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

[wingman99]

I wish you would of used XTU https://downloadcenter.intel.com/download/24075/Intel-Extreme-Tuning-Utility-Intel-XTU- Or HWiNFO https://www.hwinfo.com/ so we could of seen what was causing the VRM throttling. Like earthdog says what you have said and done makes no sense.

This is what VRM thermal throttling looks like.

 

[marjamar]

EarthDog - Not to get particular here, but I never said the VRM fan didn't do anything. I said it still throttled, which it did in both small FFT's and in blend, which is why I did adjustments to my settings. I decided to use prime95's blend not the small FFT's to see if I can get that right before going back to the small FFT's tests. Now that it is stable in the blend testing, I will consider whether or not I need to be so stable to run the small FFT test. In the real world, I don't think anything will work the CPU so hard as this small FFT test does.

As to the methods I am using, all I can say is I have a way of working things out that isn't always agreeable to everyone. I won't defend it, but I will say most of the time I get where I am trying to go, one way or the other. In this case, simply setting core voltage to 1.??? and multipliers to 50 would not let me get stable in prime95 blend, let alone small FFT's hitting 100 C. Since then, I have delidded and now I can try to get small FFT's working and temps stay about 85 C, and my clock is 5.1GHz @1.408 -1 AVX offset. Still can't get past about 4 minutes though, so I really don't know if this CPU has it in it at 5GHz to do that. Fortunately, I do have blend running, now over 2 hours, without a problem and no throttling of any kind. Highest temps hit about 75 C. with most below 65 C. and vcore pretty much at 1.408v.

wingman99 - I have been using Intel Extreme Tuning Utility, but I didn't think you guys were using it, so I never captured screens to post. It is far in a way ahead of the other utilities I have or have used, and it saves a lot of time being able to test in the OS various settings and then adjust BIOS to what is showing promise in XTU.

Here is the last test I did trying to get small FFT's to run. I did a print screen just about a minute before losing a worker and I had to stop the test. You can see there was not throttling of any kind during that short run. And on the 2 hour blend test I did before this, there was not throttling either, but I cropped out the right monitor screen on all of those print screens I captured, figuring I was not going to post them anyways.

I can run that blend test again though, just let me know.

Suggesting are welcome of course, including your setups would be nice, HA.

-Rodger

 

[EarthDog]

Hey marj,

Good luck to you. Do it the way you feel its best. Right or wrong, you will figure it out eventually.

 

[wingman99]

Looks like your doing fine. I was just confused with what you were saying.

 

[marjamar]

My new 360 fan setup is all finished and work really well by all accounts. I really like the radiator outside of the computer for a bunch of reasons. I may add quick disconnects, but I have such good flow now, I am not sure I would like any restrictions that may come with the connectors I have been eyeballing. I'm using my aquacomputer to control all fans and pumps and this new 360 radiator is no exception. Having 6 fan wires along with 2 tempature sensor wires and 1 flowmeter wire, it was going to be not so nice to bundle them all up and put them into my computer case. Luckily though, I found a nice little fan splitter made by, of all people, aquacomputer. Called Splitty9, it is simply the best way I could find to limit the huge potential pile of wires I would have to cable into the computer case. Now, I have in total 4 sets of wires that I nicely braided up and made a small hole in the rear of my case to pass it through.

Fans are fully controlled like they should be, and not loud at all. I am using very quiet fans I purchased a few years ago for my original build I bought this radiator for. So, even at full speed, nice and fairly quiet, and most of them time not noticeable to me. Here's a couple of quick pics I took.

-Rodger

 

[marjamar]

Made just a couple small changes in BIOS and got a stable Prime95 small FFT run for 31 minutes when I stopped it. I think that's long enough for me as I will never stress this computer anywhere close to this much for so long in anything I would be doing. Here's a print screen just before I stopped it.

-Rodger

 

[wingman99]

That is great. What changes did you make?

 

[marjamar]

That is great. What changes did you make?

Sorry, lost my profiles. Did an update this morning and all my profiles were lost. I am starting over from memory. I made copies through the profile settings, but they are not found when I try to restore them. Kinda wish I would have bought the Hero mb instead of this one. Seems like there is something wrong with the way this one does adaptive voltages to me -- Like it really isn't paying any attention to my settings. Guess I have to have a look at what others are doing, trying to use this adaptive voltage setting.

-Rodger

 

[wingman99]

When you flash the BIOS it erases the profiles. I write mine down for those situations. What do you think is wrong with additive voltage, I need better description to help?

 

[EarthDog]

Many boards will allow you to export profiles to usb.

 

[marjamar]

I was/am trying to have a lower clock when not working the CPU. I know at 5 GHz that may not be very low, but to run things that work the CPU hard, I find my core voltages where going above 1.4v in most of the settings I was working on.

Since I posted here yesterday, I have found my lost profiles, as I did export them from bios. They were on my usb, but as text files, which of course didn't show up when I was looking to restore them in bios. I had to manually reenter them, have them back, but that firmware update changed somehow how they work now. I can no longer run a 1/2 hour prime95 small FFT, or extended blend test using either of them.

Today I have been trying some more settings, and I have been able to get reasonably stable at 5 GHz @ 1.360 core voltage. I am tinkering with ways to keep that lower voltage and run prime95 small FFT for an extended period. So far, I can get perhaps 12 minutes. The blend test seems stable now at that voltage. I will be letting that run overnight to see.

I am not unhappy with the way it's working now and it does benchmark well, so it might be about as good as I can get it. If I raise vcore much, I will most likely see VR throttling come back. We'll see. Here my last 30 minute prime95 blend test.

-Rodger

 

[wingman99]

Does the ASUS ROG STRIX Z370-E show VRM temperature?

 

[marjamar]

Does the ASUS ROG STRIX Z370-E show VRM temperature?

Not with any software I have used, not be name anyway.

-Rodger

 

[marjamar]

Well I am satisfied with my overclock pretty much I think. I could run this 5.1 GHz clock all the time if I wanted to, but I like the 5GHz version of it more as vcore is better clocked down that 1Ghz. Turned out the latest version of Prime95 I was running was the problem all along. Don't know why, if it's something buggy with it on my system or what, but it kept dropping a worker right at 3 minutes exactly into a small FFT run. After I started noticing the same time coming up over and over, I decided to try an earlier version (v.26.6.build 3). Sure enough, it ran small FFT's just fine at 5.05 GHz and then later ran 5-1/2 hrs. @ 5.1GHz. I think I could bump the cycles up if I wanted too. Knowing me, I suppose I will be trying. Not a lot of good volts left however since I am at 1.434 vcore under load now at this 5.1GHz clock. Temps are super too, topping out at 78° C. in that 5-1/2 run. No throttling at all either. In both of these last 2 clocks I'm adding .5 or 1 GHz of BCLK overclock. I have not overclocked my ram at all yet, so there's no problem with the little bump up it gives. Actually gives a bit better XTU benchmark too. I suppose I will have to explore overclocking my ram next. I do like my aquacomputer gauges I designed doing this testing. The history of a run is visible in the bottom of the color-band in light blue on each gauge.

Here's pic of the last 5-1/2 hr. small FFT torture test.

-Rodger

 

[EarthDog]

The problem wasnt p95... its your cpu not being stable. The older versions do not use the same instruction sets so its a different, less stressful, test than the latest versions supporting the latest instruction sets. Pretty sure anything before 27.7 doesnt do avx... so it makes sense it works now as it isnt testing as hard without avx.

Edit: Wingman can help with the details as he knows those differences better than I.

 

[marjamar]

The problem wasnt p95... its your cpu not being stable. The older versions do not use the same instruction sets so its a different, less stressful, test than the latest versions supporting the latest instruction sets. Pretty sure anything before 27.7 doesnt do avx... so it makes sense it works now as it isnt testing as hard without avx.

Edit: Wingman can help with the details as he knows those differences better than I.

I think prime95 is has switched gears to provide a testing platform for AVX offset testing, more then pure overclock testing. I also think there may be settings to use to disable this feature in the newer versions. Not being very knowledgeable on things AVX, all I can really say is since I am not using AVX offsets right now, do I really need to test how well AVX works?

-Rodger

 

[EarthDog]

Avx is a real-world instruction set. Whether you personally use it or not, i wouldnt know. But p95 added it to improve its testing and speed up the gimps software, not for avx offset specifically.


Many prefer to test their system for worst case... some... dont care to.

 

[marjamar]

I think prime95 is has switched gears to provide a testing platform for AVX offset testing, more then pure overclock testing. I also think there may be settings to use to disable this feature in the newer versions. Not being very knowledgeable on things AVX, all I can really say is since I am not using AVX offsets right now, do I really need to test how well AVX works?

-Rodger

Found the setting to turn AVX testing off in Prime95, later versions. Looks to be the same testing stress as the older version as it is running now. Just have to add a line "CpuSupportsAVX=0" to the existing file "local.txt". Good to know.

-Rodger

 

[marjamar]

Avx is a real-world instruction set. Whether you personally use it or not, i wouldnt know. But p95 added it to improve its testing and speed up the gimps software, not for avx offset specifically.


Many prefer to test their system for worst case... some... dont care to.

I think it boils down to what sense of "reality" one holds to. Personally, I have to wonder why test for something that never is? I suppose if I were in the sciences, that would not be a true statement, but as a person who has done multimedia of for over 20 years, my world doesn't necessarily need the "hard knocks" prime95 seems to like to pass out with this AVX offset stuff as currently offered. Besides, AVX would seem to be a method to use high overclocks on lighter loads with a facility to downclock the heaver ones automatically. That would mean trying to stress AVX past a reasonable point would really be using an opposite philosophy to the purpose for AVX -- Sort of an antithesis to it.

I'm just glad I can do overclocking without crashing now testing with prime96 small FFT's. It also opens up the possibility of testing for higher overclocks, using prime95, which I have used since I found it some many years ago.

-Rodger

 

[EarthDog]

Avx instructions sets in the cpu are not there to be 'clocked down'. The bios has the ability to control what speed the cpu uses when running those, yes, but it is not there with the intent to do so. It CAN be clocked down when using those instruction sets BECAUSE its so hard on the cpu.. that is wjy the option exists. It can increase an overclock by running the avx instruction sets slower. But its not a feature MADE to clock up or down it CAN.

If you dont use it... dont test for it. I was simply correcting the point about the latest version of p95 being messed up.