News Leaked internal reports allegedly reveal Intel's instability problems are not over — elevated voltages could be only one of the causes of CPU crashing

[Admin]

Internal reports from alleged Intel sources reveal that Intel has not told the entire story about Raptor Lake instability. Elevated voltages are only part of the problem, and other root causes are being investigated.

Leaked internal reports allegedly reveal Intel's instability problems are not over — elevated voltages could be only one of the causes of CPU crashing : Read more

 

[rluker5]

I think 1.55v is erring on the side of performance. 1.5 max requested would probably be better and 1.4v would be safe so long as the motherboard weren't cranking the volts with some auto tuning.

I also don't think it is the only issue as the unlimited power and poor motherboard "stock" behaviors are still out there. That and motherboards can give more volts than requested by the CPU. Really need to get motherboard vendors in line.

 

[HyperMatrix]

Using the latest bios with the intel default specs shows the predicted voltage for 6GHz, which is the stock boost clock for the 14900K, is 1.609V. lol. That is nuts. In comparison, 5.3GHz is just 1.22V and 5.7GHz is 1.44V.

I decided to set a per-core max multiplier of x57 and disabled TVB. So it works like normal but doesn’t try to pull over 1.6V for an extra 300MHz in lightly threaded applications.

The “big numbers” marketing is what’s leading to all these problems. All core clocks are set to 5.7GHz anyway. 6GHz+ is only when a couple cores are being used. And it’s really not worth such a big voltage spike. Should have just advertised it as 5.7 or 5.8GHz and left it at that.

 

[diminishedfifth]

I dunno about it just being a voltage issue. Didn't Intel and MB manufacturers decrease the volts in their updates and still the issue persisted?

 

[greymaterial]

I dunno about it just being a voltage issue. Didn't Intel and MB manufacturers decrease the volts in their updates and still the issue persisted?

the newer chips like rpl has power gates located inside the core for each core, therefore a low latency when the core needs high freq requesting high voltage will happen within the core itself. MB bios is basically setting the range of voltage and power allowed to supply to the chip externally, where the power gates in core is controlled by microcode only.

 

[ThatMouse]

Shameful article. No mention of the oxidation problem.

 

[JRStern]

Is this only or mostly on unlocked chips, or other identifiable set of product numbers or features?

 

[-Fran-]

Intel not being truthful?

kermitsurprisedface.gif

Regards.

 

[d0x360]

I dunno about it just being a voltage issue. Didn't Intel and MB manufacturers decrease the volts in their updates and still the issue persisted?

It's not, it also an issue with the interconnects (probably between ecores - cache) that is being damaged or was always defective and has been since 13th gen.

Probably

The voltage issues might have made the problem much worse but it's probably not the root cause it's a bandaid to try and slow the failure rate but almost all if not all of these chips are going to self destruct earlier than they should.

 

[bit_user]

Who was the purported recipient of this memo? The way it's written like "Intel this", "Intel that", and signed "--Intel--" makes it sound like it was something sent out to some of their partners or investors.

 

[bit_user]

I decided to set a per-core max multiplier of x57 and disabled TVB. So it works like normal but doesn’t try to pull over 1.6V for an extra 300MHz in lightly threaded applications.

The “big numbers” marketing is what’s leading to all these problems. All core clocks are set to 5.7GHz anyway. 6GHz+ is only when a couple cores are being used. And it’s really not worth such a big voltage spike. Should have just advertised it as 5.7 or 5.8GHz and left it at that.

That sounds wise, but perhaps not sufficient. The article cites:

"This unsafe minimum operating voltage affects the chip even at idle conditions. Intel has observed sporadic elevated voltages when Raptor Lake chips resume low-power states to execute background operations before entering a low-power state again."
​

This could mean there are spikes not affected by that multiplier and that are too quick for you to see, but perhaps can still cause damage? Plus, the continued hunt for a root cause suggests it's perhaps not all just about voltage.

 

[Amdlova]

Installed latest bios from asrock for the H670 PG riptide (This name sux) for now my system is safe
the voltage goes from 0.669V to 1.127V I think this voltage is safe for 4.1ghz

 

[bit_user]

Shameful article. No mention of the oxidation problem.

The IgorsLab article includes a snippet about that:

"The rumour about possible oxidation was also subsequently picked up on Reddit and ruled out as the cause:

We can confirm that the via Oxidation manufacturing issue affected some early Intel Core 13th Gen desktop processors. However, the issue was root caused and addressed with manufacturing improvements and screens in 2023. We have also looked at it from the instability reports on Intel Core 13th Gen desktop processors and the analysis to-date has determined that only a small number of instability reports can be connected to the manufacturing issue.

For the Instability issue, we are delivering a microcode patch which addresses exposure to elevated voltages which is a key element of the Instability issue. We are currently validating the microcode patch to ensure the instability issues for 13th/14th Gen are addressed.

LexHoyos42 (Reddit via community.intel)

I don't know if that's why the Toms article skipped it, or if they didn't mention it just because it wasn't part of the leaked memo that was the main story.

 

[bit_user]

Is this only or mostly on unlocked chips, or other identifiable set of product numbers or features?

Seems to be mainly K-series models from i7 and i9 CPUs in Gen 13 and 14, but I've also seen mention of i5 K-series and possibly the i9-14900T.

I've not heard of cases where non-K are affected, but that doesn't mean they aren't.

 

[rluker5]

the newer chips like rpl has power gates located inside the core for each core, therefore a low latency when the core needs high freq requesting high voltage will happen within the core itself. MB bios is basically setting the range of voltage and power allowed to supply to the chip externally, where the power gates in core is controlled by microcode only.

You are thinking of FIVR. Gating is more on and off to shut things down for power savings, it isn't a gated FIVR. Here's a better explanation from: https://skatterbencher.com/2021/11/04/alder-lake-overclocking-whats-new/
"

• First, the motherboard’s BIOS tells the processor the current loadline characteristics via AC DC loadline values.
• Then, the CPU will request a voltage from the voltage controller based on its own programmed V/f curve as well as the motherboard loadline characteristics.
• Finally, the voltage that reaches the CPU is the requested voltage minus any undershoot or overshoot from the VRM loadline." This doesn't include any adjustments ofc, "stock" or user implemented. The motherboard listens to the CPU request then sends what it wants and the CPU takes it.

 

[TheHerald]

Using the latest bios with the intel default specs shows the predicted voltage for 6GHz, which is the stock boost clock for the 14900K, is 1.609V. lol. That is nuts. In comparison, 5.3GHz is just 1.22V and 5.7GHz is 1.44V.

I decided to set a per-core max multiplier of x57 and disabled TVB. So it works like normal but doesn’t try to pull over 1.6V for an extra 300MHz in lightly threaded applications.

The “big numbers” marketing is what’s leading to all these problems. All core clocks are set to 5.7GHz anyway. 6GHz+ is only when a couple cores are being used. And it’s really not worth such a big voltage spike. Should have just advertised it as 5.7 or 5.8GHz and left it at that.

Yes yes and yes again. The last few hundred mhz require a silly amount of voltage for what, a 3% performance increase? Im not even running the 14900k at 5.7, im running it at 5.5 all core. The difference in a CBR23 run is 42k for 5.7 and 41.2k for 5.5. Basically it requires around 80-100 extra watts for a 1.8% performance increase. Absolutely bonkers

 

[bit_user]

Yes yes and yes again. The last few hundred mhz require a silly amount of voltage for what, a 3% performance increase?

Where are you getting that 3% number? I've seen benchmarks where Golden Cove performance is fairly linear with clockspeed, even at the top end of the scale. If it were perfectly linear, that reduction would amount to about a 5% performance loss. Of course it's not, but probably at least 4%.

 

[TheHerald]

Where are you getting that 3% number? I've seen benchmarks where Golden Cove performance is fairly linear with clockspeed, even at the top end of the scale. If it were perfectly linear, that reduction would amount to about a 5% performance loss. Of course it's not, but probably at least 4%.

A drop from 5.7 to 5.5ghz is a 3.5% difference in clockspeeds. Even if the drop in performance is linear the cpu gets half of its performance from the ecores which aren't affected by dropping pcores clockspeeds.

 

[bit_user]

A drop from 5.7 to 5.5ghz is a 3.5% difference in clockspeeds.

Oh, I was computing 5.7 vs. 6.0, which were the speeds @HyperMatrix mentioned.

 

[jp7189]

You are thinking of FIVR. Gating is more on and off to shut things down for power savings, it isn't a gated FIVR. Here's a better explanation from: https://skatterbencher.com/2021/11/04/alder-lake-overclocking-whats-new/
"

• First, the motherboard’s BIOS tells the processor the current loadline characteristics via AC DC loadline values.
• Then, the CPU will request a voltage from the voltage controller based on its own programmed V/f curve as well as the motherboard loadline characteristics.
• Finally, the voltage that reaches the CPU is the requested voltage minus any undershoot or overshoot from the VRM loadline." This doesn't include any adjustments ofc, "stock" or user implemented. The motherboard listens to the CPU request then sends what it wants and the CPU takes it.

It seems to me LLC profiles are up to motherboard vendors and largely out of Intel's control. Extreme LLC profiles can overshoot requested voltage by a good margin.

 

[jp7189]

Yes yes and yes again. The last few hundred mhz require a silly amount of voltage for what, a 3% performance increase? Im not even running the 14900k at 5.7, im running it at 5.5 all core. The difference in a CBR23 run is 42k for 5.7 and 41.2k for 5.5. Basically it requires around 80-100 extra watts for a 1.8% performance increase. Absolutely bonkers

Monster clock speeds is the message Intel marketing needs. There can be no doubt Intel is over reaching due to competitive pressures from AMD.

 

[rluker5]

Since it seems very high voltage seems to be largely responsible for the degradation part of the instability issue, it would be nice if a tech site gave voltage control advice. Any tech site.
Just to help any readers that may have concerns.

Some sites even recommended boosting volts with failsafe settings (not this one though). Almost like they want to make the problem worse for clicks.

1.6 volts doesn't sound unheard of with RPL. Sustained exposure would degrade any chip made in the last 10 years. It isn't hard to get to 1.4v. You can even game at .75v on Raptor Lake. People probably should take measures to avoid letting their CPU run at, or frequently spike to 1.6+ volts.