News Intel issues statement about CPU crashes, blames motherboard makers — BIOSes disable thermal and power protection, causing issues

[t3t4]

The CPU manufacture has zero control on the power to the CPU, they can only write the specification then ask the motherboard makers to follow it. Voltage control and power distribution is something the motherboard does and why they like to use "X/Y/Z phase power distribution etc.." as selling points.

Uh, yes, which is why I said......

I blame the mother board manufacturers for doing an obviously stupid thing!

 

[Phaaze88]

You buy an unlocked cpu and an unlocked motherboard, you are on your own. If you don't know what the hell you are doing stick to locked parts with xmp off.

'Unlocked parts = higher performing than locked' is stubbornly ingrained in the community.
They can be so close together these days...
I can only laugh when reading that some of the 'don't really know what they're doing' unknowingly end up worse off than if they'd just bought the locked ones.

 

[thestryker]

'Unlocked parts = higher performing than locked' is stubbornly ingrained in the community.

With Intel this is actually true as the clockspeeds under stock operation are higher on K SKUs than non. Now you could argue the difference is largely negligible, but it doesn't change the fact that they are.

 

[Phaaze88]

With Intel this is actually true as the clockspeeds under stock operation are higher on K SKUs than non. Now you could argue the difference is largely negligible, but it doesn't change the fact that they are.

Not by much, and such a small difference 'shines' most on benchmarks, where my e-peen should be as big as the e-peen of the guy in the e-pr0n videos, and if it's not, something's wrong with mine.

 

[craigss]

The reputable thing for AMD to do would have been to do a full recall, their cpus were blowing up, it was a faulty product, instead they locked down your unlocked cpu, they took away part of what you paid good money for and you are happy about them stealing from you.

Also what AMD did was to blame all of the OEMs for not using in spec vsoc settings, even though those where the settings they used for years before that so AMD knew that any new CPU would have to be able to use those settings.
AMD cpus also didn't have any internal safeguards, no, actually that was what made it a faulty product, the safety measure they had was burning out because it had no safety measure to protect it from too high voltages...
so how are they better than intel?!
https://www.tomshardware.com/news/a...use-identified-expo-and-soc-voltages-to-blame

Great you have got an ignore

 

[-Fran-]

They already do that. Intel provides the chip specifications to the motherboard manufacturers and that's where it ends. It's the motherboard manufacturers job to implement those specifications. Motherboard manufacturers doing silly stuff is the fault of the motherboard manufacturer, not the chip maker.

People's brand hate needs to stop. The reason most all the board manufactures do this is to "keep up" with each other on the youtube reviews. Reviewers love benchmarks, "board A is 5% faster then board B" then everyone wants to buy board A. Since almost everything that used to be board specific has moved into IMC's and dedicated vendor chipsets, there isn't much left for motherboard vendors to differentiate themselves with so they have started to do overclocking by default.

And they follow the documents for sure. As far as Intel has provided them, all these CPUs are working "within spec".

That's the whole point:

Ian Cutress: One of the things we’ve seen with the parts that we review is that we’re taking consumer or workstation level motherboards from the likes of ASUS, ASRock, and such, and they are implementing their own values for that PL2 limit and also the turbo window – they might be pushing these values up until the maximum they can go, such as a (maximum) limit of 999 W for 4096 seconds. From your opinion, does this distort how we do reviews because it necessarily means that they are running out of Intel defined spec?

Guy Therien: Even with those values, you're not running out of spec, I want to make very clear – you’re running in spec, but you are getting higher turbo duration.

We’re going to be very crisp in our definition of what the difference between in-spec and out-of-spec is. There is an overclocking 'bit'/flag on our processors. Any change that requires you to set that overclocking bit to enable overclocking is considered out-of-spec operation. So if the motherboard manufacturer leaves a processor with its regular turbo values, but states that the power limit is 999W, that does not require a change in the overclocking bit, so it is in-spec.

https://www.anandtech.com/show/1458...ng-an-interview-with-intel-fellow-guy-therien

Regards.

 

[35below0]

Once again, non-K versions of popular intels often either cost more or are not available.

 

[Phaaze88]

Once again, non-K versions of popular intels often either cost more or are not available.

They are further refined. Fewer impurities. Even better power efficiency.
They are what the K chips should've been; but getting those good yields has gotten more difficult, so more dies have been tossed in K SKU bin.

 

[35below0]

They are further refined. Fewer impurities. Even better power efficiency.
They are what the K chips should've been; but getting those good yields has gotten more difficult, so more dies have been tossed in K SKU bin.

They were also released later, correct? The Ks were first to market? (13/14 gen)

But my point was that market prices and availability affect the decision which CPU to buy.
Glowing reviews of how much potential there is in 13th gen i5, i7 and i9 CPUs, and their rankings on "best gaming/productivity/etc" CPU lists also helps drive customers to them. Not just a blind desire to OC the hell out of them.

 

[bit_user]

Firstly, Intel has zero control over what Motherboard manufacturers do in the BIOS, zero.

When it comes to power limits, those are implemented in firmware, inside the CPU. Intel could enforce hard upper limits on them, if they wanted to.

I'm not sure how many other settings are like this, but I'd guess there are probably others.

They already do that. Intel provides the chip specifications to the motherboard manufacturers and that's where it ends.

Intel also gets to define what constitutes "overclocking" (i.e. what voids your warranty). In turn, that constrains what motherboards can do by default, since none of them should have out-of-the-box defaults that are CPU warranty-voiding.

The CPU manufacture has zero control on the power to the CPU,

The knob they use to control package power is frequency. Intel CPUs will throttle core clock frequencies, in order to stay within the configured power limits.

 

[TheHerald]

They were also released later, correct? The Ks were first to market? (13/14 gen)

But my point was that market prices and availability affect the decision which CPU to buy.
Glowing reviews of how much potential there is in 13th gen i5, i7 and i9 CPUs, and their rankings on "best gaming/productivity/etc" CPU lists also helps drive customers to them. Not just a blind desire to OC the hell out of them.

It's a chicken and egg problem, most people buy the K cpus, therefore shops don't stock the non k. I'd feel much more confident suggesting a non K to a friend that doesn't have any idea what they are doing.

 

[35below0]

It's a chicken and egg problem, most people buy the K cpus, therefore shops don't stock the non k. I'd feel much more confident suggesting a non K to a friend that doesn't have any idea what they are doing.

Well i can't disagree. I would have probably taken the 13600K for myself, but had i gone for a different CPU it's 50/50 i would have opted for the non-K. Had one been available.

Even the 13600K i bought... turned out to be unneccessary. I could have gone with a non-K.

Ultimately, if i had a more clear understanding of what the motherboard will do to an intel, i would be able to suggest CPUs more accurately. As it is, i mostly go by value for money and what's available.

 

[-Fran-]

This is an interesting read from Igor's Lab: https://www.igorslab.de/en/crash-la...d-them-and-disregards-its-own-specifications/

Regards.

 

[TheHerald]

Well i can't disagree. I would have probably taken the 13600K for myself, but had i gone for a different CPU it's 50/50 i would have opted for the non-K. Had one been available.

Even the 13600K i bought... turned out to be unneccessary. I could have gone with a non-K.

Ultimately, if i had a more clear understanding of what the motherboard will do to an intel, i would be able to suggest CPUs more accurately. As it is, i mostly go by value for money and what's available.

The i5 range is generally fine, you can't push that high wattage into them cause they will thermal throttle quite easily due to the lower amount of cores, so even if the motherboard goes full kamehameha, it's really hard to push the 13600k above 200-220w without hitting thermal limits. On the other hand, I can push 330watts on a 13/14900k - and I can actually sustain 280w for 30+ minutes on a single tower air cooler. So degrading is almost impossible for the i5s, but fairly easy for the bigger chips.

Intel just messed up completely - cause since reviewers are now focusing mostly on efficiency, their out of the box behavior doesn't show that in most segments, Intel has the more efficient products. By a lot - but they keep pushing high amounts of power for minimal gains. A 13600k - at launch - limited to 125w, would be much more efficient than the equally priced 7600x - and also much faster, actually competing and mostly beating even the much more expensive R7 7700x. But nope, they pushed it to 180w or whatever for no freaking reason.

I remember with the old corona benchmark (although it might still apply with the new one) a 13600k limited to 77w (with no undervolting) was faster than the fastest 7700x on the leaderboard. For those who don't understand, the top of the corona leaderboard is using ln2s and crap. Let that sink in for a second

 

[bit_user]

Intel just messed up completely - cause since reviewers are now focusing mostly on efficiency, their out of the box behavior doesn't show that in most segments, Intel has the more efficient products. By a lot - but they keep pushing high amounts of power for minimal gains. A 13600k - at launch - limited to 125w, would be much more efficient than the equally priced 7600x - and also much faster, actually competing and mostly beating even the much more expensive R7 7700x.

The R7 7700 (non-X) is currently selling for $281. When both limited to 65 W, it's more efficient than the i5-13600K (currently $283):

​

Yes, at 125 W, the i5-13600K does pull ahead of the R7 7700X, but it's not the sort of bloodbath you suggest.

BTW, the R5 7600X is currently selling for $208, putting it in a different bracket than the i5-13600K.

 

[TheHerald]

The R7 7700 (non-X) is currently selling for $281. When both limited to 65 W, it's more efficient than the i5-13600K:

​

​

Yes, at 125 W, the i5-13600K does pull ahead of the R7 7700X, but it's not the sort of bloodbath you suggest.

But the 7700x isn't limited to 65w (I mean, at stock), and it was a more expensive product. If Intel just matched the power limits, it would have a cheaper CPU that is faster in both MT and ST performance while using less power. Instead they went full kamehameha mode.

 

[bit_user]

But the 7700x isn't limited to 65w (I mean, at stock), and it was a more expensive product.

The non-X version is both rated at 65 W (88 W PPT) and cheaper than the 7700X.

If Intel just matched the power limits,

Not if you match against the non-X R7 7700, which is a couple bucks cheaper and more efficient at 65 W.

it would have a cheaper CPU that is faster in ... ST performance

Granted, Raptor Cove has faster ST performance, but I'm not sure it has a better ST perf/W curve.

 

[TheHerald]

The non-X version is both rated at 65 W (88 W PPT) and cheaper than the 7700X.


Not if you match against the non-X R7 7700, which is a couple bucks cheaper and more efficient at 65 W.


Granted, Raptor Cove has faster ST performance, but I'm not sure it has a better ST perf/W curve.

65w tdp isn't the same as 65w power draw. Also there was no 7700 at launch anyways.

According to techpowerUps review intel has a huge efficiency lead in st tasks. Like, up to 80% or something

 

[bit_user]

65w tdp isn't the same as 65w power draw. Also there was no 7700 at launch anyways.

The PPT for the 7700 is 88 W. At that power, the above data shows efficiency to be equal.

As for the launch timing of the 7700, I don't get why the fact that it didn't launch until Jan. 2023 is relevant, since you're talking in the abstract about efficiency and hypothetical products.

According to techpowerUps review intel has a huge efficiency lead in st tasks. Like, up to 80% or something

​

Um, not exactly. The stock R7 7700 has the i5-13600K easily beat on ST efficiency. Zen 4 is pretty efficient, if you keep it in its sweet spot.

 

[jalyst]

This is an interesting read from Igor's Lab: https://www.igorslab.de/en/crash-la...d-them-and-disregards-its-own-specifications/

Good read & wow that looks like a good site, first I've seen of it!

To those who've followed it on/off for ages, can you please confirm;
Is it a decent central repo. of knowledge for "PC/Semi-c" tech?

In the same ilk -dare i say 😝- as THG/TG...

 

[bit_user]

Good read & wow that looks like a good site, first I've seen of it!

To those who've followed it on/off for ages, can you please confirm;
Is it a decent central repo. of knowledge for "PC/Semi-c" tech?

In the same ilk -dare i say 😝- as THG/TG...

Igor used to write reviews for this site, in fact! I found his hardware teardowns & testbench analysis to be particularly comprehensive.

https://www.tomshardware.com/author/igor-wallossek
​

I haven't read much on his own site, but here's what a someone recently wrote about it:

https://forums.tomshardware.com/thr...laimed-to-be-fraudulent.3843099/post-23250072
​

 

[TheHerald]

The PPT for the 7700 is 88 W. At that power, the above data shows efficiency to be equal.

As for the launch timing of the 7700, I don't get why the fact that it didn't launch until Jan. 2023 is relevant, since you're talking in the abstract about efficiency and hypothetical products.

If launch dates aren't important then you can compare the 7700 to skylake, right? Btw, I highly doubt that even at 88w it's as fast as a 13600k - but ill let this slide.

​

Um, not exactly. The stock R7 7700 has the i5-13600K easily beat on ST efficiency. Zen 4 is pretty efficient, if you keep it in its sweet spot.

Again, you are looking at efficiency completely wrong. The 7700 doesn't have better ST efficiency, it's just slower. You need to normalize for either performance or power. The below graph demonstrates the point much better. The closest Intel CPU in terms of ST performance to the 7800x 3d is the 13400f (1796 vs 1817 score), and the Intel part has it easily beat.

Then as you move to the more high end CPUs, the 13900k absolutely blasts the 7950x and the 7950x3d WHILE being faster, which is actually insane. It's 50% more efficient - if you match the actual performance or power the difference would probably go over 100%. In other words, the 7950x needs twice the power to match a 13900k in ST tasks.

Now if you compare the 13700k to the 7900x, the AMD part needs twice the power while also being slower. If someone looks at these results realistically they'd understand that this isn't just a generational leap, that gap would take years (lots, lots of years) to cover.

How do I know it will take years? Well look at the increase between zen 2, zen 3 and zen 4. 3 generations of CPU to go from 39,7 pts / watt to 46,2 pts / watt, while Intel's 13700k hit's 92,8 while at the same time being faster. Absolutely bonkers.

 

[TerryLaze]

You're wrong on both accounts though.

AMD did acknowledge they didn't communicate with their partners correctly and fixed that, but also stated the values were never "in spec".

Relevant quote:
"The issues are known to occur either when a user adjusts the SoC voltage manually to a value beyond 1.3V or when the motherboard firmware automatically increases the SoC voltage beyond 1.3V when an EXPO memory overclocking profile is engaged (The EXPO profile itself does not increase the SoC voltage, the board vendors assign a pre-determined value of their own to support the increased speed of the EXPO profile)".

https://www.tomshardware.com/news/a...t-on-ryzen-burnout-issues-limits-soc-voltages

Their fix was just making sure motherboard makers would not push the limits beyond what AMD themselves considers "safe" (or so I read in between the lines). As this is my interpretation, I could be wrong.

Also, the CPUs were never faulty and proof of that is everyone that could have been potentially affected is still using their CPUs just fine, no?

Do you think Intel's 13 and 14 series affected by this need to be recalled or something? Because that's what your "logical" reasoning makes me believe?

Regards.

AMDs protection mechanism burns out right away at more than 1.3V , while as the title of this thread says, bioses disable intel' protection
"BIOSes disable thermal and power protection, causing issues"

A protection that is supposed to protect your CPU from overheating from things like voltage spikes gets burned out by voltage spikes which is part of what it is supposed to protect the CPU from.

AMD has a hardware fault, the protection burns out by what it should be protecting from.
Intel has a bioses problem, oems disable the protections, that's not a hardware issue.

Our sources also added further details about the nature of the chip failures — in some cases, excessive SoC voltages destroy the chips' thermal sensors and thermal protection mechanisms, completely disabling its only means of detecting and protecting itself from overheating. As a result, the chip continues to operate without knowing its temperature or tripping the thermal protections.

AMD's modern chips often run at their thermal limits to squeeze out every last drop of performance within their safe thermal range — it isn't uncommon for them to run at 95C during normal operation — so they will automatically continue to draw more power until it dials back to remain within a safe temperature.

 

[bit_user]

If launch dates aren't important then you can compare the 7700 to skylake, right?

No, because it was technically possible for AMD to have launched the 7700 when they launched the 7700X - they simply chose not to.

Btw, I highly doubt that even at 88w it's as fast as a 13600k - but ill let this slide.

That's what the data in my plot shows. If you have other data, feel free to share it. I'm sure there are some workloads where the i5-13600K is faster, but not the multithreaded ones that ComputerBase.de chose to include in their test set.

Again, you are looking at efficiency completely wrong. The 7700 doesn't have better ST efficiency, it's just slower.

Efficiency is work per unit of energy (approximated here as average power). You can define other metrics, but efficiency allows a slower CPU to win if its power usage decreases by even greater than its speed.

You need to normalize for either performance or power. The below graph demonstrates the point much better.
...

You must not have looked at them very carefully, because the only difference is that they sorted in the reverse order. The one I quoted from the Ryzen 7700 review put the most efficient at the bottom, while the one you quoted put the most efficient at the top. Also, maybe they changed the Cinebench version, because the 7700 went from 105.2 Pts/W to 112.1 Pts/W in yours!

The closest Intel CPU in terms of ST performance to the 7800x 3d is the 13400f (1796 vs 1817 score), and the Intel part has it easily beat.

Now, you're talking about performance, not efficiency.

Then as you move to the more high end CPUs, the 13900k absolutely blasts the 7950x and the 7950x3d WHILE being faster, which is actually insane.

It's not insane - it's exactly like I said. The 7950X has higher clockspeed limits than the 7700, which allows a single threaded job to push well past the point of diminishing returns, in terms of perf/W. The 7950X also has the disadvantage of the extra compute die, which is just dead weight in any single-threaded efficiency comparison.

I'm sure you can close much of the gap between the 7950X's efficiency and that of the 7700, if you just limited the former to a single-core limit of 5.3 GHz, instead of the 5.7 GHz turbo limit it has by default. This is what one should do, if they wanted the multithreaded performance of the 7950X, but didn't want to take such a hit on single-threaded efficiency.

I said this before: Zen 4 is efficient in its sweet spot. The 7950X's boost limit is well outside that sweet spot.

It's 50% more efficient - if you match the actual performance or power the difference would probably go over 100%. In other words, the 7950x needs twice the power to match a 13900k in ST tasks.

Your conceptional model is completely broken, here. Any overclocker would tell you that power increases needed to push higher frequencies grow at a nonlinear rate. You can't use an efficiency ratio to predict how much power one or the other CPU would need at a given performance level. It might be that no amount of power could extract equivalent performance from the 7950X, at a given task.

If someone looks at these results realistically they'd understand that this isn't just a generational leap, that gap would take years (lots, lots of years) to cover.

Your analysis is also deeply flawed in this part. AMD simply chose to make Zen 4 smaller. Zen 4 is primarily targeted at servers and laptops. In such cases, you don't optimize for high clockspeeds, because that requires a lot of die area and always burns a lot of power. Instead, it's targeted to be efficient in a lower window, but it's also smaller. This enables them to pack more full-size cores (or half-size Zen 4C cores, which are basically identical except for half the L3 cache), achieving strong multithreaded performance and efficiency.

In contrast, Intel opted to make Golden Cove big and complex, but added the small and simple Gracemont cores to balance it out. That hybrid strategy works okay for desktops and laptops, but not servers.

Attribute	Zen 4	Zen 3	Golden Cove	Comments
Reorder Buffer	320	256	512	Each entry on Zen 4 can hold 4 NOPs. Actual capacity confirmed using a mix of instructions
Integer Register File	224	192	280
Flags Register File	238	122	Tied to Integer Registers	AMD started renaming the flags register separately in Zen 3
FP/Vector Register File	192	160	332	Zen 4 extends vector registers to 512-bit
AVX-512 Mask Register File	52 measured + 16 non-speculative	N/A	(152 measured via MMX)	Since SKL, Intel uses one RF for MMX/x87 and AVX-512 mask registers However Golden Cove does not officially support AVX-512
Load Queue	88 (136 measured)	72 (116 measured)	192	All Zen generations can have more loads in flight than AMD’s documentation and slides suggest. Intel and AMD have different load queue implementations
Store Queue	64	64	114	A bit small on Zen 4, would have been nice to see an increase here
Branch Order Buffer	62 Taken 118 Not Taken	48 Taken 117 Not Taken	128

Source: https://chipsandcheese.com/2022/11/05/amds-zen-4-part-1-frontend-and-execution-engine/
​

How do I know it will take years?

You don't. At this point, you have no idea what's going on. You're piling invalid assumptions atop more invalid assumptions.

...
Absolutely bonkers.

What's absolutely bonkers is the real performance and efficiency advantages of Naples and Bergamo over Sapphire Rapids (or even Emerald Rapids). This is what AMD designed Zen 4 to do and it's why AMD's datacenter marketshare keeps growing. They care about that much more than the desktop performance segment.

Source: https://www.phoronix.com/review/amd-epyc-9754-bergamo
​

 

[-Fran-]

AMDs protection mechanism burns out right away at more than 1.3V , while as the title of this thread says, bioses disable intel' protection
"BIOSes disable thermal and power protection, causing issues"

A protection that is supposed to protect your CPU from overheating from things like voltage spikes gets burned out by voltage spikes which is part of what it is supposed to protect the CPU from.

AMD has a hardware fault, the protection burns out by what it should be protecting from.
Intel has a bioses problem, oems disable the protections, that's not a hardware issue.

And you should never redline an engine for extended periods of time, but you still get those redlines in the tachometer for you to go ahead. Probably not the best analogy, but take it as AMD saying "hey, there's the redline, don't cross it, please" and then the AIBs go and cross it. They made their hardware with tolerances in mind, which if you tell someone about them, you'd expect them to follow them. That is not a hardware design error, but (almost literally) AIBs ignoring AMD's guidance and doing what they want. Should AMD get ahead of AIBs not following their guidance? Hell no. Why should they?

Intel has a different way to handle this, it seems, and theirs seems more effective to the point they literally DO NOT CARE what the AIBs set, the CPU will always be in control (kind of ironic) unless the "OC" flag is switched, which seems no AIB does really?

It's two different causes which from the outside look the same, but aren't.

Neither AMD nor Intel need to recall anything. AIBs need to fix their bad habits and, in particular for Intel, be a tad more clear (and sincere) with their guidance.

Regards.