News Ryzen Burnout? AMD Board Power Cheats May Shorten CPU Lifespan

[bit_user]

So let's say someone publishes an article with the headline "Does TJ Hooker kick puppies?".

Given that AMD is asking these motherboard vendors to cease this practice, it that's not a good analogy.

It'd be more like puppies telling the newspaper that you're kicking them, and then getting someone to build a puppy-kicking detector to alert when one of them is kicked. It's at this point that the newspaper is publishing the article asking if you're kicking puppies.

The unsupported question in the headline is not whether motherboards are cheating, but just what practical implications this has on CPU longevity and long-term performance. And it's not unsupported because it's not true, but just because we don't know. Not knowing the implications of something potentially harmful isn't a reason not to report it, especially since AMD's actions suggest there is a negative consequence.

'just asking questions' phenomenon,

It's not "just asking questions", but rather calling out the elephant in the room.

 

[bit_user]

it actually is why we have so many reviews praising Ryzen for low power draw...yeah due to mobos under-reporting certain power telemetry data.

The data reported, in this case, is not absolute power draw. Anandtech repeatedly calls it a "unit-less value".

Also, good quality reviews do not rely on the telemetry of the system under test. Next time, try reading the part where reviews discuss their testing methodology. I realize it can be a little boring, but it could save you from embarrassing yourself.

 

[bit_user]

I also wonder how HWiNFO64 determines the deviation -- is it based off some internal table for the program, or is there something else it can measure to know what the "real" power is? Because if it can figure out what the CPU is really using, shouldn't AMD's software / hardware be able to do that as well?

From what I've read, it's just reporting the raw value reported by the motherboard, except rescaling it to a percentage. That's why you need to apply a multi-core load that should max it out, in order to know if it's lying.

 

[joeblowsmynose]

Well, based on how this article was reported, it's clear that Tom's was covering it an evolving story.

On Anandtech, Ian chose to address more the underlying concern, which is also appreciated. However, it should be noted that Ian didn't cite any sources or provide any data to back up his claims.

So, Tom's is playing a more journalistic role, while Ian is playing the role of an Op/Ed by an expert. Yes, differences in coverage, but I don't think the "good cop/bad cop" analogy fits or is fair to either.


No, let's not.


People are inherently tribal, and suspicious of outsiders. However, the brain is flexible in how it draws those lines. That's the part which is learned.

I guess you don't know much about high level marketing.

 

[helper800]

I guess you don't know much about high level marketing.

Just as much as you know about motherboard monitoring methodologies, and how their statistics are reported.

 

[escksu]

Given that AMD is asking these motherboard vendors to cease this practice, it that's not a good analogy.

It'd be more like puppies telling the newspaper that you're kicking them, and then getting someone to build a puppy-kicking detector to alert when one of them is kicked. It's at this point that the newspaper is publishing the article asking if you're kicking puppies.

The unsupported question in the headline is not whether motherboards are cheating, but just what practical implications this has on CPU longevity and long-term performance. And it's not unsupported because it's not true, but just because we don't know. Not knowing the implications of something potentially harmful isn't a reason not to report it, especially since AMD's actions suggest there is a negative consequence.


It's not "just asking questions", but rather calling out the elephant in the room.

Well, TDP is a number designated by CPU manufacturers. Its not a hard and fast rule to how long your cpu will live.

Most ryzen cpus are rated 65w and 95w. U series are 15w (configurable between 10-25w) etc. Why not 70/100w or 60/90 or even 66/96w?? Why its simply because amd is following intel's tdp. The 65w limit was introduced during c2d. Its 1/2 the power of pentium D .

 

[bit_user]

I guess you don't know much about high level marketing.

You're saying that as if you have some expertise in the matter. Are you speaking from first-hand experience, or why else do you think your opinion is more valid than mine?

 

[bit_user]

Well, TDP is a number designated by CPU manufacturers. Its not a hard and fast rule to how long your cpu will live.

I think temperature is the real story, here. AMD has thermal limits, sure, but they don't expect you to run the CPU right at the max temperature for the entire warranty period.

When they spec out which cooler to ship with a CPU, or what builders & partners should use, they do so with some idea of how much power the CPU needs to dissipate. However, if the CPU is being fed more power than anticipated, the CPU could run hotter than planned, which could impact its reliability & performance over time.

Most ryzen cpus are rated 65w and 95w. ... Why its simply because amd is following intel's tdp. The 65w limit was introduced during c2d. Its 1/2 the power of pentium D .

It's an interesting question, but I think you can't draw such a straight line. If you look at Intel CPUs over time, the TDP values have been shifting around, a decent amount.

Let's just take desktop i7's for example.

Model	Year	Generation	TDP (W)
i7-860	2009	Nehalem	95
i7-2600K	2011	Sandybridge	95
i7-3770K	2012	Ivy Bridge	77
i7-4770K	2013	Haswell	84
i7-4970K	2014	Haswell-R	88
i7-6700K	2015	Skylake	91
i7-7700K	2017	Kaby Lake	91
i7-8700K	2017	Coffee Lake	95
i7-9700K	2018	Coffee Lake-R	95
i7-10700K	2020	Comet Lake	125

 

[bit_user]

65w

I was going to address this part, after the table. Well, I can't figure out how to put text below a table that's at the end of a post, so I'm writing it in another.

There does seem to be a trend of non-K 65 W i7 CPUs since the i7-6700 (Skylake). I don't know the reason for this, but perhaps it was popular with partners who wanted to be able to use the same case and PSU for whatever i7 CPU model fit that bracket. It would then be natural for AMD to offer a model which fits into the same bracket, since that would ease the burden on their partners and downstream customers to use the part.

At least, that makes a lot more sense to me than assuming AMD is blindly copying off of Intel, as if they're school kids taking a test and one of them didn't do the reading. AMD isn't stupid - there's some logic behind their actions, and I'm sure it's deeper than simply looking at whatever Intel does. That's not how you get ahead, in this industry.

 

[NightHawkRMX]

If a reviewer tested system power draw with a kilowatt or they used an amp prove on the eps plug or the cpu rather than relying on software, this misreporting would not lower the power draw reported, it would increase it.

Seems even with this hack that Ryzen draws less power than intel, since most reviewers show ryzen draws less than intel core for core by using one of the above meathod's.

 

[joeblowsmynose]

You're saying that as if you have some expertise in the matter. Are you speaking from first-hand experience, or ...

That's correct. I'm not just into tech. My first CPU was a Thunderbird 1000 - to give you an idea how old I am ... I've been around the block a few times in all sorts of different expertise. I'm not an expert marketer, but have worked closely with some marketing groups.

 

[JarredWaltonGPU]

I guess you don't know much about high level marketing.

I can say unequivocally that AnandTech and Tom's Hardware were not trying to play "good cop, bad cop" on this. It was more like: Paul saw a developing story and wrote about it on Tom's Hardware, it made some waves, and then Ian/AnandTech came in afterward and asserted that, "No, this won't break your CPU and here's why we think that's the case."

There is next to zero communication between the various Future Inc. properties. Maybe at a company retreat or whatever we call those annual get-togethers, we might chat, but there's simply no way to try and coordinate between all the brands. Tom's Hardware, Tom's Guide, Tech Radar, PC Gamer, AnandTech and more all overlap on hardware coverage, and between all the brands there's probably only about 80 hours of meetings per week.

FWIW, I've read Paul's final article on the subject. It has some interesting stuff, but the conclusion is that, based on actual testing with multiple boards, things aren't too bad -- but there was at least one board with an older BIOS where the reported power was way off. As in, absurdly low (one third the actual power). The actual power however was only ~20W higher than other boards. Basically, the mobo was telling the CPU "Run as fast as you want."

Anecdotally, I talked with someone else who used liquid cooling on an X570 board (including the VRMs) and he's had nothing but trouble. It's not clear if the extra-cool VRMs are somehow causing problems, or if it's something else. I told him that's what he gets for doing a full liquid cooling loop.

 

[joeblowsmynose]

I can say unequivocally that AnandTech and Tom's Hardware were not trying to play "good cop, bad cop" on this. It was more like: Paul saw a developing story and wrote about it on Tom's Hardware, it made some waves, and then Ian/AnandTech came in afterward and asserted that, "No, this won't break your CPU and here's why we think that's the case."

...

My goodness ... it was a tongue in cheek comment (in th eform of a question, not a statement), because it fit the strategy, and I also indicated that it may not have been intentional, and requested at that time to get back to topic.

After that I was just referring to the tactic, as it it seemed the questions were objective regarding that. I'm getting sorry I even brought it up ... lol. If you were thinking that I was asserting that they absolutely were, go back and read the posts again.

AMD CPUs don't run that fast above stock behaviour anyway, so "fast as you want" probably won't be outside of what manual overclockers can and will do for sure, and not likely harmful. But I get the warranty thingy for non-overclockers.

I agree with th elikes of Steve Burke and Ian ... I don't think think anyone's Ryzen will suddenly implode ...

 

[NightHawkRMX]

This thread sure has been fun

 

[TJ Hooker]

You're oversimplifying it. There are 3 limiting factors (PPT, TDC, and EDC), and the motherboard is fudging one of them. Yes, the CPU will eventually be limited by the other two, but maybe not as quickly.

Again, the testing methodology is to apply a workload which should max the Package Power. If it doesn't then you know the motherboard is lying.

Are we assuming that if the PPT limit is fudged (raised) at all behind the scenes, this will always result in TDC/EDC becoming the limiting factor? Is it not plausible that you could raise the PPT by some amount and still have that increased PPT value be the limiting factor rather than TDC/EDC? In that case the mobo would still be reporting that the CPU is maxing out the power limit (PPT), even if the power limit has been fudged.

Alternatively, is there no scenario where you'd be limited by TDC/EDC even without a fudged PPT? I really don't know. But if that's is possible, that would then be a scenario where your mobo could potentially report a less than max power under a workload like CB, which would erronously be reported as a power limit deviation if hwinfo is using the methodology you're describing.

 

[TJ Hooker]

Given that AMD is asking these motherboard vendors to cease this practice, it that's not a good analogy.

It'd be more like puppies telling the newspaper that you're kicking them, and then getting someone to build a puppy-kicking detector to alert when one of them is kicked. It's at this point that the newspaper is publishing the article asking if you're kicking puppies.

I feel like you misinterpreted my analogy. My 'kicking puppies' headline was an analogy to the part of TH headline where they question whether Ryzens are 'burning out'. And no, the CPUs/motherboards are not reporting "burnout", nor does the new hwinfo metric detect burnout. It seems to me that your analogy is the flawed one.

The unsupported question in the headline is not whether motherboards are cheating, but just what practical implications this has on CPU longevity and long-term performance. And it's not unsupported because it's not true, but just because we don't know. Not knowing the implications of something potentially harmful isn't a reason not to report it, especially since AMD's actions suggest there is a negative consequence.

To be clear, I'm no in way suggesting that TH (or anyone) else should have refrained from reporting on this issue (mobo makers fudging numbers). I was specifically referring to the question that this might be causing Ryzen CPUs to burnout prematurely, given the lack of evidence to support this.

The fact that AMD doesn't want mobo manufactuers fudging numbers behind the scenes does not inherently imply that there must be negative effects to doing so (and certainly doesn't imply anything about CPU longevity concerns). Though there are potential negative effects that we know can occur, namely increased cooling requirements and power draw with no apparent reason why (if you aren't aware that you board is fudging the numbers). We know that AMD CPUs have a variety of safeguards in place that are not affected by what the mobo manufacturers are doing, so we really don't have any reason (at this time) to think that this could result in premature failures.

With regard to reporting on any "potential" issue on the grounds that "we just don't know", that's an incredibly low bar. Especially because you can't prove a negative. I could ask "Do AMD CPUs have hardware backdoors that leak information to government agencies?". Do we know that isn't the case? No. Do we have any good reason to think that is true? No.* Would it be irresponsible to publish an article with that headline in absence of at least somewhat reasonable evidence? Yes.

Anyway, this is starting to get away from me a bit. I have no issue with the content of this article, or the vast majority of TH articles. Part of the title (and response to criticism of the title) just happened to twig a pet peeve of mine. I don't think TH is guilty of trying to slander AMD or anything, but I do think they like to spice up an article title when they can

*In case any conspiracy-oriented people may take issue with this, I'm really not interested in getting into some tangential debate here.

 

[bit_user]

I feel like you misinterpreted my analogy.

My response was based on how it sounded to me. Maybe I missed your point, but I wasn't intentionally trying to twist anything.

To be clear, I'm no in way suggesting that TH (or anyone) else should have refrained from reporting on this issue (mobo makers fudging numbers). I was specifically referring to the question that this might be causing Ryzen CPUs to burnout prematurely, given the lack of evidence to support this.

Opinions differ. I was okay with that part, since they didn't know, didn't claim to know, and it looked to be well within the realm of possibility (i.e. as a consequence of consistently running the CPU hotter than the manufacturer expected).

With regard to reporting on any "potential" issue on the grounds that "we just don't know", that's an incredibly low bar.

It's not as black-and-white as you make it sound. Sure, people can't just publish everything the don't know to be untrue, but when a story is developing and a likely negative consequence seems to be emerging, I'm okay with putting it out there, as long as there's no misrepresentation of the supporting evidence.

Anyway, this is starting to get away from me a bit.

Yeah, I don't think continuing this is going to move either of our positions. It's a legit difference of opinion.

 

[Schlachtwolf]

How many here will have their Ryzen 2 or 3 still in their system in 10-12 years..... much ado about nothing. This story is based on sneaky Mobo makers, a programm with inaccurate readings and a healthy dollop of guesswork on top. Going to do something useful now with my life....

 

[bit_user]

How many here will have their Ryzen 2 or 3 still in their system in 10-12 years.....

I have a 10-year-old Phenom II-based fileserver that I'm about to upgrade with a Ryzen 3k.

 

[NightHawkRMX]

My server uses Xeons from ~2009 and I have an xp system using core 2 quad from ~2008. Heck i still have a pentium 4 system.

People do keep hardware for awhile, though I still do not think it matters much.

Again, both intel and amd boards run cpus out of spec all the time, which is essentially overclocking and it would potentially degade both. However, some people have ran CPUs overclocked for many years and i don't see mass destruction because of this.

 

[bit_user]

However, some people have ran CPUs overclocked for many years and i don't see mass destruction because of this.

Due to smaller process nodes, newer CPU should generally be more susceptible to degradation than older ones.

In the Anandtech article, its author cites a Sandybridge CPU that won't overclock as well. Except that was 32 nm, while these Ryzen 3000 CPUs are "7 nm".

 

[warnings007]

A CPU burn if overheat but all CPU have internal security and just can't burn at all.
But you should ask if an Intel CPU eating almost 300W of power can may be destroy motherboard due to extreme temperature near CPU and VRM who should deliver that.

You have something against AMD actual domination ?


AMD made the FIRST processor who not need any cooler. Tomshardware may be made an article with this movie pretty amazing to view inside CPU like X-Ray with a thermal camera.

View: https://www.youtube.com/watch?v=mDHuurdr67A

 

[warnings007]

Due to smaller process nodes, newer CPU should generally be more susceptible to degradation than older ones.

In the Anandtech article, its author cites a Sandybridge CPU that won't overclock as well. Except that was 32 nm, while these Ryzen 3000 CPUs are "7 nm".

It's just an unknown world the 7nm yet like all other processors generation before. These processors work well and no destroy report yet for 14, 12 and 7nm from AMD only broken pin. Most people not keep more than 3 to 5 years a processor. Some rare people run these processors during 8/11 years but I have no worry.

When I view than some people torch component for view inside silicium chip under microscope it can't burn so easily.

 

[bit_user]

AMD made the FIRST processor who not need any cooler.

If you go back in time far enough, I think it was once the norm to run CPUs without a heatsink.

I also heard a rumor that early overclockers were able to do things like squeeze 50 MHz out of a 386DX-25.

Tomshardware may be made an article with this movie pretty amazing to view inside CPU like X-Ray with a thermal camera.
https://www.youtube.com/watch?v=mDHuurdr67A

Neat, but I don't see where it says what model it is, whether it's plugged in, or how much power it's configured to use (like 15 W or 25 W).

Also, I wonder why the package is so much warmer than the chip, itself. Is that anything to do with VRMs?

 

[warnings007]

If you go back in time far enough, I think it was once the norm to run CPUs without a heatsink.

I also heard a rumor that early overclocker were able to do things like squeeze 50 MHz out of a 386DX-25.


Neat, but I don't see where it says what model it is, whether it's plugged in, or how much power it's configured to use (like 15 W or 25 W).

Also, I wonder why the package is so much warmer than the chip, itself. Is that anything to do with VRMs?

That's true all older processor haven't any cooler or just a radiator but take probably less power too due to slower clock. 486 DX2 = 6.3W maximum power listed on cpu-world that was my first computer/processor

It's an AMD Ryzen 3 4300U - TDP 15W at 0:29 you can read the model on SIV64X window.

CPU heatspreader received fresh airflow from a fan on the left of the notebook you can view more clearly the red/black cable. I think that's why heatspreader is blue when package is yellow/red.
That's permit to have an X-Ray view when CPU heating it's just amazing

AMD or the notebook manufacturer downclock enough the CPU/GPU for work without radiator/active cooler even when it launch some game benchmark.