AMD comments on burning AM5 socket — chipmaker blames motherboard vendors for not following official BIOS guidelines
Posted by imaginary_num6er@reddit | hardware | View on Reddit | 130 comments
SomeoneBritish@reddit
If possible, AMD & Intel should force motherboard manufacturers to operate CPU’s with default settings by default, unless the customers chooses to do otherwise.
FragrantGas9@reddit
Definitely agree. Yet it seems complicated when it comes to XMP settings to run RAM above default jedec speeds. Different minimum voltages needed for different memory vendors and specs of the kit, not just the memory voltage but the VSOC voltage for the memory controller on the CPU. A lot of the AMD cpu failures were from mobo makers juicing the VSOC too high to guarantee the memory is stable.
It’s possible to enforce it but a lot of effort needs to go into testing and verifying minimum voltages needed needed on every single board and every single memory kit. And it leads to more product RMAs when just a tiny bit more voltage is needed to make a certain kit stable but the board isn’t giving it.
Not saying the current situation with the mobo vendors is OK, they seem pretty lazy about setting too high voltages as and just calling it fine and shipping it. They could do better. Not to mention straight up bugs and bad code in the UEFI and interface that cause overvoltages even when it should have worked fine.
rilgebat@reddit
Personally I'd like to see the whole memory ecosystem overhauled, especially now that JEDEC is being more forthcoming in defining frequency specifications with DDR5.
We shouldn't need cruft like XMP/Expo, and the likes of AMD/Intel should be more explicit about what their platforms are capable of, rather than coy statements about "sweet spots".
a2e5@reddit
The issue here isn’t the timings and volts for the memory itself, which is the information provided by standard SPD as well as XMP/EXPO. Them information is alright.
The issue is that there is no good information on how to run the CPU IMC (NB, SA, IOD, whatever) to support such a frequency. This information can be written into some next generation of XMP/EXPO profiles given that they are, well, NOT JEDEC and directly associated with the CPU-maker, but this would be silly from an upgrade perspective.
Instead AMD and Intel should yell louder at mobo makers about standardizing what "Auto" means and put a few limits on it, limits that are not automatically removed when one selects XMP/EXPO. For "freedom" purposes you can have a switch for removing it, but point is since everyone is expected to turn on XMP/EXPO these days, you better not make the almost-default unsafe.
rilgebat@reddit
I never said it was. The problem is that JEDEC's traditional reticence to define memory specifications at or ahead of the pace of what is available on the market, has lead to an ecosystem where running at any kind of sensible speed for a high performance system required the usage of XMP/EXPO.
The likes of AMD/Intel take advantage of this situation by on the one hand talking of optimal speeds and "sweet spots", with a wink and a nudge that function at these speeds is at least broadly possible, while also washing their hands of it by labelling it overclocking.
They could solve the problem. But it would mean having to bin chips more appropriately, and that's not in the interest of maximising profits.
a2e5@reddit
They are surprisingly doing okay on the LP part with the standardization of LPDDR5X. When the manufacturers as well as them actually have a motive to make something fast and stable, they will. Sigh.
rilgebat@reddit
Note that I said traditional reticence. JEDEC have been ahead of the curve with DDR5, as the spec currently extends all the way to 8800MT/s.
But they are not the problem here per se. The responsibility ultimately lies with AMD/Intel for taking advantage of the grey area. JEDEC's inaction simply allowed for the grey area to thrive.
AntLive9218@reddit
There's no good solution for all cases, the "silicon lottery losers" simply won't work in all cases. It's especially tricky lately as even the old overvoltage solution doesn't cover all cases, as too much voltage can also cause instability, likely with overdriving increasing noise.
However there are incredibly helpful tools and solutions for power users which are either just not presented to users, or usage of them are even presented.
A lot of chip(let) to chip(let) communication is already covered by ECC or at least EDC, and error counters are commonly available, they are just not properly (especially not uniformly) exposed.
For example on modern AMD CPUs, increasing FCLK too high can result in experiencing stuttering, making it highly likely that the IFOP is protected by some EDC. EDC error counters being exposed to users could be helpful with treating errors before they turn into crashing, and they could be also useful for faster and more reliable stability tests.
Then there's that whole ECC memory issue of reliable memory not considered being important for regular users. XMP/EXPO problems would be significantly better if users could look at error counters (even better, getting error notifications) instead of just running long memtest sessions and then hoping for the best.
a2e5@reddit
DDR5, ECC (9-row) or not, also have this on chip ECC working independently of the extra row. And the problem with it is that there’s no counter for the on-chip level even on ECC sticks, exactly because it’s independent! So now guessing how much erroring is something you need to do, and something that’s actually worth a conference paper:
M. Patel, J. S. Kim, H. Hassan and O. Mutlu, "Understanding and Modeling On-Die Error Correction in Modern DRAM: An Experimental Study Using Real Devices," 2019 49th Annual IEEE/IFIP International Conference on Dependable Systems and Networks (DSN), Portland, OR, USA, 2019, pp. 13-25, doi: 10.1109/DSN.2019.00017.
(Not to belittle the authors’ work in any way. This is important work, just something that JEDEC should’ve made unnecessary.)
SomeoneBritish@reddit
For XMP, I personally think on first boot your BIOS should notify you that you have profiles which you can load to potentially improve performance, then let the customer choose whether or not to apply.
I feel bad for all the thousands of people out there with fancy memory running at default speeds :(
AssBlastingRobot@reddit
Why are people buying fancy expensive ram if they don't know how to use it?
That's an insane waste of money.
It's exactly like those douche bags that buy supercars to drive to the super market.
I mean, it's their money at the end of the day, but why buy something if you don't know how to use it?
Blueberryburntpie@reddit
Have you seen the volume of posts on computer/gaming related subreddits of people discovering their 144Hz monitor was running at 60Hz for years, or they had plugged their monitor into the IGP instead of their RTX 3090?
TenshiBR@reddit
“Think of how stupid the average person is, and realize half of them are stupider than that.”
I am no genius, but good lord... I switched to the medical field after doing computer science, the amount of people who can't even operate a computer in 2025 is staggering and these are not boomers either.
Plus, the complete lack of common sense for the most banal everyday task. Some people are alive because of God's mercy.
AntLive9218@reddit
I wouldn't put all the blame on just users though, as learning is still becoming harder since at one point it was decided that catering to the lowest common denominator means that anything confusing stupid people like silly technical details must be hidden.
Even without going into the details of hidden tech info, just consider the much simpler localization problem, users robbed of the opportunity to learn English. The degraded language skills as a result were already visible online several years ago, but I've started seeing more and more non-English posts/comments in otherwise English environments, seemingly with the expectation of others using translators (or does "new" Reddit automatically do it?).
Of course there are people who don't even try, but a lot of people learned not by actively seeking out knowledge, but just solving challenges on their way towards a goal. When an Apple phone user (think of all the tablet kids) isn't even allowed to know that files exist, or a gaming console user isn't allowed to run anything not related to gaming on what's essentially a locked down x86 PC, then how are they supposed to learn without even knowing what they are missing out on?
Hell, I'll go as far as claiming that curiosity about technology is even punished, and not even culturally like kids calling you a nerd. Can't even do a lot of tasks on a PC anymore because a lot of services are phone-only, and then it turns out that the phone crapp isn't willing to run on a non-stock OS. Why learn and experiment when anything else than just going with the flow gets you in trouble, and culturally we seem to be in a phase of the majority of the people being okay with everything getting dumbed down, users being expected to just consume without thinking too much?
We had a golden age of computing where people could (and had to) learn quite a bit about technology to get to their goals, and many people could do so with having to learn English first. Nowadays people are funneled towards entertainment/consuming with just a few clicks/taps, and even if anything goes wrong, the technical details are often hidden, only presenting a "Something went wrong" or "Oopsie woopsie" page. It's almost like knowledge is considered so dangerous, there's a lot of effort put into hiding it, and unfortunately I can actually see it scaring some kinds of people, like Apple users who seem to be proud of not even being allowed to utilize technical knowledge.
Zaev@reddit
I think new Reddit might actually do it. I did a search for a post just a little bit ago, and clicking the link unexpectedly brought me to new Reddit. It was also unexpectedly in French, though the URL was in English but, had something like
?tl=frat the end (not sure if that was it exactly, and can't check right now)TenshiBR@reddit
Since I am in Brazil atm, for 6 months now when I do a google search, there are Reddit links showing up as results and they are being translated by Reddit to Portuguese. There is a link at the OP titled "See the original post" which undo the translation
The users have no idea english is being used and make a post in their own language...
the translation is good, but most of the time, internet slangs and such are translated poorly and things get pretty weird
Strazdas1@reddit
yeah, the random different language comments are weird. At least on something like youtube theres a translate button next to it, but on reddit i always wonder if i should report it in this sub. I know some subs have english-only policy.
English is not my first language. I learned english originally by needing to communicate with others inside MMOs when i was a kid.
AntLive9218@reddit
Haven't considered reporting, especially as it seems like just a couple of reports may not even show up anywhere anymore (that gets human eyes), I just downvote and move on, but I did wonder if there's a different interface which auto-translates.
There's no need for English-only policies, it's more of a common sense matter that communication should be done in a matter that's valuable for the other parties, avoiding not just languages foreign in a specific environment, but also low effort messages like "this", or dumping a stream of thoughts with no punctuation, and no basic checks for at least gross writing errors.
Your story isn't unique, I know a lot of people who learned English a similar way. Unfortunately some of them regressed by embracing localizations, and whenever I see that they are playing cooperative games with translations, I cheekily ask them how do they intend to communicate with others without even knowing the item/spell/ability names others know and see. I occasionally get to see how that works out between friends, and short term it's somewhat hilarious as the observer, but unfortunately long term I've seen even people not wanting to use localization just giving in, and more isolated groups forming mostly based on language (and nationality).
Strazdas1@reddit
I hate the forced localization. I always choose english for software but microsoft is trying to force revert to english language every chance it gets.
Strazdas1@reddit
Want to be more depressed? the computer litaracy is getting worse. the new generation is growing up on iphones and androids and have even less knowledge how to use computers. I had to teach university educated colleagues of mine what a folder is because they never saw anything more complicated than ipad. I feel like old man yelling at clouds now.
Standard-Potential-6@reddit
Some SKUs are or have been only slightly more expensive than JEDEC RAM, they look pretty, heatsinks look confidence inspiring to newbies, and they’re inundated with advertising.
Don’t underestimate how many customers will pay a few bucks to have a good feeling buying a popular product from a known brand, and assume it’ll just work.
fullsaildan@reddit
Yeah I mean honestly the marketing and specs on ram is overwhelming for many new or relatively inexperienced builders. Even after many builds for myself, family, and friends, it’s one of the parts I hate having to choose. Throw in model names, color, and RGB types in SKU and it’s also a PITA to source sometimes.
Strazdas1@reddit
yep. even for experience builders timing can look like black magic with how complicated it can get nowdays. Does not help that many retailers flat out does not display it unless you go digging at manufacturers website.
Also love that half the manufacturers do not give physical dimensions on specification page. Have to guess if it fits under CPU cooler or not. So i just go to manufacturers that do give the detail.
Strazdas1@reddit
because advertisement works.
Illbe10-7@reddit
People get butthurt when you say this but it's true. Too many people buying things they have no business buying but get it anyway because "they can".
Proud_Tie@reddit
the 128gb of ddr5-6000 in my server can't have EXPO enabled or it becomes unstable as hell..
....its still an improvement over refusing to post with all four sticks and EXPO enabled when I first built it in January. (All four sticks worked with EXPO and no instability on another board so its not the ram)
RawbGun@reddit
Basically no Zen 4 chip can do 2DPC2R (ie 4 sticks of dual rank memory) at 6000 MT/s, generally the max recommended speed is around 5600
hexgama@reddit
Did you check if your exact 128gb kit was listed on the QVL of your motherboard with EXPO checked, and that it isn't 2x 64gb kits which will also be marked on the QVL as not supported?
Proud_Tie@reddit
One 64gb kit is supported but not two on the servers b850 gigabyte board, but is on my desktops ASRock x870 board.
Lesson learned. This was my first build with four sticks so I assumed it would work the same (and yes, I know what happens when you assume). Maybe one of these days I'll switch boards out, the server could use another nvme slot, and my desktop could use the 9950x.
FragrantGas9@reddit
100%.
The mobo makers also need to do a better job with setting voltages on XMP profiles safely, and preventing over voltage bugs.
NuclearReactions@reddit
I think what we have come to accept over time is madness, you should not sell a product that is advertised at an overclocked speed. Just advertise X frequency and let them run like that like it used to be before xmp became a thing.
Unkechaug@reddit
Truly. This world just keeps getting dumber. Don't care what the spec of RAM is and what it's rated for under ideal conditions, it should be advertised at the lowest speed it's stable on that generation CPUs (with some headroom) and people can decide to apply OC profiles if they want. This is how we did it back in the day and it worked great, now it's all bullshit marketing and problems like this happen. Worse solution. How did we get here?
Keulapaska@reddit
Who/what defines a generation for set of compatible ram? Like the official supported intel 12th gen ddr5 speed is 4800MT/s for ONE stick, you want 2 sticks? 4400MT/s, so will all 2 stick DDR5 ram boxes just have to say 4400 and nothing else?
Gonna make the packaging plain white with a photo of burning cpu to the box as well saying overclocking ram causes cancer or something next?
Strazdas1@reddit
JEDEC?
Keulapaska@reddit
JDEC is just a timing spec and it goes up to 8800MT/s for ddr5. Obviously 8800MT/s ram will not work for vast majority of cpu:s/motherboards currently, so tje JDEC spec doesn't have really anything to do with what intel/amd deem "offcial spec" ram speed for their cpu:s.
AntLive9218@reddit
That's not so simple though, because it's hard to control indirect advertisement like reviews which often go for a setup considered "realistic", not one guaranteed to be working for everyone.
FragrantGas9@reddit
Yeah. Alternatively, at least for AMD, they should accept DDR5-6000 CL30 as an officially supported configuration with 2 sticks of RAM up to 64 GB. It’s very easy to run and doesn’t need dangerously high memory or SOC voltage. It shouldn’t be a “this will technically void your warranty” type overclock to run that.
Frexxia@reddit
XMP should never be enabled by default, and even when enabled, rated voltages should be used unless specifically opted in.
FragrantGas9@reddit
It shouldn’t be enabled by default, yes. But the product performance is advertised with XMP enabled, reviewers test it with XMP enabled, and consumers expect to be able enable XMP safely.
The problem seems to be with mobo makers juicing voltage ranges for XMP profiles to try to guarantee stability for a wide range of memory lots and configurations, and that can cause problems. They are just saving time and money taking the easy way out instead of widely testing individual kits and setting reasonable voltage ranges.
And sometimes they have a semi reasonable voltage and the bios is just straight up bugged and over volts it anyways.
They have definitely been dropping the ball.
Strazdas1@reddit
which should be considered false advertisement and punished accordingly.
good ones test ate JEDEC.
You should always expect risk with overclocking.
UsernameAvaylable@reddit
Frankly i was completely put off when i finally bought a new computer a few years after chilling with my old 6 core intel for many years that XMP is overclocking "in the books". Like motherboards, CPUs and memory is sold with those speeds as advertised, but suddenly if you want to actually use it its "oh its best effort, no gurantee"?
Frexxia@reddit
Then motherboard and memory manufacturers shouldn't make promises they can't keep.
Cranking up the voltage is a band aid solution.
FragrantGas9@reddit
Yeah they take the easy route and crank the VSOC way higher than needed for the vast majority of configurations. Like waaayyy high.
Frexxia@reddit
I did, but it sounds like you're defending what they're doing.
FragrantGas9@reddit
This sounds like defending them to you? Wat
Frexxia@reddit
Well, starting your comment with "But" kind of signals that
Strazdas1@reddit
XMP is not a default expectation though. Its custom overclocking and should be left to users choice/risk.
isotope123@reddit
Easier to just set the SOC voltage to the maximum safe/stable votlage for all RAM tested per CPU family. For example, set the default SOC to 1.3V for Ryzen 9000 CPUs when enabling XMP. 1.3V was the guideline AMD set for the chips. Then overclockers can play with it and undervolt as they see fit. Solves your testing issue and stops boards frying too.
JudgeCheezels@reddit
How do you “force” them?
They’ll send a black and white document, follow this shit or else…!, manufacturers would just say ok bro chill, we will.
Who goes and check after that?
luuuuuku@reddit
You can only build a mainbord at scale when you have access to the chipsets. AMD could stop supplying chipsets when they don’t like what Asrock is doing
Netblock@reddit
Give them the boot. Stop giving them R&D help, documentation access (NDA territory), and bios cryptographic signing access; stop selling them necessary chips (southbriges, BGA CPUs, GPUs).
You may be able to reverse engineer your competitor's work and buy your competitor's products to take the chips, but you're now you're many days late and many dollars short.
Green_Struggle_1815@reddit
amd and intel profit from those companies to take the risk otherwise they perform noticeably worse in benchmarks.
AMD is in it themselves. XMP is considered overclocking by them and supposedly voids warranty.
luuuuuku@reddit
So ist PBO too which is enabled by default and PBO is the reason why AMD says it’s not on them. Doesn’t stop any reviews from using PBO and doesn’t stop amd from using it in their advertisements when they make performance comparisons. PBO is on by default, because reviews and amd marketing use it. Everyone not using it is at huge disadvantage.
goodnames679@reddit
Yep, right now they get the benefit of higher benchmark performance while having a convenient scapegoat for whenever things go wrong. Why would they give that up?
jaaval@reddit
The mobo makers wouldn't like that. It would make differentiation in their product stack more difficult. They want to be able to offer the XsuperXgamingXXX model that costs $800 for marginally improved default performance.
SomeoneBritish@reddit
I agree with you.
jaywastaken@reddit
Except it's in their interest to do the exact opposite.
If the vendors are pushing past the cpu limits to claim higher performance the cpu manufacturers get all of the kudos for the higher performance and better benchmark scores but then get to put the blame and liability for defects in the motherboard vendors.
It's a win/win for them, they aren't going to stop them.
edparadox@reddit
Where do settings like EXPO and such should stand then? Because, it's become the defacto "default" if your RAM is compatible.
BrushPsychological74@reddit
I would be okay with a first party motherboard too.
kuddlesworth9419@reddit
It would be nice if AMD actually made their own reference motherboard. Just for people that want a motherboard that just does motherboard things.
cluberti@reddit
Intel made motherboards up until about a decade ago - they got into the business to reduce the "white box junk" that was being sold around the 386 and to a lesser extent, the 486 CPU era, and also as a way to fend off AMD an Cyrix.
There's not a lot of money to be made in making desktop motherboards when the vast majority of machines sold nowadays are laptops (and servers), and the vast majority of desktops are OEM desktop models using custom designs.
I think the time of first-party motherboards has passed, although nowadays there are open-source and code-available models for system firmware that would be more beneficial to everyone than having CPU manufacturers making mainboards again.
Kezika@reddit
Yeah I was starting to see the effects of that with my recent build earlier this month. All the Gaming PC motherboards I could find just had one dealbreaker or another, or just bad reviews on quality etcetera.
Ended up going with a "Workstation" board to end up meeting my needs, which was just dual NIC, POST Code LED, and multiple PCI-e x16 slots, which like in 2016 would've been practically every gaming motherboard out there. Now it was basically a unicorn.
jrr123456@reddit
The problem is most do, until you enable XMP or Expo, at which point the board makers go nuts and start cranking power and voltages to make their board seem artificially better.
They need to be more forceful around the XMP and Expo settings otherwise the only way to guarantee the safety of the CPU from an end user perspective, is to cripple the performance with JEDEC ram speeds and timings.
1_H4t3_R3dd1t@reddit
It is very likely, but I also blame the design of the socket as any form of warping can result in arching.
diceman2037@reddit
they should not be spreading the array across the entire underside, intel has never released a cpu without some sort of spacing on the underbelly to be used for smd's on the cpu and motherboard sides, preventing such a thing from happening.
HyruleanKnight37@reddit
Considering it's mostly Asrock, it is plausible. But the fault still lies on AMD for not enforcing said guidelines or at least performing QA checks before permitting manufacturers to release the product into the market.
BlobTheOriginal@reddit
Not sure how this would even be realistic. How is AMD supposed to perform QA testing for other companies products. But I agree that they should probably be more strict on said guidelines
luuuuuku@reddit
Like Intel did it? Intel had a similar ish problem, just not as critical.
scielliht987@reddit
Run some tool to check voltages in various scenarios.
Maybe that's how the end user could find out if their motherboard is compliant.
sg3707@reddit
I have asrock and amd 9800x3d. How cccccan I avoid this issue?
luuuuuku@reddit
Disable PBO and EXPO. But that will make you lose 10-30% performance
3G6A5W338E@reddit
Ensure you're running latest bios.
teh_drewski@reddit
Update to the latest BIOS or turn PBO off.
MongoBongo25@reddit
I have an ASUS MB, how do I check if the board is not over volted and which correct settings to apply?
luuuuuku@reddit
Disable PBO and Expo.
Stennan@reddit
Contact ASUS support with the question or ask in their forums. Each MB makers have their own names for settings.
In the past ASUS had an oopsie where if you used XMP/EXPO that raised VRAM voltage it also raised SOC/CPU voltage, resulting in some fried 7800X3D CPUs which was the most voltage sensitive model. Asus fixed that after Gamers Nexus did a video about it as some customers were getting denied warranty.
If you are on default settings or only have expo enabled you should be fine. Just update BIOS to be sure as AMD can make some updates/fixes on their part.
zir_blazer@reddit
Fresh from the oven involving ASUS boards: https://gmplib.org/gmp-zen5
Pillokun@reddit
I dont care if it burns, as long as there are no external damages u just go to the store and return it as I have not damaged it and all the settings in the bios/software are there given to us to play with, ie demed safe.
Tmoney511@reddit
LOL
jezevec93@reddit
I mean... It happens exclusively to Asrock. So unlike previous intel situation it seems believable.
seansafc89@reddit
It wasn’t exclusively, but it was massively weighted towards ASRock boards.
Strazdas1@reddit
when 97% of failures is one manufacturer it can be accepted that other 3% may be user error.
Reactor-Licker@reddit
Initially, it affected all boards as far as I understand, but everyone except ASRock has fixed it with BIOS updates.
Tgrove88@reddit
A guy just posted a dead 9800x3d on Asus mobo on the Asus forums just yesterday
w_StarfoxHUN@reddit
If its just one example that could be anything, faulty board, or cpu, etc not necessarily this issue.
bizude@reddit
It could even be a faulty AIO. I had an ASUS AIO with a USB motherboard connection which was electrically defective, it caused damage to both the MSI motherboard and killed the Ryzen 9 9950X3D CPU I was using.
rstune@reddit
Curious to know if you got Asus to pay for the damage
bizude@reddit
Oh, I'm pretty sure I'm now blacklisted by ASUS.
rstune@reddit
Oh boy, it was that bad eh? I was hoping that Tech Jesus' admonishment would set them right but I guess that they just PR'ed their way through that storm and went back to their old ways.
bizude@reddit
I barely managed to not get myself fired from TH for the words I used - I was extremely angry.
I wasn't quite emotionally stable at the moment, I had just dumped my girlfriend and I was still recovering from the evil I faced while I was in Arkansas if y'all remember how I was rambling back around the end of last year and beginning of this year.
panzermuffin@reddit
I, random dude from southern Germany. I wish you all the best.
bizude@reddit
Thank you. I certainly owe a debt of thanks to many in Germany, some of y'all made reports to the US consulate out there which literally saved my life when I was in Arkansas.
rstune@reddit
Oh wow, sorry to hear that. Hope things are better now. I was more referring to Asus' RMA service.
bizude@reddit
Yeah, I've heard the products they sample often come from their RMA department. Total incompetence there, that's for sure.
dertechie@reddit
Yeah, ones and twos isn’t necessarily indicative of a problem.
ASUS is still in my doghouse after two of their X570 boards failed on me in a row. They were also IIRC the brand that was blowing up the most chips when the 7800X3D launched because they were shoving too much voltage.
_gabber_@reddit
yeah and a couple of days ago seven different people posted on the asrock sub about their dead cpus, in just one day.
Tgrove88@reddit
How many total so far?
_gabber_@reddit
who knows? not everyone posts on reddit, imagine how many get sold around the world, and then taken back to shops, because the pc doesn't work.
megathread #1 -> summary (first phrase)
megathread #2 opened just a few days ago, feel free to read through.
Sevastous-of-Caria@reddit
Old bios or just overriten power limits for Manual OC?
Tgrove88@reddit
Idk I just know it affects all the mb vendors but seems to be mainly asrock
Kougar@reddit
ASRock was just the latest. ASUS and GB were both frying chips back at AM5's launch, for multiple reasons no less. It's nuts to me that every single board maker can't seem to adhere to basic specifications anymore, on both Intel and AMD platforms.
king_of_the_potato_p@reddit
I had 2x asys boards that were way over volting the 2x 9700x i went through with them.
2 sets in a row did the exact same thing after bi9s update, pushed the baseline voltage above even the safe range for temporary boosts.
Third cpu went with the msi edge ti, no issues, voltage in the ranges.
Vaibhav_CR7@reddit
amd can get away by producing defective cpus by putting asrock under the bus same thing happened with asus and gigabyte lot less lets hope someone can replicate the dying cpus and investigate it further
Amuro__6@reddit
Huge shocker, it wasn't AMD but the only vendor, Asrock, where chips have been dying
MrDunkingDeutschman@reddit
Asrock must have a very poor clickthrough rate in the YT analytics of the usual hardware channels because you know if this was Asus, we'd have them demanding to see the CEO in person to take accountability.
HisDivineOrder@reddit
Then AMD dollars and free x3d chips might stop flowing if they talk about chips burning up no matter who's to blame.
Now if they were all buying their own hardware, they'd be eating up them views for sure.
edparadox@reddit
This is plain disinformation.
AutoModerator@reddit
Hey edparadox, your comment has been removed because it is not a trustworthy benchmark website. Consider using another website instead.
I am a bot, and this action was performed automatically. Please contact the moderators of this subreddit if you have any questions or concerns.
CatsAndCapybaras@reddit
lol, amd bucks. UBM, is that you?
Klutzy-Residen@reddit
ASrock is a subsidiary of ASUS, so being a bit more creative with the title would have solved that issue.
Lakku-82@reddit
ASRock is not a subsidiary of Asus. The co-founder of Asus spun off to create his own company which was ASRock.
Klutzy-Residen@reddit
I wasnt entirely wrong, but Wikipedia seems to be a pretty bad reference in this case.
ASRock is a subsidiary of Pegatron which Asus owns a 16.84% stake in which makes it the largest shareholder.
https://www.pegatroncorp.com/content/company-profile/73
airinato@reddit
Ya so that is otherwise known as being entirely wrong.
Capable-Silver-7436@reddit
Considering it's only happening to ass rock yeah I can believe this. They always do something fucky
Zenith251@reddit
I want to love ASRock, but they keep making it so hard.
They had/have the cheapest 9070XT in most markets, and it's not a bare bones, loud ass 2 fan solution.
They supported ECC on all of their desktop AM4 motherboards (at least the 500 series boards). All of them, so they make a handy NAS boards when users retire them. (Almost all non-APU-based Ryzen AM4 chips support ECC. And any APU with Pro In the name).
Like ... You folks do some neat stuff ... But y'all screw up a lot.
glitchvid@reddit
I really like ASRock's offerings, they used to have post code displays even on mid and low end boards for most of AM4 which was a big reason I put them on parts lists for new builders.
They also have really neat AM4/AM5 server offerings, and were basically the first to do so.
Just sucks they're burning up CPUs now and won't take responsibility.
JuanElMinero@reddit
They also had offered largest selection and cheapest entry into full PCIe 5.0 support with X670E/B650E.
Which was a nice response to the rest rolling along with AMDs crappy segmentation, only making a small selection of needlessly expensive boards.
AnshinAngkorWat@reddit
Asrock had a really good thing going. They had the best value AM4 boards toward the end of the platform's life, as well as launch AM5 boards. But then they just had to screw it up badly.
teh_drewski@reddit
They still do have by far the best value AM5 boards IMO
Hopefully they've got the frying CPU issue fixed because it would be a massive shame if they lose their market position because of one very stupid mistake.
king_of_the_potato_p@reddit
Nope, had 2x asus rog strix b850e boards that killed 2x 9700x by way over volting.
Hwmonitor showed voltage a fair bit above the safe range.
corgiperson@reddit
Ass rock is awesome lmao
Professional-Tear996@reddit
So it is not just an issue with AsRock messing up their BIOS.
It is also the only thing new that has been stated, and the rest of it is a repetition of what AMD said before.
It's funny that they're trying to somehow obfuscate this issue as harder to pin down because they say that they support more CPUs across generations on a platform vs the competition.
rUnThEoN@reddit
Does it only happend to lga or also to pga?
dfv157@reddit
Only AM5, and only started happening after 9000s cpu came out, but appear to affect 7000s too if you look through the asrock sub
rUnThEoN@reddit
I am so happy about my pga cpu and standard pcie powered gpu :-)
Codys_friend@reddit
If we all want vanilla, the same flavor, the same performance, then forcing board makers to adhere to AMD's recommendations is fine. Many enthusiasts are building pcs to tinker with them, to make adjustments to make them a little better a little different. This is why we have options in bios to make changes.The "p" in pc stands for personal, we want to make it ours. Manufacturers understand this and create products to sell into the enthusiast market to meet this need.
The manufacturers need to be mindful of the risks they take and not go to the extreme of torching cpus and mobos (a la Intel). The manufacturers are trying to balance performance vs longevity. In general they do a good job. The problems with Asrock boards effected a small percentage of boards. A significant number of boards were impacted to be sure, but still a small percentage. The estimates I've seen are perhaps 1 to 2%. Had it been as widespread as some think, 50%+, Asrock would have had to pull their boards immediately because that high a failure rate is unsustainable financially AND it will destroy the brand's credibility.
It is helpful to keep perspective on the magnitude of the problem and why the manufacturers are enticed to take risks. The board makers are appealing to builders who are looking for competitive advantages. Sometimes, the engineers fly a little too close to the sun. This is similar to the problem Nvidia has with the 12vhpwr connectors. There is no doubt there is a problem, yet the vast majority of 5090 cards run fine with no melting cables. Most people with Asrock or other boards suffered no failures. If you were someone whose board/cpu failed, it was a major problem and unacceptable. I get that. It is important to keep perspective.
I encourage the designers, engineers, manufacturers to push the envelope to wrest better and better performance from their designs and products. When they do this we all benefit. Amd occasionally, mistakes are made. The test is how people and companies respond to the mistakes. Will we own our mistakes, make amends, learn from the mistakes and move forward? Or do we want to prosecute and persecute those that made the mistake? If companies like Asrock were more forthcoming in publicly owning the mistake, then it might diffuse things. We are all in this together and need to work together.
Thus endeth today's sermon!
NewestAccount2023@reddit
It's the difference between you choosing too high a voltage yourself and the board doing it hidden inside bios code without any indication it's doing anything different or dangerous
Codys_friend@reddit
I haven't read of Asrock denying an rma for a failed board. Asrock should have gotten in front of this sooner. When it started to happen there was a dearth of data so a root cause analysis could pinpoint the cause. I would like to see manufacturers step up more quickly. It does take time to identify the real cause of the problem. As I said, if a large percentage of boards had been affected, then I believe the response would have been different.
If you were affected by this problem, then it was catastrophic. I don't dispute that. There were thousands of other Asrock board owners with 9800x3d chips who were interested but not impacted. My point is we need to have a balanced perspective. We can take a zero tolerance attitude and as a consequence we will see prices skyrocket and innovation stifle.
ChoMar05@reddit
AMD is at fault. Their CPUs are unreliable within spec. And I say that as an AM5 user. If they could identify the culprit and it weren't them, they'd at least be less vague. Oh, and AsRock? Maybe. But all we have is Reddit Data because Reddit is the last text-based Infosphere. It's the last thing that can be Googled. It's absolutely unreliable because once a Reddit thread has reached a certain threshold, it's "the truth," and everything that points in a different direction is ignored. IF it was Asrock, AMD could say so, even if it was indirectly by naming a specific BIOS-setting or anything else. The fact that they don't is a more reliable "source" than Reddit could be.
AutoModerator@reddit
Hello imaginary_num6er! Please double check that this submission is original reporting and is not an unverified rumor or repost that does not rise to the standards of /r/hardware. If this link is reporting on the work of another site/source or is an unverified rumor, please delete this submission. If this warning is in error, please report this comment and we will remove it.
I am a bot, and this action was performed automatically. Please contact the moderators of this subreddit if you have any questions or concerns.