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

[Soaptrail]

The AD's on this site are damning. The "Just Buy It" (Nvidia) was damning.
Now this article.
Is INTEL paying you to pick sides? I don't get where this article is coming from.
Bye Tom..

Clearly you don't read many articles on this site. I would argue they have been Pro AMD and bashing Intel a lot in the last year. Is it a Pro AMD stance Tom's is taking, NO! They simply point out the issues with either. I have been reading this site (off and on) for 20 years and the complaints of bias have not stopped. It does switch back and forth however. So don't complain until you have solid evidence of bias. BTW, one article is NOT bias.

 

[waltc3]

The problem with this article as written is that it states Intel pushes this practice and encourages it with mboard vendors and will cover burned out Intel CPUs under its standard warranty. The problem there of course is that Intel doesn't make any CPUs with chiplets like Ryzen--Intel is still making the old-fashioned monolithic CPUs @14nm with tons of security holes--and the same parameters do not apply (fortunately.)

However, the article makes no effort to contact AMD for verification and expends no effort in contacting even a single motherboard vendor and it presents no proof that if a CPU fails AMD will decline to cover it under its standard warranty. Therefore, I am sure that AMD will cover a defective CPU. This article, frankly, was not well thought-out.

Another problem of course is that AMD ships unlocked processors--and although they allow you to overclock they are careful to state that overclocking voids your warranty. Overclocking often involves voltage tweaks done by the user. Normal boost involves voltage tweaks, indeed, normal operation of Ryzen CPUs allows for continuous voltage manipulation per core. HWinfo measures those voltage differentials.

Last, the notion that AMD would sell x570 chipsets to motherboard vendors that purposefully burn out its CPUs--just so AMD can refuse to honor its CPU warranties-- is so idiotic I don't wish to comment on it further.

 

[RodroX]

If you read the post by The Stilt, it's pretty clear AMD is trying to do something about this motherboard manufacturer behavior. https://www.hwinfo.com/forum/thread...er-reporting-deviation-metric-in-hwinfo.6456/

"Since at least two of the largest motherboard manufacturers, still insist on using this exploit to gain an advantage over their competitors despite being constantly asked and told not to, we thought it would be only fair to allow the consumers to see if their boards are doing something they're not supposed to do. The issue with using this exploit is, that it messes up the power management of the CPU and potentially also decreases its lifespan because it is running the CPU outside the spec, in some cases by a vast margin. Also, it can cause issues when this exploit goes undetected by a hardware reviewer, since both the performance and the sofware based power consumption figures will be affected by it.

"For example, if we take a Ryzen 7 3700X CPU that has 65W TDP and 88W default power limit (PPT), and use it on a board which has declared only 60% of its actual telemetry reference current, we'll end up with effective power limit of ~ 147W (88 / 0.6) despite running at stock settings (i.e. without enabling manual overclocking or AMD PBO). While the 3700X SKU used in this example typically cannot even reach this kind of a power draw before running into the other limiters and limitations, the fact remains that the CPU is running far outside the spec without the user even acknowledging it. This exploit can also cause additional cost and work to the consumer, who starts wondering about the abnormally high CPU temperatures and starts troubleshooting the issue initially by remounting the cooling and usually, eventually by purchasing a better CPU cooler(s)."
...

Yeap I read the post, and of course AMD wont do something drastically enough to blown-up itself, like going legal against two of the bigest mobo makers, not when they have a new product line coming out soon and present products (B550 and Ryzen 1xxx, 2xxx and 3xxx) that need to be avaialable for consumers to keep buying thier CPUs.
But if you think about it, now that the info is out, isn't this worst?, for everyone involved I mean, mobo makers, and AMD too for not doing enough.
When DIY and enthusiasts PC builders realize and find out which mobo makers are doing this, I don't think they are going to keep choosing those particular brands anymore, not in the close future, not till this is fixed.

What a stupid nonsense mess.

 

[waltc3]

Yeap I read the post, and of course AMD wont do something drastically enough to blown-up itself, like going legal against two of the bigest mobo makers, not when they have a new product line coming out soon and present products (B550 and Ryzen 1xxx, 2xxx and 3xxx) that need to be avaialable for consumers to keep buying thier CPUs.
But if you think about it, now that the info is out, isn't this worst?, for everyone involved I mean, mobo makers, and AMD too for not doing enough.
When DIY and enthusiasts PC builders realize and find out which mobo makers are doing this, I don't think they are going to keep choosing those particular brands anymore, not in the close future, not till this is fixed.

What a stupid nonsense mess.

It depends on whether this information is any good...😉 It's very hard to believe in its present form. The Internet is often awash in nonsensical "facts" that aren't true at all. We will see if AMD has anything to say about it--depends. If it's just really stupid they may not give it the time of day. I don't blame them--if they had to answer every uninformed post on the Internet they'd use up all their time...😉 Sounds like total FUD to me, but we shall see, I suppose.

 

[Math Geek]

i'm not sure what else they can do to make folks happy. he article gives the info that's available that appears to come from amd indirectly. they then say


"We're spinning up a few power tests of our own to assess how well the feature works, and of course, to see which vendors are misrepresenting their power consumption figures. We're also reaching out to all the relevant players, AMD included. Stay tuned. "

they gonna look for more info and run some tests themselves to see what is true and what is not. not sure how else to go about it since this is all the info currently available. a lot of folks seem to be jumping to conclusions and/or demanding a conclusion when one is not yet available

 

[RodroX]

It depends on whether this information is any good...😉 It's very hard to believe in its present form. The Internet is often awash in nonsensical "facts" that aren't true at all. We will see if AMD has anything to say about it--depends. If it's just really stupid they may not give it the time of day. I don't blame them--if they had to answer every uninformed post on the Internet they'd use up all their time...😉 Sounds like total FUD to me, but we shall see, I suppose.

Yeah I understand that, thats why it would be great to some media to contact AMD and ask for explanation.

Even worst is that if this is true, Does it happend on every X570 board of the x-maker?, or is in some high end models only?, Does BIOS version have anything to do with it?, Runing latest AGESA may fix this, or not? What was the testing models used to report this?, What BIOS were those boards using?, Were the launch models doing this since the day 1 reviews?

Indeed there are lots of question to answer.

 

[Jimichan]

If you read the link that goes to hwinfo .com you will see that this is the motherboard vendors doing it, not AMD. Specifically named was MSI. It looks like another user reported their Gigabyte board doing it.

I doubt AMD thought to themselves in early/mid 2016 "You know what? MSI is going to juice our processors in 2020, so we better build AM4 to last!"

Separately I had a thought about running processors on the hot side, albeit at stock ratings.

Personally, I keep a spare processor on hand. I run my processors fanless, with massive heatsinks in order to get 100% silence. No fans, fanless power supply, everything. Even at full load I rarely go above the mid-hi 40C range. As I type this it's just under 20C.

Processors get cheap fairly quickly on Ebay, after a few years. So that its said, I hope the motherboard vendors issue a BIOS update to correct this, now that they have been caught and called out.

Where do you live, in the Arctic? You keep your room under 20C? For a fanless PC to run even near ambient is impressive!

 

[F-minus]

Ian Cutress just invalidated your whole article Toms, I'm happy that there are still journalist with integrity like him and don't go with sensational <Mod Edit>.

 

[jayjr1105]

Doctor Ian Cutress. Tweet here. Shame on you Tom's.

 

[joeblowsmynose]

So "some" mobo manufacturers are treating ryzen official specs like they treat Intel's official specs? Hardly news I think ... Where's the article citing the concern of Intel processors that get the same treatment? Its not like almost every reputable reviewer hasn't pointed this exact same thing out with the 9900k and 10900k.

I've been running my R7 1700(non X) at 1.375v and a 850mhz OC, 24/7, for almost three years on a B350 mobo ... still not showing any sign of degradation (and I do 3D rendering so a large sum of hours at 100% load); I accidentally ran a few Cinebench runs at 1.6v as well ... still works fine.

The only CPU I ever saw that had clear degradation was my friends old 2700k -- he got it up to 5.0ghz (game stable), but after about two years, the highest clock he could get out of it was ~4.6 - so that was pretty significant, but the OC was a bit extreme in the first place.

 

[jeremyj_83]

So admitting that the Ryzen CPU WILL burn out - even without the motherboard shenanigans. Is this called "Heritage Mode" by chance - because the Athlon XPs loved to burn up. Some things never change.

This is why AMD keeps AM4 around - since it's CPUs WILL burn out at some point - and need to be replaced. All makes sense now.

No this headline is terrible. Read Ian's article on Aanadtech about this issue: https://www.anandtech.com/show/15839/electromigration-amd-ryzen-current-boosting-wont-kill-your-cpu

 

[escksu]

If you read the post by The Stilt, it's pretty clear AMD is trying to do something about this motherboard manufacturer behavior. https://www.hwinfo.com/forum/thread...er-reporting-deviation-metric-in-hwinfo.6456/

"Since at least two of the largest motherboard manufacturers, still insist on using this exploit to gain an advantage over their competitors despite being constantly asked and told not to, we thought it would be only fair to allow the consumers to see if their boards are doing something they're not supposed to do. The issue with using this exploit is, that it messes up the power management of the CPU and potentially also decreases its lifespan because it is running the CPU outside the spec, in some cases by a vast margin. Also, it can cause issues when this exploit goes undetected by a hardware reviewer, since both the performance and the sofware based power consumption figures will be affected by it.

"For example, if we take a Ryzen 7 3700X CPU that has 65W TDP and 88W default power limit (PPT), and use it on a board which has declared only 60% of its actual telemetry reference current, we'll end up with effective power limit of ~ 147W (88 / 0.6) despite running at stock settings (i.e. without enabling manual overclocking or AMD PBO). While the 3700X SKU used in this example typically cannot even reach this kind of a power draw before running into the other limiters and limitations, the fact remains that the CPU is running far outside the spec without the user even acknowledging it. This exploit can also cause additional cost and work to the consumer, who starts wondering about the abnormally high CPU temperatures and starts troubleshooting the issue initially by remounting the cooling and usually, eventually by purchasing a better CPU cooler(s)."

You've actually got it backward. Again, referencing The Stilt:

'HWiNFO will display "Power Reporting Deviation" metric under the CPUs enhanced sensors. The displayed figure is a percentage, with 100.0% being the completely unbiased baseline. When the motherboard manufacturer has both properly calibrated and declared the reference value, the reported figure should be pretty close to 100% under a stable, near-full-load scenario. A ballpark for a threshold, where the readings become suspicious is around ±5%. So, if you see an average value that is significantly lower than ~ 95% there is most likely intentional biasing going on. Obviously, the figure can be greater than 100%, but for the obvious reasons it rarely is 😉'

So at idle, it will be way off and you shouldn't pay any heed to the deviation metric. Under a heavy load, you want the motherboard to be at 100% ideally. If it's higher than that, something is goofy -- it would theoretically mean your motherboard is underclocking or reducing performance. If it's less than 100%, 95% is probably okay, anything lower than that is intentional misreporting of current to the AMD CPU in order to boost performance. It's basically PBO without saying it's PBO. Your 90% result is probably going to be fine, but apparently at least one motherboard maker is setting the value at 50-70% lower than it should be, and that's going to result in some relatively dangerous behavior.

As an example of what can go wrong, I know someone (not at Tom's Hardware) that enabled the auto-overclock feature in the motherboard BIOS when Skylake launched. A bug in the Asus BIOS caused it to apply an extra 0.35V to the CPU and it killed the processor -- it applied a voltage offset, and then applied it again, so instead of 0.175V extra it pushed 0.35V extra. So when a motherboard maker gets fast and loose with the BIOS stuff that reports current and/or voltage, bad things can happen.

Incidentally, at stock operation this is probably only a concern if the motherboard is underreporting by more than 10%. If you overclock, however, note that this will compound the situation. AMD is already pushing its Ryzen CPUs to their limits, which is why it needs the current power information from the motherboard. Then the motherboard auto-overclocks, by lying about how much power is going to the CPU. Then you overclock, and you're trying to push things even further, and you increase voltage, power limits, etc. even more, and wonder why you're not getting much better performance than at stock, while power and temperatures jump even higher.

Hi, is this piece of software is even accurate in the first place?? Because we all know that there is a variation between whats really measured using real measuring/testing equipment and such monitoring utilities.

And then, this is just a freeware/shareware on the internet. Its not some professional software from a reputable company. Its also not calibrated for any board to ensure readings are definitely accurate.

Would be great if tomshardware can at least perform some tests to reflect actual power consumption vs reported by the software.

 

[TJ Hooker]

Something that wasn't clear to me. What is the test control? Is this "report" from the author, specifically about boards that are in "their" hands (and may or may not have been sent as review samples), or from a selection of data sent back to HWInfo?

If I'm understanding you correctly, then I was wondering the same thing. What is hwinfo using as a reference to determine how the reported power relates to actual power? Did the creators of hwinfo test every board and measure power using external test equipment while also recording the power being reported by the CPU/motherboard, to empirically come up with a value for how much each board is skewing results? And if that's the case, wouldn't that just be a fixed value, rather than one that is seemingly dynamic based on the fact that the "power reporting deviation" shown in hwinfo is variable?

 

[TJ Hooker]

During the production process, AMD tests and sorts the dies, based on what clock speed their various cores can hit. The ones which can't reach higher clocks get sold as lower-end models. And, at lower clocks, you don't need as much power.

In theory yeah, but they also need to produce sufficient amounts of each SKU. So even if a chip is good enough to be a 3600X, it's possible it could end up as a 3600 because that's what AMD happened to need. Especially if the process is improving over time.

From the reviews I've seen the 3600 and 3600X perform nearly identically and draw nearly the same power (both stock and PBO), so I'm a little skeptical how much binning is really going on there.

 

[joeblowsmynose]

Ian Cutress just invalidated your whole article Toms, I'm happy that there are still journalist with integrity like him and don't go with sensational <Mod Edit>.

Its probably a "Good cop, Bad cop" strategy.

 

[punkncat]

Its probably a "Good cop, Bad cop" strategy.

Admittedly, there isn't a whole lot to talk about right now. Particularly for Intel and their shills and fans (not accusing), but it's not been a great news cycle for them going on some time now. It's starting to look like the alternative is to try and throw shade rather than innovate.

 

[joeblowsmynose]

Admittedly, there isn't a whole lot to talk about right now. Particularly for Intel and their shills and fans (not accusing), but it's not been a great news cycle for them going on some time now. It's starting to look like the alternative is to try and throw shade rather than innovate.

Maybe, But my point was that I do notice that Anand is a sister brand to Tom's ... creating a dynamic between the two based around first, "hype/drama" then, also creating the "mature" response to "hype/drama" generates far more interest, views, and discussion (thus ad revenue) than the sum of either on their own.

Maybe not intentional, but I do know this is a known strategy in the media.

 

[F-minus]

Maybe, But my point was that I do notice that Anand is a sister brand to Tom's ... creating a dynamic between the two based around first, "hype/drama" then, also creating the "mature" response to "hype/drama" generates far more interest, views, and discussion (thus ad revenue) than the sum of either on their own.

Maybe not intentional, but I do know this is a known strategy in the media.

The difference is that Anand has actually valid, researched articles by people who actually understand the things they write about, understand, not just have a grasp on things. Tom used to be like that, but it turned into a sensational page instead of informative one, jumping on anything, regardless how small it is, that might get the clicks, and post it under the most absurd headlines, click-bait, outlandish headlines there is.

For example in this article; everyone browsing these kind of sites should know that over-volting chips reduces their life span, regardless if it is manual or automatic, however everyone also knows, that this is not the '90s anymore and technology has progress further and there are safety measure for you not to fry your chip instantly, and whilst CPU deterioration is a thing, it will take a decade or more before you will notice anything, even running out of spec, and I wager my kidneys, that nobody will run a CPU for ten years+ at least not as their main machine, and for random servers, they wont be running out of spec if set up properly so the deterioration will set after a longer time.

This articles headline however, tells me, a ZEN CPU might be dead in the next coming days, just the headline, not the article, as I'm not even bother reading it, when I see such sensational titles.

 

[Giroro]

Just so you know, trying to fit in the pertinent details in a short enough title is hard. We could have said, "AMD is frustrated that motherboard makers are using tricks to make Ryzen CPUs think they're using less power, which causes the CPUs to run hotter than expected and shortens the potential lifespan." That's about three or four times longer than desired. The important information is that some AMD motherboard makers are 'cheating' on stock clocks in order to attempt to boost performance, and that cheating is against AMD's policies.

It also reads as though The Stilt (who wrote up the lengthy explanation) either works for AMD or has direct contact with them -- which would make sense. He says "we" when seemingly referring to AMD several times. So AMD is basically taking a back channel after directly approaching motherboard makers and saying, "You are doing things that overclock the CPU and void its warranty without informing the users, and it needs to stop."

It's not that it DOES shorten lifespan, it's that it COULD shorten potential lifespan - which is just general CYA coming from a free 3rd party tool. Its very unlikely to actually be shortening the lifespan of the CPU, because power draw itself is not a typical failure mode of a CPU.
The phrasing implies that CPUs are failing because of this, and there is no evidence whatsoever that is happening. Mobo makers shouldn't be making changes without informing customers, but there isn't anything here that is that big of a deal.
CPUs can fail because of temperature, high voltage (ie static, the possible damage from tweaking the CPU by millivolts is still from heat), or eventual physical phenomena (like tin whiskers). Yes, power draw is related to voltage and temperature, but if your CPU is staying below it's thermal limits and the voltages are in the safe range, then I'm not really seeing a mechanism by which this misreporting of power draw to bypass current limits could actually damage a Ryzen CPU itself... It might be able to damage poorly designed power delvery on the mobo though, like improperly cooled VRMs. Of course this is assuming thermal throttling is still enabled on the CPU, and anyone who is disabling that probably already knows the risk.

My MSI mobo lets me change the max current going to my 3700x in the PBO options, I think it's relevant that (iirc) that value defaults to 999 Amps, and that changing that value has no effect on performance or temperatures until you go too low with it - because the PBO bottleneck is almost always going to be temperature. I understand this issue is not PBO, that's just the easiest way to play with that limit.

 

[JarredWaltonGPU]

It's not that it DOES shorten lifespan, it's that it COULD shorten potential lifespan - which is just general CYA coming from a free 3rd party tool. Its very unlikely to actually be shortening the lifespan of the CPU, because power draw itself is not a typical failure mode of a CPU.

The phrasing implies that CPUs are failing because of this, and there is no evidence whatsoever that is happening. Mobo makers shouldn't be making changes without informing customers, but there isn't anything here that is that big of a deal.

This is bunk. The phrasing does not imply anything. It says "Ryzen Burnout?' [Note the question mark.] "AMD Board Power Cheats May Shorten CPU Lifespan." Anyone trying to read anything more into it is on their own. Nowhere do we / Paul say there are CPUs failing,

The problem isn't so much with minor deviations of 5%, it's with deviations of 25-50%. Paul is doing some testing with hardware to actually check actual CPU power use, as well as what HWiNFO says, on boards from every major party. With multiple CPUs.

Ryzen 7 3700X is probably going to be fine. The bigger concern is Ryzen 9 3900X and especially Ryzen 9 3950X. Those already run at close to maximum clocks and voltage right now, so pushing them higher is more likely to cause problems. A 25% power increase on a 3700X is only going to take the chip to 81Wm and even a 50% increase is 98W. But if you take Ryzen 9 3950X from 95W to 143W, with potential higher power draw in some workloads, and it overheats and hits 90C or higher? You'll definitely shorten the lifespan. The question is by how much. Will it last 9 years instead of 10 years? Or will it last 2 years instead of 10 years? It's too early to say.

What we do know is that AMD is specifically telling motherboard makers to NOT do this, but the motherboard makers are refusing to cooperate. So now AMD has apparently worked with HWiNFO to help shed light on this practice. That right there says a lot about the behind the scenes stuff.

I wouldn't be surprised if AMD has done some internal testing that makes it think certain boards are going to cause problems sooner than later, and leave AMD holding the bag. If you enable PBO and burn out your CPU, your warranty is void. If your motherboard cheats and enables PBO-like behavior at "stock" and you burn out your CPU, it will look the same as if you had enabled PBO.

Let's wait for the follow up article showing hard numbers, though, which should hopefully be here tomorrow or the next day.

 

[mdd1963]

"may be shortening the lifespan...."

"potentially negatively affect the lifespan"

After reading all these 'maybe's' and 'might'-statements..I naturally suspect the Russians!

 

[joeblowsmynose]

...
This articles headline however, tells me, a ZEN CPU might be dead in the next coming days, just the headline, not the article, as I'm not even bother reading it, when I see such sensational titles.

I think you missed the point ...

In the example I gave to Punkncat, details don't matter, only the ability to create strong emotional reactions (Player A), then give a second player (Player B) a response that the emotionally reacted people will align with - however that plays out, doesn't matter. The dynamic is the desired result, not how people feel about one side or the other.

Anyway there's your marketing lesson for the day.

Let's get back to topic.

 

[Crolug]

Guys, stop fooling around... I'm preparing to buy X570 mobo, just give me the damn names of manufacturers I should avoid! Can be through DM.

 

[Groveling_Wyrm]

Guys, stop fooling around... I'm preparing to buy X570 mobo, just give me the damn names of manufacturers I should avoid! Can be through DM.

The two manufacturers labelled as culprits were MSI, and Gigabyte. Paul is testing others to see if perform the same way.

 

[jgraham11]

Bruh Isn't that the same thing as Intel's MCE? This is such a shill post

Agreed, this is something Intel has notoriously done but if someone looks it up now they find this article that is utterly garbage and has little basses for determining actual CPU operating frequencies.

Ok back to facts: AMD chips determine their speed based on operating tempertures not power use.
The formula is: TDP= Temp(Case)-Temp(Ambiant)
HSF(Theta)ca
where HSF(Theta) is the thermal performance of the cooler.

This is why you have a better cooler, the CPU automatically runs faster based on better thermals.

This is very different to how Intel Kaby lake and refreshes CPUs do it. They work based on power use and have short term power numbers as well as long term power numbers and a timer that is used to vary this behaviour. The thermal velocity boost is the only boost that somewhat resembles AMD's more advanced system, however it is only on or off where as AMD's system is proportional across a wide range of temperatures and operating conditions.

So where it matters
AMD cpus will back off when operating in a hot environment despite motherboard tweaks. Intel CPUs will run hotter and be throttled by their limiting temperature maximum. That's why most Intel CPUs operate in the 90 degree Celsius range most of the time. They are trying to compete against a more advanced core with a better lithography process.

The only advantage Intel systems have is memory latency and higher maximum clock speed, at the expense of power. Intel has also come up with a better way of removing heat from their CPUs but this is more out of necessity as most Intel 125W k chip run at 250W constant, but according to this article, this is encouraged!? See next paragraph...

Here is my favourite quote from the article:
"In fact, nearly every motherboard vendor makes adjustments with Intel's chips, but there's a big difference: Intel expressly approves and even encourages motherboard vendors to adjust power limits to differentiate their products, and those adjustments don't impact chip longevity within the warranty period. "

Interesting fact
Check out the Intel website:
https://www.intel.com/content/dam/w...roduct-briefs/10th-gen-core-desktop-brief.pdf

From page 7:
3 Unlocked features are present with select chipsets and processor combinations. Altering clock frequency or voltage may damage or reduce the useful life of the processor and other system components, and may reduce system stability and performance. Product warranties may not apply if the processor is operated beyond its specifications. Check with the manufacturers of system and components for additional details.

So I read this as you operate right out of the box: Outside of operating specs, waranty is gone. You decide for yourself.

These types of articles is why Toms has lost the respect and credibility of most tech savvy people following the industry. When one manufacturer is notorious to operate in a certain way, Toms publishes an article about how the other manufacturer is actually too, which everyone word may be technically true, has no barring on reality. I suppose we should just buy it!