help overclock i9 14900k

[chrcoluk]

my cpu does 5.5p 44e on 1.10v load llc3 stable all tasks

i saw with the same cpu sp98 people do 5.7p 44e on 1.17-1.18v pulls on cinbench 278waat

when i try 1.18v the cpu not stable in r23 i need more volt

if i will get beter cooler can my cpu be stable in low voltage?

These CPUs are designed to be stable at tjmax, so if you unstable at 100C, its not a stable configuration even if you lower the temp, the benefit of lowering temp is to get out of a throttling state rather than stability.

1.10v seems really low, either thats golden sample territory or a too aggressive undervolt. If its your stock voltage, then you have a very good binned chip in my opinion.

Although I have a different chip, its still same arch, and so you have something to compare to, in its original OOB config, I was above 1.3v under load, and I considered reaching 1.2's an achievement and eventually got it down to 1.168v. I dont know if I can push a little further as the only instability I have seen on this thing aside from my original experimentation is when my RAM was at incorrect voltage. But I just wanted to lower power draw a bit and left it at that.

Now days throttling is a big thing with manufacturers pushing hardware to its limits out the factory, so lowering heat/power isnt just an eco thing it can actually gain you performance.

 

[ir_cow]

This reminds me of a buildzoid video where he found out the Alder Lake had a bug where if your undervolt far enough it but also isn't unstable, the performance drastically drops.

I managed to replicate it as well. I'll see if I can find the video. So maybe that is part of the problem. I need 1.25-1.3 for 5.5GHz all-core. 1.1 is just a little suspicious.

 

[Psychoholic]

Overclock a 14900k eh? I actually undervolted and power limited mine to 180w and lost almost nothing performance wise.. less than 10% for sure.

 

[ir_cow]

Overclock a 14900k eh? I actually undervolted and power limited mine to 180w and lost almost nothing performance wise.. less than 10% for sure.

I guess 10,000 points in Cinebench R23 is nothing to you.

 

[chrcoluk]

This reminds me of a buildzoid video where he found out the Alder Lake had a bug where if your undervolt far enough it but also isn't unstable, the performance drastically drops.

I managed to replicate it as well. I'll see if I can find the video. So maybe that is part of the problem. I need 1.25-1.3 for 5.5GHz all-core. 1.1 is just a little suspicious.

You got me a little paranoid now.

But my cinebench is fine just tried it.

Buildzoid had MCE on probably not a good idea, thermal throttling disabled as well, I was thinking bios reset man and start again but he didnt do it. Looks like he has an awful chip that has pretty high stock load voltage and cant even do -50mv as well lol. I bet he brought it, not a sample.

Did he ever do this thing on a new board to see if it was ASUS weirdness? I might check his videos after this one to see if he posted an update. I also noticed his clocks never went down at idle either.

Modern versions of windows are good at pushing low threaded loads to preferred cores now so the benefit of all core clocks are massively diminished in my opinion.

 

[Psychoholic]

I guess 10,000 points in Cinebench R23 is nothing to you.

more like 2,000 points but yeah, losing 10k points wouldnt hurt my feelings much either.

 

[ir_cow]

Did he ever do this thing on a new board to see if it was ASUS weirdness? I might check his videos after this one to see if he posted an update. I also noticed his clocks never went down at idle either.

Idk if he ever did again. I only tried it in the ASUS Z690 Hero. Lowing the voltage on a static OC just above what failed cinebench before. Same results. Funky stuff.

 

[DemonicRyzen666]

it was rofl, only idiots play at 1080p on gpus more powerful than, say, 2060.

so, can you show me a game a Triple A released this year that can be played on ultra at 1080p with the RTX 2060?
I'm pretty sure most Triple A games, will need to be put on medium or lower to be playable.

 

[chrcoluk]

Idk if he ever did again. I only tried it in the ASUS Z690 Hero. Lowing the voltage on a static OC just above what failed cinebench before. Same results. Funky stuff.

Well I am on stock clocks, never touched them, and not on ASUS, so is things for you to try if you interested in trying to see if there is certain conditions that trigger it easier.

In the past I had a weird issue with my 1080ti that would occasionally happen in games, and one single reviewer had noted the same problem on their 1080ti as well.

Occasionally it would suddenly drop a ton of performance whilst reporting normal clock speeds, when this happened the power draw and temp would drop massively with it, my suspicion was it was going to some kind of failsafe mode with a number of cores disabled. I would need to reboot to put it back into its normal state.

Found this interesting article.

Adaptive Clocking in AMD's Steamroller

My favorite paper from the ISSCC processor session describes an adaptive clocking technique implemented in AMD’s 28nm Steamroller core that compensates for power supply noise. Initial results show a 10-20% decrease in power consumption from reducing the voltage, with no loss in performance. This...

www.realworldtech.com

 

[phanbuey]

This reminds me of a buildzoid video where he found out the Alder Lake had a bug where if your undervolt far enough it but also isn't unstable, the performance drastically drops.

I managed to replicate it as well. I'll see if I can find the video. So maybe that is part of the problem. I need 1.25-1.3 for 5.5GHz all-core. 1.1 is just a little suspicious.

Ive seen this as well - my 13700kf does it on this board.

 

[Dr. Dro]

Speaking of motherboards, some pretty massive update for some MSI Z690s just came out. At least for my ACE, new BIOS, new ME FW, APO support for 14900K... I just upgraded it all, working well so far. Half the code they wrote for this thing was to block non-Refresh Raptor Lakes from booting it, the KS included, it's so incredibly sad to see Intel go this low. Not only didn't they have a CPU to ship this year, but they're actually attempting to artificially segment the extremely minimal (over the 13900K) and non-existent (against the 13900KS) improvement in their "14th gen" CPU behind some custom scheduling software

 

[phanbuey]

Speaking of motherboards, some pretty massive update for some MSI Z690s just came out. At least for my ACE, new BIOS, new ME FW, APO support for 14900K... I just upgraded it all, working well so far. Half the code they wrote for this thing was to block non-Refresh Raptor Lakes from booting it, the KS included, it's so incredibly sad to see Intel go this low. Not only didn't they have a CPU to ship this year, but they're actually attempting to artificially segment the extremely minimal (over the 13900K) and non-existent (against the 13900KS) improvement in their "14th gen" CPU behind some custom scheduling software

View attachment 322463

Yeah exactly - i hope this blows up in their faces.

 

[Dr. Dro]

Yeah exactly - i hope this blows up in their faces.

Likely won't, most people don't seem to know what it is and much like around the time AMD lied about Zen 3 on X370, PC media outlets are not showing any interest in exposing Intel's dirty tricks here. It's unfortunate.

 

[ir_cow]

The X370 issue was mainly because the BIOS chip wasn't big enough for all the SKUs. Eventually Ryzen 5000 got support on the older chipset.

 

[Dr. Dro]

The X370 issue was mainly because the BIOS chip wasn't big enough for all the SKUs. Eventually Ryzen 5000 got support on the older chipset.

Which was fundamentally a lie anyway. 400 series boards with 16 MB got support and initially AMD intended to leave X470 out too until it blew up on Reddit. They backtracked fast when Alder happened.

Don't think anyone and I mean anyone looking for a Zen 3 Ryzen cared about Excavator Athlons at all, so imo that one excuse never held

 

[ir_cow]

@Dr. Dro I mean it's not really a lie. Those MBs without enough ROM space has the BIOS really stripped down with easy mode removed and graphics. Some even have notes saying Zen1 will not with this BIOS revision.

 

[AusWolf]

Why do you want to overclock a CPU that's already overclocked to the moon and back straight out of the box?

Practically speaking, like others have said, you need a much better cooler.

 

[chrcoluk]

A little more info on my daily config, I did actually dial back the UV for day to day as a buffer, something I have always done in past whenever overclocking or whatever a small dial back from what I had tested as ok, but also the vcore I am seeing when stressing all core is about 1.24v, so nowhere near the 1.168 I have previously documented. So yeah 1.10 does seem a stretch but his chip is probably more optimised manufacturing process, is on a i9 which tend to be better binned chips, and the lottery for required vcore seems highly variable based on the figures I seen posted on the net for stock behaviour on these hybrid chips.

Interesting observation hwinfo is reporting a peak overall temp of 32C on my cpu, I wish I had performed some kind of cooling miracle, but instead something on there isnt working right, the cpu package temp did hit 67C as reported by hwinfo and core temp 68c. I am also exceeding review scores whilst the chip is capped to 175w. I might drop it to 125w due to my eco obsession. Ambient temp is low with time of year at 15C. But yeah no issues with my scores, so if there is clock stretching on my chip I think its an insignificant amount.

 

[Dr. Dro]

@Dr. Dro I mean it's not really a lie. Those MBs without enough ROM space has the BIOS really stripped down with easy mode removed and graphics. Some even have notes saying Zen1 will not with this BIOS revision.

Unfortunately, it very much was a lie. Those BIOSes were just poorly designed.

Boards which had decently programmed BIOSes such as the Crosshair VI series had ~9 MB of free space back then, and once AMD allowed ASUS to update them, they had never lost any feature or had any graphical downgrade in the BIOS... For those boards, all of their stability issues could very well be pinned directly onto the AGESA, the newer versions of these BIOSes are super stable and feature complete.

What happened is that motherboard makers basically got outed for their poorly designed BIOSes and AMD used the excuse to upsell newer products at the same time. Complete mess.