What could be limiting my 12700K?

[jamieboo]

Hello folks

A tiny but irritating problem here.
I have a 12700K on an Asus Z690 Tuf Gaming Plus Wifi D4 board running Windows 10.
Everything is at stock apart from XMP1.

The thing is... my 12700K no longer has a couple of cores that boost to 5Ghz.
I thought it was because of a bios update but I rolled back and still I can't hit 5Ghz.
I recall shortly after I built it I noticed that it would occasionally hit that speed (in R23 single core for example), but now it never does. It hits a hard limit of 4901.2Mhz.

I've asked on a few other forums and folk just generally say 'Why don't you just overclock it if you want 5Ghz', but it's not that I necessarily want 5Ghz- and I know the difference of just 100Mhz is not huge - I just want it to perform to the advertised levels.
I know I could just overclock, but it irritates me because 5Ghz is an advertised stock boost speed for this CPU. I don't want to overclock to achieve a speed that it should be able to hit out of the box. Especially as it used to!

What could be limiting my CPU speed?

If anyone could help me I'd hugely appreciate it!

Thanks folks

 

[Nebulous]

5Ghz is "advertised", not a guarantee that it will. It will probably hit the 5ghz under heavy stress/workload, but not for sustained periods unless you overclock it to stay that way.

Is your rig unstable?

 

[jamieboo]

5Ghz is "advertised", not a guarantee that it will. It will probably hit the 5ghz under heavy stress/workload, but not for sustained periods unless you overclock it to stay that way.

Is your rig unstable?

Thanks for replying Nebulous!
Of course, you're absolutely right. But my CPU used to hit that speed. Yes, just a couple of cores, and yes, just for a brief time - but it would hit 5Ghz. Now it doesn't.
Any idea what could be limiting it?

Thanks

 

[Nebulous]

There's thousands of variables as to why it doesn't, but I wouldn't worry about it. If it's hitting 4.98Ghz, then technically it's 5Ghz. So long as your rig is stable, then enjoy it.

I remember I was overclocking memory and hit the fastest score ever. I tried to replicate everything to get that score again, but it never happend. Got close tho, but that was it.

Too many variables involved to give you a definitive answer

 

[jamieboo]

Thanks Nebulous!

You're right... but it's like a little splinter in my head!

I'm sure there are a million variables but, as someone who I imagine knows their stuff, what would you say are the most likely ones, given the circumstances described?

Basically I'm not quite ready to give up on it yet, I just need some good leads!

Thanks

 

[EarthDog]

If Intel says it, they'll hit it with 2-cores, it will.


That said, run Hwinfo for a while and see if you're just missing it? How are you checking to see? But DL and run Hwinfo64 and leave it running.. see if it hits 5 GHz. Rune Cine23 with it up and see if it lists throttling reasons...

Make sure Windows power mode is set to performance (shouldn't be it, but maybe it's 'stuck', lol)

 

[jamieboo]

If Intel says it, they'll hit it with 2-cores, it will.


That said, run Hwinfo for a while and see if you're just missing it? How are you checking to see? But DL and run Hwinfo64 and leave it running.. see if it hits 5 GHz. Rune Cine23 with it up and see if it lists throttling reasons...

Make sure Windows power mode is set to performance (shouldn't be it, but maybe it's 'stuck', lol)

Thanks EarthDog
Yes, I've been running HWiNFO.
It was HWiNFO that showed me occasional 5Ghz on a couple of cores in the past, and now I keep it running and it reports that now none of my P-Cores ever go above 4901.2Mhz.

Windows power is set to Balanced. But this was the setting I had before - when 5Ghz was achieved.
Anyway, I already tried switching to High Performance and it made no difference to my CPU speed.

 

[EarthDog]

So, it didn't come up with any throttling issue? Temps still good with one/two cores? Leave it up in the background for like... a day... see if it pops 5 GHz.

 

[jamieboo]

Yeah, my temps are good.

HWiNFO does mention some Performance Limit Reasons:

IA: Max Turbo Limit
RING: Max VR Voltage, ICCmax, PL4

I've asked a lot about these on other forums.
Apparently the IA: Max Turbo Limit seems to be pretty omnipresent and does really indicate anything useful.
The RING limit seems to be something no one can really figure out.

Interestingly Intel XTU does not flag any throttling issues.
But ThrottleStop has a constant EDP OTHER in the RING column.

I've asked elsewhere about all these supposed limitations but it's never lead to a solution for my issue.

I've left HWiNFO running for several hours under different loads, but never a full day. Maybe I should try that.
But again I keep coming back to the fact that it used to visibly hit 5Ghz - and reasonably often too. I wouldn't need to run a monitor for a full day just to catch a glimpse of it.

 

[EarthDog]

But again I keep coming back to the fact that it used to visibly hit 5Ghz - and reasonably often too. I wouldn't need to run a monitor for a full day just to catch a glimpse of it.

Yep.. perplexing...

What else have you asked and received an answer about? At least we can throw poo at new spots on the wall, lol.

 

[Nebulous]

Ok, so temps are good, and it boosts as it's supposed to. Might be something in the background sucking juice? I remember Joe asking about Windows power mode and switching that didn't do a lick of anything.

Maybe start from scratch? Shut down, power drain the psu by removing plug from wall, then press and hold your pc power switch. Clear Cmos via jumper or by removing battery from board. Wait like 2 minutes, and then place the battery back, plug psu back into wall and fire it up.

You'll have to set your settings back, but this will clear things up. Might get your 5Ghz shot nice and clear.

 

[jamieboo]

Glorious metaphor!

Yes, well, I guess all the usual things.

I've asked about bios upgrades and stuff, but that became a slightly redundant line of enquiry after I rolled back to my previous bios and the issue remained.
I've asked lots about those limitation/throttling flags, but nothing conclusive has resulted.
I've asked about things that can limit boost speed and looked for solutions other than simply throwing more volts at it.
I've wondered if something was limiting my cache speed / uncore as it seemed locked at 36. I know that can have a limiting effect. but since then I have seen mine go up to 46.
I spoke to unclewebb re ThrottleStop and he suggested there was probably nothing wrong with my cpu but rather there might indeed be an issue with my cache speed.
Much of my time has been trying to make folk understand that I don't necessarily want to overclock. At least not at this stage. I just want to understand what has changed to prevent my CPU hitting a boost speed that it formerly achieved.
I've asked about my results when I run Latencymon - they go into the red within 5 minutes suggesting DPC latency issues. A problem as I built this machine for audio production. Usually a driver issue is the culprit (soundcard, USB, Wifi etc). But surely a driver like that would have nothing to do with my CPU boost speed.

I do recall I updated the chipset a few weeks ago. Could that affect CPU speed?

Thanks


And thanks too Nebulous.
But does it boost as it's supposed to? I thought that was the problem. A couple of cores should boost to 5Ghz like they used to, but they don't.
Hmm, you could be right about having to do the whole clear CMOS thing. But I'd like to explore every avenue before then.

 

[EarthDog]

Yeah, a simple CMOS reset would be worth a try... but you upgraded and rolled back which is a fresh start. Still worth a try.

Chipset drivers shouldn't effect it, no...

What is your computer running? Maybe show task manager sorted by CPU use at idle.. see if anything is taking up cores that may prevent a 2-core boost? (I'm really reaching here looking for clear space on the wall, lol!!).

What about a fresh install of Windows? Do you have a spare drive? If so, unplug everything in there currently and put only the new drive in (unplug power to the rest) install winders there Hwinfo and see what happens. If it works, you know it's something software related anyway.

 

[jamieboo]

Now this is odd...

I went into the bios and changed Performance Core Ratio from Auto to By Core Usage.
This then opened a list of individual options: 1-Core Ration Limit through to 8-Core Ratio Limit all of which were set, unchangeably, to Auto. (Surely all these sub-options would have been Auto anyway if I had just kept Performance Core Ratio at Auto? Essentially I didn't change anything.)

Anyway, I saved and exited the bios, and restarted.

Opening HWiNFO I immediately saw a couple of cores hitting 5Ghz!
I also saw that PL1 and PL2 had both been raised from the default of 190W to 4095W, and Current CPU Core/Cache Current Limit raised from 364.0A to 511.75A .
I guess this is all good. But it's weird, isn't it?
Why did simply opening up the Performance Core Ratio options - and then not actually changing anything, as far as I know - have this significant change?

 

[EarthDog]

Wonky bios a cmos reset may have resolved? I dunno.

 

[ShrimpBrime]

That's all very much an Asus bios quirk.
Clear cmos was the best suggestion really.
My Asus boards do the same thing, per core usage seems to operate the cpu properly, but you can't change any multiplier past the all core boost multiplier.

When I OC, I use the per core multiplier, because the all core linked causes the cpu multi to drop to default which my chip is 3.6ghz.

Glad you figured it out though!

 

[jamieboo]

That's all very much an Asus bios quirk.
Clear cmos was the best suggestion really.
My Asus boards do the same thing, per core usage seems to operate the cpu properly, but you can't change any multiplier past the all core boost multiplier.

When I OC, I use the per core multiplier, because the all core linked causes the cpu multi to drop to default which my chip is 3.6ghz.

Glad you figured it out though!

Thanks EarthDog and ShrimpBrime.

So if this is a standard Asus thing - even yours exhibits the same eccentric behaviour - would clearing cmos really make any difference? Have you cleared your cmos?
I've never cleared cmos before so I guess I'm a tiny bit apprehensive! But even apart from that, I don't want to just clear cmos for no reason.

What do you think?

Thanks

 

[EarthDog]

I've owned plenty of them over the past few years and never ran across this issue in my use (typically overclocking not leaving it on idle though)... so I personally wouldn't call it an Asus thing. I'm also on newer boards I'd imagine where his experience lay more towards old school boards/chipsets where that may be common. When I use all core, it doesn't drop to base clocks for me (and didn't for you either)...

Just sounds like a wonky BIOS... it happens and CMOS reset typically takes care of that.

 

[ShrimpBrime]

Thanks EarthDog and ShrimpBrime.

So if this is a standard Asus thing - even yours exhibits the same eccentric behaviour - would clearing cmos really make any difference? Have you cleared your cmos?
I've never cleared cmos before so I guess I'm a tiny bit apprehensive! But even apart from that, I don't want to just clear cmos for no reason.

What do you think?

Thanks

Well mine is a different story because it's a 12400F. So I need that multiplier set as high as possible. Then I run it overclocked via fsb.

Yes, I've cleared cmos several times. Buy always set it up manually. I don't need to save power, or the very little of it.

The boost algorithm is time based. And it's shitty. I would recommend all users to set up an overclock all core steady. It'll just give better overall performance.

It's first Gen DDR5 platform. So we should expect some quirks here and there.

 

[ShrimpBrime]

Just sounds like a wonky BIOS... it happens and CMOS reset typically takes care of that.

When I post the B660-G at 6200mhz, IF (huge if) it posts to bios, it gets all scrambled. Different languages all over the place and really unstable even. It'll lock up before too long. You might have seen that post at TPU about something similar, I'd be willing to bet he pushed the memory hard on his board to produce the effect. I said nothing to call him out as he stated it was from a bios update, I've not seen this on any of the LGA 1700 boards that I've flashed thus far.... I'd hate to be the guy to call "fake news" when we probably won't read about a bios update scrambling the languages anywhere else.