- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Link: https://www.youtube.com/watch?v=TOvJAHhQKZg
When not using one of the Intel default BIOS profiles, it looks like the 0x129 microcode voltage cap fix is disabled. Why, oh why was this implemented like this? Why not at a deeper, hardcore level, that no matter what a user does in BIOS, it will still just work and cap high voltage requests.
Is this as intented/created by Intel, or is this limited to just a few motherboards? This is not a good fix. The default profiles are not efficient, push more voltage than needed and run hotter. Ultimately they cost performance.
Not all motherboards have IA VR Voltage Limit available to still cap high voltage requests when doing an undervolt and not using default profile(s). Also on some controllers, IA VR Voltage Limit doesn't even work. We need a proper fix, this is a disappointing implementation of 0x129.
Can someone from Intel explain?
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
...S
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
That's fair enough as well. On the other hand, I imagine Intel is the one dictating what the fix is and how to implement is. Has it been confirmed that this is how it is working on every motherboard, or is it just Gigabyte so far?
Ultimately Intel is the one that can probably really give the final verdict on this, as well as any other details about inner workings of 0x129 besides a hard voltage cap. It's not all that transparant unfortunately.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello Janitorus,
Thank you for reaching out.
As a general best practice Intel recommends users adhere to Intel Default Settings on their desktop processors. Intel is not changing the tuning capabilities of existing K SKU processors. Users who desire to overclock or utilize higher power delivery settings than recommended can still do so at their own risk as overclocking may void warranty or affect system health (learn more at www.intel.com/overclocking).
The microcode update (0x129) will limit voltage requests above 1.55V as a preventative mitigation for processors not experiencing instability symptoms. This microcode update will primarily improve operating conditions for K/KF/KS processors, with less impact on non-K Series processors. Analysis is ongoing to address other scenarios that can result in this Vmin shift on Intel Core 13th and 14th Gen desktop processors (list of vulnerable processors is at Additional Warranty Updates on Intel Core 13th/14th Gen Desktop Processors - Intel Community). Intel will share a progress update on this analysis by end of August.
If you have any other questions, kindly let us know.
We look forward for your response!
Best Regards,
Robbie R.
Intel Customer Support Technician
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Well, sorry, but that just looks like a standard copy-paste that doesn't answer the question(s) in OP and the post you replied to.
People are mainly looking to run these CPU's at sensible voltages in order to literally save them, not overclock. And in order to do so, sometimes (most times) users need to disable Intel default profiles, otherwise the required options are either locked out/grayed out or at insane values. Things like AC load line, which is at an 1.1 or 0.9 mOhm by default on Intels profile. While 99.9% of CPU's still run stable at 0.5 mOhm at way more sensible voltages, far away from Intels 1.55V cap and at way better performance.
Why is impedance and the complete load line not better tuned by now, a year+ later? Intel has tools for this, why is it not tuned and enforced more strictly?
Intel default profile costs 2000-3000 Cinebench 23 points. Runs hotter. Runs higher voltages at lower frequencies. Slams into thermal throttling.
Powerlimits, iccMax, Multicore Enhancement are still not correct on every Intel default implementation.
There's just no way I am using the Intel default profile when my 14900K runs at 1.280Vcore (gaming) on average when setting it all the right way, with IA VR Voltage Limit at 1400mV. There's margins for out of the box stability and then there's this ... not even sure what to call it. I do not feel comfortable running this architecture at 1.55V. Which is what people end up with, during gaming, at high temperatures. Has there been long term testing done on Intels part? Or are end-users the long term testers?
Reddit thread regarding all tweaks for anyone interested:
https://www.reddit.com/r/intel/comments/1eebdid/1314th_gen_intel_baseline_can_still_degrade_cpu/
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I'm sitting here with my 14700K which for some reason doesn't adhere to the 1.55V limit even when I'm on BIOS defaults (which sets it to Intel Default Settings) on my STRIX Z790-E Gaming WiFi II. I'm getting VID peaks as high as 1.565V in Cinebench R23. HWiNFO confirms I'm on Microcode version 129.
I started a thread but nobody is able to tell me why this is happening.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
1.565V requests for a 14700K are absolutely insane and most likely because of 1.1 mOhm AC/DC LL defaults. Set your SVID behavior to "typical" instead of auto or, even worse "intel fail-safe" as that overvolts it severely.
Other than that, just undervolt and don't waste your time trying to find out why this is happening. AC LL at 0.5 mOhm will be perfect, you can lower it even more if you want with LLC on 4. Or go the adaptive offset method so you can leave IA CEP enabled that way, if you want to.
Do not expect anything but "adhere to intel defaults" copy-paste in here.
Also, do not use XTU. Completely uninstall it. It causes weird behavior.
Fully reset BIOS if you haven't yet.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
My LLC is automatically set to Level 4 and IA AC/DC LL is 1.000 mOhm.
SVID is on Auto which seems to set it to 'Trained'.
Yes, there are ways to undervolt but I feel that is besides the point and might give Intel a reason to deny an RMA - the microcode should work at stopping VIDs above 1.55V and it's not, which means there might be a deeper issue here that needs fixing.
By now I've come across at least 2 other people on this forum alone whos chips are also not adhering to the 1.55 cap despite applying the update. That means there might well be lots of others who may not even know to check things like HWiNFO.
I don't know if this is the voltage accumulation bug that Intel talked about in the latest official communication.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Understood. Intel should know about this potential bug. It doesn't really instill confidence... Also worrying is that on some MSI boards, the cap doesn't activate until user gets in Windows. Not sure what's up with that.
LLC4+AC/DC LL 1.0 mOhm is a lot of voltage, with proper working 0x129 but especially with a potential bug still not limiting voltage requests. Feels like playing with fire even though (extended) warranty is given for that. I'll leave it at that.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
The VIDs are only the requested voltage, the actual voltage supplied to the CPU is a single value, Vcore. HWINFO64 lists this under its motherboard section. I have a 14th gen system (ASUS motherboard) using the 0x129 firmware update and have seen a 13th gen system with the same update (ASRock board) so two different motherboard vendors. No sign of Vcore getting close to 1.55v in either case.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I appreciate it, but I know the difference between the two.
The issue Gessler reported is that the requested voltage with 0x129 should not even be able to surpass 1.55V because Intel's fix specifically should prevent that. Even when incorrect DC LL value is set and Vcore and VID's are seriously out of balance, it still shouldn't request anything higher than 1.55V on 0x129. But apparently for some people, it still does.
Personally I haven't witnessed this, because I think 1.55V is absolutely crazy, whether requested or given. So I undervolt my chips and they all run far, far below that.
"The latest microcode update (0x129) will limit voltage requests above 1.55V as a preventative mitigation for processors not experiencing instability symptoms"
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
It is the difference between limit and prevent. To me limit means there can be voltage requests above 1.55V but they will be reduced to no more than 1.55V. Vcore is the value that matters. But there is probably room for at least some more improvement on this issue and we may see at least one further microcode update before it is fully resolved.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I don't know why we're going into this direction, below again is literally what Intel says about 0x129:
"The latest microcode update (0x129) will limit voltage requests above 1.55V as a preventative mitigation for processors not experiencing instability symptoms"
So when users observe that 1.55V request limit is surpassed, the limit simply isn't working.
This can have serious consequences, depending on the amount of Vdroop happening or not happening. Sure, at 1.6V requests, Vdroop might have you end up at 1.55V at the CPU. But that differs from board to board and LLC level etc. And end of the day, it means the fix still simply isn't working.
This is not something we should just accept and say that these are "just" the requests and that only Vcore matters. It's a balance of all things involved.
I don't see how limit and prevent in this story can be two different things. The limit is literally put in place to prevent damage. These are $600 CPU's so can we please hold Intel/vendors accountable when things don't seem to work as promised.
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page