My new Intel 8th Gen Core i7-8700K Processor

[marjamar]

When your doing the delidding temperature comparison try to note the ambient temperature before and after delidding.

I planed on that.

Today I was running prime95 blend again at 4.8GHz without offsets and after about 15 minutes, and a worker dropped off. Temps peaked at about 90 C. a few times for just a second. I have seen this run at 95 C. for a few seconds before I stopped the run without dropping a worker. So, I think I had better do something like a 15 and a 30 minute run using OCCT CPU:OCCT small data set, not prime95. OCCT stops the run automatically at 87 C. I seem to recall. But even the small data set seems more mild then prime95 blend, so I'm pretty sure I can get both a 15 & 30 minute run. OCCT also gives a pretty full set of graphs that show sensor readings for the entire run time, which will no doubt be significant for comparison.

I also changed my water loop a bit today, removing all the adjustable elbows (4 of them), reducing internal restrictions. Now, my pumps push between 1.9 & 2 gallons per minute @ 100%, whereas it was moving 1.7 gpm max. That should give me a bit more cooling ability I should think, which would be useful if I can get 20% or more thermal transfer after the delidding.

I guess it it won't be until this weekend or even after before all the parts get here for that. Just a waiting game for a bit.

-Rodger

 

[EarthDog]

If you drop a worker, it isnt stable.

OCCT stops at 85C by default. You can change that to whatever.

1.5gpm flow is around the sweetspot. More flow is a diminishing return.

 

[marjamar]

Got my CPU delidded. It was easy enough to crack it open. Cleaning it up was OK too, but it's a little tricky to get that liquid metal to stick very well. Tried brushing it slow, fast, just a very little bit, more then a little bit and maybe a few more ways. Got it on and it is covered, but seems like it took more material then I was it expecting too use. I also used the same material on the external side of lid and the CPU waterblock. None to spare at all, but just enough to finish it.

Still testing and will be for a good bit I expect. I wound up doing 2 sets of test, before and after for both, using CPU:OCCT small data set. Did 4 screen grabs total, one each before starting the test and one just as the test was about to finish. All were using the same profile, 4.8 GHz as that was the fastest I could get a full run with before delidding. Ambient temps were 64.5 when tests were done. The top 2 were before I delidded the CPU and the bottom 2 were after I delidded. I think I have about 20C or so more headroom now, but take a look and see if I am about right. As I said, I will be doing some overclocking now and I will see where that leads.

-Rodger

 

[marjamar]

Playing with my new delidded CPU this morning. This delidding really makes a huge difference. I am surprised CPU manufactures aren't using liquid metal in the production of these CPU's yet. Bet they are going to shortly though. Anyways, I am at 5.1 GHz @ 1.390v. in BIOS, which becomes 1.392v.~1.408v. vcore when running. Also have AVX set to 0 offset. It would boot this clock at 1.360v. and try to run prime95 blend, but once it's just about completed the first set, it would drop a worker. So, I just bumped it up to 1.390v. to see what I can do with it. Ran Cinebench and it completed fine with a pretty good score considering I am using an older EVGA GTX 980, overclocked a bit. Interesting the CPU temps never got over 55C. running Cinebench. Going to trying some other stress tests/benching, but thought I'd post this one to start with.

-Rodger

 

[wingman99]

Looks good. I'm glad to see your having good results.

 

[DaveB]

Playing with my new delidded CPU this morning. This delidding really makes a huge difference. I am surprised CPU manufactures aren't using liquid metal in the production of these CPU's yet. Bet they are going to shortly though. Anyways, I am at 5.1 GHz @ 1.390v. in BIOS, which becomes 1.392v.~1.408v. vcore when running. Also have AVX set to 0 offset. It would boot this clock at 1.360v. and try to run prime95 blend, but once it's just about completed the first set, it would drop a worker. So, I just bumped it up to 1.390v. to see what I can do with it. Ran Cinebench and it completed fine with a pretty good score considering I am using an older EVGA GTX 980, overclocked a bit. Interesting the CPU temps never got over 55C. running Cinebench. Going to trying some other stress tests/benching, but thought I'd post this one to start with.

-Rodger

You really should delid that i7-8700K and call it done. I delidded mine and the 100% load temps dropped more than 20C.

Finally! I told you to delid the damn thing 3 weeks ago!!!!!!!!!!!!!!!!!!!

 

[EarthDog]

I am surprised CPU manufactures aren't using liquid metal in the production of these CPU's yet. Bet they are going to shortly though.

They won't. They moved away from the indium solder a few generations ago here in the mainstream segment and a couple of generations ago in HEDT.

The TIM isn't the best for overclocking and low temps, but, that isn't something Intel technically supports in the first place anyway. They run plenty fine at stock as well as allowing overclocking.

Anyway, congrats, and enjoy that thing!

 

[marjamar]

Thanks for the replies all. Seems it's still going to be tough to get the low volts I have been hearing about @ 5 GHz and above. I still cannot run stable @ 5 GHz using prime95 small FFT's for some reason. CPU temps are plenty low now, so it's not heat related. I will continue trying and researching what BIOS settings may need to be adjusted, but having already tried many suggested settings for 5 GHz clocks, it may be this particular CPU isn't going to be stable above 4.8 GHz or so unless I run vcore above 1.4v which I'd rather not have to do. We'll see.

-Rodger

 

[wingman99]

Thanks for the replies all. Seems it's still going to be tough to get the low volts I have been hearing about @ 5 GHz and above. I still cannot run stable @ 5 GHz using prime95 small FFT's for some reason. CPU temps are plenty low now, so it's not heat related. I will continue trying and researching what BIOS settings may need to be adjusted, but having already tried many suggested settings for 5 GHz clocks, it may be this particular CPU isn't going to be stable above 4.8 GHz or so unless I run vcore above 1.4v which I'd rather not have to do. We'll see.

-Rodger

As of 3/22/18, the top 88% of tested 8700Ks were able to hit 5.0GHz or greater. 1.40V Vcore -2 AVX Offset https://siliconlottery.com/collections/coffeelake/products/8700k50g

 

[marjamar]

I've managed to get what would seem to be a stable 5 GHz prime95 blend test, but having some motherboard VR throttling in select portions of that test come up. I am still working on my voltages so it may go away if I get them lower. Currently I am holding a solid 1.440 vcore, even under max loading. I'm also running a 5.1 clock with 1 AVX offset. Makes for a faster computer in some ways and having 1 offset for the AVX seems to make prime95 happy for some reason. So, running prime95 at 5 GHz and some other things at 5.1GHz. I can also run prime95 small FFT's with this setup, but it has quite a bit of VR throttling. All temps are usually under 65C with an occasional 70 or 71C for a bit. Like the delidding a huge bunch. Still learning these BIOS settings, but getting there.

-Rodger

 

[wingman99]

What is causing the VRM throttling?

 

[DaveB]

I can also run prime95 small FFT's with this setup, but it has quite a bit of VR throttling.

Probably need some airflow over the VRMs. Since you're water cooling you might need to add a fan blowing air to the VRM area.

 

[EarthDog]

At that voltage, either the board cant handle it, or temps for the vrm are too hot. Either way, those are voltages i wouldn't want to run 24/7 anyway. Does the strix-e show vrm temps?

 

[Johan45]

There are two areas that allow you to set higher current limits in that BIOS
1. CPU Core/Cache Current Limit> set to 255.5 which is max
2. In Digi Plus section there's CPU Current Capability which will add to that in percentages.

 

[marjamar]

Thanks again for the advice/replies.

I have done a few more things which seems to be working, at least working better then before. Been running prime95 blend for about 1/2 hour I'd suppose now and everything looks stable. About 10 minutes ago, my Macrium backup started a scheduled backup and I let it run for a few minutes as I was curious. Decided to stop the backup and prime95 kept plowing away.

I did find my small fan for the VRM's and put that on. The heatsinks are still hotter then I'd like so I might want a bit larger then 40mm fan on there. Did a number of changes in BIOS (I bet over a hundred or more since the delidding) and although vcore is still higher then I'd like, I don't really think they are much too high as I am running a 5.1 clock, which must of course boot. Highest vcore I have seen with this setup right now is 1.424 under load. At idle, it's 1.392v and generally goes up to 4.08 under a 12K FFT, which seems stable to me. I post a few screen grabs I did as this prime95 is running. I think I will let it run until it drops a worker or at least an hour or more. I'm not done yet with this overclock. I really would like to keep it under 1.4, but in all honesty, I don't see that happening with a 5.1 clock. I'm pretty sure I can use most of my settings where they are and just drop to a 50 multiplier and I should be able to be well under 1.4 vcore. I will be trying that, once I believe I for sure have a stable setup at 5.1GHz. No throttling of any kind yet and all cores remaining at 5GHz @ 100% load.

-Rodger

 

[EarthDog]

Cool... does that mean you eliminated the throttling????

A stable system, to me, is a system which does not throttle for any reason either. Throttling is there for a reason to protect something...

 

[marjamar]

Cool... does that mean you eliminated the throttling????

Your a stable system, to me, is a system which does not throttle for any reason either. Throttling is there for a reason to protect something...

Over an hour now running prime95 blend and no throttling. I change a number of settings and added the fan. Actually I add the fan and tested, still throttled, so I then started eliminating, one by one, what I suspected might be contributing to the VR throttling. Actually, it was a combination things that finally worked. Having been absent from overclocking much at all for a few years, this new motherboard is taking me time to learn the ins and outs and relationships. I guess you can add some age factors with the gray matter and it gets even more interesting... Or sleepy sometimes.

-Rodger

 

[wingman99]

Over an hour now running prime95 blend and no throttling. I change a number of settings and added the fan. Actually I add the fan and tested, still throttled, so I then started eliminating, one by one, what I suspected might be contributing to the VR throttling. Actually, it was a combination things that finally worked. Having been absent from overclocking much at all for a few years, this new motherboard is taking me time to learn the ins and outs and relationships. I guess you can add some age factors with the gray matter and it gets even more interesting... Or sleepy sometimes.

-Rodger

What combination resolved the VRM throttling?

 

[marjamar]

What combination resolved the VRM throttling?

The most effect came from the VRM fan I'd expect. But I did lower voltage offset a good bit and raised CPU Load-line to reduce voltage droop. Also lowered CPU Current Capability and CPU Power Phase to Standard.

I think I will see about getting this AVX to zero. I could not run prime95 blend without dropping workers with it AVX at zero before. But there are a number of different settings in play right now that were not the same before, so I might get lucky.

I also want to play with substituting some extra buss speed for CPU multiplier. I had success doing this with my AMD chips, and I'd like to see if it works with this 8700K in any good way.

-Rodger

 

[EarthDog]

The most effect came from the VRM fan I'd expect.

But above you said the VRM fan didn't do anything?

Actually I add the fan and tested, still throttled

Once LLC is set, other adjustments, like offset or manual/static voltages shouldn't change things. Lowering CPU current is counterintuitive to what you want it to....

I also want to play with substituting some extra buss speed for CPU multiplier. I had success doing this with my AMD chips,

These do not work that way. The other busses are not tied to BCLK as they were in the far past with Intel and with AMD.

I also notice in post 68 you mention things wont pass small FFT, now you switched to blend? Mar, I really don't think there is a method to this madness friend. Blend tests memory and CPU and isn't nearly as hard on the CPU as Small FFT is. When testing the CPU alone, you should run small FFT. When you want to test more memory, then run Blend. But yeah, I can't make heads or tails of a direction, feels like watching a rabid wolf chase its own tail.