4790k will not overclock at all.....

beekermartin

Distinguished
Jan 22, 2008
580
0
19,010
I started another thread about this here: http://www.tomshardware.com/answers/id-2464234/having-issues-overclocking-gigabyte-z97x-ud5h-4790k.html

Basically if I leave everything at auto and only change the ram, select XMP and change the voltage to the recommended 1.60, I am stable with all cores boosting to 4.4 ghz. It passes 10 runs of IBT and vcore tops out around 1.26. I then found the lowest stable vcore which ended up being 1.10.

I would like to think that since I can run with all cores boosting to 4.4 at only 1.1 vcore that hitting 4.5-4.8 at much higher voltage would be possible. Nope! If I leave the vcore at auto and just select 4.5 the vcore will hit 1.40+ and still not be stable. I tried manually selecting 1.4, 1.35, 1.30, 1.25, etc and nothing works. As soon as I start IBT it hard resets about 20 seconds in. I have tried this with the memory running at the XMP profile 1 - 2133mhz and with XMP disabled and the ram running at 1600mhz with very loose timings.

So either I have the worst 4790k ever released or I am missing something???
 
Solution
Great were on the same page :) lol

We have identical issues which aren't temp related . I can't get to 4.4ghz even if i try 1.35v (it's stable at 4.3Ghz @ 1.25v)

I've tried tweaking the Ring multiplier and the Ring voltage - this had no impact even pushing ring voltage to 1.2v!

Here's my results during stress:

screenshot_43.png


Have you tried tweaking your ring multiplier/voltages?
What were your CPU peak temps during these attempts? At a guess, it's probably your temps that are causing the shutdowns, especially if you're trying to run 1.4V vcore with an H60 cooler. 1.3V is the recommended safe ceiling, assuming you can keep your CPU proper cooled. Anything over that and you're going to need a much, much better cooling solution.

You'd be much better advised to start at the 1.16V-1.28V range and shoot for a stable 4.5GHz clock with temps at 100% CPU loading at 75C or lower. If you can get the clock and the peak temp and be stable, start lowering your vcore until you find the lowest voltage you can maintain stability at (should be in the area of 1.15V-1.16V, or thereabouts). Once you've got a stable 4.5GHz and temps are good, bump it to 4.6GHz and do the same process. At the point that you can't keep it stable, and at 1.3V or lower, AND your temps at 75C or lower (you can let it go higher, even to mid-80s, but 70s would be better), you've reached your OC limit, at least without a better cooler.

The key here is take baby-steps. Leaps and bounds will earn you a fried CPU. There are a number of guides on how to OC "safely" - if you haven't read any of them, I'd recommend you take the time before experimenting more.
 
My temps are excellent at 1.1 vcore. Highest core hit 67c while running 10 passes of IBT at the extreme setting. At 1.26, or auto, highest core hit 82c. Still very good temps. The H60 in push pull is a great cooler contrary to what most people think. In most reviews the H60 is only @3-6c warmer than the H100i.

What is causing the hard resets is not temperature related. I have been overclocking for years. I know more than the average joe on these boards. This chip refuses to overclock at all. I know the issue has to be a setting in the bios but I can't figure it out. If I can run completely stable with all cores boosting to 4.4 with only 1.1 volts then 4.5 should be easy to hit with higher voltage.
 
Could wish you'd mentioned experience level in original post - with an experienced OCer, I can (more or less) assume any blank spots are filled with appropriate numbers. With a beginner, I automatically assume that the blank spots are filled with...well, blank spots:) Not that I claim any real expertise myself, just a little experience. I was having visions of 100C+ temps with no attempts at monitoring.

Have you tried any other test? I've stopped using Intel BT and Prime, at least with my Haswell and DC chips. It may just be me, but I'm convinced that IBT has arbitrarily decided to not even pretend to test for stability and just goes straight to trying to turn my rig into a bonfire. Prime isn't much better - at 4.6GHz, I can get it so it's running fine for hours, but if I stop and restart the test, it'll crash. Forget 4.8.

I tested my DC with AIDA64, stable at 4.8 on air at 1.28V and 4.6 at 1.22V (don't quote me on that one - might be 1.24V, am on the road so I'm going off memory). I've been running it at one or the other for six months now, pretty much 24/7 and benchmarking regularly, with nary an issue. That's a valid OC in my book. Did okay with OCCT also, but I like the AIDA UI better so that's what I've stuck with.

I would give it a shot with something other than IBT and see how it does. And keep your fingers crossed. Getting a lottery loser sucks -my 4770K will barely do 3.9, and that's only if I hold it's hand and promise it ice cream. I'm getting ready to either take it out and shoot it or solder it into my garage door opener remote.
 
Hello,

I'm in the same boat with a 4670k 4.3ghz @ 1.2v - if i try to get to 4.4Ghz i cant get it even with 1.35v - i'm simply not prepared to push more voltage through it so settled for 4.3 - As you already know o/c results are chip dependent.

Have you tried removing the XMP profile and using 'stock' settings - then trying for 4.5? i found that dialling in my ram AFTER the stable overclock helped.

i use Prime95 v26.6 for stress testing - give it a try. (dont get the latest version :) )
 
Lol. Thanks for the responses Volcanoscout. I appreciate it.

IBT is stable with all cores boosting to 4.4 at 1.1 vcore. If it can pass IBT at those settings I would like it to pass it at any overclock level.

What is strange is that if I manually set the the multi to 44 and leave XMP profile 1 it won't pass. If I leave the multi at auto and select XMP it passes.

Something odd is going on in the bios. I am going to post this in the motherboard section. Maybe another Z97X UD5H owner can help me out.
 


I have tried. With the ram xmp set to disabled and at stock speeds, 1600 mhz with loose timings, I can manually select the multiplier of 44 and it will pass IBT. If I leave XMP profile 1 selected it will not pass IBT with the same 44 multiplier. What is odd that if I leave the multiplier at auto it boosts to 4.4 on all cores with XMP profile 1 selected and the ram running at spec of 2133 mhz and better timings.
 
Might try browsing this thread - http://www.tomshardware.com/forum/id-1722630/intel-god-quick-dirty-guide-4ghz-haswell.html

It's got some settings tweak ideas that aren't mentioned in this thread or your previous one. Speaking of which, having multiple threads open on the same problem is considered bad form - may want to close out your current threads before you start another. If you do that, in a new thread you can still post links to the first two to save yourself some typing and Q&A time. Another option is to PM a moderator and ask them to consolidate them into a new thread in Motherboards.
 
No sweat. The three things that caught my eye with Intel God's thread were the multiple references to the use of IBT and Prime by multiple posters (successful and unsuccessful), the number of posters who were hitting the wall at relatively low clock speeds, and the adjustments to settings other than vcore and multiplier.

Some other thoughts:
Are you using group tuning, or is each of your cores clocking independently? If you're currently using group tuning, you could try per core. If you're using auto features, that might allow some improvement.
What is your vcore offset set to? I use +.01 (.001?) - when I leave it on auto, even if the peak voltage jump is still within safe levels, the jump itself can cause stability issues. I'd have to check, but I think my default offset was something like +.25 which is a pretty big jump.
 
I will have to check when I get home. I left everything set to auto except for the vcore, ram xmp, and multiplier. Most reviews I've read about the Z97X UD5H that is what was recommended.

With everything set to auto and the vcore set to 1.10 in the bios CPU-ID reported 1.098 vcore and HWmonitor reports 1.16 while running IBT.

I know with my X58 rig it was called Vdroop. I forget the exact way it worked but I believe with it enabled the voltage would jump up under load and with it disabled the voltage would drop slightly under load. I don't know what that setting is called on the UD5H. Load calibration or something like that. Anyway, it is set to auto regardless of what it is called.
 
When you set the multiplier to AUTO, your CPU is designed to TURBO BOOST to 4.4Ghz - That's +100mhz per core, NOT all 4 cores +400mhz.

The 4790k supports the following memory types: DDR3-1333/1600 NOT 2133Mhz - Your motherboard does!

This may explain why your CPU crashes @ 4.4 with XMP enabled - Your CPU's memory controller that sits inside your CPU is heating up to the point of instability. (See image below on the right hand side)

screenshot_40.png


PLEASE PLEASE PLEASE read this thread: (the guys a TH legend)

http://www.tomshardware.co.uk/forum/id-2450639/overclocking-cpus-memory-controller-effect-raised-multiplier-cpu-overclock.html
 
Cave, some of your details are correct, but your conclusions are not. Just take a glance at some overclocking forums and articles. These CPUs are certified by Intel to support 1333MHz and 1600MHz, but that does not mean that they can't handle much higher rated DRAM. If you look at my sig, you'll see that my system runs 4.8GHz with 2400MHz Tri-X's running at full rated overclock speed - and that's a daily driver, not just for benchmarks. If your conclusions were correct, there would be absolutely zero market for unlocked K-chips and O.C. capable DRAM, which is obviously not the case.

To get the max performance out of a system, it's a matter of finding the right mix of components and settings and, when it comes to CPUs, a large dose of luck. Some CPUs come off the line with a huge amount of overclocking headroom, while others have almost none.
 
As regards the article, it's a good one - read it before. But you should give it another read. It's not saying don't overclock or that overclocking is just unfeasible marketing hype - it's saying that if you're going to do it, it's important to know the details of what it is that you're asking of your computer to do, to have appropriate components (particularly cooling), and to do it in a logical order.
 
Hi Both,

The article essentially says running anything > 1600mhz ram adds heat to the CPU memory controller, and this heat is part of the whole CPU temperature, and temperature may inhibit the maximum multiplier.

I haven't concluded, i said "may" be the reason - sorry for the confusion.

Getting back to your situation, what are the temps of the CPU when it BSOD's with XMP enabled? i guess at this stage its a matter of elimination to find the source of the BSOD and i for one, will pay close attention to this thread as i may be able to get my chip over the 4.3 ceiling i believe it currently has.

i'll remove the last line of my previous post as i don't want to confuse people :)
 
Just for the record my cpu runs at 4.4, for all 4 cores, when I leave the cpu multiplier on auto. It is also stable with my memory set to XMP which makes it run at 2133 mhz with its rated timings. The only thing I change is the ram voltage to 1.60 because it defaults to 1.65 and G.SKILL recommends 1.60 for Haswell rigs. If I leave the Vcore at at auto it will run at @1.25 volts while running IBT. I have found that it is stable as low as 1.10 volts.

So basically all I change is the vcore voltage to 1.1, select xmp profile 1 and then change the ram voltage to 1.60. With those settings and leaving everything else set to auto in the bios, besides the usually settings (raid, boot order, turn off cpu video, etc), I am stable with all 4 cores boosting to 4.4. Well stable enough to pass 10 runs of IBT and pass 3dmark11.
 
Correct Cave! Even with xmp disabled and the ram running at 1600 mhz and very loose timings I can't get to even 4.5 ghz. I never manually tried 1.40 volts. That is what it hit when I left the vcore at auto and set the multiplier to 45. I did try 1.25, 1.30 and 1.35 and they all did the exact same thing. Restart @20 seconds of starting IBT.

I am positive my issue is not temperature related.
 
Great were on the same page :) lol

We have identical issues which aren't temp related . I can't get to 4.4ghz even if i try 1.35v (it's stable at 4.3Ghz @ 1.25v)

I've tried tweaking the Ring multiplier and the Ring voltage - this had no impact even pushing ring voltage to 1.2v!

Here's my results during stress:

screenshot_43.png


Have you tried tweaking your ring multiplier/voltages?
 
Solution
Could be worth a try, but if you can't get it to 4.5Ghz upto 1.35v without touching anything else it could be down to your chip.

What PSU do you have?

i am sure there will be a solution! - there are plenty people out there getting 4.8!

You may need to adjust the bus clock, but this effects everything else in your PC and it's not something i'd personally recommend.

screenshot_44.png
- @ http://www.anandtech.com/show/8227/devils-canyon-review-intel-core-i7-4790k-and-i5-4690k/2
 
What is so hard for me to understand is that my chip can run at 4.4 on all 4 cores as low as 1.10 vcore. How can it not handle at least 4.5 with more voltage? It doesn't make sense to me. I guess it could be a chip that can handle 4.4 on all cores at very low voltage but can't be overclocked at all. That is always part of the silicon lottery. Usually a chip that runs on such low voltage as mine does at stock speeds but with all cores boosting to 4.4 should be a great overclocker. That is why I feel I am missing something.