AMD Ryzen 7 7800X3D CPU Burns Up

[Brizzles]

EXPO in itself is not the issue, it just sets the ram timings and dram voltage.

The issue is mobo vendors making enabling EXPO raise the SOC voltage ( to unsafe levels* ) to help stability.

*edit

 

[Quartz]

The issue is mobo vendors making enabling EXPO raise the SOC voltage to help stability.

The issue is someone not putting a voltage limiter on the CPU, chipset, or motherboard to prevent damage.

 

[Dave2150]

Things need to be put into some sort of perspective.

Gamers Nexus, by their own admission had to force and actively try to trigger the failure and it took a lot of effort even though they new what to look for.

We have had possibly 10 failures from Reddit, OcUK alone sold/pre-sold 500 units in April ( approx figures from Gibbo ) - so if the entire Reddit community worldwide bought them from Overclockers - thats a failure rate of 2%.

I'm not saying this situation is acceptable, but you can estimate the likelyhood of it happening are pretty slim.

All hardware has a failure rate.

(I'm currently building a 7800X3D and Asus B650E-E pc at the moment, we'll see if this post has jinxed me )

POP!

 

[Dave2150]

No. Unless you return a board for RMA and its enabled.

Not impossible for the sun's radiation to flip a bit in the BIOS while in transit....

 

[doobyscoo]

How on earth can EXPO, a memory profile created by AMD, that HAS to be turned on in order to reach the memory speeds of the RAM you bought, cause the warranty to be void ?
It's crazy.

I would imagine that a change like this would be legally problematic. Motherboard manufacturers advertise EXPO as a feature and detail validated ram kits.

No. Unless you return a board for RMA and it’s enabled.

Thanks. That makes sense. I have turned EXPO off for now. I don’t need the extra performance.

 

[z10m]

GM failure lab claims there is a "unknown amount of damage to the cpus out there in the field already" (degradation) and they are working on a new video.

 

[delta0]

No. Unless you return a board for RMA and its enabled.

Could probably just flash the bios again. The computer doesn’t need to be on or even have a CPU to do that. It would put everything back to default.

 

[Dave2150]

GM failure lab claims there is a "unknown amount of damage to the cpus out there in the field already" (degradation) and they are working on a new video.

I'd recommend all Ryzen 7000 owners to inspect their CPU's contact pads and motherboard socket pins, check this video to see the angle/lightning you need to spot the visible burn marks (will likely be much less noticeable if your CPU is damaged but still working)

 

[Andy D]

And if there’s burn marks, what then?
Send CPU back to Overclockers for refund or replacement?

 

[eddiew]

GM failure lab claims there is a "unknown amount of damage to the cpus out there in the field already" (degradation) and they are working on a new video.

Increasingly grateful that I haven't bought a motherboard+CPU yet... and increasingly feeling like the investment in the AM5 upgrade path isn't worth it

Basically they have to convince me that the platform actually has longevity before I manage to get my hands on a Fractal North to build in

 

[modez]

That is the end of line for CPU and motherboard factory. For how long they want to continue making products which is no longer required when we have previous generations already?

 

[Minusorange]

Gamers Nexus, by their own admission had to force and actively try to trigger the failure and it took a lot of effort even though they new what to look for.

We have had possibly 10 failures from Reddit, OcUK alone sold/pre-sold 500 units in April ( approx figures from Gibbo ) - so if the entire Reddit community worldwide bought them from Overclockers - thats a failure rate of 2%.

I'm not saying this situation is acceptable, but you can estimate the likelyhood of it happening are pretty slim.

All hardware has a failure rate.

GN also mentioned in their latest vid that it's quite possible that chips have been running "fine" but will have suffered some degredation due to the issue, so it may have just shortened the lifespan of the chip where it won't fail tomorrow, but is more likely to fail in the future

GM failure lab claims there is a "unknown amount of damage to the cpus out there in the field already" (degradation) and they are working on a new video.

Yes this is the vid, I'd be real paranoid if I owned one of these chips now and be expecting it to fail at some point

 

[jigger]

The MTBF of silicon chips is pretty well understood. We have tons of papers on the effects of voltage and electro migration.

Video summary. More voltage = more bad. For longevity run your electronics at the lowest possible voltage and as close room temperature as possible. Power down when not in use.

 

[Silent_Scone]

The issue is someone not putting a voltage limiter on the CPU, chipset, or motherboard to prevent damage.

This is an excellent example post of individuals conflating the many things happening here. Quite a few pages worth at this point of the same half-accurate truths, and it's not that surprising. Unfortunately for ASUS, the OCP issue made for some theatrics and detracted from a much deeper problem. In Steve's shoes, I would have personally covered this separately in order to avoid confusing his audience, as the saga and its makeup have enough legs to be covered in this way.

The TLDR is it would have made more sense to focus on why it happens, rather than what happens after. This is also applicable to the thermal trip not working as intended which didn't get as much airtime. All of these subsequent failures do not directly address why the CPU is failing in the first place, and without AMD's direct interjection, I don't think we will find out.

To draw back to why I singled the above quote out, giving partners some control comes down to wanting to make the CPU competitive. AMD EXPO endorses 6000MT, after all, and had pushed for these kits to be used with review kits. Whilst memory vendors are able to programme a multitude of rails into the SPD profile, the majority if not all decided to forfeit applying a manual VSOC voltage. Does that mean the memory vendor expects the kits to be able to run at stock CPU voltages? No. This isn't uncommon, and the memory vendor knows the motherboard vendor's auto rules kick in, which look for the worst case in terms of silicon quality. The media has clearly ascertained that ASUS is applying more VSOC than the competition, but what they haven't been able to ascertain is at what voltage this issue is averted. As we know, there have been failures across multiple board vendors, one case claiming to be run at stock.

There isn't a great deal of documentation on what's defined as stock. For instance, there are 5200MT memory kits with timing adjustments - is this stock? EXPO is still very new, after all. It is also very important to realise from an engineering perspective, that frequency and current are intrinsically related. To give an example, we can claim that 1.25v is a "safe" level of voltage for 5600MT, but is it also safe for 6000MT? As we increase the frequency, the current also increases. With Intel, for example, the whitepapers only offer voltage guidelines at stock frequencies. Naturally, some CPUs will need more voltage than others when at the upper platform limits, so if 1.25v isn't enough we have to either back off the overclock or apply more if able to do so. Some users are already finding this out for themselves. When we also consider the VSOC being fed for the I/O die - it may potentially transpire that you really can't push the voltage that high at all. How else does one explain multiple failures on various boards, not all of which would have been applying circa 1.3v+. Ultimately, there's blame to be shared everywhere here IMO.

 

[z10m]

The MTBF of silicon chips is pretty well understood. We have tons of papers on the effects of voltage and electro migration.

Video summary. More voltage = more bad. For longevity run your electronics at the lowest possible voltage and as close room temperature as possible. Power down when not in use.

failure is one thing and degradation is another
it doesn't take that much to degrade a chip to a point where it doesn't boost as high or becomes unstable with the same voltage as before and needing more.

 

[Dave2150]

And if there’s burn marks, what then?
Send CPU back to Overclockers for refund or replacement?

Yeap

 

[Jamin280672]

Could probably just flash the bios again. The computer doesn’t need to be on or even have a CPU to do that. It would put everything back to default.

Just hit the bios reset button on the back of the board......if it doesn't have one, use the jumper, default settings.....here I come.

 

[jigger]

failure is one thing and degradation is another
it doesn't take that much to degrade a chip to a point where it doesn't boost as high or becomes unstable with the same voltage as before and needing more.

Degradation starts as soon as the CPU is powered on, maybe even as the chip leaves the production line.

Properly explaining silicon degradation is beyond the scope of a YouTube video. Steve isn’t going to be able to put a number on this question for you.

 

[Rawr]

Degradation starts as soon as the CPU is power on, maybe even as the chip leaves the production line.

Properly explaining silicon degradation is beyond the scope of a YouTube video. Steve isn’t going to be able to put a number on this question for you.

Steve even talks about it in that pcworld podcast he was on the other day. How you would have needed to have limit tested when you first got the CPU to see how far it would push to then figure out if/how much the cpu would have degraded by now.

Hindsight is 20/20 as they say.

 

[doobyscoo]

If these issues cause chips to fail prematurely and in significant enough numbers, it won’t be a good look. I smell a class action.