Wednesday, December 13th 2023

Threadripper Overclocking Blows a Hidden Fuse, AMD confirms: Warranty not Voided

According to Tom's Hardware, today we are finding out that AMD Ryzen Threadripper 7000 series processors, codenamed Storm Peak, including Pro and non-Pro SKUs, blow a fuse on the chip when overclocking is enabled. Modern microprocessors have dozens of fuses that are used to store information inside the chip. For example, the factory stores the per-processor default voltage information in the fuses. On downgraded graphics chips, the shaders get disabled through such fuses, too. These fuses are not like your household circuit breakers—they will blow only when a specific command is sent to the processor, there is no way for them to break accidentally through system crashes or power spikes. In the case of Ryzen Threadripper 7000, the BIOS code will blow a fuse when the user enables overclocking in the BIOS settings, it reacts only to the user-initiated UI change, not to any kind of measurement. Before that happens a warning is shown. AMD uses this mechanism to see any indications if any kind of overclocking has been done to the processor.

While the messaging might suggest otherwise, just enabling overclocking does not void all warranties. In a statement to Tom's Hardware an AMD representative confirmed: "Threadripper 7000 Series processors do contain a fuse that is blown when overclocking is enabled. To be clear, blowing this fuse does not void your warranty. Statements that enabling an overclocking/overvolting feature will "void" the processor warranty are not correct. Per AMD's standard Terms of Sale, the warranty excludes any damage that results from overclocking/overvolting the processor. However, other unrelated issues could still qualify for warranty repair/replacement," noted the spokesperson. Ultimately, overclocking and overvolting by themselves will not cause the owner of AMD's Ryzen Threadripper 7000 series CPU to lose the right to repair and seek AMD's help. Other factors, such as damages induced by overclocking, will be a warranty-voiding factor though. These can occur from constant overheating, which significantly lowers the life expectancy of the CPU.
Source: Tom's Hardware
Add your own comment

73 Comments on Threadripper Overclocking Blows a Hidden Fuse, AMD confirms: Warranty not Voided

#51
tabascosauz
RJARRRPCGPThey got signs, if it's degraded, especially severely:

The most suspect errors would be: (If any of the same errors continue, even at stock core clocks)

1: A machine check exception error (known as a WHEA error by Microsoft) (WHEA is another term of Microsoft alphabet soup. It stands for Windows Hardware Error Architecture, even though, I though it was "Windows Hardware Error Assessment")

2: "0xA"/"0xa" BSOD ("IRQL_NOT_LESS_OR_EQUAL" BSOD, which is a known CPU fault error, only seen with an unstable core OC, usually. You can get that error code during a failed Linpack core test)
1. WHEAs happen all the time on AM4 with zero user input. Just a bad CPU with bad cores (Cache Hierarchy) or bad uncore (Bus/Interconnect).
2. IRQL_NOT_LESS_OR_EQUAL has nothing to do with OC. Dying CPUs, unstable memory setups and physically defective boards do it aplenty.

Bottom line, AMD RMA staff are the last people on earth to be considered qualified enough and giving enough of a shit to enforce any kind of general policy about PBO/OC = no warranty. I'm pretty sure most of them are in some kind of call center in Asia anyway. Just firmly feed them enough technical jargon to get past the one or two canned responses about "troubleshooting", and the fast RMA process gets going (iirc even cross shipped sometimes).
Posted on Reply
#52
RJARRRPCGP
tabascosauz1. WHEAs happen all the time on AM4 with zero user input. Just a bad CPU with bad cores (Cache Hierarchy) or bad uncore (Bus/Interconnect).
2. IRQL_NOT_LESS_OR_EQUAL has nothing to do with OC. Dying CPUs, unstable memory setups and physically defective boards do it aplenty.

Bottom line, AMD RMA staff are the last people on earth to be considered qualified enough and giving enough of a shit to enforce any kind of general policy about PBO/OC = no warranty. I'm pretty sure most of them are in some kind of call center in Asia anyway. Just firmly feed them enough technical jargon to get past the one or two canned responses about "troubleshooting", and the fast RMA process gets going (iirc even cross shipped sometimes).
I usually see "Cache Hierarchy Error" on Ryzens! It's the meme of Ryzen, it seems!
Posted on Reply
#53
LabRat 891
tabascosauz1. WHEAs happen all the time on AM4 with zero user input. Just a bad CPU with bad cores (Cache Hierarchy) or bad uncore (Bus/Interconnect).
2. IRQL_NOT_LESS_OR_EQUAL has nothing to do with OC. Dying CPUs, unstable memory setups and physically defective boards do it aplenty.

Bottom line, AMD RMA staff are the last people on earth to be considered qualified enough and giving enough of a shit to enforce any kind of general policy about PBO/OC = no warranty. I'm pretty sure most of them are in some kind of call center in Asia anyway. Just firmly feed them enough technical jargon to get past the one or two canned responses about "troubleshooting", and the fast RMA process gets going (iirc even cross shipped sometimes).
That was a/the symptom w/ my 1st R5 5600. Disabling a core (5-core), resolved all issues.

Regardless of detailed-cause, after the tiniest amount of (expected) 'hassle', verifying my purchase/ownership
AMD replaced my R5 5600 in a very reasonable amount of time, and gave me a BNIB 'retail' replacement while specifying that I not include my stock cooler.

No cross-shipping but, I did get a free Display Box, Stock Cooler, and best of all Case Badge :D
Posted on Reply
#54
AsRock
TPU addict
bugMaybe so, but how can the user prove damage was or wasn't caused by overclocking/overvolting? You send in a defective CPU, AMD sees the blown fuse and claims it broke because of overclocking/overvolting. There's no way for the end user to argue.

And to be crystal clear, I'm not saying AMD plans to abuse that. I'm just saying that fuse opens the way for at least some distributors to go that route. At the same time, it's probably a useful tool in diagnosing, because users sending in defective parts are very unlikely to admit they overclocked before the damage happened.
Maybe the user did not overclock it and ASUS just thought what the hell and spiked your chip HAHA.
Posted on Reply
#55
Redwoodz
Well, you can thank two parties for this. The one who melted a 5800X3D and the mobo vendors who overclocked their SoC voltages with XMP.
Posted on Reply
#56
Jism
I wonder if there's a software based tool we or consumers can use to determine if a fuse has bin blown off or not on TR's / Epyc CPU's.
Posted on Reply
#57
Eskimonster
RedwoodzWell, you can thank two parties for this. The one who melted a 5800X3D and the mobo vendors who overclocked their SoC voltages with XMP.
Don`t you mean 7800X3D ?
Posted on Reply
#58
bug
AsRockMaybe the user did not overclock it and ASUS just thought what the hell and spiked your chip HAHA.
Read W1zzard's post/explanation ;)
AquinusEhhhh, I think you're being too kind. I seriously doubt that they did this just to gather statistics. At best, it tells them to not even bother doing testing on an RMA'ed CPU where that bit has flipped. At worst it could deny you an RMA. If AMD was serious about it not impacting RMA, they'd remove all the verbiage regarding this, but they didn't. This feels like smoke and mirrors.
I said "statistics at first". Depending what those reveal, they can work with manufacturers than have a higher than average failure rate or something like that.
To me, this is about visibility: most users won't admit they overclocked it, once it goes belly up.
Posted on Reply
#59
W1zzard
Count von SchwalbeI have no idea what size they are; I know there are some SMD components on the package (hence the odd shape of the Ryzen 7000 IHS) but I do not know if they are resistors, capacitors, fuses, magic smoke containers, or nuclear reactors. If they are part of the die, then which die is it? The CCD is universal - used in Ryzen as well, and AFAIK the same BIOS warnings pop up about voiding warranty on standard consumer boards.
There are package fuses, which are just resistors on the green substrate, these can be switched during manufacturing, by adding/removing the resistor
Another variant is laser-cut fuses, which are just a line on the substrate that a laser cuts through, that you can fix with conductive paint
and there's die-fuses, which are inside the silicon, not user-visible, can only read them with an electron-microscope, or if you know how to read them through software
Posted on Reply
#60
trsttte
W1zzardThere are package fuses, which are just resistors on the green substrate, these can be switched during manufacturing, by adding/removing the resistor
Another variant is laser-cut fuses, which are just a line on the substrate that a laser cuts through, that you can fix with conductive paint
and there's die-fuses, which are inside the silicon, not user-visible, can only read them with an electron-microscope, or if you know how to read them through software
These are written at runtime as a consequence of user action, i.e. overclocking, so the third kind
Posted on Reply
#61
AsRock
TPU addict
bugRead W1zzard's post/explanation ;)


I said "statistics at first". Depending what those reveal, they can work with manufacturers than have a higher than average failure rate or something like that.
To me, this is about visibility: most users won't admit they overclocked it, once it goes belly up.
Sorry i did not mean this specific user but everyone else including my self were ASUS were pumping high volts by default.
Posted on Reply
#62
THU31
What about undervolting? What action actually blows the fuse? Going above max boost multiplier, or just changing anything from auto?
Posted on Reply
#63
Redwoodz
EskimonsterDon`t you mean 7800X3D ?
Yes, thank you.

It's kind of funny when you think of it. I mean what do you think when you press the "I accept" button in the bios warning when overclocking is enabled? Save it in memory? A burned up cpu can have no memory. No, you have a "fuse". ;)
Posted on Reply
#64
Aquinus
Resident Wat-man
AleksandarKOther factors, such as damages induced by overclocking, will be a warranty-voiding factor though. These can occur from constant overheating, which significantly lowers the life expectancy of the CPU.
Will people read these two sentences again and stop saying that you can't void your warranty? To me this says, "overclocking won't void warranty, but breaking something because you were overclocking will."

Again, this is idiotic smoke and mirrors I'm calling out and others should too.
Posted on Reply
#65
R-T-B
Count von SchwalbeI have no idea what size they are; I know there are some SMD components on the package (hence the odd shape of the Ryzen 7000 IHS) but I do not know if they are resistors, capacitors, fuses, magic smoke containers, or nuclear reactors. If they are part of the die, then which die is it? The CCD is universal - used in Ryzen as well, and AFAIK the same BIOS warnings pop up about voiding warranty on standard consumer boards.
efuses are typically on silicon.
FoulOnWhiteI reckon the actual fuses are on one of the dies and so tiny you'd need a microscope to see them.
This.
Posted on Reply
#66
Count von Schwalbe
R-T-Befuses are typically on silicon.
So these are definitely present on consumer Ryzen chips. Or maybe on the TR IOD?
Posted on Reply
#67
R-T-B
Count von SchwalbeSo these are definitely present on consumer Ryzen chips. Or maybe on the TR IOD?
Possibly the IOD (that would frankly make sense). But you won't be able to replace them for sure.
Posted on Reply
#68
Count von Schwalbe
R-T-BPossibly the IOD (that would frankly make sense). But you won't be able to replace them for sure.
No doubt. But if it was in the CCD, then there is no reason AMD couldn't check on Ryzen just as much as TR.
Posted on Reply
#69
trsttte
Count von SchwalbeNo doubt. But if it was in the CCD, then there is no reason AMD couldn't check on Ryzen just as much as TR.
This are usually written by firmware, they're called fuses but they don't really work as fuses. My understanding is Threadripper is programed to set a "fuse" when overclocking happens, Ryzen maybe is not.

We're spitballing but they could well be on the CCD but while that flipped bit indicates overclocking in threadripper, in ryzen it indicates something entirely different.
Posted on Reply
#70
Abby Normal
W1zzardNo, this is not possible, we've updated the article accordingly


The BIOS code will blow the fuse when the user selects "overclocking: enabled" in the BIOS and saves, it does not measure anything or react to system parameters
Yeah, unfortunately it seems if I want to apply EXPO timings, that sets the BIOS flag off according to AsRock's Bios disclaimer. The only thing I've tried is setting EXPO the XMP profiles, and it gives me the overclocking mode nag message. I did not set anything in the AMD Overclocking section that is shown below the message, I only set the memory to an EXPO profile, which doesn't seem to set anything to enabled in the AMD Overclocking section in the BIOS either, it just triggers the pop up on pressing F10 and saying yes to saving. Pressing no on the popup just leaves it at DDR5-4800 though it will show the EXPO profile timings box, it just ignores it.

The wording definitely makes it sound like a permanent action.
Posted on Reply
#71
W1zzard
Abby NormalYeah, unfortunately it seems if I want to apply EXPO timings, that sets the BIOS flag off according to AsRock's Bios disclaimer. The only thing I've tried is setting EXPO the XMP profiles, and it gives me the overclocking mode nag message. I did not set anything in the AMD Overclocking section that is shown below the message, I only set the memory to an EXPO profile, which doesn't seem to set anything to enabled in the AMD Overclocking section in the BIOS either, it just triggers the pop up on pressing F10 and saying yes to saving. Pressing no on the popup just leaves it at DDR5-4800 though it will show the EXPO profile timings box, it just ignores it.

The wording definitely makes it sound like a permanent action.
That sucks, maybe contact AMD and ASRock support to get some confirmation on your warranty status, just in case.

If you do, please share here what they wrote
Posted on Reply
#72
Bond008
bugMaybe so, but how can the user prove damage was or wasn't caused by overclocking/overvolting? You send in a defective CPU, AMD sees the blown fuse and claims it broke because of overclocking/overvolting. There's no way for the end user to argue.

And to be crystal clear, I'm not saying AMD plans to abuse that. I'm just saying that fuse opens the way for at least some distributors to go that route. At the same time, it's probably a useful tool in diagnosing, because users sending in defective parts are very unlikely to admit they overclocked before the damage happened.
100% right.
Posted on Reply
#73
ThrashZone
Abby NormalYeah, unfortunately it seems if I want to apply EXPO timings, that sets the BIOS flag off according to AsRock's Bios disclaimer. The only thing I've tried is setting EXPO the XMP profiles, and it gives me the overclocking mode nag message. I did not set anything in the AMD Overclocking section that is shown below the message, I only set the memory to an EXPO profile, which doesn't seem to set anything to enabled in the AMD Overclocking section in the BIOS either, it just triggers the pop up on pressing F10 and saying yes to saving. Pressing no on the popup just leaves it at DDR5-4800 though it will show the EXPO profile timings box, it just ignores it.

The wording definitely makes it sound like a permanent action.
Hi,
Pretty much setting xmp profile voids warranty message lol okay as rock :kookoo:
Just having the option there is as rock legally liable hehe
Posted on Reply
Add your own comment
May 29th, 2024 18:31 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts