Intel "Coffee Lake" Builders Thread

[Campy]

Riot55, re-mount the cpu cooler, it's probably not done correctly. That cooler is quite heavy, so make sure it's tight enough that it can't move due to it's own weight. If you want to be very careful, after re-mounting, test temps while the case is still laying on it's side. If the temps are more normal, try setting the case upright again and re-test to check if they go up again.

 

[Ferzerp]

This thing is virtually silent except benchmarking. It's awesome. The coil whine on the GPU has really toned down since I first did the rebuild. I don't even notice it now.

 

[weevilone]

Next step is making sure that these fans are installed right and I guess taking out the Noctua and cleaning off/reapplying thermal paste and making sure it's tight I guess.... the cooler is always my scariest part so maybe I was too gentle though I felt like I tightened it down mostly. When do I know I need to stop? lol

Every Noctua heatsink I've ever installed, you work the screws down and they simply stop. There's no wiggle room or gray area. There isn't anything scary about it.

It sounds like you didn't secure it all the way, which would be bad both for temps and movement over the CPU.

 

[psolord]

I just did some preliminary runs with Ashes of the Singularity Escalation, on the new 8600k.

Here are the results in contrast to my old 2500k.

Spoiler: 2500k@4.8Ghz+1070

Spoiler: 8600k stock (+multicore enh)+1070

Ok,ok, the 2500k is on a little older version (the system is dismantled for now-I will retest both systems with the new 2.60 soon), but the difference is telling for stock vs OC, no? xD

Sorry for busting your chaps with AotS again, but I'm excited!

I did a new run with 5Ghz on the 8600k this time, 1070@2Ghz again and this is what I got for DX12.

Spoiler: 8600k@5Ghz GTX 1070@2Ghz

130fps while the 2500k@4.8Ghz was giving 63fps, lol. With fast 2133 DDR3 ram too!

Speaking of the RAM, for this run I set my RAM of 3200, 15-15-15-35 to 3200, 14-14-14-35 and it was OK. Also lowered the vcore another 10mv. It's now at +60mV and I started with +100mV.

For now I know it's stable for the jobs I need it to be. Temps are also very good at 5Ghz with the Noctua D14, as seen with the Intel Power Gadget.

I have set the memory controller to 4.5Ghz. Should I go lower/higher? 4Ghz is the norm.

Also should I set the forth timing of the RAM lower?

This chip is crazy man. Absolutely nuts.

 

[fastamdman]

I put all fans to performance in BIOS, and it's already super loud as it is. I recorded a little video, and keep in mind this is the lowest volume (it ramps up the RPM twice at different temps beyond this). I also took some pictures. I don't THINK I have the fans in the wrong direction... it seems like it's pulling air from the front the way the fins are, but I notice the two front fans have the text on the opposite side... but the fans and the directional arrows seem to be pointing the right way. I also can't tell if it's my imagination or not that the Noctua heatsink is sagging down a tiny bit, like maybe one side isn't screwed down as much? That might explain some spikes but again, might be my imagination. I didn't REALLY capture it in the video but sitting here on desktop downloading a game on steam and typing this, temps around 40 and every couple seconds a single core will jump up to 60 or so and then back down.

For reference, I have the two Fractal 120mm fans that came with the case at the rear exhaust and I moved the front one down a tiny bit to fit the extra Noctua 140mm fan for the top of front of the case. I also have the two 140mm fans that came with the Noctua DH15 on the heatsink itself. I don't have anything on the top because I heard that was counterproductive. I really appreciate any and all help. Next step is making sure that these fans are installed right and I guess taking out the Noctua and cleaning off/reapplying thermal paste and making sure it's tight I guess.... the cooler is always my scariest part so maybe I was too gentle though I felt like I tightened it down mostly. When do I know I need to stop? lol

https://imgur.com/a/dJSjl

https://youtu.be/PJlBGLCEM0I

Just took a look and it appears that your fans on the heatsink and the exhaust of the case are pointing in the correct direction. Also your front case fans are pointed in the correct direction. Just remember that where the sticker points, is where the air flows. So you have 2 front intake fans, 2 fans blowing towards the rear on your heatsink and 1 exhaust case fan. So all of your fans are blowing towards the back of your case. If you have any fans on the top of your case, I recommend blowing them as exhaust, aka out the top of the case As far as your temps that you showed on the video however, your max temps are REALLY high. So if you didn't do any stress testing when you ran that, I would take off the heatsink, reapply thermal paste and put it back on. Those temps are very very high, especially for not stress testing. Goal for me personally is to keep temps under 90C DURING stress testing, which will result in 50-60c during gaming generally. To give you an idea, I mine crypto currency, which is something I just got into recently and I run my cpu and gpu at full load constantly. My temps for cpu are at high 50s low 60s, which fluctuate slightly to mid to high 60s at times. GPU temps are at a peak of 58C.

 

[fastamdman]

Sorry for busting your chaps with AotS again, but I'm excited!

I did a new run with 5Ghz on the 8600k this time, 1070@2Ghz again and this is what I got for DX12.

Spoiler: 8600k@5Ghz GTX 1070@2Ghz

130fps while the 2500k@4.8Ghz was giving 63fps, lol. With fast 2133 DDR3 ram too!

Speaking of the RAM, for this run I set my RAM of 3200, 15-15-15-35 to 3200, 14-14-14-35 and it was OK. Also lowered the vcore another 10mv. It's now at +60mV and I started with +100mV.

For now I know it's stable for the jobs I need it to be. Temps are also very good at 5Ghz with the Noctua D14, as seen with the Intel Power Gadget.

I have set the memory controller to 4.5Ghz. Should I go lower/higher? 4Ghz is the norm.

Also should I set the forth timing of the RAM lower?

This chip is crazy man. Absolutely nuts.

Congrats again on such a nice 8600k!! Welcome to the 5ghz club as well If you are going to play with ram timings I suggest using prime95 with a lot of ram testing to make sure its stable with tighter timings. LinX or intel burn test, you can set really high ram usage amounts as well, for a quick test, but I find prime95 to be a bit better with ram than linx or intel burn test personally. Once I get my delidding tool I will be running 5ghz+ 24/7 as well, for now my chip is a little to power hungry to want to be able to do it comfortably on AIO cooling lol. But, soon, soon I will be pushing insane voltages through this chip. I think around 1.4V or 1.425 I will be able to get 5.2 or 5.3 completely stable, but again I gotta delid to get there Waiting on my tool to get here from the great German himself before I start running super high volts though.

 

[fastamdman]

@Riot55 Also if you want better air flow, you can remove the dust filter that is in the front of your case. This will allow your intake fans to suck air in a lot easier, just remember though that you will have to clean them occasionally. You don't HAVE to do this by any means though, it's just a way to increase air flow and to lower over all temps. Just an idea For example on my corsair 900D, I leave the front panel off, but I leave the dust filter on, a mix of best of both worlds type of thing.

 

[Aikouka]

Speaking of temperatures, I can say that adding the extra fan inside the case has done some wonders for my M.2 temperatures. As a recap, back when adding games to the Crucial M.2 drive, I saw it hit what appeared to be its internal temperature limit of 63C (70C is its on-paper max). With the EKWB M.2 heatsink, I saw about 53C max, but to be fair, I wasn't doing constant writes that would be expected when installing nearly 80GB of games. So, what I did was stand up a Corsair AF140 (1500 RPM) fan on top of the HDD cage (freestanding, there is no mounting point) and have it point in the general direction of the M.2 drives. I checked yesterday, and the Crucial drive was at about 35C and the Samsung M.2 drive at 31C. So, it definitely helps a lot; however, it also raises a good question.

Essentially, leaving that fan there isn't really viable given that there's no mounting point for it. A good shake will topple the fan over, and I'm not a huge fan of that. Albeit, I'm not a huge fan of a good shake of my PC either. I was thinking about designing and 3D printing a 40mm fan mount for the EKWB M.2 heatsink, which would likely remove the need for the 140mm fan. The more extreme solution, and arguably the better one is to find a different case that isn't so large that it has pressure problems.

I've also been disappointed in EVGA's SuperNOVA G3 PSU. I've been using their PSUs for my last couple builds, and I kept hearing this fan whir up randomly while playing games. Well, it turns out that it's the SuperNOVA's ECO mode that's doing it. Rather than gradually raise the fan RPM to combat temperatures, it just jacks up the RPM, and it's quite loud comparatively to everything else. It's strange, because I normally use ECO mode on my PSUs, and I own another SuperNOVA G3 850W, and I don't recall having that problem, which means I must not be using ECO mode... or maybe this new one has a problem? I'm not entirely sure as I forgot to look.

@Riot55 Also if you want better air flow, you can remove the dust filter that is in the front of your case. This will allow your intake fans to suck air in a lot easier, just remember though that you will have to clean them occasionally. You don't HAVE to do this by any means though, it's just a way to increase air flow and to lower over all temps. Just an idea For example on my corsair 900D, I leave the front panel off, but I leave the dust filter on, a mix of best of both worlds type of thing.

In the Gamers Nexus video on case design flaws, they noted that about a 1" distance between the solid front panel and the front of the fans is the optimal amount to not completely hinder their pressure. (They saw something like a 10-12C drop when removing the front panel on Cooler Masters new H500P.) I never thought much about it until I watched the video, and now, whenever I look at cases, I keep seeing these poor designs everywhere. Although, a solid panel will reduce pressure regardless, so like you mention, I'll pull the front panel if I'm doing something like mining.

 

[Riot55]

I just wanted to say thank you for all of your help about my temperatures. I just cleaned thermal paste and remounted cooler and made sure it was all the way tightened down this time. I had a hard time catching the screw with the heatsink screw which I also remember during the build process, so I bet I left it not all the way tight. Currently idling in the mid 20s on desktop, and only peaked in the low 60s when I ran Cinebench and was in 40s while playing Destiny 2. Currently letting it do a CPU-Z CPU stress test and temperature are staying right at 60. This is SO much weight off my mind, you guys have no idea.

 

[StinkyPinky]

Sorry for busting your chaps with AotS again, but I'm excited!

I did a new run with 5Ghz on the 8600k this time, 1070@2Ghz again and this is what I got for DX12.

Spoiler: 8600k@5Ghz GTX 1070@2Ghz

130fps while the 2500k@4.8Ghz was giving 63fps, lol. With fast 2133 DDR3 ram too!

Speaking of the RAM, for this run I set my RAM of 3200, 15-15-15-35 to 3200, 14-14-14-35 and it was OK. Also lowered the vcore another 10mv. It's now at +60mV and I started with +100mV.

For now I know it's stable for the jobs I need it to be. Temps are also very good at 5Ghz with the Noctua D14, as seen with the Intel Power Gadget.

I have set the memory controller to 4.5Ghz. Should I go lower/higher? 4Ghz is the norm.

Also should I set the forth timing of the RAM lower?

This chip is crazy man. Absolutely nuts.

What voltage are you using on your 8600k? I was actually thinking of going for 5.1 on my 8700k but I suspect I will need to go over 1.35 for that and I don't feel comfortable. You're welcome to try first though....

 

[fastamdman]

I just wanted to say thank you for all of your help about my temperatures. I just cleaned thermal paste and remounted cooler and made sure it was all the way tightened down this time. I had a hard time catching the screw with the heatsink screw which I also remember during the build process, so I bet I left it not all the way tight. Currently idling in the mid 20s on desktop, and only peaked in the low 60s when I ran Cinebench and was in 40s while playing Destiny 2. Currently letting it do a CPU-Z CPU stress test and temperature are staying right at 60. This is SO much weight off my mind, you guys have no idea.

Glad I could help man! However I HIGHLY recommend using a real stress test. Cinebench is a great start, but cpu-z isn't really a benching utility. Download Prime 95 26.6 (thats non avx) or download the newest prime 95 and let them run for a few hours or over night. Another great option is Intel Burn Test or LinX, they are essentially the same programs with a different skin. These are real benchmarks that will push your cpu properly. Just keep an eye on temps, although from the sounds of it, you figured out the solution. Great job and I am glad I could help. I knew remounting that cooler would solve the issue Glad you are back on track!! Just do some proper stress testing and enjoy it

 

[Campy]

I just wanted to say thank you for all of your help about my temperatures. I just cleaned thermal paste and remounted cooler and made sure it was all the way tightened down this time. I had a hard time catching the screw with the heatsink screw which I also remember during the build process, so I bet I left it not all the way tight. Currently idling in the mid 20s on desktop, and only peaked in the low 60s when I ran Cinebench and was in 40s while playing Destiny 2. Currently letting it do a CPU-Z CPU stress test and temperature are staying right at 60. This is SO much weight off my mind, you guys have no idea.

That's awesome man, and much more in line with what I expect from that cpu cooler

 

[psolord]

Congrats again on such a nice 8600k!! Welcome to the 5ghz club as well If you are going to play with ram timings I suggest using prime95 with a lot of ram testing to make sure its stable with tighter timings. LinX or intel burn test, you can set really high ram usage amounts as well, for a quick test, but I find prime95 to be a bit better with ram than linx or intel burn test personally. Once I get my delidding tool I will be running 5ghz+ 24/7 as well, for now my chip is a little to power hungry to want to be able to do it comfortably on AIO cooling lol. But, soon, soon I will be pushing insane voltages through this chip. I think around 1.4V or 1.425 I will be able to get 5.2 or 5.3 completely stable, but again I gotta delid to get there Waiting on my tool to get here from the great German himself before I start running super high volts though.

Thanks mate. Personally I will stick with 5GHz. I like round numbers!

My RAM is 15-15-15-35, 1.35V. They have samsung B dies. Is it safe to increase the voltage a bit in order to run them at 14-14-14-35, in case they are not stable? If so, how much?

 

[psolord]

What voltage are you using on your 8600k? I was actually thinking of going for 5.1 on my 8700k but I suspect I will need to go over 1.35 for that and I don't feel comfortable. You're welcome to try first though....

Since HT is missing from the 8600k, voltage requirements are quite lower.

I reduced the offset by another +10mV today. I am now at +50mV overall.

In high loads like Prime95, it still spikes up to 1.328v but it operates quite often at 1.312v too. At load that is.

I will try even lower with longer testing, but it's gaming stable I care mostly about and that is a completely different story, because cores do not go that high for such long periods, on gaming. For my encodings I will keep the stock speed +multicore enhancement. It's more than enough for now.

In xmedia recode, I was getting 11-14fps with my 2500k@4.3Ghz and I'm getting 27-33fps with the 8600k+multicore enhancement which also means 4.3Ghz. This chip is crazy! xD

 

[DrMrLordX]

Is it safe to increase the voltage a bit in order to run them at 14-14-14-35, in case they are not stable? If so, how much?

If it's actual b-die (hope you've confirmed this with Thaiphoon Burner; you may have already, I'm just not doing a great job paying attention) then you can push up to 1.45v vDIMM with reasonable confidence. Personally my limit is 1.42v for daily use. BUT! Ask around about the IMC since some Intel IMCs like to burn out if you feed too much RAM voltage. I remember that being an issue with Haswell in some extreme cases, but I'm not sure about Skylake/Kabylake/Coffeelake.

 

[Campy]

If it's actual b-die (hope you've confirmed this with Thaiphoon Burner; you may have already, I'm just not doing a great job paying attention) then you can push up to 1.45v vDIMM with reasonable confidence. Personally my limit is 1.42v for daily use. BUT! Ask around about the IMC since some Intel IMCs like to burn out if you feed too much RAM voltage. I remember that being an issue with Haswell in some extreme cases, but I'm not sure about Skylake/Kabylake/Coffeelake.

Honest question; Does dimm voltage affect the IMC? I was under the impression that was more related to vccio/vccsa voltages(For the lakes at least, I have no idea how anything worked on Haswell). Hope someone can clear this up.

 

[ozzy702]

Honest question; Does dimm voltage affect the IMC? I was under the impression that was more related to vccio/vccsa voltages(For the lakes at least, I have no idea how anything worked on Haswell). Hope someone can clear this up.

That was my understanding as well but I could be wrong.

 

[psolord]

Honest question; Does dimm voltage affect the IMC? I was under the impression that was more related to vccio/vccsa voltages(For the lakes at least, I have no idea how anything worked on Haswell). Hope someone can clear this up.

Having both a Lynnfield and a Sandy I remember this being a thing for them.

http://www.tomshardware.co.uk/forum/286748-30-memory-sandy-bridge-processors

I don't know if it's something to worry about Coffeelake, but I don't intend overvolting the RAM much. :S

 

[psolord]

If it's actual b-die (hope you've confirmed this with Thaiphoon Burner; you may have already, I'm just not doing a great job paying attention) then you can push up to 1.45v vDIMM with reasonable confidence. Personally my limit is 1.42v for daily use. BUT! Ask around about the IMC since some Intel IMCs like to burn out if you feed too much RAM voltage. I remember that being an issue with Haswell in some extreme cases, but I'm not sure about Skylake/Kabylake/Coffeelake.

I hadn't confirmed myself per se, but I had read that thread in reddit, explaining ram for Rysens and my sticks were reported as Samsung Bs.

Thanks for the reminder. They are indeed Samsung B.

Regarding the IMC, you made as worry! We need someone with deep knowledge to verify, just to be safe, but I don't intend to go crazy.

The thing is that my XMP is already running the RAM at 1.35 and maybe that's the upper limit. I remember something similar with my Sandy's G.skills. They were borderline OK with their voltage on the XMP in order to reach 2133Mhz.

 

[stockwiz]

I'm gonna delid mine with those 3D printed units you can get for 12 bucks on ebay along with some Thermal Grizzly Conductonaut and some liquid electrical tape to protect the capacitors and then I'll finish my my overclocking. What the heck.. it's a hobby.

 

[rondocap]

If I overclock my 8700 K, what’s the best cooling solution I can yet but uses 120 mm, as I am limited by my case. Currently h60


Would an ek open loop be better?

 

[ozzy702]

I hadn't confirmed myself per se, but I had read that thread in reddit, explaining ram for Rysens and my sticks were reported as Samsung Bs.

Thanks for the reminder. They are indeed Samsung B.

Regarding the IMC, you made as worry! We need someone with deep knowledge to verify, just to be safe, but I don't intend to go crazy.

The thing is that my XMP is already running the RAM at 1.35 and maybe that's the upper limit. I remember something similar with my Sandy's G.skills. They were borderline OK with their voltage on the XMP in order to reach 2133Mhz.

Considering Gskill specs all their 4200mhz and faster memory kits at 1.4v or higher, I'd say it's probably "safe" on QVL mobos and their corresponding chips. That said, I'm curious what the upper limit for memory voltage is for the IMC in Skylake and newer Intel chips. The vccio/vccsa voltage limits are written about but I don't ever remember hearing anything about max memory voltage like in SB.

 

[coercitiv]

Considering Gskill specs all their 4200mhz and faster memory kits at 1.4v or higher, I'd say it's probably "safe" on QVL mobos and their corresponding chips. That said, I'm curious what the upper limit for memory voltage is for the IMC in Skylake and newer Intel chips. The vccio/vccsa voltage limits are written about but I don't ever remember hearing anything about max memory voltage like in SB.

The max Intel XMP allowed voltage is 1.5V - I can't give you an official link, but back when CFL launched I started researching this a bit, looking around in forums and articles. With that in mind, most recommendations I've seen so far are to dial back even more and only consider 1.4 - 1.45V for daily usage.

Personally I run my kit at 1.4V, would have gone higher to 1.42-1.45 if the chip had shown potential to scale up in frequency, but it's a Hynix based kit so I'll take that 3466 CL16 and say thx for now.

[EDIT] Turns out the values were one Google search away:

1.5v is the absolute max we allow for XMP certifications. However, good DDR4 memory will run at 1.35v up to 3200. Technically, no “safe” (guaranteed) OC over-voltage but 1.35v or lower is best.

link

 

[psolord]

Good info guys thanks. For now I will stick with 3200-14-14-14-35,1.36V.

 

[DrMrLordX]

Honest question; Does dimm voltage affect the IMC? I was under the impression that was more related to vccio/vccsa voltages(For the lakes at least, I have no idea how anything worked on Haswell). Hope someone can clear this up.

Short answer is yes. I learned this the hard way back when I had my x2-3600+ and I uh . . . let's say, put a little extra sizzle on the IMC by trying to run it at stock vcore but with enough vDIMM to run DDR2-800. Doing that even for a little while made the IMC exhibit really weird behavior. x2s like my Brisbane had the IMC and cores on the same basic voltage plane (more or less) so the "safe" thing to do was to up vcore at least a little bit to compensate for the increase in vDIMM.

So according to my own personal experiences, if you are going to increase voltage to anything OTHER than vcore, it's a good idea to up voltage to every connected system at least a little bit. So if you're going to go over the RAM's rated spec (1.35v) it's a good idea to up vcore and . . . well whatever voltage(s) are related to the plane that controls the IMC. Sorry been on AMD systems too long, but I'm just gonna call it uncore voltage and hope I'm mostly correct.

Regarding the IMC, you made as worry! We need someone with deep knowledge to verify, just to be safe, but I don't intend to go crazy.

The thing is that my XMP is already running the RAM at 1.35 and maybe that's the upper limit. I remember something similar with my Sandy's G.skills. They were borderline OK with their voltage on the XMP in order to reach 2133Mhz.

I think we have the answers we need thanks to coercitiv et al. Anything over 1.35v may or may not shorten the life of the IMC on modern Intel chips. I *think* you can buy yourself a little extra safety if associated voltage planes get a little extra juice to go along with the RAM. The surest way to burn out your IMC would be to jack up vDIMM without changing anything else.

I'm gonna delid mine with those 3D printed units you can get for 12 bucks on ebay along with some Thermal Grizzly Conductonaut and some liquid electrical tape to protect the capacitors and then I'll finish my my overclocking. What the heck.. it's a hobby.

Heck yeah. Let us know how the Conductonaut does under the lid. I want to hear about whether it pumps out like MX-4 or stays put like CLU.