AMD Ryzen 7 7800X3D CPU Burns Up

keyser van someone · 10 May 2023 at 12:16

Now we just have to test every motherboard soc voltage, with a meter, after telling it what we want and we're all good. Repeat for every future BIOS update.

eddiew · 10 May 2023 at 12:19

That new GN video is fascinating. Suddenly I wish I'd gotten into electronics failure analysis as a career :cry:

Does make me wonder though; we think that 1.45v will kill them fast, and 1.4 might do so only slightly slower - so exactly how confident is AMD that the new 1.3v cap is absolutely fine?

Also I guess I won't be buying an Asus board after all. The whole point of the Asus tax is that it's well made and backed with a good warranty. If it's in fact neither of these things, I'll just buy something else. Maybe an MSI board because I just plain like the dragon logo.

Silent_Scone · 10 May 2023 at 12:22

keyser van someone said:
Now we just have to test every motherboard soc voltage, with a meter, after telling it what we want and we're all good. Repeat for every future BIOS update.

And measure from the correct place.

jigger · 10 May 2023 at 13:07

eddiew said:
That new GN video is fascinating. Suddenly I wish I'd gotten into electronics failure analysis as a career

Does make me wonder though; we think that 1.45v will kill them fast, and 1.4 might do so only slightly slower - so exactly how confident is AMD that the new 1.3v cap is absolutely fine?

Also I guess I won't be buying an Asus board after all. The whole point of the Asus tax is that it's well made and backed with a good warranty. If it's in fact neither of these things, I'll just buy something else. Maybe an MSI board because I just plain like the dragon logo.

I’ve been very impressed with ASRock and Gigabyte of late.

Troezar · 10 May 2023 at 13:16

jigger said:
I’ve been very impressed with ASRock and Gigabyte of late.

I always used to have MSI motherboards but so far I've been pleased with the Gigabyte boards.

FredFlint · 10 May 2023 at 13:19

eddiew said:
That new GN video is fascinating. Suddenly I wish I'd gotten into electronics failure analysis as a career

Does make me wonder though; we think that 1.45v will kill them fast, and 1.4 might do so only slightly slower - so exactly how confident is AMD that the new 1.3v cap is absolutely fine?

Also I guess I won't be buying an Asus board after all. The whole point of the Asus tax is that it's well made and backed with a good warranty. If it's in fact neither of these things, I'll just buy something else. Maybe an MSI board because I just plain like the dragon logo.

I would go for MSI if buying now. My PC has been fine so far, I went with 5600 RAM as it was less likely to have issues and I plan to get a 64GB kit at some point. The SOC has always been < 1.28v, have set it to 1.2v now and it seems stable. Its like AMD rushed to get AM5 out and did not fully know the safe voltage as it’s a new architecture, nodes, and RAM. Most companies would not do all the changes at once as it increases risk, AM5 changed everything so it was never going to be risk free.

Harlequin · 10 May 2023 at 13:19

So TLDR from GN is Asus being Asus and burning up cpu`s since 2000.

Silent_Scone · 10 May 2023 at 13:42

Harlequin said:
So TLDR from GN is Asus being Asus and burning up cpu`s since 2000.

No. The CPU was long dead at this point. Applying in excess of 1.4 to 1.5v to VSOC and then detaching the cooler, subsequently, PROCHOT doesn't work as intended and the CPU doesn't shut down; GN's entire focus has then since been on what happens after the failure occurs as that's where the theatrics happen on ASUS due to the excessive runaway current and high OCP tripping point.

Whatever has happened inside the CPU for it to fail initially, only AMD can discern. The lab analysis, albeit interesting, doesn't actually tell us anything other than inspecting elements that have already failed - so the title is misleading.

Rawr · 10 May 2023 at 13:48

Silent_Scone said:
No. The CPU was long dead at this point. Applying in excess of 1.4 to 1.5v to VSOC and then detaching the cooler, subsequently, PROCHOT doesn't work as intended and the CPU doesn't shut down; GN's entire focus has then since been on what happens after the failure occurs as that's where the theatrics happen on ASUS due to the excessive runaway current and high OCP tripping point.

Whatever has happened inside the CPU for it to fail initially, only AMD can discern. The lab analysis, albeit interesting, doesn't actually tell us anything other than inspecting elements that have already failed - so the title is misleading.

Well it's potentially possible it was sped up a bit on a Asus board because they were supplying 1.4v soc while saying it was 1.35v in the bios.

Over all the most likely reason for failing seems to be getting unlucky with the silicon lottery and getting one which couldn't tolerate voltage as well.

Grim5 · 10 May 2023 at 13:50

According to MLID this whole story of the high voltage and burning CPUs is a "nothingburger". He formed this opinion after speaking to one best buy store who told him there was no increase in 7800x3d RMA in the last week.

He obviously didn't watch the GN video as he has no understanding of what has actually happened. Which is not surprising; he claims the media on this subject is just fishing for content and not worth paying attention to. He is in denial as an amd fanboy

Silent_Scone · 10 May 2023 at 13:53

Rawr said:
Well it's potentially possible it was sped up a bit on a Asus board because they were supplying 1.4v soc while saying it was 1.35v in the bios.

Over all the most likely reason for failing seems to be getting unlucky with the silicon lottery and getting one which couldn't tolerate voltage as well.

That's assuming it was even measured from the right place, which I don't think it was. If you're measuring off-board there will be potential losses across the power plane, which would account for at least some of the differential. How much of a differential will depend on where on the power plane you are measuring from as well as the measuring equipment. The correct measurement is from the correct pads on the CPU pin pad. If so, the end result is ASUS is [or was] still applying more, just not as much as the video would make out.

Harlequin · 10 May 2023 at 13:54

Silent_Scone said:
No. The CPU was long dead at this point. Applying in excess of 1.4 to 1.5v to VSOC and then detaching the cooler, subsequently, PROCHOT doesn't work as intended and the CPU doesn't shut down; GN's entire focus has then since been on what happens after the failure occurs as that's where the theatrics happen on ASUS due to the excessive runaway current and high OCP tripping point.

Whatever has happened inside the CPU for it to fail initially, only AMD can discern. The lab analysis, albeit interesting, doesn't actually tell us anything other than inspecting elements that have already failed - so the title is misleading.

So thats yes then, Asus have been feeding too much voltage into cpus for the last 20 years, so *their* boards are the best.

Silent_Scone · 10 May 2023 at 14:00

Harlequin said:
So thats yes then, Asus have been feeding too much voltage into cpus for the last 20 years, so *their* boards are the best.

Seeing as failures have occurred on boards from multiple vendors, seems like a bit of an unobjective blanket statement. At this point, who is even sure if 1.25v is still excessive.

Harlequin · 10 May 2023 at 14:04

Silent_Scone said:
Seeing as failures have occurred on boards from multiple vendors, seems like a bit of an unobjective blanket statement. At this point, who is even sure if 1.25v is still excessive.

GN said that Asus in the biggest culprit , wait for part 3 and you`ll understand. Those `multiple vendors` are single or low number cases and require stupid steps and disabling protections to replicate; with Asus boards they just go bang out of the box, because asus doing what asus always has done. Asus boards overvolting isnt a new thing.

Rawr · 10 May 2023 at 14:05

Silent_Scone said:
That's assuming it was even measured from the right place, which I don't think it was. If you're measuring off-board there will be potential losses across the power plane, which would account for at least some of the differential. How much of a differential will depend on where on the power plane you are measuring from as well as the measuring equipment. The correct measurement is from the correct pads on the CPU pin pad. If so, the end result is ASUS is [or was] still applying more, just not as much as the video would make out.

I mean it was GN who measured the ASUS ones. I think I trust them enough to get the correct value. Level1tech measured the others, and well I trust him too to be fair. He also knows his stuff.

This wasn't just some random youtubers.

Brizzles · 10 May 2023 at 14:05

Rawr said:
Over all the most likely reason for failing seems to be getting unlucky with the silicon lottery and getting one which couldn't tolerate voltage as well.

I agree, Reddit reports seems to have died, and no-one here ( in what is probably the largest PC enthuiast forum in the UK ) has reported one.

It was possibly a few of the really weak chips, that just scrapped into the 'Pass' category, combined with high 1.4+soc voltages.

Brizzles · 10 May 2023 at 14:09

Harlequin said:
GN said that Asus in the biggest culprit

Out of maybe 10 reported failures worldwide.

Easily put down to % of sales - I would be more concerned that Asrock had a failure given the far lower market share than Asus.

Rawr · 10 May 2023 at 14:12

Brizzles said:
Out of maybe 10 reported failures worldwide.

Easily put down to % of sales - I would be more concerned that Asrock had a failure given the far lower market share than Asus.

Funnily enough Asrock were also reasonably close to Asus with their... under-reporting of SoC values being supplied. I doubt it's a coincidence really.

Troezar · 10 May 2023 at 14:13

I'll be building at the weekend and other than updating to the latest BIOS I won't giving it a second thought. It's interesting, and makes for a good video or two, but doesn't to be a real issue to the average user from what I can tell.

GoogalyMoogaly · 10 May 2023 at 14:15

Didn't GN say they didn't even test ASRock or MSI boards?
So really we have no idea what they're like.

This is part of my issue, I'd send my Asus back, but what do I get instead?