News Report: TSMC To Start 3nm Volume Production In 2022

[Admin]

According to JK Wang, the TSMC's senior vice president for fab operations, the company is on track to begin volume production of 3nm chips in 2022.

Report: TSMC To Start 3nm Volume Production In 2022 : Read more

 

[srimasis]

Meanwhile Intel is going to release another generation of 14nm CPUs

 

[bit_user]

It's long been known that TSMC fudges their node size numbers, when they name a manufacturing process. I wonder what Intel would actually consider their 3 nm to be.

Also, I wonder about the longevity of these chips. According to AMD, 7 nm Ryzens will degrade with use, causing gradual clock speed decline (but only because the microcode is smart enough to measure the degradation & compensate). I wonder if chips made on this 3 nm node will have to be replaced on an annual basis, just from wearing out. ...talk about "planned obsolescence"!

 

[nofanneeded]

According to AMD, 7 nm Ryzens will degrade with use, causing gradual clock speed decline

source ?

 

[photon123]

Also, I wonder about the longevity of these chips. According to AMD, 7 nm Ryzens will degrade with use, causing gradual clock speed decline (but only because the microcode is smart enough to measure the degradation & compensate). I wonder if chips made on this 3 nm node will have to be replaced on an annual basis, just from wearing out. ...talk about "planned obsolescence"!

All chips degrade with use. This is one of the main reasons chips are rated below their true maximal frequency and can be overclocked. AMD however made a strange choice of clocking their Ryzen 3000 CPUs at their maximal frequency. This seems like the recipe for disaster, as the CPUs will degrade and will no longer support their rated frequency. I really hope AMD has a fail-safe system of reducing clock speed with time and also that they can avoid any lawsuits on this. Otherwise this will not end well for AMD.
What did they actually announce about this?

 

[bit_user]

source ?

I will try to dig up some references on this, but I don't have time just now.

All chips degrade with use. This is one of the main reasons chips are rated below their true maximal frequency and can be overclocked. AMD however made a strange choice of clocking their Ryzen 3000 CPUs at their maximal frequency. This seems like the recipe for disaster, as the CPUs will degrade and will no longer support their rated frequency.

I'm not sure we actually know that. Their internal microcode is what's truly managing the CPU's clock frequency. So, just because you can't overclock it any higher, that doesn't mean that the bare silicon couldn't go any faster.

 

[hannibal]

TSMC 3nm is most likely bigger than Intel 7nm... those marketing names in Computer technology Are getting really stupid. Both Intel and TSMC Are really close to 40-50nm at this moment...
https://en.wikichip.org/wiki/7_nm_lithography_process
It is like buying TSMC 100hp car vs Intel 70hp car, where Intel actually have more real horse powers... we should really stop looking those marketing nm ratings, the really has been thrown out of Windows Many years ago...

 

[bit_user]

TSMC 3nm is most likely bigger than Intel 7nm... those marketing names in Computer technology Are getting really stupid.

Just curious: do you have any source on this?

I would've thought TSMC's 5 nm = Intel's 7 nm.

 

[bit_user]

source ?

Found it. Here's the key bit:

With electromigration, there are two solutions. One is to set the frequency and voltage of the processor low enough that over the expected age of the CPU it won’t ever become an issue, as it happens at such a slow rate – alternatively set the voltage high enough that it won’t become an issue over the lifetime. The second solution is to monitor the effect of electromigration as the core is used over months and years, then adjust the voltage upwards to compensate. This requires a greater level of detection and management inside the CPU, and is arguably a more difficult problem.

What AMD does in Ryzen 3000 is the second solution. The first solution results in lower-than-ideal performance, and so the second solution allows AMD to ride the voltage/frequency limits of a given core.

However, I recommend reading the entire last 4 paragraphs of this page:

https://www.anandtech.com/show/1487...ing-perception-with-amds-frequency-metrics-/3

 

[TJ Hooker]

Found it. Here's the key bit:



However, I recommend reading the entire last 4 paragraphs of this page:

https://www.anandtech.com/show/1487...ing-perception-with-amds-frequency-metrics-/3

There's nothing in there about frequencies declining over time though.

 

[bit_user]

There's nothing in there about frequencies declining over time though.

Oh, sorry. It doesn't really make sense that they would say that - it would certainly have created quite a lot of commotion, if they had.

Perhaps I meant to phrase that as a question. Let me try that, again.

Also, I worry about the longevity of these "3 nm" chips. Reportedly, 7 nm Ryzens will degrade with use, which the chip's microcode is smart enough to measure and compensate for. However, I wonder if we'll eventually start to see gradual clock speed declines on those or future CPUs, with use. I wonder if chips made on this 3 nm node will have to be replaced on an annual basis, just from wearing out. ...talk about "planned obsolescence"!​

Better?

 

[nofanneeded]

Oh, sorry. It doesn't really make sense that they would say that - it would certainly have created quite a lot of commotion, if they had.

Perhaps I meant to phrase that as a question. Let me try that, again.

Also, I worry about the longevity of these "3 nm" chips. Reportedly, 7 nm Ryzens will degrade with use, which the chip's microcode is smart enough to measure and compensate for. However, I wonder if we'll eventually start to see gradual clock speed declines on those or future CPUs, with use. I wonder if chips made on this 3 nm node will have to be replaced on an annual basis, just from wearing out. ...talk about "planned obsolescence"!​

Better?

It depends on the rate of degrade , I am okay if the Chip lives 10 years with minimum speed loss ... also , you can allways replace a CPU after 10 years of use cheap.

 

[bit_user]

It depends on the rate of degrade , I am okay if the Chip lives 10 years with minimum speed loss ... also , you can allways replace a CPU after 10 years of use cheap.

Depends on how you use it. If you do a ton of heavy gaming, use PBO, and run toward the hotter end of the range, then I think you might do well to maintain steady boost clocks only for the warranty period.

Related: https://www.tomshardware.com/news/amd-ryzen-3000-boost-clock-controversy-intel-attack,40231.html

My uneducated guess is that we're already past the point of heavily-used CPUs routinely lasting for 10 years.

 

[nofanneeded]

Depends on how you use it. If you do a ton of heavy gaming, use PBO, and run toward the hotter end of the range, then I think you might do well to maintain steady boost clocks only for the warranty period.

Related: https://www.tomshardware.com/news/amd-ryzen-3000-boost-clock-controversy-intel-attack,40231.html

My uneducated guess is that we're already past the point of heavily-used CPUs routinely lasting for 10 years.

well maybe then The Turbo concept will have to die sooner as we shrink the CPU more and more...

 

[InvalidError]

My uneducated guess is that we're already past the point of heavily-used CPUs routinely lasting for 10 years.

Were early CPU failures for reasons other than user abuse or something else upstream failing and taking the CPU down with it ever a significant concern? All of my old PCs that I haven't ditched yet are still working and the oldest one is 22 years old, keeping it in case I want to play FFVII-PC again since that thing is so picky about the hardware it'll run on

 

[bit_user]

All of my old PCs that I haven't ditched yet are still working and the oldest one is 22 years old,

I was talking about CPUs being made today.

At work, we have two dual-CPU Westmere servers that run a pretty good load for about 8+ hours/day, and they're still going after about 9 years. Granted, one has random ECC errors that even swapping out the corresponding DIMM didn't resolve, but it still works (single-bit errors are correctable).

 

[nofanneeded]

Were early CPU failures for reasons other than user abuse or something else upstream failing and taking the CPU down with it ever a significant concern? All of my old PCs that I haven't ditched yet are still working and the oldest one is 22 years old, keeping it in case I want to play FFVII-PC again since that thing is so picky about the hardware it'll run on

I have an Apple //e still WORKING ...