Are binned chips reliable?

[blinkie]

Hey, l’m trying to parse the seemingly small differences between M1pro 8 and 10 core laptops. I’m not sure there’s going to be a huge difference in performance for my uses. But I did wonder if the 8core binned chips might be slightly less reliable being that they are already ‘defective’. I plan on using this computer for a very very long time so maybe some discrepancies might show over the lifetime. Have you guys got any thoughts on this? I’d love to hear them if you do.
Thanks in advance,
B

 

[chabig]

I don't believe the 8 core chips are a result of binning. There are so many other parts to these chips that it would be vanishingly improbably that two GPUs (and only two GPUs) wouldn't work after production. These chips are manufactured with either 8 or 10 working GPUs. That's all we know. There will be no difference between the two 8 and 10 core versions over their lifespan other than performance.

 

[blinkie]

I don't believe the 8 core chips are a result of binning. There are so many other parts to these chips that it would be vanishingly improbably that two GPUs (and only two GPUs) wouldn't work after production. These chips are manufactured with either 8 or 10 working GPUs. That's all we know. There will be no difference between the two 8 and 10 core versions over their lifespan other than performance.

That’s interesting. Seems to be an accepted phrase applied to the 8core chips. You’re the first person I’ve heard reject it. Thanks chabig

 

[JPack]

It's artificial binning with purposefully disabled cores, in all likelihood, to fill the product stack. As mentioned above, it's extremely unlikely two GPU cores would be defective and none of the CPUs are affected. In addition, true binning would result in limited supplies of Apple's most popular base model MacBook Pro. And we never see that happen. Combined with a mature N5 process, it doesn't make sense for binning to occur except to satisfy marketing requirements.

 

[chabig]

Very well said, JPack!

 

[blinkie]

It's artificial binning with purposefully disabled cores, in all likelihood, to fill the product stack. As mentioned above, it's extremely unlikely two GPU cores would be defective and none of the CPUs are affected. In addition, true binning would result in limited supplies of Apple's most popular base model MacBook Pro. And we never see that happen. Combined with a mature N5 process, it doesn't make sense for binning to occur except to satisfy marketing requirements.

You know what? I’m glad I asked this question. I wonder why the phrase is so widely used in reviews. Maybe it sounds cool.

 

[Xiao_Xi]

It's artificial binning with purposefully disabled cores,

Would it be possible to enable these cores? Has anyone tried it?

 

[JPack]

Would it be possible to enable these cores? Has anyone tried it?

Given Apple's complete control over the hardware/software, I would say no. For AMD, Intel, NVIDIA, it's been possible in some limited situations.

 

[Technerd108]

The base model M1 Pro on the 14" is a very powerful processor and so far has been very reliable. The 14" base model is a heck of a deal right now. Probably the best value for a pro machine from Apple I have ever seen from 3rd party vendors. If I was in the market for one I would buy the base model in a heartbeat because it is a very capable machine even when compared to the 10 core model.

 

[theorist9]

You know what? I’m glad I asked this question. I wonder why the phrase is so widely used in reviews. Maybe it sounds cool.

It's not being used b/c it's cool. It was speculated that Apple was actually doing binning on the M1 chips when one of the GPU cores was bad, resulting in both 7-core and 8-core versions (https://www.anandtech.com/show/1623...-2020-macbook-air-13inch-macbook-pro-mac-mini ). So that language simply carried over when distinguishing the low-spec vs. normal-spec M1 Pro. As to whether they're actually doing binning there or not, we don't know.

I would guess there are at least *some* defects, so there is probably some true binning, in combination with deliberate core disablement to meet sales needs.

I don't believe the 8 core chips are a result of binning. There are so many other parts to these chips that it would be vanishingly improbably that two GPUs (and only two GPUs) wouldn't work after production.

Since the OP was referring to 8-core vs. 10-core in the M1 Pro, that's CPU core count reduction, not GPU core count reduction (separately, there is also GPU reduction, from 16-core to 14-core).

As to the probabilities, it could be that Apple's analysis indicated a reasonable probability of one CPU core being defective, with a small probability of two. Thus, to simplify things, Apple decided to only offer a single 8-core binned model, which covers both possibilities, but where most of those 8-core binned chips actually have only one defective CPU core rather than two. I.e., it's not required that two CPU cores failed for it to be a legit binned chip.

In addition, true binning would result in limited supplies of Apple's most popular base model MacBook Pro. And we never see that happen. Combined with a mature N5 process, it doesn't make sense for binning to occur except to satisfy marketing requirements.

If they need more of the low-spec chips than they actually see due to defects, they could always take non-binned chips and disable the cores to supplement the binned chips. So the supply issue doesn't mean true binning isn't actually going on.

 

[blinkie]

Thanks guys that was very interesting.

B

 

[dumastudetto]

One of my MacBook Airs has a binned chip and I've had no issues with it. I must admit, the "binning" terminology is a bit off-putting to say the least.

 

[blinkie]

One of my MacBook Airs has a binned chip and I've had no issues with it. I must admit, the "binning" terminology is a bit off-putting to say the least.

Yes! Isn’t it? I’m in the process of replacing a smashed laptop so trying to get my head around the variants.

 

[dumastudetto]

Yes! Isn’t it? I’m in the process of replacing a smashed laptop so trying to get my head around the variants.

Yeah as others have said though, there's no reason to be concerned. Just go with whatever laptop best fits your needs and budget.

If there was any kind of real issue with this concern you raised, #binninggate would have been all over the internet now. Everyone in the media seemingly loves a bad news Apple story.

 

[wilberforce]

Remember Apple does not do any binning, TSMC does.
Apple contracts with TSMC to provide x number of 10-core chips at price xx, and y number of 8-core chips at price yy. (Probably more complicated contract terms than that, but you get the point.)

How TSMC provides these is up to TSMC. It may be a combination of real binning and artificial binning.

 

[theorist9]

Remember Apple does not do any binning, TSMC does.
Apple contracts with TSMC to provide x number of 10-core chips at price xx, and y number of 8-core chips at price yy. (Probably more complicated contract terms than that, but you get the point.)

How TSMC provides these is up to TSMC. It may be a combination of real binning and artificial binning.

I'd imagine the conversation between Apple and TSMC is much more collaborative than that. You're painting a picture in which Apple says: "Here's the details on what we want, now you figure out how to do it and what it will cost us". Instead, I think Apple probably works collaboratively with TSMC to determine an optimum binning strategy, based on costs and defect rates, and that in turn determines the SKU's that Apple offers.

 

[JPack]

I'd imagine the conversation between Apple and TSMC is much more collaborative than that. You're painting a picture in which Apple says: "Here's the details on what we want, now you figure out how to do it and what it will cost us". Instead, I think Apple probably works collaboratively with TSMC to determine an optimum binning strategy, based on costs and defect rates, and that in turn determines the SKU's that Apple offers.

Absolutely agree. Given Apple’s business model depends on yields and performance from TSMC, both parties will be far more involved. TSMC co-develops their processes based on Apple’s requirements.

 

[1096bimu]

I don't believe the 8 core chips are a result of binning. There are so many other parts to these chips that it would be vanishingly improbably that two GPUs (and only two GPUs) wouldn't work after production. These chips are manufactured with either 8 or 10 working GPUs. That's all we know. There will be no difference between the two 8 and 10 core versions over their lifespan other than performance.

This is completely wrong, They would NEVER manufacture 2 different designs, that's just SO MUCH more expensive.

Firstly if any of the critical components don't come out right then obviously it wouldn't be a working chip so no, it's not that ONLY the 2 GPU cores are faulty, this is simple survivor bias stuff.

Secondly, the GPU is by far the largest part of the chip so statistically it's just more likely to be faulty.

Thirdly, you don't have to have exactly 2 cores to be not working, it could be one not working, it could be one that's working but requires very high voltage, they would they bin all of them into 8-core versions.

Lastly, on the much larger M1 Pro chip you see there's all kinds of bins, you get 8 or 10 core CPU and 14 or 16 core GPU. The two disabled CPU cores are obviously the big cores because they have more chance to come out faulty like previously explained.

 

[leman]

There has been speculation that binned chips are “power binned”, that is, chips that draw more current have some GPU cores disabled to bring their power consumption back to some design target. If I understand these things correctly, this kind of binning is common on mobile where one tries to match the power target rather than performance, but I’m not convinced in how far this applies to Apple Silicon. Most likely it’s revenue optimization move by Apple. These GPU cores are certainly not defective though. There is very little chance that a defect will be constrained to a minuscule GPU core.

 

[1096bimu]

There has been speculation that binned chips are “power binned”, that is, chips that draw more current have some GPU cores disabled to bring their power consumption back to some design target. If I understand these things correctly, this kind of binning is common on mobile where one tries to match the power target rather than performance, but I’m not convinced in how far this applies to Apple Silicon. Most likely it’s revenue optimization move by Apple. These GPU cores are certainly not defective though. There is very little chance that a defect will be constrained to a minuscule GPU core.

chip defects are nanometers in scale, of course they'll be constrained in a GPU core, in fact it could be a SINGLE defective transistor.

 

[theorist9]

There has been speculation that binned chips are “power binned”, that is, chips that draw more current have some GPU cores disabled to bring their power consumption back to some design target. If I understand these things correctly, this kind of binning is common on mobile where one tries to match the power target rather than performance, but I’m not convinced in how far this applies to Apple Silicon.

Interesting.

These GPU cores are certainly not defective though. There is very little chance that a defect will be constrained to a minuscule GPU core.

Note that the OP, in asking about 8-core vs. 10-core M1 Pro chips, was referring to CPU binning rather than GPU binning.

 

[theorist9]

chip defects are nanometers in scale, of course they'll be constrained in a GPU core, in fact it could be a SINGLE defective transistor.

That's my understanding as well. Indeed, defects can often be too small to affect a single transistor. But if they are large enough to do so, the entire chip may fail:

"One problem that can hinder yield rates from semiconductor wafers is point defects. During the semiconductor production process, individual points in the single crystal can be defective for several reasons, including a void, a contaminant, or a dislocation in the crystal structure. Large-scale transistors are rarely affected by point defects due to the small size of the defect relative to the transistor. However, transistors whose size is similar to a defect will be severely affected and may not function correctly.

This may not be a challenge for a chip with ten transistors, but modern devices can now have transistors numbering in the billions. If just a single transistor operates incorrectly, the entire design may not function, and as the number of transistors increases, the chances of just one failing increases too."

NIST develop method for detecting faulty transistors on silicon chips

What challenges do large scale integrated circuits face, what did the NIST develop, and how can it help semiconductor producers in the future?

www.electropages.com

 

[theorist9]

This is completely wrong, They would NEVER manufacture 2 different designs, that's just SO MUCH more expensive.

The poster who whom you were responding didn't mean two different designs. He meant a single design in which some cores either were or were not disabled.

 

[wilberforce]

I'd imagine the conversation between Apple and TSMC is much more collaborative than that. You're painting a picture in which Apple says: "Here's the details on what we want, now you figure out how to do it and what it will cost us". Instead, I think Apple probably works collaboratively with TSMC to determine an optimum binning strategy, based on costs and defect rates, and that in turn determines the SKU's that Apple offers.

Sure, it is in Apple's interest to collaboratively determine the optimum contract. But in the end, there is a contract.

It's not like TSMC says to Apple: we'll see what we can do in production and let you know, let's hope for the best. That puts all the risk on Apple (who don't control the fab process), and none on TSMC (who do).

 

[1096bimu]

The poster who whom you were responding didn't mean two different designs. He meant a single design in which some cores either were or were not disabled.

that's two different designs...
if you etch a chip with a core disabled, it's not the same design, even if it's just a single disconnected wire.

This is even more retarded than just designing another chip with 2 less cores, because the disabled cores take up wafer space, on top of being a different and non-interchangeable design.