Intel has a Pretty Big Problem..

[TrixP10]

Now seeing adverts like this :
"If you've been put off getting a new budget gaming PC because of concerns over stories about Intel's stability issues then how about one that is still very affordable but uses a processor that's as stable as they come"

No more - Intel Inside

 

[tommoT.Striker]

I don't think Intel would be in panic mode if it was just 1-2 sources. I'm sure there is a lot info that Intel won't make public about the extent of the problem.

 

[Minstadave]

I don't think Intel would be in panic mode if it was just 1-2 sources. I'm sure there is a lot info that Intel won't make public about the extent of the problem.

The problem is potentially massive. The high end parts are brcoming unstable in a matter of weeks in specific use cases, if the low end parts follow suit over the coming months they're looking at big problems. If they knacker performance to try and slow down the problem they're going to have some unhappy customers.

 

[Rroff]

I don't think Intel would be in panic mode if it was just 1-2 sources. I'm sure there is a lot info that Intel won't make public about the extent of the problem.

I'm not saying they aren't in panic mode, but it is people saying they are in panic mode, Intel themselves are saying (and sadly doing) almost nothing aside from a couple of fairly vague statements with acknowledgement of *some* issues and the original rollout of Intel/baseline profiles. And some actions which might be interpreted in context but we don't have evidence of that like the supposed changes with Bartlett CPUs.

Be interesting to find out what the deal is with the Bartlett CPUs - some are claiming the P core only ones are just re-released Raptor Lake, but others are saying they are using 2x 6P Redwood Cove tiles and others that they are failed production from Granite Rapids and using the mesh bus architecture.

 

[Rroff]

Dunno if this has been posted here, trying to follow this on different forums but supposed leak info:

What Intel didn’t write on Reddit but thinks internally - The search for the solution to the Raptor Lake S instabilities continues (Leak) | igor´sLAB

I already posted yesterday in a news item and briefly commented on what Intel officially published on the night of 22 July 2024 via Reddit as a kind of interim report. However…

www.igorslab.de

 

[Silent_Scone]

Your point being?

I didn't mention any names, not sure why you think it was aimed at you. Do you somehow think it's related to your tangent about > 1.5v MT load voltage in Cinebench?

This is in context to how component failure happens - which is principally related to current and power dissipation, but is a bit more complicated than simply voltage goes up = current goes up because you also get constant power load situations around things like voltage regulators and control/feedback loops, etc. where lower voltage = higher current and in droop situations can result in current outside of spec.

One of the problems, though not sure if the fundamental issue here, seems to be a bug in the microcode resulting in excess voltage incompatible with the thermal situation, which potentially also violates current and power regulation. What isn't clear is whether the CPUs actually need that excess voltage or not - one situation is an easy fix which may or may not involve a small performance drop, the other means Intel is going to have to replace a lot of CPUs (as some won't be able to meet spec even at a small performance drop).

It's in the context of the amount of current passing through the die being the catalyst for accelerated electromigration. The rest is mansplaining different workloads producing different swings in current where we see a voltage drop and you trying to describe the negatives on capacitance. Motherboard VRMs handle these things just fine, no need to go into a power debate. Although it's a little funny as I am wondering how this factors into claims of seeing very little vdroop in Cinebench.

 

[Rroff]

I didn't mention any names, not sure why you think it was aimed at you.

I dunno, though I've 1-2 posters on ignore but I've not seen anyone else posting anything relevant and I'm the only one really who has touched on it - but it is quite a bad misunderstanding of what I've posted if it was aimed at me. Likewise with this bit below which you seem to be reading a whole load in my post which isn't there - nothing I've said before is relevant to vdroop in Cinebench and I'm talking in context of the silicon level (i.e. FIVR or whatever Intel uses now and other implementations involving on die power management as well as non-power management related control circuits) not motherboard VRMs.

Most of what I'm talking about though is where things aren't working as intended, outside the spec scope whereas you seem to be trying to interpret these issues within the scope of how things are "supposed" to work when things are working to spec.

It's in the context of the amount of current passing through the die being the catalyst for accelerated electromigration. The rest is mansplaining different workloads producing different swings in current where we see a voltage drop and you trying to describe the negatives on capacitance. Motherboard VRMs handle these things just fine, no need to go into a power debate. Although it's a little funny as I am wondering how this factors into claims of seeing very little vdroop in Cinebench.

 

[Rroff]

I can't confirm if there is anything to it but the supposed leak as linked above:

– Intel observes a significant increase to the minimum operating voltage (Vmin) across multiple cores on returned affected processors from customers.
– This increase is similar in outcome to parts subjected to elevated voltage and temperature conditions for reliability testing.
– Factors contributing to this Vmin increase include elevated voltage, high frequency, and elevated temperature.
– Even under idle conditions at relatively cool temperatures, sporadic elevated voltages are observed when the processor is resumed from low power states in order to service background operations before entering a low power state again.
– At a sufficiently high voltage, these short-duration events can accumulate over time, contributing to the increase in Vmin.
– Intel analysis indicates a need to reduce the maximum voltage requested by the processor in order to reduce or eliminate accumulated exposure to voltages which may result in an increase to Vmin.
– While Intel has confirmed elevated voltages impact the increase in Vmin, investigation continues in order to fully understand root cause and address other potential aspects of this issue.

 

[Grim5]

What percentage failiure rate is acceptable 0.1% - 10% - 50%?
So do we just wait till we know

1-2% is typical failure rate for consumer electronics, anything more is excessive

 

[Grim5]

Dunno if this has been posted here, trying to follow this on different forums but supposed leak info:

What Intel didn’t write on Reddit but thinks internally - The search for the solution to the Raptor Lake S instabilities continues (Leak) | igor´sLAB

I already posted yesterday in a news item and briefly commented on what Intel officially published on the night of 22 July 2024 via Reddit as a kind of interim report. However…

www.igorslab.de

Most interesting for me is Intel saying:

1) The August Microcode will lower CPU performance

2) The microcode update may not give you stability if your CPU is too far damaged due to either oxidisation or exposure to high voltages

 

[Minstadave]

Most interesting for me is Intel saying:

1) The August Microcode will lower CPU performance

2) The microcode update may not give you stability if your CPU is too far damaged due to either oxidisation or exposure to high voltages

Yup it doesn't look great.

 

[humbug]

Now seeing adverts like this :
"If you've been put off getting a new budget gaming PC because of concerns over stories about Intel's stability issues then how about one that is still very affordable but uses a processor that's as stable as they come"

No more - Intel Inside

Company's don't chose their own CPU's, they wouldn't even know where to start, so they hire supply consultants, those consultants are always going to opt for Intel, even if they use twice the power and have half the performance, you see its never about what's best for their client, that's irrelevant, its about competency perception, its about their job security, Intel is the safe choice because it has the brand image, no one is going to blame you for choosing Intel, you're not going to get fired for choosing Intel.

Until they are fired for incompetence for choosing Intel that's not going to change, the more Intel's name gets pulled through the mud the more likely that is to happen, this is the war that AMD should be waging, it need's to be convincing the companies, not the suppliers but the companies being supplied that using Intel is a bad idea, this is what Intel have been doing to AMD for decades.

 

[humbug]

AMD can make a very good case for its self, and it should, it should be investing time and money to do that, it should have people going out all over the place making its case. Ruthlessly i might add...

 

[humbug]

Don't take my word for it... here's a free server and a few guys to help you plug it in to your infrastructure, do what you will with it, put it through its paces, use it, its yours.

Oh and here's my card...

 

[LtMatt]

Don't take my word for it... here's a free server and a few guys to help you plug it in to your infrastructure, do what you will with it, put it through its paces, use it, its yours.

Oh and here's my card...

Spoiler

 

[humbug]

Spoiler

Not that kind of card you plumb