Intel says their 13th/14th Gen instabilities are from 'elevated operating voltages stemming from a microcode algorithm'

[Karthanon]

Quote

Based on extensive analysis of Intel Core 13th/14th Gen desktop processors returned to us due to instability issues, we have determined that elevated operating voltage is causing instability issues in some 13th/14th Gen desktop processors. Our analysis of returned processors confirms that the elevated operating voltage is stemming from a microcode algorithm resulting in incorrect voltage requests to the processor.

 

The news release/report came from Overclock3d.net.  So, Intel says the issue is from elevated operating voltages due to microcode, so microcode fixes shall kill the problem - is that for all *new* chips, or have CPU SKU's that were affected running under the bad algorithm received possible long-term damage?

Based on Intel's past track history (user-created issue, hurr durr) of blaming the problem on users (or anyone else but themselves), my response would be to wait until people get the new microcode and test it over a 3-6 month period...so in other words, not buy a 13th/14th gen.

 

Sources

https://overclock3d.net/news/cpu_mainboard/intel-uncovers-the-root-cause-of-13th-14th-gen-raptor-lake-cpu-instability-fix-incoming/
https://community.intel.com/t5/Processors/July-2024-Update-on-Instability-Reports-on-Intel-Core-13th-and/m-p/1617113#M74792

[Tetras]

4 minutes ago, Karthanon said:

is that for all *new* chips, or have CPU SKU's that were affected running under the bad algorithm received possible long-term damage?

Elevated operating voltage would suggest they mean permanent degradation, so I doubt the microcode update would fix the worst examples, though it may help prevent further degradation.

[podkall]

38 minutes ago, Tetras said:

Elevated operating voltage would suggest they mean permanent degradation, so I doubt the microcode update would fix the worst examples, though it may help prevent further degradation.

OC voltage raise might not always be cranking at full,

 

but microcode algorithm voltages?

 

I'm making stuff up, but to me that sounds like accelerated degradation,

 

like components running slightly-moderately-considerably higher wattage, than they need, but not because of OC settings...

[Tetras]

5 minutes ago, podkall said:

like components running slightly-moderately-considerably higher wattage, than they need, but not because of OC settings...

My suspicion is that what Intel are referring to is the TVB boost algorithm which was leaked as a cause awhile ago, which is arguably a form of 'stock' OC. Though, it was noticed quite awhile ago that a lot of K CPUs were running very high and possibly unsafe voltages.

[LinusTech]

This might explain why this flew under the radar on 13th gen for so long and is suddenly cropping up now.

 

That's the piece that hasn't made any sense to me up until this point.

[vanpuffelen]

1 minute ago, Tetras said:

My suspicion is that what Intel are referring to is the TVB boost algorithm which was leaked as a cause awhile ago, which is arguably a form of 'stock' OC. Though, it was noticed quite awhile ago that a lot of K CPUs were running very high and possibly unsafe voltages.

My gut thinks this TVB thingamabob is related to my old i7-3770K automatically getting OC'd to 3.90GHz, even though it had a shoddy Intel stock cooler connected to it. 

[Dabombinable]

7 minutes ago, vanpuffelen said:

My gut thinks this TVB thingamabob is related to my old i7-3770K automatically getting OC'd to 3.90GHz, even though it had a shoddy Intel stock cooler connected to it. 

Your motherboard probably had a setting enabled by default that ran it at full turbo across all cores. That's what the Asrock board currently in my brother's PC (i7 3770S) does right now. It is the same for the PC I built for a friend with my old i7 4790K - board by default runs all cores at 4.4GHz.

TVB is only relatively recent - your i7 just has oldschool turbo. Different set boost clocks depending on the load of 1, 2 and 4 cores.

[porina]

Random thoughts:

Voltage demands can vary a lot, so it doesn't necessarily mean the wrong voltage was present all the time.

Depending on the magnitude of error and how often it happened, it does sound like it could cause degradation.

Mobo makers pushing beyond spec certainly wouldn't have helped at all, so that still needed to be addressed regardless.

Have all microcode for affected CPUs been affected since release, or was it introduced with a mid-lifespan update?

 

Presuming the '900 parts are most affected since they're sailing closest to the wind. '700 and lower get progressively less likely to be affected as they might have a bit more margin to play with.

[vanpuffelen]

1 minute ago, Dabombinable said:

Your motherboard probably had a setting enabled by default that ran it at full turbo across all cores. That's what the Asrock board currently in my brother's PC (i7 3770S) does right now. It is the same for the PC I built for a friend with my old i7 4790K - board by default runs all cores at 4.4GHz.

It was mounted on an MSI B75MA-P45. It was the default setting. Painful.

[podkall]

12 minutes ago, LinusTech said:

This might explain why this flew under the radar on 13th gen for so long and is suddenly cropping up now.

 

That's the piece that hasn't made any sense to me up until this point.

omg it's Sebastian himself

 

13 minutes ago, Tetras said:

My suspicion is that what Intel are referring to is the TVB boost algorithm which was leaked as a cause awhile ago, which is arguably a form of 'stock' OC. Though, it was noticed quite awhile ago that a lot of K CPUs were running very high and possibly unsafe voltages.

so the "unlocked" ones huh?

[Tetras]

Just now, podkall said:

so the "unlocked" ones huh?

The non-K 13900/14900 do have TVB so presumably they'd be suspectable too if Intel is referring to the TVB bug leaked earlier, but I don't think any enthusiasts bothered monitoring/buying any non-K CPUs anyway, so I don't know if they generally had high voltages too.

[SpookyCitrus]

@Karthanon @podkall @LinusTech

 

So my 14900K is one of the affected units, I upgraded from a 12700K to it a month after release. It was fine for three months then whamo bamo, Cinebench no longer running, DX12 issues with my 4090 being "out of video memory", etc. I used the Intel XTU fix to lower the performance core ratio from 57x to 55x and it works just fine. I have only had issues when for some reason XTU stops auto applying my settings on start up when it needs an update. Would it be wise to RMA my CPU while I still have the option or wait to see if this fix, when it comes out, alleviates the issue first? To be clear mine is not one that has gotten worse over time like some others who have posted about this issue. 

[podkall]

5 minutes ago, Tetras said:

The non-K 13900/14900 do have TVB so presumably they'd be suspectable too if Intel is referring to the TVB bug leaked earlier, but I don't think any enthusiasts bothered monitoring/buying any non-K CPUs anyway, so I don't know if they generally had high voltages too.

yes, but K versions have higher TDP, no?

 

[Skiiwee29]

6 minutes ago, SpookyCitrus said:

@Karthanon @podkall @LinusTech

 

So my 14900K is one of the affected units, I upgraded from a 12700K to it a month after release. It was fine for three months then whamo bamo, Cinebench no longer running, DX12 issues with my 4090 being "out of video memory", etc. I used the Intel XTU fix to lower the performance core ratio from 57x to 55x and it works just fine. I have only had issues when for some reason XTU stops auto applying my settings on start up when it needs an update. Would it be wise to RMA my CPU while I still have the option or wait to see if this fix, when it comes out, alleviates the issue first? To be clear mine is not one that has gotten worse over time like some others who have posted about this issue. 

RMA it, its degraded. 

[Tetras]

1 minute ago, podkall said:

yes, but K versions have higher TDP, no?

True, but a lot of boards just ignore the stock PL1/PL2 and non-K CPUs don't have that setting locked.

[Lunar River]

8 minutes ago, SpookyCitrus said:

@Karthanon @podkall @LinusTech

 

So my 14900K is one of the affected units, I upgraded from a 12700K to it a month after release. It was fine for three months then whamo bamo, Cinebench no longer running, DX12 issues with my 4090 being "out of video memory", etc. I used the Intel XTU fix to lower the performance core ratio from 57x to 55x and it works just fine. I have only had issues when for some reason XTU stops auto applying my settings on start up when it needs an update. Would it be wise to RMA my CPU while I still have the option or wait to see if this fix, when it comes out, alleviates the issue first? To be clear mine is not one that has gotten worse over time like some others who have posted about this issue. 

either RMA now or you can wait to see what kind of remediattion Intel might offer now that the issue has been indentified.

[podkall]

2 minutes ago, Tetras said:

True, but a lot of boards just ignore the stock PL1/PL2 and non-K CPUs don't have that setting locked.

K versions, still mean they pump even more wattage, and hover at higher wattage even on idle, which can degrade considerably faster than on non K version most logically.

[SpookyCitrus]

Just now, Skiiwee29 said:

RMA it, its degraded. 

Will do, might be a dumb question but is there a way for me to get the serial number for it without removing the cooler? Or am I going to need to wait for my project to finish? I have a box for one in my office but I've built a couple computers with them so I'm not positive if it's the one from my work computer or one I built for a client. Figured I'd see if there was a program that showed that type of info before I pause or cancel my project and take the cooler off to get it.

[Tetras]

10 minutes ago, SpookyCitrus said:

So my 14900K is one of the affected units, I upgraded from a 12700K to it a month after release. It was fine for three months then whamo bamo, Cinebench no longer running, DX12 issues with my 4090 being "out of video memory", etc. I used the Intel XTU fix to lower the performance core ratio from 57x to 55x and it works just fine. I have only had issues when for some reason XTU stops auto applying my settings on start up when it needs an update. Would it be wise to RMA my CPU while I still have the option or wait to see if this fix, when it comes out, alleviates the issue first? To be clear mine is not one that has gotten worse over time like some others who have posted about this issue. 

If it was like that from the factory we could have said something like "Intel binned it wrong", or "your motherboard doesn't supply enough voltage", but you said it changed after 3 months which suggests your CPU is degrading. It is possible that it uses a lower voltage at 55x and will not degrade further.

 

1 minute ago, podkall said:

K versions, still mean they pump even more wattage, and hover at higher wattage even on idle, which can degrade considerably faster than on non K version most logically.

Not necessarily. If non-K CPUs are run 100% power unlocked then they can boost to whatever their max turbo clock is (usually just 1 or 2 hundred Mhz lower than the K) and if they're a poorer quality bin, then they might even need more voltage than a K CPU to achieve the same clocks.

[podkall]

1 minute ago, Tetras said:

Not necessarily. If non-K CPUs are run 100% power unlocked then they can boost to whatever their max turbo clock is (usually just 1 or 2 hundred Mhz lower than the K) and if they're a poorer quality bin, then they might even need more voltage than a K CPU to achieve the same clocks.

if they'd need more voltage, they'd be called and branded K most likely

[Tetras]

1 minute ago, podkall said:

if they'd need more voltage, they'd be called and branded K most likely

Possibly, I suppose it would depend on what the limiting factor is with the bin and how far they can scale with voltage.

 

When they're power-locked, they might hardly ever sustain those clocks across all the cores, or they'll just hit the thermal limit and throttle themselves anyway.

 

E.g. 219 watts PL2 on a 13900 non-K, I'd think is too low to sustain high enough clocks that TVB would have enough headroom to even do anything.

 

In TPU's review of the 10700 non-K, for example, when it was power unlocked it was massively faster at Blender, because it allowed it to boost.

Quote

"Just like in all our CPU reviews, we measured the maximum boost clocks of the Core i7-10700, and I have to say I'm a little bit disappointed. While the CPU is marketed with 4.8 GHz boost clock, it tops out at 4.7 GHz nearly all the time, even with a single-threaded load. Once you dial up the threads, clock speeds drop a lot when twelve or more threads are active. Intel rates the Core i7-10700 with a 2.9 GHz base frequency, an extremely conservative estimate. We typically saw around 4.2 GHz with all cores loaded."

https://www.techpowerup.com/review/intel-core-i7-10700/22.html

[Salted Spinach]

Ok then release a microcode fix for performance testing on a brand new 14900k

[StDragon]

1 hour ago, SpookyCitrus said:

Will do, might be a dumb question but is there a way for me to get the serial number for it without removing the cooler? Or am I going to need to wait for my project to finish? I have a box for one in my office but I've built a couple computers with them so I'm not positive if it's the one from my work computer or one I built for a client. Figured I'd see if there was a program that showed that type of info before I pause or cancel my project and take the cooler off to get it.

SN is laser-ed on the heat spreader or printed on the original box.

For those that have a 13th gen boxed processor (released fall of 2022) you have just over a year left on your 3 year warranty. If there's an issue after a fix is released to MB vendors, you'll want to RMA that ASAP before it runs out.

https://www.intel.com/content/www/us/en/support/articles/000024255/processors.htm

[BachChain]

Intel overvolts their processors to within an inch of it's life and they're surprised it caused problems

[starsmine]

1 minute ago, BachChain said:

Intel overvolts their processors to within an inch of it's life and they're surprised it caused problems

yes? because its within an inch.