Well, you’re not me and I’m in no mood to be playing musical CPUs.
I want Intel to replace my quite-possibly-damaged CPU with a brand new one from an unaffected batch and I will install it after their microcode fix has been deployed and stabilized. I am super unhappy with them and with this industry in general for being so incredibly unprofessional in their testing/validation.
There’s no excuse for zoinc-ups this big in 2024.
Fair enough, but it’d be helpful to make sure the issue IS the cpu.
Anyway, good luck!
Not sure if this detects your problem?
No.
It would be highly risky to run any kind of stress test on any Intel Core 13th and 14th Generation desktop processors with 65W or higher base power given the warning of permanent damage.
The release notes for version 4.1.9.41 of that diagnostic tool are dated January 10,2024, therefore predates the confirmation by Intel of the issue affecting 13th and 14th gen CPUs.
I dont think so especially if you are using factory defaults for current and power settings. As CPU voltage will decrease a lot in all-core workloads due to vdroop and LLC. High voltages are expected on single core or idling… And how you will be sure if CPU is configured properly if not running benchmarks and stress tests at least at bare minimum amount of time (I mean real world stress tests and not synthetic crap like Prime95 which will cook your cpu). Just needs to monitor temps, power, current and voltage when stress testing…
But I understand your concerns regarding this - as there is a huge possibility that Intel dont know themselves which are safe long-term settings and frequencies for Raptor Lake… I think its biggest PC tech fail in decades now… And Im pissed a lot as a 13900KS owner…
Anyway unfortunately knowing the all this disaster we cant trust Intel and board vendors 100% regarding ‘‘correct default settings’’ and especially voltages for Raptor Lake - its best to monitor and check them in HWINFO64 or similar soft during everyday usage and try to configure system in a way to not exceed 1.45-1.5v for now…
In Intel Spec Sheet for RP-L the maximum operating voltage stated as 1.72v - but the problem is that at such a voltage this cpu will degrade in couple of days… I guess its the maximum voltage in which CPU will boot and work - but its not supposed to be a 24/7 safe voltage, which could be I guess in a range of 1.45-1.5v…
If any one is interested I will upload latest Intel specs as well…
Yeah, I’m gonna wait for Intel’s August rollout. I already updated to my motherboard’s latest BIOS (Asus) which they claim includes microcode fixes, but I’m not gonna push this thing harder than I already have.
It’s almost certainly been permanently destabilized, though: I’ve hit it quite hard in the past year or so.
My level of ire at Intel over this fiasco is pretty hard to overstate.
Latest news:
Microcode (0x129) Update for Intel Core 13th and 14th Gen Desktop Processors
“The latest microcode update (0x129) will limit voltage requests above 1.55V as a preventative mitigation” My understanding is that overclockers can still destroy their CPUs, should they choose to do so.
Hi,
I can report MY PERSONAL experience with Cubase 13 and a i7-14700K.
It can happen that the system runs stable with the BIOS standard setup for a while, and then after some weeks or months I experience totally weird behavior of Cubase. Tracks are not played correctly (only some notes are played) and even these notes are wrong in timing.
As soon as I deactivate the E-cores in the BIOS, then Cubase and all plugins runs fine.
I just experienced this some days ago. Worked on a small test project with standard BIOS setting and all was ok. About 1 week later I reopened this small project and nothing worked well. I changed the BIOS setting (deactivation of E-cores) and Cubase ran fine again.
I informed Steinberg about this issue already some months ago, but they said that “this is my system”.
Hi,
Cubase 13.0.41, Intel I9-13900K.
Maybe I’m the lucky guy, but my system is very stable.
No tweaking e-cores in the bios, just normal system settings.
The only thing I noticed that the latency in Windows 11 has improved.
But again, no problems with Cubase.
Windows updates can, and usually do , reset the power profile back to Balanced or whatever the default setting is. Make sure whe you update to check your power profile settings again and put them back to high performance.
This has caught me out a few times in the past.
M
Hi Norbury,
It is on high performance.
Any chance you can share your computer components and what you have done to achieve such low latency for 20 minutes. After tweaking forever I have my cubase 13 running very stable on windows 11 with i13900K but it spikes a lot every 15 minutes. It’s very strange. I tried everything and just can’t figure that out.
Thanks
If you’re curious about the ‘15 minute’ problem, have a read here. Some folks did a deep dive trying to find the reason for it, and it appears they were successful.
I haven’t tried their solution so can’t comment on it, but if you’re having issues you might give it a go.
Also, have you tried turning off Core Parking? This is the one setting that I’ve found makes a real difference on 12th, 13th and 14th gen Intel machines.
As mentioned above I disabled Diagnostic Policy Service in Services - it disabled that every 15min ntoskrnl.exe spike.
My config is: 13900KS, 64GB DDR5, ASUS ROG Strix Z790-E
I had disabled core parking for both P and E cores, using High Performance windows power plan (it do not downclocks CPU). Also I had disabled SysMain service aswell. In BIOS I had disabled Enhanced C-States and limited package C-States to level C0/C1.
Other settings are 400A ICCMAX, 320W PL1 PL2, latest BIOS with 129 microcode, I limited my voltage SVID requests to max 1.5v in IA VR Voltage Limit by setting it to 1500mV. In real world usage VCore is not going above 1.475v. I dont use Intel Default Settings, I limited voltage and ICCMAX by myself to Intel spec.
I’m glad these steps resolved the issue for you. But be aware of potential downsides for your OS. Disabling the Diagnostic Policy Service will prevent you from viewing Network % utilization in Task Manager, and diagnostic scenarios won’t run. Similarly, disabling SysMain will turn off memory compression, potentially increasing RAM usage… a concern if you work with sample libraries, for example. Keep these in mind if you notice any unusual behavior in the future.
I did not found any issues for my workflow disabling those windows services. Regarding RAM - I tested and did not have absolutely any difference at all between SysMain enabled and disabled - both have same around only 3GB consumed when idling in desktop and no difference when even usind DaVinci Resolve Studio heavy projects with up to 50GB of RAM usage…
I nor recommend nor not anyone to use tweaks I am using, its up to you to decide. For me it works quite perfectly and I am enjoying low latencies and excellent responsiveness on my 13900KS system. 
YOU ARE AMAZING, thank you so much for your help. This helped. After struggling for so long I finally have amazingly low latency over 15 minutes. I did both sysmain and Diagnostic policy service. Love it now. Beer on me to you!!!
Thank you
@jackthepilot were you actually having audio dropouts though?
reason i ask is I had a 12900 laptop I used for a year, it was great, no issues at all. After a year of use I one day decided to do the DPC latency check just out of interest. It failed miserably , it was in the red and saying that this machine wouldn’t do real time audio.
I’d just finished doing a weeks recording at 32/96k on this machine without an issue so that surprised me.
From this I learned DPC latency checker is only really useful ‘IF’ you have a problem and with a high core count CPU high DPC latency doesn’t meant your computer will have problems as a DAW.
M
Good Question, I actually didn’t but it was bothering me to see such spikes, now it’s perfect. I agree with you if it ain’t broken don’t fix it, but it was always in the back of my mind. I always strive to make my system as light as possible and are always opened to some further enhancements.
Thank you
Wow. We have the exact same motherboard. Surprising how similar our system specs are, actually. Thank you for sharing your settings.
Intel has agreed to replace my i9-13900K free of charge. I’m actually getting a slight upgrade to the 14900K, so I guess that’s nice. We’ll see.
This whole thing has just been such a nightmare, but thank you all sincerely for chiming in and sharing your settings & tweaks. Some of these conclusions I don’t know how we arrive at, so it’s always appreciated to share your reasoning behind them as much as possible, too, so we can evaluate whether or not they might make sense in our case.
But again, I continue to stand by my original assertion that these tweaks and technical deep-dives into our system settings should not be on us to figure out: these should be on these monster corporations with the resources to hire people to do this full time. That includes Yamaha, Intel, Microsoft, and NVIDIA.
I think it would be wise for all of us to continue to find ways to unify our voices in asking for this, in demanding this: they are getting so, so much free engineering time from us, and they know it. It is not right that the onus to work around their mistakes and oversights is on us when our job should be just getting music done.
Some of you may enjoy such tinkering. There was a time long ago when I used to, but it’s eaten into so much of the time I could’ve been doing being creative that I’ve come to deeply resent the need to do so–because I know, again, that these extremely wealthy corporations feel no pressure from us to shape up and solve these problems amongst themselves.
They continue to sell us extremely expensive software and computers with the implied promise that stuff will actually work as intended. And they continue to break this promise as they “innovate” their brains out at the expense of stability–which is a forever-moving target these days. It’s enervating and maddening.
No more free engineering time for large corporations by the consumer: it should be on them to get together and figure out where things are breaking down, and why. Hire musicians who are actually using the software to be part of this effort, if necessary. But the status quo is untenable.
