News AMD records its highest server market share in decades — Intel fights back in client PCs

[Admin]

AMD now controls near a quarter of server CPU shipments, but Intel maintains undisputed lead in both client and datacenter processors.

AMD records its highest server market share in decades — Intel fights back in client PCs : Read more

 

[Pierce2623]

To be fair, Intel only has access to “vast production capacity” if they can get back to producing their chips in their own foundries. 18A is holding all their hopes and dreams but supposedly it’s pretty good. Lets hope it turns out that way.

 

[TesseractOrion]

Best hope is that Intel fails and other, more innovative and market disruptive companies get traction.

 

[oofdragon]

So.. AMD dominates high end and Intel low end... guess the 75% low end market share doesn't care about defective chips 🤷

 

[rluker5]

Best hope is that Intel fails and other, more innovative and market disruptive companies get traction.

What could Intel do to be more innovative?

 

[Kamen Rider Blade]

What could Intel do to be more innovative?

Sell off it's Design side and become "TRULY INDEPENDENT".

Pledge to never have a Design Division EVER AGAIN, just like TSMC.

Become a "For Contract" Semi-Conductor Foundry

Let Intel's Design side become a independent Fabless company, just like AMD.

 

[thestryker]

Sell off it's Design side and become "TRULY INDEPENDENT".

Pledge to never have a Design Division EVER AGAIN, just like TSMC.

Become a "For Contract" Semi-Conductor Foundry

Let Intel's Design side become a independent Fabless company, just like AMD.

None of this is in any way innovative it's just something that would benefit shareholders.

 

[thestryker]

To be fair, Intel only has access to “vast production capacity” if they can get back to producing their chips in their own foundries. 18A is holding all their hopes and dreams but supposedly it’s pretty good. Lets hope it turns out that way.

They're having to make over 100 million CPUs every year on Intel 7 capacity itself isn't a problem at all. This is still a DUV process node whereas everything newer is EUV so the biggest machines used in fabrication aren't the same. The key will be the transition point (probably 2025-2026) and whether or not they nail that execution. The way they did MTL is a great example of how messy this has been where they shipped fabrication tools from Oregon to Ireland and had to spin up immediately. So while it sounds like this all went smoothly that isn't the way you want to have to do things. Once they're able to cut a lot of the Intel 7 volume this should make things simpler since Intel 3 and 18A are both refined EUV processes based on their respective predecessors.

 

[Amdlova]

This new ryzen zen 5 is a server cpu they will hit the server market

 

[dalek1234]

What could Intel do to be more innovative?

Nothing at this point. Intel started screwing up more than 10 years ago. It has been slowly catching up with them since. It's a slowly sinking Titanic. They couldn't save the Titanic; Intel is no different now.

 

[artk2219]

Nothing at this point. Intel started screwing up more than 10 years ago. It has been slowly catching up with them since. It's a slowly sinking Titanic. They couldn't save the Titanic; Intel is no different now.

I wouldn't say they're doomed, they still have plenty of capital and talent. But they are going through a transition period, they haven't been in the position they are in for at least 20 years, if ever. So it will take some time for it to stabilize, that is unless the shareholders or management runs the boat into the ground. That said, it wouldn't surprise me if we end up with a Coke and Pepsi situation, with one holding 50% of the market, the other 45, and everyone else taking a small piece of what's left.

 

[rluker5]

Sell off it's Design side and become "TRULY INDEPENDENT".

Pledge to never have a Design Division EVER AGAIN, just like TSMC.

Become a "For Contract" Semi-Conductor Foundry

Let Intel's Design side become a independent Fabless company, just like AMD.

SOCs can be bought and reverse engineered through physical methods. Having the schematics is just cheaper from a time perspective but is easier to catch and has greater legal risk. If Intel could reverse engineer designs, how about AMD, Apple, Qualcomm, Huawei, Samsung, Texas Instruments, IBM, MediaTek, etc? What would be the point if you can get caught either way? Intel isn't like AMD, they don't have a history of reverse engineering their client's CPUs and selling them as their own designs. Stop projecting.

Chopping off their vertical integration isn't going to do anything new, it is just a different business model that many have chosen in the past. Having control of both their own design and fabs give them better feedback on how changes on one affects the other which can lead to more innovation.

 

[OneMoreUser]

So.. AMD dominates high end and Intel low end... guess the 75% low end market share doesn't care about defective chips 🤷

It is early days, the Intel problems is hardly visible in the data we have here. What it looks like in the next couple of quarters is much more interesting - sales numbers combined with how much money Intel earns should tell a lot.

 

[Kamen Rider Blade]

SOCs can be bought and reverse engineered through physical methods. Having the schematics is just cheaper from a time perspective but is easier to catch and has greater legal risk. If Intel could reverse engineer designs, how about AMD, Apple, Qualcomm, Huawei, Samsung, Texas Instruments, IBM, MediaTek, etc? What would be the point if you can get caught either way? Intel isn't like AMD, they don't have a history of reverse engineering their client's CPUs and selling them as their own designs. Stop projecting.

But making it that much harder, more expensive to do, more time intensive to accomplish is more of a barrier than just simply having the schematics.

You could easily understand why so many of TSMC's customers are wary to use Intel's fabs.
The proof is in the pudding, the vast majority of customers want TSMC, not Intel's fabs.

It's not a capability issue, it's a TRUST issue.

Chopping off their vertical integration isn't going to do anything new, it is just a different business model that many have chosen in the past. Having control of both their own design and fabs give them better feedback on how changes on one affects the other which can lead to more innovation.

But then they're married at the hips and Intel Foundary is basically only tied to one customer.

Where as if IFS was independent, it would have access to the entire world for customers.

 

[rluker5]

But making it that much harder, more expensive to do, more time intensive to accomplish is more of a barrier than just simply having the schematics.

You could easily understand why so many of TSMC's customers are wary to use Intel's fabs.
The proof is in the pudding, the vast majority of customers want TSMC, not Intel's fabs.

It's not a capability issue, it's a TRUST issue.



But then they're married at the hips and Intel Foundary is basically only tied to one customer.

Where as if IFS was independent, it would have access to the entire world for customers.

So how long has it been now that Intel has had the superior node?
I think that is what is keeping customers at TSMC currently. I bet some of the contracts currently being fulfilled over there were signed prior to IDM 2.0 being a thing.

If the customers were staying away from Intel's foundries due to Intel also being one of the big chip designers with their hands in many baskets, then why are they staying away from Samsung? Samsung just makes memory and a few ARM smartphone SOCs. But their nodes also underperform TSMC's. Intel is just leaving DUV and are still selling DUV desktop chips. By the time Intel's foundries get to their fifth node in four years and start pulling ahead of TSMC is when interest will pick up. And Intel says their foundry and design businesses aren't tied at the hip. They even have separate reporting now.

 

[honeystinger]

But making it that much harder, more expensive to do, more time intensive to accomplish is more of a barrier than just simply having the schematics.

You could easily understand why so many of TSMC's customers are wary to use Intel's fabs.
The proof is in the pudding, the vast majority of customers want TSMC, not Intel's fabs.

It's not a capability issue, it's a TRUST issue.



But then they're married at the hips and Intel Foundary is basically only tied to one customer.

Where as if IFS was independent, it would have access to the entire world for customers.

Make no mistake the Manufacturing branch of Intel WILL be spun off at some point, or at least that is the plan. But right now is not a good time as they don't have enough customers to be profitable, as you can see from the earnings reports.

Customers don't "want TSMC", they want the best process at the best price they can get. These are BUSINESSES not crying babies like in the movies or in your vision quests.

It's not just a trust issue, it's a price, cost, performance, power, yield and everything issue. Nature abhors a vacuum and right now there is not enough competition in the Foundry space. TSMC can charge whatever they want, because "where you gonna go? Samsung? PFFT! Intel? PFFT". So customers will go to Intel, at first just to try and get leverage with TSMC and their pricing, and Intel will have to give discounts to get paying customers. But eventually, if they succeed, Intel will become a big player in the Foundry space, and they can fill their full capacity with their own designs as well. Somewhere in between now and full profitability, IFS will be spun off as its own company.

 

[honeystinger]

SOCs can be bought and reverse engineered through physical methods. Having the schematics is just cheaper from a time perspective but is easier to catch and has greater legal risk. If Intel could reverse engineer designs, how about AMD, Apple, Qualcomm, Huawei, Samsung, Texas Instruments, IBM, MediaTek, etc? What would be the point if you can get caught either way? Intel isn't like AMD, they don't have a history of reverse engineering their client's CPUs and selling them as their own designs. Stop projecting.

Chopping off their vertical integration isn't going to do anything new, it is just a different business model that many have chosen in the past. Having control of both their own design and fabs give them better feedback on how changes on one affects the other which can lead to more innovation.

This is simply not true. You cannot reverse engineer a complex soc with 100s of function blocks, and 100s of millions of logic gates totaling 10s of billions of transistors with a microscope and a pen and paper. And, no, you can't "do it with AI" either (yet).

Also nobody uses "schematics" anymore except for a very small portion of the design with custom circuits.

You are just making stuff up to hear yourself talk at this point.

AMD never REd any of Intel's designs either. What they did was to buy other companies that had good x86 designs, and those companies had been founded by people that left Intel and other prominent CPU design companies. AMD has a perpetual license to manufacture x86 processors from back in the days when Intel could not supply enough of the demand for 386 and 486 chips themselves, so they licensed to AMD.

 

[thestryker]

AMD never REd any of Intel's designs either.

Yeah about that they absolutely did reverse engineer Intel designs when the agreement from the 80s for their second sourcing expired. There was a pretty ugly legal back and forth between the two in the 90s which ended up being resolved in a settlement and the two have been effectively tied together since and x86-64 cemented it.

 

[rluker5]

This is simply not true. You cannot reverse engineer a complex soc with 100s of function blocks, and 100s of millions of logic gates totaling 10s of billions of transistors with a microscope and a pen and paper. And, no, you can't "do it with AI" either (yet).

Also nobody uses "schematics" anymore except for a very small portion of the design with custom circuits.

You are just making stuff up to hear yourself talk at this point.

AMD never REd any of Intel's designs either. What they did was to buy other companies that had good x86 designs, and those companies had been founded by people that left Intel and other prominent CPU design companies. AMD has a perpetual license to manufacture x86 processors from back in the days when Intel could not supply enough of the demand for 386 and 486 chips themselves, so they licensed to AMD.

Interesting story: https://www.asianometry.com/p/intel-and-amd-the-first-30-years

 

[Kamen Rider Blade]

So how long has it been now that Intel has had the superior node?

Superior Node on paper, what products are being made on it right now?

I think that is what is keeping customers at TSMC currently. I bet some of the contracts currently being fulfilled over there were signed prior to IDM 2.0 being a thing.

Nobody trusts IFS, for good reason. TSMC isn't trying to compete with their customers by making chips.

Intel has every reason to cheat/copy designs.

If the customers were staying away from Intel's foundries due to Intel also being one of the big chip designers with their hands in many baskets, then why are they staying away from Samsung? Samsung just makes memory and a few ARM smartphone SOCs. But their nodes also underperform TSMC's. Intel is just leaving DUV and are still selling DUV desktop chips. By the time Intel's foundries get to their fifth node in four years and start pulling ahead of TSMC is when interest will pick up. And Intel says their foundry and design businesses aren't tied at the hip. They even have separate reporting now.

We'll see if they catch up, or is it all marketing.

Why did Intel have their Mobile chips manufactured at TSMC?
Is there any good "Logical Reason" as to why Lunar Lake or Meteor Lake was fully manufactured at TSMC?

I thought IFS was "Good Enough" or "The best in the world" according to your logic.
So why have their #1 Rival make the Mobile CPU's?

They cancelled Meteor Lake-S for DeskTop, so IFS has nothing to make there.

Arrow Lake is the only DeskTop / Mobile Parts that are currently going to IFS.

IFS will be spun off as its own company.

I'll believe it when I see them officially sell it off into it's own company like AMD did with Global Foundry.
Until that day comes, they're tied at the hip.