[SOLVED] Advice on Overclock: 9700K

FurryVengence

Commendable
Jul 26, 2019
198
25
1,640
After about 6 months running this new build I thought it was time to finally take advantage of the extra money I spent for the unlocked 9700 I7 I purchased. Bare in mind I've never done much with overclocks before. I know very little outside of what the tutorials pointed out to me.

Took a look over at youtube and found a few basic overclocking guides, and did the "basic" approach. I am currently running the following (Anything I dont mention its left on auto or whatever the bios defaulted to):

5 GHZ clock across all 8 cores.
Using standard XMP2 profile for RAM
SVID behavior 0
running 1.29 on voltage
CPU VCCIO 1.10
system agent 1.10
Level 6 CPU Load line
170 % current capability
Long duration package is 4095
package power time 127
short duration power 4905
cpu core cache 255.75

My current PC build:
INTEL i7 9700k (5.00G)
Asus TUFF z390-PLUS GAMING (WIFI)
EVGA GEFORCE RTX 2800 TI
NZXT h700 case (with the 4 stock case fans)
DARK ROCK PRO 4 cooler
16 GB DDR4 RAM 2666

So far things seem to be running fine. But I'm wondering if its enough, did I just hit a dial or two . Whats really good ways to test I truly have stability? Currently my CPU temps even under load appear to max out at 60 C. Any advice or input would be very helpful. I put a lot of money into this machine already I dont want to burn out randomly and me not know it.
 
Solution
Aidia64 can test most of your system as well. Noctua NH-D15 is a good choice in air cooling. The main issue is that voltages like vcore are specific to the chip being overclocked. 9700k will run hot and will hit +90c when overclocked in stress tests. Temperatures are much lower when running games. Stress tests, just push the limits so that you can find out if you are stable enough.

Many reviews have 5.1GHz for the 9700k and 5GHz for the 9900k. Review overclocks don't have to be 100% stable or use vcore voltages that are long term 24/7 safe. For example, you can be stable 5.2GHz @1.4 volts but you can never cool that in stress tests without delidding and going direct die water cooling...

FurryVengence

Commendable
Jul 26, 2019
198
25
1,640
Update found a program called OCCT and it provided lots of errors in minutes. Pushed my computer to pre-oc. No more errors. Any advice on where I could push my overclock would be appreciated.
 

Panzerbjorne39

Commendable
Jan 14, 2020
27
3
1,545
Based on you mentioning occult and the specific setting I’m guessing you followed that one guy video who also does car aim stuff? I followed that too and it wasn’t stable. It really needs to be tailored more to your specific chip requirements which vary from chip to chip.
It’s kinda taboo with computer nerds but give ASUS AI over clock mode a try. It’s a learning algorithm based on your cooling capability. Works pretty well. It puts my 9700 at 4.9 all core and is very stable. Gives a nice performance boost over stock.
 

FurryVengence

Commendable
Jul 26, 2019
198
25
1,640
Thanks for the reply, if you don't mind me asking what kind of temperatures do you get ?

My worry with using anything AI generated is thermals. I'm using a hybrid at the moment, using recommend settings from reddit (at a 5ghz clock) , minus the core/cache voltage. The recommended 1.30 and 1.29 are causing instabilities. Even going to 1.35 didn't help. Leaving that on auto seems to keep it stable. However on 100% utlization I'm at 92 degrees at the highest. While while it mostly sat at 82 degrees C. I'm not running liquid cooling and I'm not really planning on moving to it. I've been quiet pleased with my high 60 C pre overclock. And id like to keep it as cool as can be.
 
I'd speculate many folks have the '5 GHz or it's useless' mindset....; it certainly is a popular target.

It would seem a reasonable suspicion there was not much of a load if temps were not much above 60C, as a 9700K I built for a friend running at stock clocks (1,2 cores at 4.9 GHz, 7,8 cores at 4.6 GHz) hit 68C on a Noctua NH-D15.

A decent intermediate step would be to merely enable MCE within the BIOS, which should induce all cores to boost to 4.9 GHz if cooling/power budget allows...

One can do thermal testing with Prime95/small FFTs/all AVX disabled for 15 minutes...; if temps stay at 80C or below, you'd never see those temps in gaming anyway....
 

Panzerbjorne39

Commendable
Jan 14, 2020
27
3
1,545
Thanks for the reply, if you don't mind me asking what kind of temperatures do you get ?

My worry with using anything AI generated is thermals. I'm using a hybrid at the moment, using recommend settings from reddit (at a 5ghz clock) , minus the core/cache voltage. The recommended 1.30 and 1.29 are causing instabilities. Even going to 1.35 didn't help. Leaving that on auto seems to keep it stable. However on 100% utlization I'm at 92 degrees at the highest. While while it mostly sat at 82 degrees C. I'm not running liquid cooling and I'm not really planning on moving to it. I've been quiet pleased with my high 60 C pre overclock. And id like to keep it as cool as can be.
The ASUS technical marketing guru JJ is telling me adaptive/dynamic vcore is the way. I’m tending to agree. It’s logical, makes much more sense. My temps are great in a 360mm aio, idled high 30’s, max load 84.
 

zx128k

Reputable
Aidia64 can test most of your system as well. Noctua NH-D15 is a good choice in air cooling. The main issue is that voltages like vcore are specific to the chip being overclocked. 9700k will run hot and will hit +90c when overclocked in stress tests. Temperatures are much lower when running games. Stress tests, just push the limits so that you can find out if you are stable enough.

Many reviews have 5.1GHz for the 9700k and 5GHz for the 9900k. Review overclocks don't have to be 100% stable or use vcore voltages that are long term 24/7 safe. For example, you can be stable 5.2GHz @1.4 volts but you can never cool that in stress tests without delidding and going direct die water cooling https://www.kitguru.net/wp-content/uploads/2019/10/5200MHz-OC-CPUZ-1-1024x1024.jpg . You can still run that voltage for bench runs in a review. https://www.kitguru.net/wp-content/uploads/2019/10/Temps-Cine-OC-XTU-1.jpg core max 95c.

We managed to hit a 5.2GHz all-core overclock with no AVX offset using 1.375V and Turbo LLC in the Gigabyte motherboard UEFI. This resulted in a software-read voltage of 1.380-1.392V when under load, which was just manageable with our Corsair H100X 240mm AIO cooler with 2435 RPM fans (not enough cooling).
We were forced to upgrade our 240mm Corsair H100X AIO from 1700 RPM fans to 2435 RPM Corsair SP120L alternatives in order to keep stock 9900KS temperatures in check.
SP120L fans running at 2435 RPM. https://www.corsair.com/uk/en/Categ...IES-LED-CONFIG/p/CO-9050014-WW#tab-tech-specs £33 ($40)for two. They are pushing extreme air flow because they can't cool 5.2 GHz enough to bench it. chart shows highest fan speed top cooling

My 3800x with the EDC bug is ~5300 which is an extreme overclock. With just the RAM overclock 5200 maximum. Both are faster than this 9900ks in cinbench r20 and I can cool the 3800x correctly. https://www.kitguru.net/wp-content/uploads/2019/10/cine-R20-multi.jpg

All this to hit 11472 in Time Spy CPUhttps://www.kitguru.net/components/cpu/luke-hill/intel-core-i9-9900ks-5ghz-8c16t-cpu-review/7/. 3800x with the same extreme mindset to cooling https://www.3dmark.com/spy/11990955 11600 (if you have 5c outside and the window open to let the cold air in, normal stock core boost with RAM overclock, normally 11400-11500). With a good 9900ks overclock they should be higher than 12k cpu. In Aida64 the maximum temperature is 98c which is unstable (because temperature is unsafe. A 2c variation in ambient temperature is all that is needed for throttling to occur as it too small a margin, making this OC not 24/7) because temperature is too high. CPU package power of just under 210W.

What you are not told is that the more you push an overclock the harder it can be to keep stable. You start having to mess with settings you never touched before. RAM overclocking is a real pain for this reason, you have so many settings to play with and each one can affect stablity. In the end, some point, "the wall" will be unstable no matter what you do. Accept this limit.

Given that CPU chips and RAM kits all have their own characteristics. You can't blindly follow a guide, at some point you have to find the balance that works best for your build. At the start of an overclock you can just put in a the max safe vcore and hit some frequency. The more you push things, the more you have to make changes. Like at some point having maximum vcore causes you to have high temperatures, so to reach that next 100Mhz you start tuning vcore lower until you find the minimum value you are stable. View: https://www.reddit.com/r/intel/comments/9qvb2y/my_9700k_aint_boosting_to_49ghz/e8c2fcb/?st=jnnlbfy6&sh=c77f4c38
Hoping to reduce the total watts required to hit that frequency and thus the cooling required. If that works, you get an extra 100MHz but if your temps are still too high then you have to back off the overclock. Also you can use an AVX offset to reduce frequency with AVX loads. This can help you overclock higher for instructions like SSE which will affect benchmarks like Time Spy and many games.

There are other settings other than voltage and sometimes they can help too. Like switching frequency of the VRMs, LLC etc. The only way to make sure you hit the frequency you want is to bin for it. This requires buying lots of CPU's and picking the fastest for your overclock. You sell the losers on ebay. The data you get from the process gives you useful information that you can turn into a frequency voltage graph.

intel-9th-3.jpg


From the graph you can make general recommendations: 1.32v for OC i9-9900K 5GHz, 1.37v for i7-9700K 5GHz and 1.43v for i5-9600K 5GHz.

So to overclock the 9700K, you first need to start with the CPU Ratio. The target can be 5GHz, so just type in “50” for the CPU Ratio. Then as a starting point you try 1.37 volts which you can increase or reduce as required. If you can't reach 5GHz within safe voltage and temperature you reduce the CPU ratio and start again tuning the voltage.

Then you change Ring Ratio to 47. You can try your own Ring Ratio, but MSI for example suggest a Ring Ratio that is 3x less than the CPU Ratio. Ring frequency is the frequency of non-core parts of the CPU, such as memory controller and cache. Higher Ring frequency is helpful for better benchmark performance. Too high a ring frequency can also affect the stability of the non-core parts of the CPU. Vcore voltage is also the CPU ring voltage. https://www.msi.com/blog/intel-9th-cpu-overclocking-5ghz-with-z390-motherboards https://www.gigabyte.com/FileUpload/Global/multimedia/2/file/525/946.pdf

Remember that you have the option to overclock both RAM and CPU. CPU overclocking will always hit some limit, temps or voltage quickly. On the other hand RAM overclocking is a huge time sink which can take weeks of testing and fiddling with settings (there are a lot of settings to play with) to find the limit. You can overclock both RAM and CPU to reach the target performance you want.

Just treat overclock like a project, it takes time and you need to reseach and understand what you are changing. This way you can reduce the risks of overclocking.

Just putting maximum settings like LLC etc is not the best way to keep your CPU cool. You can limit the amount of power or current your cpu uses. You can try offset with adaptive voltage instead of manual and keep c-states enabled.

For example you can try other methods of reaching 5GHz, https://forums.bit-tech.net/index.php?threads/9900k-5ghz-1-2v-guide-gigabyte-z390-master.353729/

There are lots of overclocking guides and videos. It's worth reading and watching as many as possible.

Example of a video guide,


Just be careful some guides and videos get some things wrong, the video above is done by Asus. You need to reseach enough to understand whats happening and be able to catch mistakes. Remember its your CPU you will break if you make a big mistake.
 
Last edited:
  • Like
Reactions: Phaaze88
Solution

FurryVengence

Commendable
Jul 26, 2019
198
25
1,640
Currently I am doing something similar to the post you have linked. I left a few settings, particularly those for voltage set to auto that most guides told me to push. Including leaving things like speed step on. At idle I'm in the 30s. And in 4 game tests I ran the max temp I saw was 75C. Granted that never stuck, it always immediately dropped. But the stress test I ran (didn't try aidia yet) did push it to 90 max over 10 minutes. I'll have to keep playing with settings , seeing if I can get better thermals. If that's even really a concern at those core temps. Just glad I got away from the instabilities.

Luckily this is my dedicated gaming machine and it doesn't get used for anything else. I also turn it off after use, so I'm not as worried as some maybe for CPU life.

Guess I just have a lot of research ahead of me. Thank you for the advice everyone.
 
First of all, realize that the potential overclock of a chip can vary.
Here is what you can expect:

As of 2/6/2019
What percent can get an overclock at a somewhat sane Vcore in the 1.337 to 1.375 range.
And AVX offset = 2.

I7-9700K

5.2 10%
5.1 35%
5.0 78%
4.9 100%

You could try the intel performance maximizer app that will do an overclock for you:

https://www.tomshardware.com/reviews/intel-cpu-auto-overclock-performance-maximizer,6179.html

Stress testers are just that, they create stress.
Some generate heat, like the intel IBT.
some use stressful instructions like prime 95.
OCCT can use AVX instructions which are particularly stressful, but not that commonly used.

How good do you really need to be?
My simple approach would be to leave all voltages on auto and gradually raise the multiplier.
As the multiplier is increased, the needed voltage will increase.
You could use a simple stress tester like the one in cpu-Z.
Or something like OCCT.
Monitor the voltage with cpu-z. something in the 1,3 to 1.35 is ok, not more than 1.4
Also monitor the temperature with something like HWmonitor.
A stress test should probably be stopped around 85c. That is,I think, the default for occt.
The processor will monitor temperatures and will slow down or shut off if it detects a dangerous temperature.
That is around 100c.

Set the avx offset to 2. That reduces the multiplier when avx instructions are present.

Lastly, implement speedstep and adaptive voltage.
That reduces the multiplier and voltage when the processor has little to do.
 

FurryVengence

Commendable
Jul 26, 2019
198
25
1,640
So leaving yesterday I had what I felt was a overall stable overclock. Only concerns was a stress test that touched 92C within 10 minutes. And game temps hitting 75 at max. Watched some videos, more guides and tried making small tweaks. Forcing an manual voltage. 1.30 (nothing insane) and an XMP2 ram config. And some how it ended up corrupting my OS. Not a huge deal but frustrating. The changes I made I feel fall all within the average of where people say I should attempt my voltage. And it made it pretty fiercely crap out.

I will mostly end up trying this AI tweak solution, which im sure will go clock per thread. But Im wondering if there are issues with my CPU to begin with. I feel like 1.30 to 1.35 should be achievable at 5 GHZ or below? And while I dont have water cooling this Dark Rock Pro 4 has done an admirable job with the i7 so far. All the stress test at PRE OC had no stability issues whatsoever thus far. Guess Ill have to do some checks using CPUID to see what my voltages are pre-overclock.
 

FurryVengence

Commendable
Jul 26, 2019
198
25
1,640
After more tweaking I have a stable cpu at 4.9 GHZ with a max core temp hitting 83 C. Its never sustained thankfully, mostly averaging 70-75 C . But I am currently running at 1.40 Volts. Maybe hoping to see if I can get stability at 1.35 V that should lower the temps to the point I'd be really happy.
 

zx128k

Reputable
You need to boot from an extrenal drive to test overclocks. CPU and RAM errors are common when a system is unstable. This can and does lead to data corruption. Sometimes you are forced to reset the OS install. Other times its a fresh install from a CD-ROM disk.

Overclocking is about changing one setting and testing. It can try ones patience when trying to find the one RAM setting to make a RAM overclock pass 32 passes of memtest test 5 (errors after 24 passes). Then find out it procODT that needs changing. So all the changes to timings was pointless.

Just take your time, be methodical and patient. Research everything.
 

FurryVengence

Commendable
Jul 26, 2019
198
25
1,640
I've been figuring that out slowly as I go, a lot of newbie mistakes for sure. Got my overclock down to 1.35 volts and temps maxing out at 80C on one core. Most didn't go above 75C. Whats odd is the gap from 4.9 GHZ to 5.0 GHZ. I can hold stable overclock at 1.35 volts (maybe less) for 1.35 volts. Yet It appears I will have to push 1.5 volts to maintain a stable overclock at 5GHZ.

I'm wondering if this one of those 'silicon lottery' type situations? Or maybe I am missing something?

While I definitely cold attempt to push volts at that level, given my current circumstances with cooling I don't think its worth the risk. I had it at 90C at one point and that even for a short time is past my comfort level. To much money spent on this rig already.
 
  • Like
Reactions: zx128k
Do not worry too much about temperatures.
You have a very good cooler.
The processor will monitor it's temperatures and will throttle or shut down to protect itself if it detects a dangerous temperature. That is around 100c.
Running close to 100c for a short time peak is not alarming.
Under normal gaming, up to 75c. is ok and peaks of 85c. should be ok.

With a good cooler, your OC limit will be determined by the voltage you will tolerate under load.
Continuous running @ 1.4 is probably not good for processor longevity. 1.35 would be more normal. Exceeding 1.5v will likely damage your chip.
CPU chip binning measures what multiplier a given voltage is needed for stable operation.
Not much that can be done about that unless you pay more for a well binned chip.
I think your chip is about average.

I would be interested in your results if you tried the Intel ITM overclock app.
It takes an hour or more and tests all aspects of your particular chip and cooler.
My understanding is that an expert overclocker can do a bit better.
But, the ITM has the benefit that it is an intel sponsored app if anything should happen.

I might add that shutting down a pc may not be helpful for cpu longevity.
There is stress on electronics starting from cold.
As a practical matter, I suggest you try using sleep to ram instead.
That puts your pc and monitor into a very low power state, not much different from a full power off. A benefit is that sleep/wake is only a handful of seconds.
 

FurryVengence

Commendable
Jul 26, 2019
198
25
1,640
I
Do not worry too much about temperatures.
You have a very good cooler.
The processor will monitor it's temperatures and will throttle or shut down to protect itself if it detects a dangerous temperature. That is around 100c.
Running close to 100c for a short time peak is not alarming.
Under normal gaming, up to 75c. is ok and peaks of 85c. should be ok.

With a good cooler, your OC limit will be determined by the voltage you will tolerate under load.
Continuous running @ 1.4 is probably not good for processor longevity. 1.35 would be more normal. Exceeding 1.5v will likely damage your chip.
CPU chip binning measures what multiplier a given voltage is needed for stable operation.
Not much that can be done about that unless you pay more for a well binned chip.
I think your chip is about average.

I would be interested in your results if you tried the Intel ITM overclock app.
It takes an hour or more and tests all aspects of your particular chip and cooler.
My understanding is that an expert overclocker can do a bit better.
But, the ITM has the benefit that it is an intel sponsored app if anything should happen.

I might add that shutting down a pc may not be helpful for cpu longevity.
There is stress on electronics starting from cold.
As a practical matter, I suggest you try using sleep to ram instead.
That puts your pc and monitor into a very low power state, not much different from a full power off. A benefit is that sleep/wake is only a handful of seconds.

I really appreciate the advice. I had fears about pushing past 1.4 particularly because no one mentioned that voltage ever. I may at some point attempt to push down to 1.30 I think it maybe possible. Right now I'm quite content with 1.35V. I do want to try the intel overclock tool. I am curious to see what they say. Sadly I have yet to have a chance to test against any games. After my last debacle with my OS corrupting. I hadn't even installed steam yet. Rather I just went straight for all my drivers and some stress tests. Given my last game tests, (with what I learned was 1.57 V , on my initial overclock with the ASUS auto settings ) I was still getting at max 75C with the few games I tested. COD, Planet Coaster and Elite Dangerous. Those went relatively smoothly, so I have little concern for this voltage level.

Currently I am using manual voltage, if I switch to adaptive voltage, what would you suggest I put increment at?
 
It has been a while since I set up my overclock.
I do not remember how I set adaptive voltage; sorry.
Then, also the bios nomenclature differs in different motherboard bios.
The essential thing about adaptive voltage is that it lets the processor lower voltage when it is not needed. That is a good thing.
 

zx128k

Reputable
I've been figuring that out slowly as I go, a lot of newbie mistakes for sure. Got my overclock down to 1.35 volts and temps maxing out at 80C on one core. Most didn't go above 75C. Whats odd is the gap from 4.9 GHZ to 5.0 GHZ. I can hold stable overclock at 1.35 volts (maybe less) for 1.35 volts. Yet It appears I will have to push 1.5 volts to maintain a stable overclock at 5GHZ.

I'm wondering if this one of those 'silicon lottery' type situations? Or maybe I am missing something?

While I definitely cold attempt to push volts at that level, given my current circumstances with cooling I don't think its worth the risk. I had it at 90C at one point and that even for a short time is past my comfort level. To much money spent on this rig already.

With some cpu's there is a wall say 4.9MHz were getting that extra 100MHz takes more and more volts. This is the point were vcore no long scales like it did before reaching the wall. You may find that the voltage scales by 0.x voltages to the wall then its huge increases to get to the next bump in frequency. Sometimes you get luckly and tuning some other setting helps you scale at bit again. Like increasing VRM switching frequency or LLC sometimes helps. The more you push the voltage the hotter its going to become.
 

FurryVengence

Commendable
Jul 26, 2019
198
25
1,640
I think I finally hit the sweet spot. Have a lot of it on auto, but forced 5ghz on all cores with a small negative offset. Gets me passed the benchmarks and pushes 5GHZ solid in games. All while keeping low temps (avg of 60 or below) So I can't complain.

Thanks everyone for helping me out. I could always dig harder. But I will take what I can get.
 

FurryVengence

Commendable
Jul 26, 2019
198
25
1,640
I'll have to double check that, good point. Temps have been rock solid -- which shocked me. Yesterday I had no instabilities but kept on seeing temps hit 100C+. Turns out I was pushing a setting a bit too hard. Moved that back to auto and forced manual voltage for the moment. I had it set to 1.35 V. I'll run a few more tests, but overall I consider this a decent success. Even if during anything with AVC it dials it down a bit. 5GHZ rock solid in games works for me!
 
  • Like
Reactions: zx128k