Alan 2020 年 7 月 27 日 上午 7:46
My core frequencies are fluctuating on load from an overclock
My Specs -
CPU - I7-9700k stock
MOBO - Z390 A pro
RAM - 32 vengeance clocked to 3000
GPU MSI - 2070S
PSU - 600w Evga White plus
AOI - Arctic Freezer 2 240

In short - My core frequencies are fluctuating on load from an OC. I do not know why.

Longer version - I have a i7 9700k which I have Overlocked to 4.9 over its stock of 4.6 many times.

I had to remove the OC as my case and cooler could not deal with the CPU heat. I just got a new case and got a AOI so I can OC.

I OC the CPU to 4.9 using the default option to do that on the MOBO - I'm very used to doing it.

I OC to 4.9 and then once I was testing while gaming, using MSI afterburner, I can see my core frequency fluctuating on load, dropping back down to 4.6 then it would go back up to 4.9 and it just keeps on doing until its idle then its back up to 4.9.

I used Hwinfo64 and CPU z confirming the drops.

My Temps are stable and under 80c at all times - Mostly 60s to 70s tops.

I have tried manually OC to 4.9 and putting it on fixed and set my own votage to 1.25.

I can't work out whats going on, I don't see anything about overclocked cores dropping down to stock on load only.
< >
目前顯示第 76-89 則留言,共 89
crunchyfrog 2020 年 7 月 29 日 上午 8:17 
引用自 Englishmen
引用自 Monk
Again crunchy frog, you don't seem to follow that it cannot be an airflow issue, his individual core temps are all in check, this isn't a motherboard se dor showing the heat around an area, it's the chip itself giving very accurate constantly updating results, which even record a max and minimum temperature, so even if it spikes for a fraction of a second, it's recorded.

Beyond that he is using an aio, so airflow isn't really as important as long as its pulling air in through it directly or its being used as an exhaust (my personal preference), if the airflow was bad at any point, I can assure you, his temps would be an obvious problem.

As an idiot who spent £1400 watercooling his system and has another £700 of parts I never installed into another (going on with the 3090 most likely as the blocks for my 2080ti took too long to get made), I can assure you, I have spent alot of time looking at and testing cooling and airflow and unless you have a crazy restrictive case, it's not going to be an issue and 80c under load tells me, his case isn't super restrictive, at best, dropping a few degrees could be possible, but his issue does not seem to be temperature related 8n the slightest.

I get what you are sayi g that it doesn't hurt to test, but, when it has no potential link to the problem, it's almost as daft as saying the colour leds could be causing it.

Part of the problem here is we have people coming on that do not build High spec gaming rigs and do not know the subject all that well or the hardware being used..

Nope, as I made CLEAR this is merely a side step as I'm well versed in troubleshooting in electronics of ALL household kind.

All I offered was a bit of lateral thinking which is EASILY carried out, is useful for the future, and can easily be discounted by process of elimination.

So nice strawman. That's all it was.
Alan 2020 年 7 月 29 日 上午 10:08 
Voltage dropping under load is normal. This is VDROP/VDROOP. It occurs because there is often a spike in voltage when load raises, and rather than spiking up to really unsafe levels, it drops so the spike isn't much higher than idle. Also, and may people don't know this, but voltage is not at all equal; a given voltage is more dangerous at higher temperatures (which is often the case at load). There's things like "LLC" which is loadline calibration (or whatever else modern stuff may term it) to counteract this, but I'd really recommend to either leave it be unless you know what you are doing, or if you do use it, use on of the lower/milder settings. Otherwise, you may get idle/load voltages that are near parity but you'll have spiking ABOVE the voltage point you set and otherwise have consistently.

I remember during the Core 2/early Sandy Bridge days, many were using aggressive LLC settings under the preference for "voltage I set in BIOS is consistent or load voltage" and it often actually had them stable under load, but BSODing under light load/idle, and the solution was to give it more voltage.

Just learn to set your voltage target for idle with this (dropping behavior) in mind. You may simply be reaching your temperature limit which is going to limit you, but glad to hear you made progress.

I see - I hear you, i tend to leave the LLC alone as I don't know what I'm doing. I have used it while using MSI guilds. But nearly all the MSI guilds are very simple and for good reason.

With this MOBO and Chip, all you need to do, is turn the turbo mode on and lower the voltage and your done and it works or has done. Its just not working anymore which is the problem, i'm getting alot of advise on how to overclock or people telling me not to.

But none of that matters, the issue is, the MOBO or whatever it is isn't doing what it was doing beforehand and what it was designed to do.

I just cant work out why its not working now when i have better temps than i have ever done.. And why is it working on Auto only when it used to work before on a set voltage. I JUST bought the Prime PRIME Z390-A which is top end and has AI turning which I hope tell me whats going on.
Illusion of Progress 2020 年 7 月 29 日 下午 1:21 
If you were using some sort of "auto" setting for overclocking, those usually don't get you very far. It doesn't mean there's anything wrong with it if it's not allowing you an even higher overclock. Again, you might just be near your limits (or have to manually do it to get further).

If you're saying it was stable at a setting before and isn't now, you may have been right on the very edge and it degraded a bit since (though overclocking is usually safe if not done blindly, this IS still a thing).

Add to all this that modern CPUs are closer to their ceilings than before. This is more true with AMD, but it is the case with Intel as well. In the Core 2 to Sandy Bridge days, you could go from ~3 GHz (plus or minus) to nearly 4 GHz (or more), but CPUs just come clocked higher by default and boost on top of that (with Intel, some go near or a bit above 5 GHz now), so the headroom is already gone. This is not necessarily a bad thing; it means you don't have to put in the work unless you want to fine tune it even more.
Magma Dragoon 2020 年 7 月 29 日 下午 1:28 
Reset BIOS
Leave xmp off for now to rule out any RAM instability issues
Disable turbo boost
Disable multicore enhancement of it exists
Make sure BCLK is exactly 100.0
Try the vcore, max multiplier and AVX offset from silicon lottery. 1.337v, 49, 2 is achievable by 100% of units they tested, unless the auto OC you've been using has degraded your chip it should work.
https://siliconlottery.com/pages/statistics
For LLC, leave it at low or medium.
Enable xmp once you know things are stable, if it becomes unstable at this point you'll know it is the memory.

Asus Prime boards are not top end they are goodish, similar to the MSI A Pro you already have. AI Suite is an auto OC feature, quite conservative with multiplier but aggressive with BCLK and voltage when I used it on Z170. Really you're better off plugging in the numbers from SL and tweaking from there.

I'm not an expert overclocker but this has worked for me.
Alan 2020 年 7 月 29 日 下午 2:24 
So you guys think the chip could have degraded, thats sad news.. Never done that much with it really. Its odd as I can lower the Vcore to 1.28 and be stable and run stress tests with no issues.

I just tested CPU z - Stressed the CPU and the clock cores stayed at 4.9. But as soon as i put more load on, the cores go back down to 4.6.. I really hope its no downgraded over a years use.

I'll try some of the advise given here.. I have just seen the Mobo i got, its med ranged. The VRMs are not the best on it but yes.. Just wanted a new MOBO to make sure its not the Mobo doing as I cant afford a new 9th gen chip.
最後修改者:Alan; 2020 年 7 月 29 日 下午 2:48
crunchyfrog 2020 年 7 月 30 日 上午 9:01 
引用自 Englishmen
So you guys think the chip could have degraded, thats sad news.. Never done that much with it really. Its odd as I can lower the Vcore to 1.28 and be stable and run stress tests with no issues.

I just tested CPU z - Stressed the CPU and the clock cores stayed at 4.9. But as soon as i put more load on, the cores go back down to 4.6.. I really hope its no downgraded over a years use.

I'll try some of the advise given here.. I have just seen the Mobo i got, its med ranged. The VRMs are not the best on it but yes.. Just wanted a new MOBO to make sure its not the Mobo doing as I cant afford a new 9th gen chip.

The cardinal rule you need to remember for ANY semicondcutors - chips, transistors, anything, is that the more you push them power-wise the quicker they will degrade.

The thing is, most people tend to think that's linear, you know kind of, "if I give it 10% more power, I'd expect something like 10% less on the lifespan".

Sadly, it does not work like that, It's probabilty.

So, what you actually get is a MORE increased chance of failure with such an increase, and it ain't linear. You could have had trouble free use for 10 years without overclocking, yet it'd fail within a year with just a 10% increase. You'll never know of course, as it's almost impossible to test.

So yeah, it's entirely likely just a small increase has caused the chance of failure up. Them's the risks.

Monk 2020 年 7 月 30 日 上午 9:21 
It's very unlikely, and any ammount of power you can put through it unless you disable safety features, won't harm it, hell, you could set it to 5v, boot it and nothing will happen, it simply won't boot up.

So I am pretty confident it hasn't degraded from that little use.
Illusion of Progress 2020 年 7 月 30 日 上午 10:31 
引用自 Englishmen
I just tested CPU z - Stressed the CPU and the clock cores stayed at 4.9. But as soon as i put more load on, the cores go back down to 4.6.. I really hope its no downgraded over a years use.
I'm not 100% familiar with the newer-ish Intel CPUs that have turbo boost, but is that not how they operate? I know that's how AMD Ryzen does it; the frequency will be higher on low core loads, but as soon as it's using many or all cores, it drops more. You have to manually set an all core frequency to keep it consistent, but then you lose out on the slight single/low core boosting beyond that.
Monk 2020 年 7 月 30 日 上午 11:37 
That's exactly how it works, 4.9 is a 1 or 2 core boost speed only
Alan 2020 年 7 月 30 日 下午 4:04 
引用自 Englishmen
I just tested CPU z - Stressed the CPU and the clock cores stayed at 4.9. But as soon as i put more load on, the cores go back down to 4.6.. I really hope its no downgraded over a years use.
I'm not 100% familiar with the newer-ish Intel CPUs that have turbo boost, but is that not how they operate? I know that's how AMD Ryzen does it; the frequency will be higher on low core loads, but as soon as it's using many or all cores, it drops more. You have to manually set an all core frequency to keep it consistent, but then you lose out on the slight single/low core boosting beyond that.

The chip does boost to 4.9 but only 1 or 2 cores and you wont ever see it. My vcore is always showing 4.9 on stock. And I have tried doing an OC at 4.9 setting all cores, and doing an OC for 5ghz. I found out that unless I turn unto vocre on i cant OC where it does not underclock.

Did some benchmarks and Im getting results still so i dont thinks its degraded. New Mobo coming.

Any OC of any kind just does not hold guys, at all.
Illusion of Progress 2020 年 7 月 31 日 上午 9:16 
引用自 Englishmen
The chip does boost to 4.9 but only 1 or 2 cores and you wont ever see it. My vcore is always showing 4.9 on stock. And I have tried doing an OC at 4.9 setting all cores, and doing an OC for 5ghz. I found out that unless I turn unto vocre on i cant OC where it does not underclock.
With AMD CPUs, it takes the voltage and temperature into consideration when doing it's boosting thing at all core loads (nVidia GPUs also do something similar, and AMD GPUs might as well I would imagine). Basically, CPUs and GPUs are "smart" in a way today. If Intel CPUs works similarly, then the drops in your case are happening because it can't reliably boost high unless it has more voltage.

So, you say it's dropping down to 4.6 GHz under higher loads, and that the 4.9 GHz boost for lesser core loads is almost never seen, meaning your workflow pushes you above that. At that point, if you want improvement, you're better off setting an all core overclock to above 4.6 GHz (if possible and within the limits of voltage and temperature for your PC), but keep in mind you will lose some performance IF the load is ever lowly threaded UNLESS your all core overclock is 4.9 GHz or above. If you can't, you'll have to decide if the trade-off is worth having a better all core boost.
最後修改者:Illusion of Progress; 2020 年 7 月 31 日 上午 9:19
Alan 2020 年 7 月 31 日 下午 4:47 
引用自 Englishmen
The chip does boost to 4.9 but only 1 or 2 cores and you wont ever see it. My vcore is always showing 4.9 on stock. And I have tried doing an OC at 4.9 setting all cores, and doing an OC for 5ghz. I found out that unless I turn unto vocre on i cant OC where it does not underclock.
With AMD CPUs, it takes the voltage and temperature into consideration when doing it's boosting thing at all core loads (nVidia GPUs also do something similar, and AMD GPUs might as well I would imagine). Basically, CPUs and GPUs are "smart" in a way today. If Intel CPUs works similarly, then the drops in your case are happening because it can't reliably boost high unless it has more voltage.

So, you say it's dropping down to 4.6 GHz under higher loads, and that the 4.9 GHz boost for lesser core loads is almost never seen, meaning your workflow pushes you above that. At that point, if you want improvement, you're better off setting an all core overclock to above 4.6 GHz (if possible and within the limits of voltage and temperature for your PC), but keep in mind you will lose some performance IF the load is ever lowly threaded UNLESS your all core overclock is 4.9 GHz or above. If you can't, you'll have to decide if the trade-off is worth having a better all core boost.

Yes - 4.6 on a I7 9700k though - I don't think there isn't anything better got gaming even now. I just could not work out why it was not doing something it had done beforehand. So I will just stick with stock.

I bought a new MOBO and now its holding, I can even run 50 ghz.. I could prob go higher, maybe 5.2 but i will just go back to 4.6. I did use a mini guild for the OC i have tested on the new Mobo but no matter what OC i used on the last Mobo it would not hold and it did for a year so... Must have been the Mobo mate?!
最後修改者:Alan; 2020 年 7 月 31 日 下午 4:48
Illusion of Progress 2020 年 7 月 31 日 下午 6:49 
I have no idea, but if the new motherboard is working better, then it's possible. I'm confused as to why you'd get a new motherboard though if you're not even going to push it beyond what the old one was already capable of. Why not just stay with the old one then?
Alan 2020 年 8 月 1 日 上午 8:37 
I have no idea, but if the new motherboard is working better, then it's possible. I'm confused as to why you'd get a new motherboard though if you're not even going to push it beyond what the old one was already capable of. Why not just stay with the old one then?

Because I bought an unlocked chip and will Overlock later on as time goes on to get more from the chip. For the future, I like to OC every now and then.
最後修改者:Alan; 2020 年 8 月 1 日 上午 8:37
< >
目前顯示第 76-89 則留言,共 89
每頁顯示: 1530 50

張貼日期: 2020 年 7 月 27 日 上午 7:46
回覆: 89