i5-8600k [email protected] Questions about Voltage and Stability.

[alpha9101]

Hi,

It has only been a month since I got this build and after reading and viewing many guides I got myself to overclocking. This is also my first post and if I'm posting in the wrong section, please forgive me and move my post to an appropriate section. I'm basically making this long post to get some experienced opinions on my overclock and if I should change something else to fine tune. Apologies in advance, for the long post.

Specs:-

CPU: i5-8600k
Motherboard: Gigabyte Z390 Aorus Elite(rev-1.0)
PSU: Thermaltake Toughpower 750W 80+ Gold
RAM: G-Skill Trident Z 16GB(8x2) @3200Mhz/Timings: 16 18 18 38(XMP Profile)
GPU: EVGA 1080ti FTW3(No OC)
Casing: Cooler Master MB 500
Air Cooler: Thermalright True Spirit 120M(Will be upgrading soon)
Windows 10 64-bit

BIOS:-

BCLK:100Mhz
Core Clock Ratio: 49
AVX Offset: -3
Vcore: 1.20V
LLC : Turbo
Uncore Ratio: 40
DRAM Voltage 1.35V(Set by XMP)
C-States/EIST: Off
Intel Speed Shift: Off
Power Limit(Short & Long Duration): Max(4095W)
Internal Graphics/VT-d: Off
Windows Power Plan: High Performance
Turbo: Off

How I OC'ed:-

My goal was to reach my max clock speed for 1.2Vcore and keep it there because my cooler is not so great, in fact, it's awful and I will be upgrading soon. First, I did Memtest86 loop of all tests on stock(2133) and XMP(3200) for 12 hours each establishing memory stability. I disabled Integrated Graphics, C-States, Intel Speed Shift and EIST. Maxed out power limits. According to the guides, I went up in increments. I set a manual Vcore of 1.2V which I believe is the stock voltage showing on the right side of my Gigabyte BIOS and started from my max Turbo Clock at 4.3Ghz and went up in 100Mhz increments. At each increment I ran 30 minutes of Prime95 26.6(Non-AVX) Small FFT with temperatures staying around 75 degrees Celcius on individual cores at 1.2Vcore/81 degrees on AVX load though. I then ran an hour of Realbench Stress Test with half my RAM(8gb) at each increment after the Prime95 run. At 5Ghz, one of the workers on Prime95 failed so went back to 4.9Ghz and started my long-term stability tests for a 24.7 rig. Vcore was dropping far too much during loads so I set LLC directly to Turbo as Gigabyte's own Z390 OC guide suggested and the voltages stayed around what I set. XMP was on during the following stability tests.

Stability Tests:-

1. Prime95 26.6 - 2 hours - Small FFT.
2. RealBench - 8 hours - Stress Test(Full 16GB RAM selected)
3. Cinebench R15 - Many many runs, CPU Score around 1190
4. Prime95 29.4 Build 8 - 2 hours - AVX Small FFT
5. Prime95 26.6 - 24 hours - Blend Test
6. IntelBurnTest - 10 Passes(Ran Twice) - Maximum Stress - Temps max 82 Degrees
7. Gaming - Hours of Assassin's Creed Odyssey and Shadow of Tomb Raider on current settings.
8. 3D Mark - Multiple Firestrike Runs(Highest Score: 22684)/Multiple Time Spy Runs(Highest Score: 9116)

All these tests passed quite nicely with no ridiculous Vcore spikes or unacceptable temperatures on any cores. Vcore almost always stays around 1.2V which I had set because of the Turbo LLC I'm guessing. I tried to lower LLC but even one step lower LLC either makes the voltages drop way too low or one of the cores just fail. AVX instruction loads were failing stress tests above 4.7Ghz so I set an AVX Offset to -3(4.6Ghz) just to be safe. During load my VID fluctuates around 1.150v to 1.220v. On Idle it's always close 1.2V.

Questions:-

1. Is 1.2Vcore the stock voltage for the 8600k like my BIOS shows? Is it possible to go lower than 1.2V for lower clock speeds if I pass stability tests?
2. Is AVX offset safe? (Clock speeds sometimes go down to 4.6Ghz for like half a second and back to 4.9Ghz on idle)
3. Is it necessary to OC Uncore as well? If so, are there specific stability tests I can run for Uncore OC? If not specific, what tests should I run after OC'ing Uncore?
4. Any other tests I should run to deem my system stable?
5. Since I'm on the verge of stability at 1.2Vcore on 4.9Ghz, should I clock down to 4.8Ghz to be absolutely sure the system is stable?
6. Finally, Have I won the silicon lottery by any chance? 4.9Ghz at 1.2V seems too good to be true

Conclusion:-

Unfortunately, I did not take screenshots while running all the tests. The one screen I have is of the 24h Blend test taken at the 24 hour mark(on load) which I have added below. Any suggestions to improve my stability will be greatly appreciated since I just got in the game of OC. I wish to keep this OC for good, if possible.

Thank you.

Screen:-



The Vcore showing on HWinfo of 1.176V is a very brief drop from 1.2V, probably for just a second as you can see from the graph. But what is that Vcore overshoot of 1.284V showing on Maximum column? Is that bad? Probably also happened for a second.

 

[trents]

Hi,

It has only been a month since I got this build and after reading and viewing many guides I got myself to overclocking. This is also my first post and if I'm posting in the wrong section, please forgive me and move my post to an appropriate section. I'm basically making this long post to get some experienced opinions on my overclock and if I should change something else to fine tune. Apologies in advance, for the long post.

Specs:-

CPU: i5-8600k
Motherboard: Gigabyte Z390 Aorus Elite(rev-1.0)
PSU: Thermaltake Toughpower 750W 80+ Gold
RAM: G-Skill Trident Z 16GB(8x2) @3200Mhz/Timings: 16 18 18 38(XMP Profile)
GPU: EVGA 1080ti FTW3(No OC)
Casing: Cooler Master MB 500
Air Cooler: Thermalright True Spirit 120M(Will be upgrading soon)
Windows 10 64-bit

BIOS:-

BCLK:100Mhz
Core Clock Ratio: 49
AVX Offset: -3
Vcore: 1.20V
LLC : Turbo
Uncore Ratio: 40
DRAM Voltage 1.35V(Set by XMP)
C-States/EIST: Off
Intel Speed Shift: Off
Power Limit(Short & Long Duration): Max(4095W)
Internal Graphics/VT-d: Off
Windows Power Plan: High Performance
Turbo: Off

How I OC'ed:-

My goal was to reach my max clock speed for 1.2Vcore and keep it there because my cooler is not so great, in fact, it's awful and I will be upgrading soon. First, I did Memtest86 loop of all tests on stock(2133) and XMP(3200) for 12 hours each establishing memory stability. I disabled Integrated Graphics, C-States, Intel Speed Shift and EIST. Maxed out power limits. According to the guides, I went up in increments. I set a manual Vcore of 1.2V which I believe is the stock voltage showing on the right side of my Gigabyte BIOS and started from my max Turbo Clock at 4.3Ghz and went up in 100Mhz increments. At each increment I ran 30 minutes of Prime95 26.6(Non-AVX) Small FFT with temperatures staying around 75 degrees Celcius on individual cores at 1.2Vcore/81 degrees on AVX load though. I then ran an hour of Realbench Stress Test with half my RAM(8gb) at each increment after the Prime95 run. At 5Ghz, one of the workers on Prime95 failed so went back to 4.9Ghz and started my long-term stability tests for a 24.7 rig. Vcore was dropping far too much during loads so I set LLC directly to Turbo as Gigabyte's own Z390 OC guide suggested and the voltages stayed around what I set. XMP was on during the following stability tests.

Stability Tests:-

1. Prime95 26.6 - 2 hours - Small FFT.
2. RealBench - 8 hours - Stress Test(Full 16GB RAM selected)
3. Cinebench R15 - Many many runs, CPU Score around 1190
4. Prime95 29.4 Build 8 - 2 hours - AVX Small FFT
5. Prime95 26.6 - 24 hours - Blend Test
6. IntelBurnTest - 10 Passes(Ran Twice) - Maximum Stress - Temps max 82 Degrees
7. Gaming - Hours of Assassin's Creed Odyssey and Shadow of Tomb Raider on current settings.
8. 3D Mark - Multiple Firestrike Runs(Highest Score: 22684)/Multiple Time Spy Runs(Highest Score: 9116)

All these tests passed quite nicely with no ridiculous Vcore spikes or unacceptable temperatures on any cores. Vcore almost always stays around 1.2V which I had set because of the Turbo LLC I'm guessing. I tried to lower LLC but even one step lower LLC either makes the voltages drop way too low or one of the cores just fail. AVX instruction loads were failing stress tests above 4.7Ghz so I set an AVX Offset to -3(4.6Ghz) just to be safe. During load my VID fluctuates around 1.150v to 1.220v. On Idle it's always close 1.2V.

Questions:-

Yes, you posted in the correct section. By the way, welcome to OCF!

1. Is 1.2Vcore the stock voltage for the 8600k like my BIOS shows? Is it possible to go lower than 1.2V for lower clock speeds if I pass stability tests?
Your true core voltage ("vcore") under load is actually 1.284. You have it circled in your picture. Different power management technologies are supplementing your base vcore. Did you use a "fixed" value or an "offset"? Normally, we set the LLC (Load Line Calibration) to a level that counters "vdroop" so as to keep the vcore value under stress to the approximate base value you enter in bios. With most motherboards, the next to highest LLC setting seems to be about right.

It might be possible for you to be stable at a lower vcore. You can only know by experimentation.

2. Is AVX offset safe? (Clock speeds sometimes go down to 4.6Ghz for like half a second and back to 4.9Ghz on idle)
Certainly it is safe. An AVX offset of -3 means the CPU will run at .3 ghz slower when it encounters AVX instruction sets in software. Many or most stress testing software these days utilize AVX. Earlier versions of Prime95 (like 26.6) do not. A lot of real world applications do not use AVX but it is becoming more common.
3. Is it necessary to OC Uncore as well? No, not necessary but you can overclock it if you wish. Not much to be gained by doing so and it makes it harder to get the overclock stable. Generally, it will not overclock as high as the cores. If so, are there specific stability tests I can run for Uncore OC? If not specific, what tests should I run after OC'ing Uncore? No separate special tests. Same as you use for the cores. When you stress the CPU you stress both the cores and the uncore at the same time. They work as a unit together.
4. Any other tests I should run to deem my system stable?
No. You have run more than enough.
5. Since I'm on the verge of stability at 1.2Vcore on 4.9Ghz, should I clock down to 4.8Ghz to be absolutely sure the system is stable?
You have thoroughly demonstrated stability already at 4.9 with a 3x AVX offset with both dedicated stress testing apps and real world apps. Don't get OCD about this.
6. Finally, Have I won the silicon lottery by any chance? 4.9Ghz at 1.2V seems too good to be true
Not really and for two reasons. First, you are not really running the chip at 1.2 vcore. I explained that already. Second, you are actually running at 4.6, not 4.9, when apps that use AVX are run. Having said that, you might be able to push it to 5.0 stable in non AVX applications with more voltage. I think a 1.345 vcore is considered safe if temps permit. You seem to have at least an average chip.

Conclusion:-

Unfortunately, I did not take screenshots while running all the tests. The one screen I have is of the 24h Blend test taken at the 24 hour mark(on load) which I have added below. Any suggestions to improve my stability will be greatly appreciated since I just got in the game of OC. I wish to keep this OC for good, if possible.

Thank you.

Screen:-

View attachment 203109

The Vcore showing on HWinfo of 1.176V is a very brief drop from 1.2V, probably for just a second as you can see from the graph. But what is that Vcore overshoot of 1.284V showing on Maximum column? Is that bad? Probably also happened for a second.

If I were you I would probably look at increasing your base vcore and reducing your LLC to bring your declared vcore in bios and your max vcore under stress testing more in line with each other. I would also probably look an setting the AVX offset to 2X instead of 3X. 3X is a lot and gives the false impression that you are stable at a higher frequency than you actually are. I would see if you can get the system stable at 5.0 ghz with a 1x or 2x AVX offset.

I think you are overdoing the stress testing. The goal is to have the system stable in real world use. Stress testing is just a tool to help get there quickly. Passing four hours of Realbench and 6 hours of AIDA64 Extreme should be sufficient testing. Many have abandoned Prime95 because it drives temps up unrealistically on the newer generations of Intels.

 

[alpha9101]

Thank you for the reply trents, much appreciated. This is exactly why I was seeking experienced opinions.

Yes, my Vcore is set to a "fixed(Manual)" value of 1.2V not offset. I'm not sure how to work the offset voltage on Gigabyte's BIOS tbh. To use offset for Gigabyte you have to set Vcore to Normal, which unlocks DVID setting(offset). I don't understand how much Vcore is set when I use a +0.00 offset. How do I find this out? Is the Vcore set according to the VID?

LLC(Load-Line Calibration) is on Turbo which is next-next to the highest meaning there are two levels above my current Turbo setting. (Next setting is Extreme, and Ultra Extreme after that) on the Gigabyte BIOS. Are you suggesting I go forward setting LLC to Extreme? Extreme and above settings are said to be increasing Vcore during load most of the time I heard, and Gigabyte's own OC guide suggests to put it on Turbo. (On Extreme LLC, with 1.2V set on BIOS, Vcore goes upto 1.224-1.248v on load. UltraExtreme LLC, even higher on load)

Clarify this for me please, what do you mean by 1.284v being the "true" voltage during load? Does this mean I should be instead setting my Vcore around 1.284v and try lowering LLC? I have tried lower than Turbo setting and Vcore never stays around what I set in BIOS. Turbo keeps it exactly on what was set and fluctuates around it occasionally, and Extreme increases Vcore a bit, during load and idle.

About AVX, I have passed stability tests such as RealBench and IBT on -2 AVX offset as well. I just put a -3 thinking it's safer.

One last noob question about VID. From what I understand, VID is what the voltage a CPU requests to run stable right? So for example, if I set 1.2V but VID shows around 1.150V, does this mean I can set Vcore to 1.150 or something like that?

Here's the thing though, since my cooler is not that great(Crossing 80 degrees on individual cores on AVX loads at 1.2Vcore), I would really like to stay on 1.2V or as close as possible even if it means I have to reduce clock speeds to 4.7Ghz for example, where I would not have to set an AVX offset either. At least until I get a better cooler, maybe a 240mm AIO or a premium Air cooler I'd prefer not to increase Vcore at all. I hope it's okay to ask all these questions. I'm sure the info I need is out there and I have checked many forums and guides, but at this point I feel like I need some direct assistance.

 

[caddi daddi]

I would reduce the llc, it looks like your vcore is overshooting.

 

[alpha9101]

Any lower LLC, and the Vcore goes down to 1.150v(minimum 1.140v) during load. On idle it's 1.2v.

This screen is on High LLC, one step lower than Turbo.

 

[caddi daddi]

so, is it stable at 1.152 vcore?

 

[alpha9101]

I'm not sure about that, have not run tests for long at lower LLC. I used to think any Vcore under 1.2v is bad as I thought it was stock voltage and you can't go lower than that but I was wrong. I'll test it out with RealBench see if it's stable. What if it turns out to be stable at 1.152Vcore on load? could I lower my Vcore on Bios to around 1.150v? Maybe 1.160?

 

[caddi daddi]

that would be correct.
remember to run a benchmark, sometimes we run a touch unstable and a stress test will not show it but a benchmark will by giving a reduced score.

 

[The Coolest]

Any lower LLC, and the Vcore goes down to 1.150v(minimum 1.140v) during load. On idle it's 1.2v.

This screen is on High LLC, one step lower than Turbo.

Core Temp displays the VID, not the vcore. These are two different values.
From the average column in your first screenshot, it seems that 1.2v is the average, and the 1.284v in the max column is indeed a product of spikes.
I don't think that's something I would be worried about.

 

[alpha9101]

Yes, I understand this. I always follow Vcore on HWinfo, CoreTemp is only for keeping an eye on temps. If the short spike of 1.284v is nothing to be worried about, could I keep this setup for 24/7 usage? I have tested for days on several programs, games and benchmarks, no issues so far. This feels alright lol. I will ofcourse, keep experimenting with lower LLC as trents had suggested to try to eliminate these overshoots. Lets see.

 

[trents]

To explain further, VID is the factory voltage setting for the cores. And I think it not smart to ignore the spikes altogether. Depending on how long they last it could cause high temps and throttling. To get an idea on the duration of the spikes one could keep an eye on the far left column of HWINF64 Info for what the current voltage is and how fast it changes.

 

[alpha9101]

Understood. I'll go AVX Offset -2 like you suggested at 4.9G. Will see if these overshoots can be dealt with by lowering LLC and increasing Vcore if necessary. As far as I've seen, I'm not the only one having Vcore overshoot issues with Gigabyte's Turbo LLC.

In BIOS, where the LLC settings are, there is a thing called the Internal AC/DC loadline. I saw this guide where it was suggested that if the voltages drop a bit more than you'd prefer with LLC around High or Medium you can set AC/DC loadline to Turbo and it would help with that. Should I try it?

 

[trents]

Yes, try it. Overclocking is all about experimenting to find what works and is safe and sustainable for your particular combination of components.

 

[alpha9101]

Aida64 is amazing, temps on all cores are so much more reasonable. Never passed 70C on any core during the tests even on AVX instruction loads. Even at 1.3Vcore I'm under 80C on all cores. Whereas, Prime95 would drive temps close to 85C on AVX loads even on 1.2V. Thanks trents. I used to be afraid to increase Vcore above 1.2V cuz of temperatures and my crap cooler on Prime95 AVX loads.
You guys think I could deem my system stable if I pass RealBench and Aida64 with these nice temperatures? If so, I'll go for higher Clock Ratios and Vcore and test.

 

[trents]

I think with those two stress testers, if you pass them when run for several hours, you can deem the system stable. If in real life applications you ever find some random instability just give your vcore a small bump (.005-.01) and it should take care of it.

 

[wingman99]

Iv been running my i5 8600k at 5.0GHz LLC AUTO, DVID +0.080 = 1.260v to 1.332v for a year. Don't worry about the voltage these processors when the heat is in check will last a useful life time at 1.4v.

 

[alpha9101]

Thanks guys. This has been very helpful.

 

[Rainwater]

The CPU-Z reading from my understanding has always been the true/correct vcore reading. I have always used the cpu-z reading for vcore. I use hw monitor pro for temps and cpu-z for vcore. HW Info imo is crap and I don't use it ever. HW monitor and cpu-z are the only two I use as they are much more accurate. If he set his vcore manually in bios at 1.2 and cpu-z reads it at 1.2 then his chip is running at 1.2v. Which is good to know because I just started overclocking my 8600k and haven't oc'd a chip since my 6700k. I had it stable at 4.5 Ghz with 1.295v. So I started with my 8600k at 4.8 with 1.295 then 4.9 with 1.295v. LLC set to turbo on my gaming 7 and cpu-z reads it at 1.296 and it'll drop to 1.284 under load. Never lower than 1.284 never higher than 1.296. My temps are rather high tho, so I'm checking threads looking for stable settings and it's goo to see someone stable with 1.2v.

 

[EarthDog]

CPUz can be wrong... I've seen it do that plenty of times. It is always best to check the BIOS, CPUz, Hwmonits, etc and see what is what. Ideally, the board has voltage read points and you can get it true.