Some Intel CPUs lost 9% of their performance almost overnight

erek

erek

[H]F Junkie
Joined
Dec 19, 2005
Messages
10,923
"Over the past few weeks, we’ve seen an increasing number of reports of instability on high-end Intel CPUs like the Core i9-14900K. Asus has released a BIOS update for its Z790 motherboards aimed at addressing the problem, but it carries a performance loss of upwards of 9% in some workloads."

1713844400758.png

Source: https://www.digitaltrends.com/computing/intel-baseline-profile-tested/
 
Honestly, this seems like a positive and IMO the "Intel Baseline Profile" should be the default. No non-XOC mobo (Intel or AMD) should ship with unlimited (or effectively unlimited) PL being the default setting.
The situation with Intel mobos the past few generations has been kinda out of control. Imagine for a moment if this had happened with GPUs- say NV AiBs saw how competitive the 6900XT was and responded by shipping 3090 and 3080Ti boards with uncapped PL to look better, that would've been a shitshow, and frankly it reflects poorly on Intel that they've been allowing their mobo partners to just do whatever tf they want with stock settings...
 
I know someone who has seemingly been bitten by those crashes.

Worse is it seems to degrade the processor over time and they get worse and worse. Eventually it's basically unusable under any kind of load.

I think they're on their 3rd processor now, having also replaced the mobo. The first 2 were a 13 series, this one is a 14 and hasn't run into issues... yet.

Bonkers
 
Asus was one of the AIBs who was shipping the default profile with a 4096w 512a max power setting. Instead of the much lower 340w and 350a power targets they should.

They are essentially OC’ing the crap out of the chip as a default and letting Intels built in protections keep it from catching fire.

It didn’t end well for early AM5 adopters when they did it there, and I’m surprised that it’s only 9% in limited loads.
 
socK said:
I know someone who has seemingly been bitten by those crashes.

Worse is it seems to degrade the processor over time and they get worse and worse. Eventually it's basically unusable under any kind of load.

I think they're on their 3rd processor now, having also replaced the mobo. The first 2 were a 13 series, this one is a 14 and hasn't run into issues... yet.

Bonkers
Click to expand...
Yeah caught my friends system on that early, he built his in late Feb and we changed the Auto settings to limit the power curve by March. So it’s unlikely that it did lasting damage, but others out there been running for a year plus those are gonna be in for a bad time.
 
This story has been coming to this point for months now. When this first kicked off a couple of months ago with reports of instability it was said it likely has something to do with outrageous motherboard presets so its no surprise that this is where we'd end up, board vendors simply having to adhere to Intel's base spec.

Board vendors got greedy and this is what happens.
 
criccio said:
This story has been coming to this point for months now. When this first kicked off a couple of months ago with reports of instability it was said it likely has something to do with outrageous motherboard presets so its no surprise that this is where we'd end up, board vendors simply having to adhere to Intel's base spec.

Board vendors got greedy and this is what happens.
Click to expand...
They did it to AMD, so you know they had to be doing it to Intel too.

Makes you wonder if maybe Nvidia is right when they call those AIB’s dirty grubbers who will happily damage your brand for a quick buck.
 
socK said:
I know someone who has seemingly been bitten by those crashes.

Worse is it seems to degrade the processor over time and they get worse and worse. Eventually it's basically unusable under any kind of load.

I think they're on their 3rd processor now, having also replaced the mobo. The first 2 were a 13 series, this one is a 14 and hasn't run into issues... yet.

Bonkers
Click to expand...
I believe it. It seems the stock V/F curves for 12th/13th/14th Gen Intel (especially the latter two) will absolutely blast the CPU with voltage if you let it. Gotta get that 6Ghz to compete with 7800X3D, power consumption and degradation be damned!
 
One of the first things I did when I built my 13900 system was to set back the Intel defaults in place. It shipped with much higher limits and it was always getting limited by heat. I was a little concerned about it running so hot under load, and it also seemed pretty pointless as there was little if any clockspeed gain. So I flipped it to enforce all limits and I've left it that way ever since.

To me, overclocking seems like not something that useful these days, unless you have a special use case or just like seeing how hard you can push things. Like back in Ye Olde Days sure, I had a Celeron 300a that I ran at 450mhz with 100mhz RAM because that was an AMAZING speed uplift for the price. That did require some hefty extra voltage though, and did burn out the CPU after like 3 years.

Even more recently on my 8700k I messed with the clocks because it had a pretty significant negative AVX offset and with some tweaking I was able to get it to basically do its single core speed at all cores, even with AVX.

However the 13900? Nah man, it handles itself real well. It seems to run at as high a speed as a given work/thermal load allows, very often hitting 5.7-5.8ghz on two cores with stock settings. It slows things down, as needed, if there's a heavy multi-core workload, but its dynamic adjustment just works really well. I don't see the need to try and squeeze it.

Also might explain why I've never seen any of the crashes people are talking about. I played a lot of Hogwarts Legacy and it didn't have stability issues.
 
Lakados said:
Asus was one of the AIBs who was shipping the default profile with a 4096w 512a max power setting. Instead of the much lower 340w and 350a power targets they should.

They are essentially OC’ing the crap out of the chip as a default and letting Intels built in protections keep it from catching fire.

It didn’t end well for early AM5 adopters when they did it there, and I’m surprised that it’s only 9% in limited loads.
Click to expand...
This isn't something new, and has been going on for over a decade. The problem is that AMD and Intel don't hold motherboard manufacturers accountable for playing with their CPU's in order to get consumers to "buy the better motherboard". If AMD and Intel sell them a chipset then they better obey their chips limits or don't sell them anything.
socK said:
I know someone who has seemingly been bitten by those crashes.

Worse is it seems to degrade the processor over time and they get worse and worse. Eventually it's basically unusable under any kind of load.

I think they're on their 3rd processor now, having also replaced the mobo. The first 2 were a 13 series, this one is a 14 and hasn't run into issues... yet.

Bonkers
Click to expand...
It's technically overclocking the CPU's, which does degrade them. It's one of the reasons I stopped overclocking certain CPU's because I've noticed they don't hold that clock speed forever and will need their clocks to be lowered to keep things stable. My Ryzen 1700 was able to hit 4Ghz just fine for a while, but as time went on I had to lower it to 3.9Ghz to keep things stable. That's not so bad, as some CPU's I had to really lower their clocks down to almost stock to keep things stable. Now if motherboard manufacturers have been degrading CPU's to the point where they can't run stable at all, then I feel they owe their customers a new CPU. I would say a class action lawsuit is needed at this point.
 
DukenukemX said:
This isn't something new, and has been going on for over a decade. The problem is that AMD and Intel don't hold motherboard manufacturers accountable for playing with their CPU's in order to get consumers to "buy the better motherboard". If AMD and Intel sell them a chipset then they better obey their chips limits or don't sell them anything.
Click to expand...
They aren’t really allowed to, they can wave a finger at them and say “No! bad AIB!!!!” But they can’t fine them, they can’t withhold product, they can’t take legal actions, and they can’t revoke their ability to sell their product. So what can they do?
Intel like AMD before it will ask them to cover some of the RMA costs associated with chips damaged in this process, meanwhile users of those affected devices are left wondering “Why is my system like this”.
Intel has too many past legal presidents against it to be able to do anything beyond a stern letter and a public statement, and AMD can’t afford to risk their partnerships with the AIB’s because they are too low volume to significantly matter. The AIB’s have them both by the balls.

We can give Nvidia a rough time for how they deal with their AIB “partners” but it’s because they know they would do this kind of stuff to their products too and ultimately hurt the Nvidia brand if they had the chance. So Nvidia makes darned sure that they know they don’t have that chance, but sadly it’s only their absolutely dominant market position and pure volume of sales that make them capable of that. But a few anti competitive legal rulings is all it would take to render Nvidia as toothless as Intel in this type of situation, preparing for that sort of ruling is why Nvidia is gradually trying to build their first party brand in availability.
 
Last edited:
I personally ran into issues on my 13900k, had to run intel fail safe which gives the CPU 1.6v for it to be stable. I had the PL1/2 set to 300w since day one as well.

Swapped to a 14900k and the system is stable again but intel is giving me the run around to get it replaced.
 
LOCO LAPTOP said:
I personally ran into issues on my 13900k, had to run intel fail safe which gives the CPU 1.6v for it to be stable. I had the PL1/2 set to 300w since day one as well.

Swapped to a 14900k and the system is stable again but intel is giving me the run around to get it replaced.
Click to expand...
Sadly this sort of crap will go on for years while Intel tries to pass the buck onto the AIBs and they say no, customers are stuck in a legal limbo while neither party wants to take the responsibility.
Intel will say the AIB’s are at fault for badly programming the power curves, the AIB’s are going to say that if Intel didn’t want the power curves messed with they should have hard locked they should have hardcoded a maximum into the CPU or the Chipsets so the numbers couldn’t go higher than a known safe amount.
Round and round they go while people in the same situation as yourself are stuck circling the bowl.

I hate to say it but things like this are why when my friends ask me to “Build them a PC” I often redirect them to one of my OEM partners for a decent deal on a Dell, or Lenovo. At least then there is a clear chain of ownership for these sorts of things.
Dell does better desktops, Lenovo has more Laptop options for gaming.

Hell my next rig will likely just be a gaming laptop, if I can get my existing performance plus 5-10% then it’s all I really need, I don’t game nearly as much as I used too, and 1440p isn’t the hardest to push any more. Story might change if I upgrade to a 4K gaming monitor but that’s not happening any time soon.
 
Lakados said:
They aren’t really allowed to, they can wave a finger at them and say “No! bad AIB!!!!” But they can’t fine them, they can’t withhold product, they can’t take legal actions, and they can’t revoke their ability to sell their product. So what can they do?
Click to expand...
In the past, there were 3rd party chipsets. I don't remember what it was like for Intel, but the ones for AMD's platforms were universally bad compared to the bog-standard AMD reference chipsets. Now we only have 1st party chipsets, which generally work much better. Motherboard companies have proven to be universally bad at their job, so perhaps it's time to do the same for the entire motherboard? I would by an Intel or AMD motherboard if it was made to be reliable instead of made to be shiny and cheat benchmarks.
 
Mr Evil said:
In the past, there were 3rd party chipsets. I don't remember what it was like for Intel, but the ones for AMD's platforms were universally bad compared to the bog-standard AMD reference chipsets. Now we only have 1st party chipsets, which generally work much better. Motherboard companies have proven to be universally bad at their job, so perhaps it's time to do the same for the entire motherboard? I would by an Intel or AMD motherboard if it was made to be reliable instead of made to be shiny and cheat benchmarks.
Click to expand...

nVidia has great chipsets for AMD back in the day, they were the way to go with 939.

Intel used to have 1st party boards, they moved away from it but still had the NUC series up until this year. I think the last enthusiast board they made may have been the Z77 chipset.


View: https://www.youtube.com/watch?v=uqVYmtM0xCU
 
Mr Evil said:
In the past, there were 3rd party chipsets. I don't remember what it was like for Intel, but the ones for AMD's platforms were universally bad compared to the bog-standard AMD reference chipsets. Now we only have 1st party chipsets, which generally work much better. Motherboard companies have proven to be universally bad at their job, so perhaps it's time to do the same for the entire motherboard? I would by an Intel or AMD motherboard if it was made to be reliable instead of made to be shiny and cheat benchmarks.
Click to expand...
They were bad for everybody, Via, and Nvidia blamed Intel for not giving better documentation, and Intel blamed them for not being able to understand the documentation that was provided and cutting corners so I can see both being true.
Intel used to provide first-party motherboards and they were fantastic, stable, had good OC options, top-tier components, and were price competitive, they did cost more but you could physically see the board components and could verify that they were from higher-quality manufacturers with better tolerances.
But then the AIBs sued because it was anti-competitive for Intel to be selling first-party motherboards for their products, as they had better internal pricing to their chipsets, could guarantee supply to themselves, and had better first-hand knowledge of the components, and all the other nitpicky things.
Intel was then forced to stop selling first-party boards and that has gradually led us to where we are today, now the environment has changed and Intel might be able to do so again, and similarly AMD might be able to as well, but 100% the AIBs would sue and as they have won that argument in the past it would not be hard to argue it a second time.
So is the problem big enough for Intel to want to voluntarily take on that lawsuit?

Just as a note:
Say chipsets are your full-time job, you live and breathe them all day every day, the documentation you make for yourself and your team who similarly live inside those chipsets will be vastly different than somebody who visits on the weekend.
So I can easily see Intel saying F-you, our documentation is fantastic what the hell are you complaining about, for the other guys to look at it and go this is crap simply because they don't have the same background or experience as the Intel team.
 
Last edited:
To give the story on my issue and what intel has sent.
On 4/8/2024 I opened a ticket at Intel stating:
For the past month Discord always crashing, WD Dashboard won't open every time, CRC errors trying to install Nvidia drivers and sometimes (but rare) my games would crash. Figured it was a OS issue or ram as it progressively got worse Ran memtest for 12 hours, no errors. After the memtest I reloaded windows, kept looping back to "install now" screen after "copying windows files" after the 4th try it worked. Cinebench froze once out of the 3 runs. Windows Event logs would give a handful of: WHEA-Logger Error 19. A corrected hardware error has occurred. Reported by component: Processor Core Error Source: Corrected Machine Check Error Type: Internal parity error Processor APIC ID: 0

Specs: i9-13900K, ASUS ROG Strix Z790-E Gaming, G.Skill RipJaws S5 64GB DDR5 6000Mhz, WD Black SN850x 1TB and SN850x 4TB NVMe SSD, MSI RTX 4090 Gaming X TRIO, EVGA Supernova 1000W T2 Since this is my work machine and can't really afford downtime during the week, ordered a 14900k off Amazon and swapped it. All issues are solved. Reloaded Windows again as I had issues with the 13900k and that was also fine. Nothing else was changed.|
Click to expand...

On 4/12/2024 after no response from Intel:
On the day I put in this ticket, I also ordered a Asus Z790-PLUS and another identical set of Gskill 6000mhz ram since I figured you would want to troubleshoot this but I haven't heard back, so started my own testing to figure out where the issue really is. Set PL1/2 to 253W and left everything else default. Using OCCT I was getting CPU errors in 15-20 minutes. Started to turn off P cores one by one. No changes. Enabled them again and turned off Hyperthreading. After 2 hours I stopped the test as it didn't fail anymore.

If you have any suggestions or any other test you want me to run, I'm all ears.
Click to expand...

They responded a few hours later:

Thank you for reaching out to Intel Customer Support about an issue with your Intel® Core™ i9 processor 13900K. We understand the frustration this may have caused and apologize for any inconvenience.



To assist you further, we recommend following these steps to isolate the issue:

  1. Adjust the BIOS settings as follows:
  • Access Advanced (F7) settings.
    • Navigate to AI Tweaker.
    • Set SVID Behavior to "Intel’s Fail Safe".
      • "Long duration power limit" -> reduce to 125W
      • "Short duration power limit" -> reduce to 253W
    • Reboot the OS and test your system to see if the issue persists.


We understand your time is valuable, and we appreciate your patience as we work through this. Would it be convenient for us to follow up with you after 2 business days to ensure you received this email and were able to complete the steps? If this date doesn't suit you, please let us know your preferred date and time for us to reach out.



Thank you for choosing Intel®. We're here to assist you every step of the way. Have a fantastic day!"
Click to expand...

Now I was a bit puzzled, but I tried and it passed just fine, besides the fact that Cinebench gave me a 27,394 instead of the 37,364 that it was getting. I did some more testing and found the only thing that made a difference was "Intel Fail Safe" The power limits could be set to 300w like before and it ran perfectly fine for days. Downside with that is, the CPU voltage is showing was showing 1.607v in the Bios, so I asked intel about this.

Adjusting the settings helped. It hasn't crashed during a few hours of testing with OCCT, so that's an improvement. I did noticed it's running 1.607 V core. which seems a bit high?

I did try other SVID settings besides "Intel’s Fail Safe" after the testing to see if it would be stable, it is not. Auto was the default setting which cause the errors in 15-20 minutes. The other settings caused lock ups and crashes within 30 seconds at the start of a stress test.
Click to expand...

They responded with:
Good day! and I hope you are doing okay today. To answer your question about why the voltage is a little bit high, it is because running with “Intel Fail Safe,” increasing Dynamic Vcore, or adjusting LLC settings is either Intel default behavior or within specification. It is not increasing the voltage beyond specification; it is setting the behavior back to what Intel’s specifications say it should be. This will not “burn out” the processor and will not impact the CPU lifecycle.
Click to expand...

Then I said:
Based on that response, that's it? Nothing else needs to be done troubleshooting wise as it's not crashing at this time and will just have to run at a higher vcore (and remember to set intel fail safe on each bios update) on a 16 month old processor (January 8, 2023)?

That doesn't seem right to me.
Click to expand...

Now I won't make this post much longer with the back and forth but because the CPU is working 'as design" which I guess it is, they simply won't replace it as of now. However I am concern about the possibly that it degraded to the point that maybe another year or so I will start having issues again. So I'm stuck with something I don't trust and would be a paper weight.


Anyway I tried the BIOS update, these are the settings the Asus TUF Z790-PLUS set for the intel base line.

1713910100405.png


After two hours, it appears to be stable. 33499 Cinebench score, which is lower than the factory settings from 37,364, it pulls a max of 227 watts now while it is set to 253w PL1/2 and the base ratio is now 51 (5.1Ghz) instead of 55 (5.5Ghz). I'm still testing and playing around with it. But IMO this is a band aid.
 
Lakados said:
They aren’t really allowed to, they can wave a finger at them and say “No! bad AIB!!!!” But they can’t fine them, they can’t withhold product, they can’t take legal actions, and they can’t revoke their ability to sell their product. So what can they do?
Click to expand...
Could they not stop selling them chipset past the current licensing agreement (which I would imagine cover design limit clause) ? If they start and modify the Intel reference model (like I imagine they all do now), they would be using something protected in some way.

But there is a bit of assumption that Intel is angry that AIB are pushing the product a little bit, part of them would like them to find limits and push performance up and have nice benchmark versus AMD, and you have to live with the it was too much from time to time that is inevitable that come with it.

Didn't AMD simply did put an hard voltage limit in their part of the BIOS that all AIB seem to use to fix the issue, something Intel could also do.
 
Glad I stuck with my 13700kf. I had to also reel in the voltage that MSI was dosing my cpu with at default.. it made the cpu shoot up over 100c during just the normal intel bench marking tool.

EDIT: I wonder if the higher default P core clocks on the i9's are just too on the cusp of being stable for some cpus.. and they're falling out of that sweet spot of voltage+frequency?
 
Last edited:
LukeTbk said:
Could they not stop selling them chipset past the current licensing agreement (which I would imagine cover design limit clause) ? If they start and modify the Intel reference model (like I imagine they all do now), they would be using something protected in some way.

But there is a bit of assumption that Intel is angry that AIB are pushing the product a little bit, part of them would like them to find limits and push performance up and have nice benchmark versus AMD, and you have to live with the it was too much from time to time that is inevitable that come with it.

Didn't AMD simply did put an hard voltage limit in their part of the BIOS that all AIB seem to use to fix the issue, something Intel could also do.
Click to expand...
Sadly no, to comply with the previous rulings Intel doesn't sell any of its components directly, they sell them at a fixed price to their authorized distributors who then sell them to the AIBs and everybody else so Intel doesn't technically sell their parts directly to any source anymore. The independent distributors are just that, Independent, and Intel has little to no control over who they sell those components to. If the various distributors got together and decided to not sell them the parts that would be a different story though, but they won't, they are just middlemen and they don't deal with any of the return or RMA processes, just the sales.

AMD provides their partners with the AGESA packages that all their BIOSs are built from, Intel similarly hands out their UEFI updates to vendors which they could include a default profile or limiter too. I suspect they issue an update with an Intel Approved profile which the vendors can similarly rename to be more inline with their brand.

Ultimately and unfortunately Intel should have been hard locking those values to begin with, their not doing so was just leaving the door open for the AIBs to try and get away with whatever they could for that extra 0.1%, and the customers are left holding the bag. While I do believe Intel will begrudgingly replace those CPUs it will still absolutely suck for any customer who has to do it, and the AIBs won't care, and they won't apologize.
 
Last edited:
Lakados said:
Actually yes, that is the symptom of the voltage issue. It really becomes noticeable in games using Oodle compressions.
Click to expand...
Yah Oodle can hit the CPU very hard in short bursts. Newer versions of the DLL are highly multithreaded too. Cause of much angst for me running modded Cyberpunk on overclocked CPUs.

Actually, huh, I'm just now wondering if this Intel issue is part of the reason a handful of people have been saying my CPU-murdering traffic density mod for Cyberpunk keeps crashing their game but I'm unable to reproduce the crashes :cautious:
 
Sycraft said:
To me, overclocking seems like not something that useful these days, unless you have a special use case or just like seeing how hard you can push things. Like back in Ye Olde Days sure, I had a Celeron 300a that I ran at 450mhz with 100mhz RAM because that was an AMAZING speed uplift for the price. That did require some hefty extra voltage though, and did burn out the CPU after like 3 years.
Click to expand...
Not to derail. It should not have required extra voltage never mind a HEFTY one. That was the whole magic of the 300a was it was a hobbled chip for market segmentation and simply just needed to be set to 100MHz bus. Maybe poor paste or lacking cooling which was the norm back in the day. Poor Celly :(
 
  • Like
Reactions: erek
like this
NattyKathy said:
Yah Oodle can hit the CPU very hard in short bursts. Newer versions of the DLL are highly multithreaded too. Cause of much angst for me running modded Cyberpunk on overclocked CPUs.

Actually, huh, I'm just now wondering if this Intel issue is part of the reason a handful of people have been saying my CPU-murdering traffic density mod for Cyberpunk keeps crashing their game but I'm unable to reproduce the crashes :cautious:
Click to expand...
Could be, I first started hearing about this issue in the Unreal forums for Oodle and they were telling Intel users to drop their max clocks by 200 mhz to keep things stable.

https://www.radgametools.com/oodleintel.htm
 
funkydmunky said:
Not to derail. It should not have required extra voltage never mind a HEFTY one. That was the whole magic of the 300a was it was a hobbled chip for market segmentation and simply just needed to be set to 100MHz bus. Maybe poor paste or lacking cooling which was the norm back in the day. Poor Celly :(
Click to expand...
Well it depended on how good of a sample you had. While they had good yields and plenty of stuff ran faster, they only tested 300a units at 300. So if you had particularly bad silicon, it could still pass in that bin, even if it would have failed in a higher speed. My roommate had one that ran at 450 just by changing the bus speed, it was a good sample. Mine required a lot more voltage, I can't remember how much, I wanna say 200mV but it was a long time ago. It worked though and was good value for the money, even if I did eventually burn it out.

That's the other thing with OCing: Sometimes you get lucky, sometimes you don't. My 5930K was a good sample, I just set the all-core clock speed to 4Ghz and it said "OK" no voltage increases or anything, I'm sure I could have pushed it farther but I decided that was plenty.

Just easier to leave the limits in place these days. Both Intel and AMD seem to be pushing their high end chips very close to their limits out of the box. Particularly since turbo boost, thermal velocity, and whatever other fancy names they want to come up with for variable speed are so good. The CPUs do a good job of throttling up and down as needed, and allocating their power to cores intelligently.
 
DukenukemX said:
It's technically overclocking the CPU's, which does degrade them. It's one of the reasons I stopped overclocking certain CPU's because I've noticed they don't hold that clock speed forever and will need their clocks to be lowered to keep things stable. My Ryzen 1700 was able to hit 4Ghz just fine for a while, but as time went on I had to lower it to 3.9Ghz to keep things stable. That's not so bad, as some CPU's I had to really lower their clocks down to almost stock to keep things stable. Now if motherboard manufacturers have been degrading CPU's to the point where they can't run stable at all, then I feel they owe their customers a new CPU. I would say a class action lawsuit is needed at this point.
Click to expand...
My 8700k, which is roughly the same era as the 1700x, has been running @ 4.8Ghz all cores synced with the power limit maxed out and AVX offset disabled since day dot - No problems at all. In game, all cores/threads hit a full 4.8Ghz and stay there.
 
Lakados said:
Sadly this sort of crap will go on for years while Intel tries to pass the buck onto the AIBs and they say no, customers are stuck in a legal limbo while neither party wants to take the responsibility.
Intel will say the AIB’s are at fault for badly programming the power curves, the AIB’s are going to say that if Intel didn’t want the power curves messed with they should have hard locked they should have hardcoded a maximum into the CPU or the Chipsets so the numbers couldn’t go higher than a known safe amount.
Round and round they go while people in the same situation as yourself are stuck circling the bowl.

I hate to say it but things like this are why when my friends ask me to “Build them a PC” I often redirect them to one of my OEM partners for a decent deal on a Dell, or Lenovo. At least then there is a clear chain of ownership for these sorts of things.
Dell does better desktops, Lenovo has more Laptop options for gaming.

Hell my next rig will likely just be a gaming laptop, if I can get my existing performance plus 5-10% then it’s all I really need, I don’t game nearly as much as I used too, and 1440p isn’t the hardest to push any more. Story might change if I upgrade to a 4K gaming monitor but that’s not happening any time soon.
Click to expand...


In this case AIB's are clearly wrong. Intel has clearly defined the max safe operating conditions but instead of following them motherboard manufacturers by default allow those CPU's run at unlimited power draw, which pushes temperatures to the max until the CPU temperature throttles. That is not healthy! As Jays2cents pointed out, changing one setting in the BIOS enforces the Intel limits and they work just fine. It does not draw stupid amounts of power and temperatures stay at manageable, healthy levels. I do not believe in forcibly limiting the power levels like Nvidia does because I am still a tinkerer and overclocker at heart. There is still a place for unlimited power draw but that place is not in supposedly stock systems with small factory coolers and hell, even most tower coolers are pushing it considering how hot and power hungry these Intel cpu's can be.
 
MaZa said:
In this case AIB's are clearly wrong. Intel has clearly defined the max safe operating conditions but instead of following them motherboard manufacturers by default allow those CPU's run at unlimited power draw, which pushes temperatures to the max until the CPU temperature throttles. That is not healthy! As Jays2cents pointed out, changing one setting in the BIOS enforces the Intel limits and they work just fine. It does not draw stupid amounts of power and temperatures stay at manageable, healthy levels. I do not believe in forcibly limiting the power levels like Nvidia does because I am still a tinkerer and overclocker at heart. There is still a place for unlimited power draw but that place is not in supposedly stock systems with small factory coolers and hell, even most tower coolers are pushing it considering how hot and power hungry these Intel cpu's can be.
Click to expand...
For sure, I don't mind that boards CAN disable the limits. That's long been a thing with enthusiast boards is you can push limits, often beyond what is reasonable. However it shouldn't be that way by DEFAULT. The default should always be manufacturer's recommendations, then let the user disable those if they choose, and accept the consequences.

If I OC a CPU and I'm getting instability and crashes, I'm ok with that. I mean I don't love it, but hey, I was pushing the limits, looks like maybe I pushed them too far. But if my shit is crashing stock? Now I'm mad, that shouldn't happen. It should be an informed risk, not some shit that just happens and you don't know about it.
 
Sycraft said:
For sure, I don't mind that boards CAN disable the limits. That's long been a thing with enthusiast boards is you can push limits, often beyond what is reasonable. However it shouldn't be that way by DEFAULT. The default should always be manufacturer's recommendations, then let the user disable those if they choose, and accept the consequences.

If I OC a CPU and I'm getting instability and crashes, I'm ok with that. I mean I don't love it, but hey, I was pushing the limits, looks like maybe I pushed them too far. But if my shit is crashing stock? Now I'm mad, that shouldn't happen. It should be an informed risk, not some shit that just happens and you don't know about it.
Click to expand...

Precisely my point. The option to go over limits should be there but it should be an OPTION, not factory default.
 
LOCO LAPTOP said:
To give the story on my issue and what intel has sent.
On 4/8/2024 I opened a ticket at Intel stating:


On 4/12/2024 after no response from Intel:


They responded a few hours later:



Now I was a bit puzzled, but I tried and it passed just fine, besides the fact that Cinebench gave me a 27,394 instead of the 37,364 that it was getting. I did some more testing and found the only thing that made a difference was "Intel Fail Safe" The power limits could be set to 300w like before and it ran perfectly fine for days. Downside with that is, the CPU voltage is showing was showing 1.607v in the Bios, so I asked intel about this.



They responded with:


Then I said:


Now I won't make this post much longer with the back and forth but because the CPU is working 'as design" which I guess it is, they simply won't replace it as of now. However I am concern about the possibly that it degraded to the point that maybe another year or so I will start having issues again. So I'm stuck with something I don't trust and would be a paper weight.


Anyway I tried the BIOS update, these are the settings the Asus TUF Z790-PLUS set for the intel base line.

View attachment 649717

After two hours, it appears to be stable. 33499 Cinebench score, which is lower than the factory settings from 37,364, it pulls a max of 227 watts now while it is set to 253w PL1/2 and the base ratio is now 51 (5.1Ghz) instead of 55 (5.5Ghz). I'm still testing and playing around with it. But IMO this is a band aid.
Click to expand...

Update to this. Last response from Intel:

Good day! and sorry for the late response. From the information I gathered from our engineers, by running “Intel Fail Safe," increasing Dynamic Vcore, or adjusting LLC settings, is it either Intel default behavior or within specification? It is not increasing the voltage beyond specification; it is setting the behavior back to what Intel’s specifications say it should be. This will not “burn out” the processor and will not impact the CPU lifecycle. Our specifications are designed to allow the processor to run by design for the life of the processor. ASUS’ “Auto” setting for SVID is undervolting the processor.
Click to expand...

So basically Asus, MSI, Gigabyte, ect... are/were undervolting the 13 and 14th gen CPUs? And says that 1.6Vcore is normal.
 
LOCO LAPTOP said:
Update to this. Last response from Intel:



So basically Asus, MSI, Gigabyte, ect... are/were undervolting the 13 and 14th gen CPUs? And says that 1.6Vcore is normal.
Click to expand...
Yeah, I saw a spec sheet that said those high end skus top out at 1.7, iirc - crazy.
 
This is no different that Diesel gate with VW. Do WHATEVER you need to do to win. Even if it means doing something ridiculous like zero power limits. It's sad that Intel's last place of dominance was on the upper end desktop CPUs. Now? They have even tarnished that.

If Dell Latitude's came in AMD I wouldn't even buy Intel at work at this point.
 
And here I thought I was slowly switching back to Intel.... (sigh)

Updated the Asus BIOS's on both my Intel boxes. Good to see that my results seem normal compared to everyone else's updates. Slightly slower and still stable at the same temps as before the update.

Sounds fair.....
 
You must log in or register to reply here.
Back
Top