Who will be next Intel CEO?

[DrMrLordX]

I . . . dunno about that, unless Intel wants to give up on the CPU business and just provide Optane memory/storage products for everyone.

 

[cbn]

I . . . dunno about that, unless Intel wants to give up on the CPU business and just provide Optane memory/storage products for everyone.

If Intel can remove (or eliminate most of) the memory gap (via lowered latency of future Optane generations) that will affect how CPUs are designed.

In fact, speculative execution is a result of CPU design that compensates for memory gap.

http://personals.ac.upc.edu/mpajuelo/papers/MEDEA04.pdf

 

[UsandThem]

I . . . dunno about that, unless Intel wants to give up on the CPU business and just provide Optane memory/storage products for everyone.

If Intel can remove (or eliminate most of) the memory gap (via lowered latency of future Optane generations) that will affect how CPUs are designed.

In fact, speculative execution is a result of CPU design that compensates for memory gap.

http://personals.ac.upc.edu/mpajuelo/papers/MEDEA04.pdf

Post #45

This thread is who the next Intel CEO will be.
For technical discussion on CPU tech (or discussing things like FD-SOI),
post in the appropriate thread.

AT Mod Usandthem

 

[DrMrLordX]

If Intel can remove (or eliminate most of) the memory gap (via lowered latency of future Optane generations) that will affect how CPUs are designed.

In fact, speculative execution is a result of CPU design that compensates for memory gap.

http://personals.ac.upc.edu/mpajuelo/papers/MEDEA04.pdf

Be that as it may (and I do not think Optane will do what you think it will; it eliminates the gap between storage and memory, not the gap between memory and cache), it's unclear how or why the CEO would have to be an Optane evangelist.

Right now, Intel's major source of revenue is the server room. They have to protect their Xeon market. The previous two CEOs have attempted to break into other markets with little success. The first order of business for any incoming CEO (not counting interim CEOs) is to execute on existing projects perfectly, so as to buy them enough time to recover from their current design/fab woes. In general, they have to move their server lineup onto Sapphire Rapids, and they have to move desktop onto Ice Lake. ASAP. As a logical extension, that means cleaning up the fab mess by hitting targets for 10nm+ and 7nm. Everything else is either a stop-loss measure or a side-project in comparison.

Once they have their core business stabilized, they can maybe shift focus to one of their other projects.

The logical choice would be a CEO that can help them pull all that together. Not someone who is going to dither off chasing some new market or some less-significant source of revenue.

Which means John Schnatter Colin Kaepernick.

 

[NTMBK]

I do wonder what will happen to BK's side projects when the new CEO comes in. Which ones will get binned in the great clear out? I suspect that at least a couple of the AI acquisitions will get "streamlined", and I don't know if dedicated GPUs will still get the support and funding they need to compete with NVidia.

 

[cbn]

Right now, Intel's major source of revenue is the server room. They have to protect their Xeon market. The previous two CEOs have attempted to break into other markets with little success. The first order of business for any incoming CEO (not counting interim CEOs) is to execute on existing projects perfectly, so as to buy them enough time to recover from their current design/fab woes. In general, they have to move their server lineup onto Sapphire Rapids, and they have to move desktop onto Ice Lake. ASAP. As a logical extension, that means cleaning up the fab mess by hitting targets for 10nm+ and 7nm. Everything else is either a stop-loss measure or a side-project in comparison.

Once they have their core business stabilized, they can maybe shift focus to one of their other projects.

The CPU performance is tied to the memory though......so if Intel can reduce the memory gap (even somewhat) that will make their future Xeons more efficient. Just think about reducing the need for speculative execution on performance per watt as one example.

So I think future versions of Optane ( a form of phase change memory) is not really separate from CPU. In fact, I have to wonder if it is through memory that Intel will functionally extend restart* Moore's Law (ie, not literally, of course, but from a performance and efficiency standpoint).

*Restart because Moore's Law has already stopped.

This is why I think the next CEO will take a more integrated approach. (ie, treating future CPUs and future generations of Optane memory as something that needs to be thought of as one unit.)

 

[NTMBK]

The CPU performance is tied to the memory though......so if Intel can reduce the memory gap (even somewhat) that will make their future Xeons more efficient. Just think about reducing the need for speculative execution on performance per watt as one example.

So I think future versions of Optane ( a form of phase change memory) is not really separate from CPU. In fact, I have to wonder if it is through memory that Intel will functionally extend Moore's Law (ie, not literally, of course, but from a performance and efficiency standpoint).

This is why I think the next CEO will take a more integrated approach. (ie, treating future CPUs and future generations of Optane memory as something that needs to be thought of as one unit.)

Optane is still considerably slower than DRAM.

 

[DrMrLordX]

I do wonder what will happen to BK's side projects when the new CEO comes in. Which ones will get binned in the great clear out? I suspect that at least a couple of the AI acquisitions will get "streamlined", and I don't know if dedicated GPUs will still get the support and funding they need to compete with NVidia.

Good question!

They've still got Raja onboard, so I would consider the dGPU project to be full steam ahead until he finds an outbuilding at the Santa Clara campus, declares it his own holding, walls everyone off, and starts to pout. If they nix Raja then you can probably bet the dGPU work (such as it is, at this point) to have been rolled into one of their AI initiatives. Everything else? Who knows. Intel throws a metric ton of resources at R&D compared to a lot of other firms in the semiconductor world. Will they stop? That may be one way the new CEO gets Intel's investors through "the lean times" that are to come, by propping up profits by cutting R&D on speculative projects.

Cut the wrong one though, and there goes lots of potential future revenue.

I expect they'll leave Keller untouched to do his own thing. Of course, he may be on Sapphire Rapids or something else core to the business.

So I think future versions of Optane ( a form of phase change memory) is not really separate from CPU.

Well hey, if you can get the job, maybe you can do something about that.

Optane is still considerably slower than DRAM.

Yup, it is. Last time I checked, the only real advantage it had over NVMe anything is in rapid small file transfers.

 

[krumme]

They ought to buy amd and get Lisa as CEO.
Amd seems to be more innovative all over but lacks the power Intel have.

 

[cbn]

Optane is still considerably slower than DRAM.

Yes, the current generation of Optane (which is used in the Apache Pass DIMM) is slower than DRAM.

But I am thinking of the future generations of Optane, not the current generation.