After 20 years, I had finally decided to build a PC myself (for my kids, really) and this was a star component in the shopping list.[1]
Now I will wait until these more serious bugs on the platform are ironed out. It doesn’t seem right to spend hundreds on supposedly higher end motherboards when it’s clear they lack basic protections. (And don’t seem to have been through QA for basic things?)
Unfortunately there is always a risk of ending up with defective components. In same ways this is a better outcome than getting a random defective chip, because the manufacturer/vendor cannot make the claim of user error as easily.
For what it's worth, these kind of manufacturer errors are very rare these days. Most PCs people build just work right away. And everyone involved is almost certain to get their full money back.
You could probably go with an AM4 board and a 5800x3d instead. Pretty close performance, a lot lower price, not many complaints about lack of magic smoke retention.
Sure, you lose out on future proofing, with no pci-e 5 and no ddr-5, but eh. Might be better to get into something at the end of the platform, rather than the beginning of a platform. Or you can stay on the wait and see train and maybe next year's systems will be a better choice. AM4 had issues for the first round or two, and Zen3 was pretty nice; it's a pattern that repeats (which is probably why AMD keeps sockets longer, Intel shows this sometimes too, but it's not so pronounced)
First, one of the boards failed with stock settings.
Second, EXPO is not really overclocking in the traditional sense. It's basically the AMD equivalent of enabling an XMP profile. It's not even supposed to modify the SOC voltage that caused the failure.
And if you watched the video, you'd know that not only did the ASUS board set the SOC voltage way beyond spec, but it had a complete lack of protection for the CPU in event of a failure.
* Failure of over-temperature protections
* Failure of over-current protections
* A motherboard shoving current into a CPU which clearly is dead (or at the least not booted) because it has not sent a "power good" signal.
* The motherboard setting voltages for vSOC that are way beyond AMD's recommended specs
* The motherboard setting the voltage for vSOC well beyond what it shows in the BIOS (ie, you think you're getting 1.3V, actually getting 1.4V+)
This is not "doofus overclocks CPU, kills it". This is "motherboard manufacturer completely inept at basic job, kills parts". And it's not just ASUS, though ASUS is the primary offender here.
I'm not sure we watched the same video. In the video I watched, the host explained how it was possible for this to happen entirely with default settings.
First, the only "overclocking" happening is the enabling of EXPO, which is similar to an XMP profile from Intel. It's not really overclocking. Also, it is not supposed to modify the voltage values seen. Also, one of the failed systems did not have EXPO on at all.
Second, this is a failure on multiple levels:
* Failure of over-temperature protections
* Failure of over-current protections
* A motherboard shoving current into a CPU which clearly is dead (or at the least not booted) because it has not sent a "power good" signal.
* The motherboard setting voltages for vSOC that are way beyond AMD's recommended specs
* The motherboard setting the voltage for vSOC well beyond what it shows in the BIOS (ie, you think you're getting 1.3V, actually getting 1.4V+)
The video explains that this is a serious failure of several motherboard manufacturers, though primarily ASUS. It's also a failure of AMD to have more stringent requirements on its partners.
I am by no means an expert, but I was under the impression modern motherboards and CPUs had protections/limits specifically to prevent this from happening (or at least make it very difficult).
I'm not sure if it's "worthy" of an HN front page or not, but it was at least interesting to me that those protections seemed to either not exist or fail in this case.
Right, back in the days AMD CPUs lacked thermal protection and would die quickly if the heatsink wasn't mounted properly[1]. Intel had thermal protection and would just clock down the CPU.
These days there's a lot of protection built in. For example, the voltage regulator ICs they use on the motherboard should all have thermal, over-current, under-voltage and other protection built-in.
So yeah I do think it's newsworthy that a modern CPU can burn itself up. I would not expect that, even overclocked.
The investigation in the video shows that on a number AM5 motherboards many of these protections don't work properly to the point of the motherboard creating a literal puddle of molten silicon inside the CPU. In addition the firmware of many AM5 boards is buggy and sets incorrect voltages for the system, for example certain ASUS boards set the SoC voltage way too high (1.4x V instead of 1.0 V). That seems to cause excessive power dissipation in some cases but can also end up triggering what seems to be a latchup (which would almost assuredly kill the CPU anyway, but is not supposed to start a fire).
I have been unlucky with my apple devices, and the fixes made by apple have all been sub-par. From poor soldering and lack of underfill to complete lackluster solutions and expensive fixes that should have been $20, but cost ten times more.
I though I just had had bad luck, but then I discovered that rossmanm guy's channel and was amazed they were actually doing a worse job on some laptops.
M2 is amazing wrt performance per watt, but I have sworn to never again buy apple.
My average gaming laptop comes with a cpu ranked identically with an m1 pro (my cpu is older than the m1), has 64 gb of ram and a dedicated gpu with 6 gb vram. Runs any os except macos, i can easily upgrade my ram, wifi and hdd, has a 140 hz refresh rate and loads of connectivity options, replaceable battery and a working keyboard with a numpad. Bought it for fun as it was dirt cheap by comparison. I am not convinced apple is top notch.
I mean benchmarks place it right next to an m1, but what you are saying is that marketing trumps real numbers? Also by old i mean 3 years or so. Do you really believe that apple silicon is the absolute best thing out there?
Now I will wait until these more serious bugs on the platform are ironed out. It doesn’t seem right to spend hundreds on supposedly higher end motherboards when it’s clear they lack basic protections. (And don’t seem to have been through QA for basic things?)
[1] https://uk.pcpartpicker.com/user/hnarayanan/saved/#view=VWWB...