Intel has a Pretty Big Problem..

[spooks]

It did seem to be (from the way he talked) that he was just assuming that W680 was ran 100% at stock and there were incorrect assumptions made by other news outlets too about what W680 can actually do. It is true that Supermicro usually use the Intel defaults, but W680 as a chipset does not have any restrictions and as you have noted, can overclock both the CPU and the memory.


We're not 100% sure of anything, both voltage and power draw being the cause are just speculation.

Agreed, and this is why I find it a little frustrating that these YouTubers are stating that this is basically fact and there’s definitely an issue (with the cpu) without going into a proper problem analysis.

Everyone is revering Wendell for his assumptions come facts but personally I prefer to stick to the evidence and wait for a proper diagnosis. I guess they want the clicks. ‍

 

[Tetras]

Everyone is revering Wendell for his assumptions come facts but personally I prefer to stick to the evidence and wait for a proper diagnosis. I guess they want the clicks. ‍

I hardly ever watch Wendell and I know he's generally interested in light server topics, so I don't want to disparage him unfairly.., but it did seem like he was very lackadaisical on providing enough detail for us to verify anything from the data he has.

The way he talked about the memory clocks also seemed like he didn't even know what Intel's stock memory rating is for these CPUs and that at least some of the systems were being overclocked using XMP or manual tuning.

If he did have this data and could 100% confirm what settings were being used (he said he was working with some of the server people, so I'd guess he had enough access to it) then it should have been provided, because otherwise there is a lot of misinformation doing the rounds due to this video.

The buyers of these CPUs to run servers are already taking shortcuts in terms of stability guarantees by buying regular Core i7/i9 CPUs, so I don't think it is unreasonable to conclude these systems might not be running entirely at Intel defaults.

 

[G J]

Frame Chasers put out a intresting video yesterday which I can't link here as it has swearing in it though Jufus says that the single core/two preferred cores boosting is giving the CPU too much voltage is the issue thats causing the CPU's to degrade and locking the cores to stop this boosting should stop it from happening.

The video also goes into detail of an degraded 13900KS he had recieved from a discord member that crashes after loading bios optimised defaults and put under a little load/pretty much idle in Windows. To make the CPU stable one has to lock the cores and up the voltage to make the CPU stable to account for the degradation that has already happened though some people in the comments said this needs more extensive testing.

This could explain why even with power limits in place this would still happen though its still not clear if this is an issue from the factory or a motherboard issue or a combination off both and/or other issues.

 

[jigger]

What a palaver.

I hardly ever watch Wendell and I know he's generally interested in light server topics, so I don't want to disparage him unfairly.., but it did seem like he was very lackadaisical on providing enough detail for us to verify anything from the data he has.

The way he talked about the memory clocks also seemed like he didn't even know what Intel's stock memory rating is for these CPUs and that at least some of the systems were being overclocked using XMP or manual tuning.

If he did have this data and could 100% confirm what settings were being used (he said he was working with some of the server people, so I'd guess he had enough access to it) then it should have been provided, because otherwise there is a lot of misinformation doing the rounds due to this video.

The buyers of these CPUs to run servers are already taking shortcuts in terms of stability guarantees by buying regular Core i7/i9 CPUs, so I don't think it is unreasonable to conclude these systems might not be running entirely at Intel defaults.

The buyers of at least some of these servers are looking to use them for cloud gaming.

Intel have simply been playing too lose and fast and we’re now seeing the results of cobbling together chip designs that have been pushed well outside of their power and frequency envelope.

 

[chaparral]

Are we 100% sure this isn't the unlimited power draw that is causing this? The argument against this is that 'Enterprise boards' have had the same issue. Yet, the one board they give an example of has had the same microcode / profile updates that all the Z790 boards have. Link here to the ASUS W680 Wendell uses as his example. W680 BIOS Release notes

As we can see:

"1. Introduce the ""Performance Preferences"" with options for Intel Default Settings (Performance/Extreme) and ASUS Advanced OC Profile.
2. Redefine the factory defaults based on Intel’s new ""Intel Default Settings"" for various CPU SKUs.
3. Change F5 from ""Load Optimized Defaults"" to ""Reset to Defaults"".
4. Add warnings when users switch from the defaults to other settings.

Furthermore, looking in the BIOS manual for that motherboard (W680) there are the same overclocking and AI tweaker settings so they can't be wildy different from consumer boards.

When he talked about servers failing with 13th/14th gen CPU's he said that the DC engineers "Were not sure if the problems were now fixed". To me I am super sceptical of a hardware issue and unless its categorically proven I will remain that way.

Obviously if mine fails I will become a believer but I've only ever run it at stock (inc an undervolt) with recommended intel settings (not default bios ones).

The other problem i say that could easliy cause problems is what CPU cooler is being used with these power hungry CPU's that had problems
As even with my front mounted 420mm AIO some my 14900k cores would still hit 100c with the PL2 power draw set to unlimited

You also get some people that will buy these really hot cpu's and then stick them in a tiny PC case
(Done that myself years ago with shuttle cases and had nothing but trouble)

 

[cliffy]

 

[Aretak]

"We can't recommend Intel CPUs right now. Not until there's either some level of first-party transparency and a support assurance from Intel itself, or until we or some other third party are able to verify why its CPUs are not stable in some situations, which CPUs might be affected, what those conditions are and how you might resolve it."

"If there is no statement and we're just left in the dark on what it might be, then we have no assurance or confidence in the product or the company behind it. And that means right now our blanket statement is we cannot comfortably recommend Intel CPUs and it's gonna be that simple. Until they say something so that we know a little bit more about what the scope of this is."

Steve's joined the ranks of the AMD uberfans.

 

[Yulaw]

"We can't recommend Intel CPUs right now. Not until there's either some level of first-party transparency and a support assurance from Intel itself, or until we or some other third party are able to verify why its CPUs are not stable in some situations, which CPUs might be affected, what those conditions are and how you might resolve it."

"If there is no statement and we're just left in the dark on what it might be, then we have no assurance or confidence in the product or the company behind it. And that means right now our blanket statement is we cannot comfortably recommend Intel CPUs and it's gonna be that simple. Until they say something so that we know a little bit more about what the scope of this is."

Steve's joined the ranks of the AMD uberfans.

Why would he recommend a potential ticking time bomb, seems rather sensible to me ?

 

[RavenXXX2]

Intel stops being shady and they might go back on potential buy list.

 

[TrixP10]

"We can't recommend Intel CPUs right now. Not until there's either some level of first-party transparency and a support assurance from Intel itself, or until we or some other third party are able to verify why its CPUs are not stable in some situations, which CPUs might be affected, what those conditions are and how you might resolve it."

"If there is no statement and we're just left in the dark on what it might be, then we have no assurance or confidence in the product or the company behind it. And that means right now our blanket statement is we cannot comfortably recommend Intel CPUs and it's gonna be that simple. Until they say something so that we know a little bit more about what the scope of this is."

Steve's joined the ranks of the AMD uberfans.

Nope. He's stating the bleeding obvious.

For whatever reason these CPU's don't work as intended. Also, all these views on why it doesn't work are pointless - it does not work as intended. It's not for the customer/user to find a fix.

 

[Rroff]

It's not for the customer/user to find a fix.

It may yet be a simple fix, no one really knows what the problem(s) is. I've found some interesting behaviour with some 13th and 14th gen CPUs and FFMPEG but I've not extensively tested it enough yet to know if it is a software/OS environment issue or this issue or whether it happens with other CPUs or not. Hoping to get my hands on a couple of AMD systems later to see if I can reproduce it there.

EDIT: Though even if it does turn out to be a 13th/14th gen problem that doesn't necessarily help indicate what and probably beyond my ability to narrow down.

 

[TrixP10]

It may yet be a simple fix, no one really knows what the problem(s) is. I've found some interesting behaviour with some 13th and 14th gen CPUs and FFMPEG but I've not extensively tested it enough yet to know if it is a software/OS environment issue or this issue or whether it happens with other CPUs or not. Hoping to get my hands on a couple of AMD systems later to see if I can reproduce it there.

EDIT: Though even if it does turn out to be a 13th/14th gen problem that doesn't necessarily help indicate what and probably beyond my ability to narrow down.

IMHO I very much doubt if there is a simple fix. However, if you do find out what the problem is maybe you should let Intel know as presently their silence is disturbing.
Even better, find an AMD error and notify AMD of the problem. That should help mitigate Intel's broken CPU's.

 

[uscool]

"We can't recommend Intel CPUs right now. Not until there's either some level of first-party transparency and a support assurance from Intel itself, or until we or some other third party are able to verify why its CPUs are not stable in some situations, which CPUs might be affected, what those conditions are and how you might resolve it."

"If there is no statement and we're just left in the dark on what it might be, then we have no assurance or confidence in the product or the company behind it. And that means right now our blanket statement is we cannot comfortably recommend Intel CPUs and it's gonna be that simple. Until they say something so that we know a little bit more about what the scope of this is."

Steve's joined the ranks of the AMD uberfans.

Lol saying he is pro AMD .

 

[Rroff]

IMHO I very much doubt if there is a simple fix. However, if you do find out what the problem is maybe you should let Intel know as presently their silence is disturbing.
Even better, find an AMD error and notify AMD of the problem. That should help mitigate Intel's broken CPU's.

When I mentioned AMD CPUs what I mean is if the error also happens there then what I'm seeing is either a software bug in FFMPEG itself, or less likely but a CPU hardware bug in general.

 

[Naff8000]

Steve's joined the ranks of the AMD uberfans.

He can't compare 9000 to 13/14 series, even at super gimped speeds with an image of a trash basket. No one knows if these products are any good.

 

[Journey]

Steve's joined the ranks of the AMD uberfans.

That's a funny take on it, from a consumer protection perspective stating you are unable to recommend something, as you'd don't know what is broken, or if it will be fixed or even if the vendor will admit to a fault is pretty solid consumer advice.

I'm sure you'd love your travel agent to recommend you go to holiday to "Lovely sunny island" but they couldn't comment if the civil unrest would be dangerous to you specifically, so instead of recommending something else they send you on your merry way to potentially enjoy your vacation, or not.

 

[CaptainRAVE]

It’s hardly a fanboy issue when the hardware doesnt work and they can’t recommend it.

 

[hartgeh3]

Shares tanking -6% keep going...lol

 

[StarShock]

Do you think intel will cover people's motherboard costs if they do end up recalling stuff? I have a feeling they won't want to do that.

 

[tommoT.Striker]

This is a disaster, I wonder if there has been any precedence of this happening before? If the manufacturing fault theory is true, did they only identify there was an issue when they got reports of instability this year, or was it a case of we know but hopefully nobody will notice / chances of issues are remote? I will send the info of the failed 14900KF I had to GamersNexus as they look like they are trying to identify a date range of affected CPUs.