AMD says overclocking blows a hidden fuse on Ryzen Threadripper 7000 to show if you've overclocked the chip, but it doesn't automatically void your CPU's warranty
tomshardware.com
AMD says overclocking blows a hidden fuse on Ryzen Threadripper 7000 to show if you've overclocked the chip, but it doesn't automatically void your CPU's warranty::AMD explains the hidden fuses behind Ryzen Threadripper 7000 processors and how it will handle warranty claims.
So they'll arbitrarily decide what to cover and always have a reason for denying coverage, got it.
They only have that option if you run the cpu outside of design spec. Rambunctious o/c’ers no longer get a free replacement at AMD’s expense, and helps amd figure out if there’s a problem with cpus if they are failing and are not o/c’d.
Yeah, I don't really see much of an issue here. If you get a defective chip back, it's probably a good data point to know if it was "abused". Even if it's just so you can ask more questions, or prioritize problems that show up on non-OC'd chips rather than flat rejecting an RMA.
The design spec of a CPU is the clock speed it runs at coming from the factory, overclocking by definition means going above it - that's why it's called overclocking.
I don't o/c my 7700x. I have no need to and I want longevity. I'd have even less of a need to o/c a thread ripper!
No, only if you've overclocked and the fault is one that could be caused by overclocking. Honestly, I think it's absolutely fair for them to not replace a CPU if you killed it yourself by running it outside of spec. Most manufacturers would just say you're SOL if your CPU dies and it's ever been overclocked, even if the issue is unrelated, so I think this is a pretty good policy for them to have.
Granted, it's possible that they could always say that it was the overclocking that killed it whenever the fuse is blown, but considering how heavily AMD leans into customer good-will and positive sentiment about the company, they will probably try their best to honor any legitimate warranty claims.
That's a bit shitty but hopefully they don't just use it as a trap to deny any warranty coverage on an overclocked CPU.
Meanwhile Intel will void your warranty if you've enabled XMP. I don't know if they have a way of telling if you did so or not but they will try to trick you into admitting it when you're asking for an RMA.
Reviewers really should say "ok, well if it's not covered by warranty then we'll just do CPU benchmarks at the minimum JDEC speeds, as the manufacturer recommends"
Any good reviewer should already be doing a typical non-OC’d benchmark and an OC’d benchmark anyway.
The majority of people don’t overclock so would only care about the stock performance anyway. And overclockers should recognize that if you damage the chip by pushing it too far, it shouldn’t be covered.
Most people don't consider enabling the advertised memory clock speeds as an overclock.
We aren't talking about taking your CPU and overclocking it. We're talking about a simple UEFI checkbox that everyone is told to do.
Who the fuck is “we” here? Because the article is about CPU overlocking. I don’t give a fuck about the parent comments offhand comment about Intel. Intel is irrelevant here.
Your comment I replied to was about reviewer CPU benchmarks.
Who the fuck is we is literally the entire industry. Intel, AMD, every reviewer I have ever seen. Everyone.
Seriously, look at ANY review. They're all done with XMP or DOCP profiles set, just as the CPU manufacturer, motherboard manufacturer, and memory manufacturer recommends.
I don't give a fuck about what your offhand opinions are, I'm taking a out the reality. And the reality is that everyone is told to enable XMP.
...
You know memory speed impacts CPU performance right?
Enabling XMP isn’t overclocking the CPU. It wouldn’t blow this fuse.
So have you just abandoned what you said above, or are you just ignoring it?
AMD doesn't consider it CPU overclocking, no. Intel does. That's what I was replying to, as you very well know.
Intel is not in the article. Literally nothing about this post is about Intel other than an offhand remark about XMP.
The article is about AMD CPUs. I could not give a flying fuck about Intel.
I'm replying to the comment. In case you haven't noticed this is a public discussion, and when talking about AMD CPUs, it's common to bring up their only competitor, Intel CPUs.
If you didn't want to talk about intel CPUs, don't reply to a comment about intel CPUs. You inserted yourself into this discussion about Intel's practices then got angry that we were talking about Intel. Amazing.
It's a very reasonable thing to happen in a discussion. Do you know how those work? Talk about AMD leads to talk about Intel. It's not rocket science. You don't need to come in and start screaming that Intel is a forbidden topic.
Seems to me like you're just grasping for something after I dismissed your comment as nonsense.
Well what we're talking about here is just memory speeds, not core overclocking. If you're building a computer and you're paying for RAM that is rated at a certain speed, you need to enable XMP to have it run at that speed. Since the memory controller is now integrated into CPUs, intel considers that overclocking so it voids your warranty. I think most people who are buying CPUs to build their own PCs know this and will not run at base JEDEC speeds.
That's definitely not common knowledge for people who build their own PC.
It definitely is.
Every single review and YouTube video, even from channels with broad appeal like LTT and the like always talk about the need to enable XMP and talk about it having to be enabled to get the advertised performance.
It gets advertised on memory kits and motherboards and they provide easy instructions on how to do it.
It's common knowledge to enable it.
Eh, yeah maybe you're right but it's such a tremendous amount of performance to lose out on for a couple keystrokes. Any halfway decent guide for beginners should be mentioning it but I don't know how people outside my circles build computers. Do they read/watch guides? Do they just plug shit together and pray that it works? 🤷♀️
It's very prominent in any build guide, on even casual PC youtuber videos, in motherboard manuals, on ram kits.
It's absolutely common knowledge to enable XMP, I dunno what that guy is smoking.
There is literally nothing in the article about memory speeds
It’s entirely about overlocking the CPU .
The only thing about memory is your offhand comment about Intel and XMP which is entirely irrelevant to the article.
Hah yeah actually, that should become the standard for Intel CPU reviews.
It's not shitty, it's fair. If damage is caused by the overclock why should the manufacturer foot the bill? You modified the product to run outside the specs!
The "shitty" part of it is it's a binary one time feedback. If the fuse blows that's it. It doesn't matter if the CPU failed for something else the fuse can't unblow. I don't know what type of fuse they're using, would it blow with any level of over clocking, or with an extreme amount, is it a time delayed fuse that requires a bunch of time over clocked or is it instant? If i want to over clock just a bit but test it at a higher clock rate before setting my desired speed will that blow the fuse? The only point of the fuse is to determine if the user "missused" their cpu at any point.
It's probably a collection of fuses instead of just a single one. One for xmp, one for each of the pbo options, various ones for manual OCs. I'd guess there's tiers of how aggressive the OC is, maybe a counter for how many times it was booted with that OC enabled.
I wonder how much extra cost that would add to CPU production. There's probably some cost benefit analysis looking at the saving from denying warranties to the cost of extra components on the chip.
I think they'd scale well so it wouldn't have that big of an impact. Like it could be one set of fuses for the entire chip. Even a KB of those fuses wouldn't take up much area on modern chips. That's if they are detecting settings or overall chip power.
If they are detecting OC damage to circuitry, that might involve a lot more fuses throughout the chip along with circuitry to read them (or at least detect their state), which could be more involved. Though there is already circuitry to test the functionality of the chip at a fine level for binning and QC, and it might be trivial to add some fuses to that.
It's a bit shitty because we then have to trust that they won't use this as an excuse to void the warranty on chips that had a fatal defect to begin with. Overclocking is pretty safe unless you're doing extreme overclocking and they won't say how they determine if a failure was caused by an overclock or not.
It's definitely "more fair" for AMD than Intel to do it since they don't charge a premium for unlocked processors but I still don't like it. They developed PBO, it's a feature included with the CPU I bought, I want to be able to use it without fear of losing my warranty, but even just enabling that will trip that fuse.
If they start selling new ones with the proper terms of sale ("overclocking voids warranty") then there's nothing wrong with that.
It would definitely be within their rights to do so.
Not in every market. That wouldn't fly in the EU. They'd only be able to deny warranty claims if they could prove that the overclock is what broke the chip
Dunno whether it's uniform all over the EU but in Germany the burden of proof shifts from the manufacturer to the consumer a year after sale, that is, if you want to rely on AMD having to prove that it was the overclock you better break the thing fast.
Probably not, it was just a way of saying that there is absolutely something wrong with that.
It's just like a car having an odometer. This would come in handy when buying second-hand, remember all the uncertainty about the condition of used GPUs?
(That is assuming they make the state user-readable though.)
That's actually a nice idea, if it's, as you said, user readable in eg. CPU-Z
this should be industry norm there is way too many bad users
A lot of people in the comments seem to not understand that overclocking IS running the hardware outside the specs, unless that hardware was specifically meant to do that.
It's exactly like overclocking GPUs in the late 90's, for them to fry themself after a month. You went outside ther specs for doing that. Even if I replace the speaker from a telephone, and the keypad stops working a month later, I have voided the warranty already by doing the speaker change as they can't know if the now not working keypad was done because of you or a failing unit.
This is false. They have to PROVE that the repair that you made caused the keypad to fail in other to void your warranty, at least in the USA. Most people are misinformed as you are however, and they'll TELL you that's the case to make you go away, despite it being illegal.
https://en.wikipedia.org/wiki/Magnuson-Moss_Warranty_Act
warrantys are optional buy it or dont or get out of my mom and pop Microcenter
eating a piece of bread after the best by date is also using hardware outside the specs
The bakery does not know when the bread will mold, but they know it probably won't start to for a certain number of days. The best by date is that day minus four to seven.
exactly
Do you even need to overclock a Threadripper?
Why not if you've got thermal space to use? Overclocking will help with single-core applications where it would struggle more otherwise. It's also just a general boost to performance for free.
I kinda wish it were possible to overclock a single core and be able to direct single-threaded processes to it. I understand how CPUs and clock speed works, I'm just saying it would be cool.
That said, as I sit here thinking about it, it might he possible to have a core that uses a higher-frequency harmonic as part of the architecture of the chip. It might need a larger L1 and some special transport architecture to step the processed data back to the lower clock speed, but I don't think there are any physical or computational reasons it shouldn't be possible.
Is the Auto OC covered by this or are they talking manual OC?
I hope they don't screw the auto OC. I have never OCed but my motherboard did decide once that my normal default core frequency of 3.2GHz was too low, and did an automatic unstable OC which resulted in a nice and high 1,2 GHz -__-. I had to switch it to a manual OC, and then back to automatic and it was fixed.
Man the Intel Core Duo 2 was a snapshot in time where overclocking was acceptable.
Those chips could go, and that rerelease just kept that train going.
Mmm. Raising prices. Implementing anti-consumer tactics. Where have I seen this before....?
Oh, I know. When a company becomes the bad guy. Just like the Intel monopoly from 2011-2018. Or Nvidia. Or microsoft in the 90s. Or Google or Amazon now.
Remember when checked bags were free on airlines and they didn't nickel and dime you to death? When seats had room? When exit rows were free? This happens all the time, and it's never a good change.
Your airline example fails when you account for ticket price trends and access to flights for the poor.
Is it really that equitable, though? Some airlines hold a monopoly on certain routes, and people fly much more often now, to the tune of a couple hundred dollars round trip. Cars and planes doomed our already-faltering rail lines, which were our best shot at low cost, low carbon transportation. We can still do it, but we've coughed up a ton of money to a few air carriers when we used to have a booming consumer rail network.
Anti-consumer tactics of having mechanisms to detect when a customer has abused the product they might try to get replaced for free, that they say in this article won't automatically mean the warranty is rejected?
You're free to overclock your computer hardware, but you take the risk on yourself when you do so. It's always been this way. If you want to be covered by warranty, keep it in spec, hit "no" when the warning comes up that enabling this feature can void your warranty.