succeeded by Whiskey lake U and not Icelake U or Cannon Lake U.

Whiskey lake U is apparently a 4+2 part, production starting ww27-34'18.

The person that posted it are apparently one of the more reliable posters.

DQS9YVcWkAInLjQ.jpg:large


Source


EDIT: Seems to be CFL U with 4+2 instead of 4+3e.

Whiskey Lake doesn't replace Icelake-U/H, it replaces KBL-R.
 
Whiskey Lake doesn't replace Icelake-U/H, it replaces KBL-R.

so we not getting icelake in 2018? that latest intel slide shows cascade lake end of 2018 but nothing about CFLS being replaced by icelake 10nm+ in 2018.
 
so we not getting icelake in 2018? that latest intel slide shows cascade lake end of 2018 but nothing about CFLS being replaced by icelake 10nm+ in 2018.

Mobile or desktop? Quite the difference.
 
Picked up my 8700K on Sunday along with a gigabyte Gaming 7. It's barely stable at 4.9Ghz, and it's hot as fuck with a Mugen 5 on it. 85-90c under Realbench stresstests. Think I might RMA this back to microcenter when they get new stock.

Also... My Mugen 5 is kind of... Sagging? It's really odd. https://imgur.com/a/K2Ji3 Never seen a heatsink do this, but it's on pretty solid.

Now I just need to find someone with a delid tool in NJ. I'd hate to pay $50 and buy my own for a one-off delid.
Gigabyte tends to set auto voltage to 1.34v on z170/z270 so that could be why its hot
 
there are some news on the z370-drama-front. apparently some guys in poland managed to get an i3-8350k running on a z170 board. they used a custom BIOS which runs with everything except the IGP.

source in polish.
 

Attachments

  • intel582.jpg
    intel582.jpg
    66.4 KB · Views: 9
there are some news on the z370-drama-front. apparently some guys in poland managed to get an i3-8350k running on a z170 board. they used a custom BIOS which runs with everything except the IGP.

source in polish.

I think it's safe to assume an ie-8350k will not stress much in the system the same way an 8700k does. I don't find this terribly surprising at all.
 
I think it's safe to assume an ie-8350k will not stress much in the system the same way an 8700k does. I don't find this terribly surprising at all.

Why ia that a problem, isn't the pin layout the same?

intel's excuse has been that the pin layout changed for z370. from my understanding it was never about "stressing the system". besides, the 8th gen i3/i5/i7 chips have comparable thermal envelopes. at stock at least.
 
intel's excuse has been that the pin layout changed for z370. from my understanding it was never about "stressing the system". besides, the 8th gen i3/i5/i7 chips have comparable thermal envelopes. at stock at least.
Inconceivable! Maybe it was easier to make more money with a new board? :)

Reading about it states it works, but since the bios is custom there are things that don't work.
 
intel's excuse has been that the pin layout changed for z370. from my understanding it was never about "stressing the system". besides, the 8th gen i3/i5/i7 chips have comparable thermal envelopes. at stock at least.

Pins which were reserved are now defined to have grounds and power supplied. So what the modder actually proved is they are used - only the i3 worked, and not with the IGP on.

Comparable thermal envelopes is not the same as peak current, or ability to clock even remotely above stock.
 
intel's excuse has been that the pin layout changed for z370. from my understanding it was never about "stressing the system". besides, the 8th gen i3/i5/i7 chips have comparable thermal envelopes. at stock at least.

it did change, mostly they activated reserved power pins to expand power delivery but they also removed a few non power related pins and turned them into power input.

pinoutclose.jpg


more over a big issue is some high end overclocking motherboards were already using some of these "rsvd" (reserved) pins for added power delivery and its been said that if one was to put coffee lake into one of these better z170/270 mobos it would actually fry the CPU.

lastly the hurdle which is artificial and likely what those people that got the quad core to run in z170 did is they disabled intel's management engine via the bios (this is same hack that enabled non K bclk overclock on z170 motherboards)

the big issue with the IME hack is you loose IGPU, AVX, and apparently with the 8350k they lost the primary x16 PCI-E slot as well that is probably pin missmatch related IMHO or just some funky bios quirk from the hack.

EDIT: on the "power envelope" thing I really wish people would get it through their thick skulls that TDP ratings are simply for OEM use for stock operation on cooling solutions. its a meaningless figure and your tripping if you think that a 8350k uses the same power as a 8600k/8700k even at stock operation. more cores more power period, not to mention they are also clocked even higher out of the box to boot along with that.
 
Care to elaborate? How does this predict when ICL comes out?
If you are a big customer, you always get CPU's before release...for validation, testing ect.
It makes it quite simple to predict when things will get release, based on the dates you get new "beta" servers form vendors.
Many times when I read CPU-rumors on the web I chuckle...because most often that not...I have the same hardware at hand...way before any release date or consumer variants are out for sale.
 
I probably recieved the worst possible 8600K sample yet, 4.8GHz would require north 1.35v to be stable, maybe ~1.37~1.38v or so and the temps are hot as such on air borderline acceptable without delid (cores hitting 100C with avx and 90~92C AVX-less). I settled for 4.7GHz 1.31v (my luck of silicone draw as usual) and hitting temps around 80C max AVX-less stress testing. Here's hoping the upcoming 8 core versions would be Z370 compatible so I could upgrade quickly.
 
I probably recieved the worst possible 8600K sample yet, 4.8GHz would require north 1.35v to be stable, maybe ~1.37~1.38v or so and the temps are hot as such on air borderline acceptable without delid (cores hitting 100C with avx and 90~92C AVX-less). I settled for 4.7GHz 1.31v (my luck of silicone draw as usual) and hitting temps around 80C max AVX-less stress testing. Here's hoping the upcoming 8 core versions would be Z370 compatible so I could upgrade quickly.
Hmm silicon lottery said 98% could do 4.9. you either got one of the worst or it doesn't like your settings. Have you tried it with low ram speed. What test are you using for stability?
 
Hmm silicon lottery said 98% could do 4.9. you either got one of the worst or it doesn't like your settings. Have you tried it with low ram speed. What test are you using for stability?

I've tried several different tests, Prime95, LinX, Aida64. Even in Aida64 I tend to get comp freezing when running for a while at 4.8GHz or the temp starting to get near to throttle cpu level. I did start out with lose ram settings, I've got the G.Skill Trident Z 3200 CL14 kit. There's a big difference in stability between 4.7 -> 4.8GHz. 4.7GHz 1.31v will run as long as I can withstand testing but going up to 4.8GHz at 1.31v which already would be on the very bad sample of silicon lottery draw side, the computer freezes after 1 sec of starting prime95. It more commonly just freezes than gives any BSODs. 1.35v will run a couple of mins perhaps.

I suspect something is very fishy about the cpu cooler or the the CPU IHS though.... because it doesn't seem truly secure somehow. the Cooler/fans seems to vibrate too much (putting my hand on the heatsink and I can feel the whole thing wobblying a bit). It's a Phanteks TH-TC14PE that I also used for the last build (I bought a new same cooler due how happy I've been with it and being cheaper than the Noctua over here with similar performance, was planning to sell my old stuff to a workmate so).

What's more the CPU cooler doesn't get hot which again is odd, even the exhaust air is almost as cool as in front of the case, only if trying to reach with the fingers underneath the heatsinks I start feeling slight bit of heat, dafuq? Again if there were some serious contact issues, I doubt even 4.7GHz would be possible though? And with stress test doesn't get past 80C after 30 mins+? It's odd though. Just now I'm testing prime95, temps around 75C and putting my hand inside the case just behind the cooler and the air is very cool.... I did try reseat, reapply tim once already though. Screws tightened as far as they go.

Perhaps I should install my older Phanteks cooler that I know for sure works normal just to see if there's any difference? Phanteks does state Socket 1151 compatible.
 
Last edited:
I've tried several different tests, Prime95, LinX, Aida64. Even in Aida64 I tend to get comp freezing when running for a while at 4.8GHz or the temp starting to get near to throttle cpu level. I did start out with lose ram settings, I've got the G.Skill Trident Z 3200 CL14 kit. There's a big difference in stability between 4.7 -> 4.8GHz. 4.7GHz 1.31v will run as long as I can withstand testing but going up to 4.8GHz at 1.31v which already would be on the very bad sample of silicon lottery draw side, the computer freezes after 1 sec of starting prime95. It more commonly just freezes than gives any BSODs. 1.35v will run a couple of mins perhaps.

I suspect something is very fishy about the cpu cooler or the the CPU IHS though.... because it doesn't seem truly secure somehow. the Cooler/fans seems to vibrate too much (putting my hand on the heatsink and I can feel the whole thing wobblying a bit). It's a Phanteks TH-TC14PE that I also used for the last build (I bought a new same cooler due how happy I've been with it and being cheaper than the Noctua over here with similar performance, was planning to sell my old stuff to a workmate so).

What's more the CPU cooler doesn't get hot which again is odd, even the exhaust air is almost as cool as in front of the case, only if trying to reach with the fingers underneath the heatsinks I start feeling slight bit of heat, dafuq? Again if there were some serious contact issues, I doubt even 4.7GHz would be possible though? And with stress test doesn't get past 80C after 30 mins+? It's odd though. Just now I'm testing prime95, temps around 75C and putting my hand inside the case just behind the cooler and the air is very cool.... I did try reseat, reapply tim once already though. Screws tightened as far as they go.

Perhaps I should install my older Phanteks cooler that I know for sure works normal just to see if there's any difference? Phanteks does state Socket 1151 compatible.
Did you start with low ram speed? I've seen at least for z270 that over 3000mhz ram system agent and vicco voltage need to be raised.
 
Did you start with low ram speed? I've seen at least for z270 that over 3000mhz ram system agent and vicco voltage need to be raised.

Yes I did start with very low ram speed at first, DDR4-2100 or something like that, now I tested just to be sure 2600 and no change in stability. 1.31 4.8GHz almost immediatly freezes (stuttering observed in mouse pointer already when getting into windows). 1.33v 30 sec ~ 1 min prime95 stable, 1.35v a couple of minutes stable, maybe 5~10 min or so. Not worth pushing this sample past 4.7GHz it seems, at least there I can keep things cool n quiet. I've never experienced that much of an abrupt stability scaling from a 100MHz clock jump before.

I did go through all other volt settings one by one pushing them up 3-4 steps one by one at 4.8GHz at a setting that linX crashed roughly at the same time (within 1 min and saw no increase in stability).
 
Last edited:
Where did you buy it from? You could always replace it, if you're feeling morally dubious.


Let's not get this into people's heads. Must everyone return their cpu because it won't get 5.2 GHZ at 1.3v? You can' tell me that doing this isn't jacking up the price for everyone else do to logistics and having to sell an opened product now.
 
Let's not get this into people's heads. Must everyone return their cpu because it won't get 5.2 GHZ at 1.3v? You can' tell me that doing this isn't jacking up the price for everyone else do to logistics and having to sell an opened product now.

Word, I'm not that kinda person, neither would it be worth the hazzle to me being without a workstation for so long and take it out and install a new one, takes just forever to get those CPU cooler heatsink screws tightened down! But I hope my luck would turn at least once. I always get slightly below average samples, this one then would belong in the far worst end that I've had though as I imagined worst samples of 8600K would do 4.8GHz at 1.32~1.33v (4.9GHz at ~1.4v'ish) or so perhaps meanwhile the best 8600K one I saw so far went 5.2GHz 1.35v stable on air or 5.3GHz 1.41v (high temps).

Maybe one day I'll get the average+ chip, one day!!
 
Last edited:
Let's not get this into people's heads. Must everyone return their cpu because it won't get 5.2 GHZ at 1.3v? You can' tell me that doing this isn't jacking up the price for everyone else do to logistics and having to sell an opened product now.
If an enthusiast overclocker gets a chip in the, let's say, bottom 10th percentile (4.8 GHz at 1.35V), I'm cool with them returning it if they want to. There's a certain point where a k-series chip loses its value, people deserve to get what they paid for. And if someone isn't happy with their purchase they're allowed to return it, usually with a 10-15% restocking fee. Newegg Premier and Amazon Prime exist for situations like this, too. The cost of convenience is already covered in your membership.

If they want to go through a dozen CPUs to get a golden sample, then its the retailer's responsibility to cut them off.

Also if I had a chip that bad, I'd worry about early failure. That's some weak silicon.
 
Last edited:
How long do you guys stress your CPU and under what program until you declare it stable?
 
How long do you guys stress your CPU and under what program until you declare it stable?
I do 3 or 4 hours of RealBench Encoding + Multitasking. However, all of my unstable OC's have failed in less than 1 hour (all of my OC's that survived more than 1 hour have been guaranteed stable). The new version of RealBench supports AVX if that's something you want to test.

Prime95 hasn't been reliable for me since Haswell.
 
Thanks guys, I have been doing 3-4 hours of both latest versions of pirme95 small fft + realbench to get my stable overclock. Maybe it's overkill but I find I need 1.360 vcore to get prime95 stable for that long with my 8700K @ 5ghz de-lidded. Realbench seems to need less vcore for me and maybe it's a more realistic load. My chip seems pretty average but I am not using any AVX offset and I am also enabling all the power-savings feature too. (Idles in the mid 20s and load is high 70s at these settings)
 
Back
Top