Why does AMD have such little development outside of x86 and GPUs? APUs and the IPs were neat but they need to transfer their expertise to innovation and rapidly growing markets like this

fancuckerWhy does AMD have such little development outside of x86 and GPUs?

Compare the size and R&D of each of the companies (intel to AMD) and that'll give you the full picture.

Xx Tek Tip xXCompare the size and R&D of each of the companies (intel to AMD) and that'll give you the full picture.

How will they encourage investment and growth with limited innovation and an anemic product portfolio? I really want them to grow and EPYC simply won't do it

fancuckerHow will they encourage investment and growth with limited innovation and an anemic product portfolio? I really want them to grow and EPYC simply won't do it

Considering they already can't fully compete in the GPU space, they don't need to spread themselves thinner.

moproblems99Considering they already can't fully compete in the GPU space, they don't need to spread themselves thinner.

Quite funny how people, after one generation of good products, start to believe AMD is the master of all chips and they need to go everywhere, nobody else can do better, end of game.
Chill guys, Intel is stuck using an ancient old process and respinning 2015 parts while AMD has now caught up in IPC, a full 4-5 years after Skylake launched. AMD had a good idea with chiplets but that is it, Zen core is nothing special.

fancuckerWhy does AMD have such little development outside of x86 and GPUs? APUs and the IPs were neat but they need to transfer their expertise to innovation and rapidly growing markets like this

First it comes to size. AMD is a relatively small company compared to Intel (1/10th the size) so they don't have the man-power to devote to these ventures. Not only this, but they are also competing against Nvidia who is larger than AMD, but able to focus their resources entirely on GPU/GPU related technologies. The fact that AMD is able to stomp Intel and compete very well against Nvidia speaks volumes to their competency, but there is a limit to how much you can do, and AMD seems to be at that limit.

www.diffen.com/difference/AMD_vs_Intel
www.owler.com/company/nvidia

Second, they are probably very wise for not getting into the modem business, Intel has lost around 20 billion dollars in their disastrous attempt to compete in this market (wirelessone.news/10-r/1383-intel-s-us-20b-loss-on-5g). AMD can't afford to just throw away that kind of money.

yeeeemanZen core is nothing special.

I mean it is pretty much as good as anything Intel has. So does that mean that CPUs in general are nothing special? Generally speaking, Intel and AMD are neck and neck single core with AMD able to handily win multicore in many cases.

Additionally, the fact it is chiplet makes it special.

yeeeemanAMD had a good idea with chiplets but that is it, Zen core is nothing special.

I bet there is some executive back at Intel that keeps saying exactly that. I don't know if it's funny or sad, it wont end well for anyone either way if that's what people really think.

yeeeemanQuite funny how people, after one generation of good products, start to believe AMD is the master of all chips and they need to go everywhere, nobody else can do better, end of game.
Chill guys, Intel is stuck using an ancient old process and respinning 2015 parts while AMD has now caught up in IPC, a full 4-5 years after Skylake launched. AMD had a good idea with chiplets but that is it, Zen core is nothing special.

WIn the same way that processors in general are nothing special, Intel has nothing special, AMD has nothing special, what is the point of this post?

somethinggenericFirst it comes to size. AMD is a relatively small company compared to Intel (1/10th the size) so they don't have the man-power to devote to these ventures.

Of course they do.

Not only this, but they are also competing against Nvidia who is larger than AMD, but able to focus their resources entirely on GPU/GPU related technologies. The fact that AMD is able to stomp Intel and compete very well against Nvidia speaks volumes to their competency, but there is a limit to how much you can do, and AMD seems to be at that limit.

Or processors are limited by science and technology more than anything else and it's not that hard to design them once you gather 50 talented people and give them tools and IP? :)

Your main mistake is that you compare number of employees like if they were all responsible for developing the chips.

The reality though is that in Intel and AMD, much like in most other companies, the team responsible for developing the product is relatively small.
We're talking about maybe a few dozen engineers designing the CPUs.
They are surrounded by thousands that test the product and run the business (administration, HR, sales, IT, marketing, analytics, client support etc.).
AMD is expected to have relatively low number of people in sales and support/consultancy. The rest is likely quite limited as well. But the engineering team may be just as large as Intel's.

And keep in mind that a lot of Intel's staff is responsible for semiconductor process and fabrication (anything from material science to physical work at the plant).
By comparison: TSMC employs 48,000 people (Intel: 110,000; AMD: 10,000).

Second, they are probably very wise for not getting into the modem business, Intel has lost around 20 billion dollars in their disastrous attempt to compete in this market (wirelessone.news/10-r/1383-intel-s-us-20b-loss-on-5g). AMD can't afford to just throw away that kind of money.

No offense, but why do you believe in (let alone: cite) a guess from a bloke put on an icky website?
He explicitly says that a "reputable firm" estimates this to be $14B, but he thinks it's $20B (because why not?).

That's even worse than youtube vloggers. What's wrong with the world?

ZoneDymoWIn the same way that processors in general are nothing special, Intel has nothing special, AMD has nothing special, what is the point of this post?

I would have been amazed if they beat Intel IPC at Out of Order Execution, but they don't, because the chiplet. It adds latency and decreases IPC. If they had a monolithic die it would be faster, but they don't due to budget, and die costs.

So their chiplets are good, but not great. They managed to best Intel at core counts, and in order execution.

Let's not call that a chocolate pie when it still has corn in it.

On topic, Intel doesn't have the foundry to support what smartphone manufactures want, so better to Apple who can lease it to Samsung with their fans than to Samsung to hold for a couple years before leasing to Apple.

fancuckerHow will they encourage investment and growth with limited innovation and an anemic product portfolio? I really want them to grow and EPYC simply won't do it

Lol. I hope you're getting paid to post this anti-amd rhetoric all the time.

SteevoI would have been amazed if they beat Intel IPC at Out of Order Execution, but they don't, because the chiplet. It adds latency and decreases IPC. If they had a monolithic die it would be faster, but they don't due to budget, and die costs.

So their chiplets are good, but not great. They managed to best Intel at core counts, and in order execution.

Let's not call that a chocolate pie when it still has corn in it.

On topic, Intel doesn't have the foundry to support what smartphone manufactures want, so better to Apple who can lease it to Samsung with their fans than to Samsung to hold for a couple years before leasing to Apple.

Everything can always be called undoable due to budget but the truth is that yields would be extremely low if everything was 7nm, chiplets are a practical answer to that, it works.

Nobody is calling anything chocolate pie, you are making things up, AMD is just beating Intel right now but to you the jump is not big enough, ok.... now what?
According to you:
Intel has nothing special
And AMD has nothing special sooooo you want some made-up processor that is better then everything that actually currently exists then?
Great...again...what is the point of this post?

ZoneDymoEverything can always be called undoable due to budget but the truth is that yields would be extremely low if everything was 7nm, chiplets are a practical answer to that, it works.

Nobody is calling anything chocolate pie, you are making things up, AMD is just beating Intel right now but to you the jump is not big enough, ok.... now what?
According to you:
Intel has nothing special
And AMD has nothing special sooooo you want some made-up processor that is better then everything that actually currently exists then?
Great...again...what is the point of this post?

Intel has nothing special, AMD has a revolutionary design that finally caught them up with Intel on some fronts, and behind on others.

My point was about the AMD posts in a thread about Intel IP, not X86 CPU's. AMD sold off their mobile division long ago.

The chocolate pie was reference to others, that calling the AMD design "special" may be correct, but its no more special than when Intel did it, and then AMD touted their "real cores that real men" use, and how they didn't need glued together chips. Whatever works is great, but it does by fact add latency which hurts performance in some areas. I understand why they did it, and would have made the same call.

notbOf course they do.

They don't and they frankly shouldn't. AMD is competing VERY favorably against Intel and has had good success against Nvidia and entering into a market with juggernauts like Qualcomm and Huawei where they have zero experience is an extremely bad idea.

Or processors are limited by science and technology more than anything else and it's not that hard to design them once you gather 50 talented people and give them tools and IP? :)

This reminds me of the recipe for bear stew where the first step is to kill a bear.
Finding these people, the tools and the IP will not be cheap, easy, or fast coming. This is why I brought up the company size issue because a company like Intel either has the people, tools and IP's on hand, or can throw money at this problem much more effectively than AMD and steal established talent away from companies, buy IPs and buy tools.

The reality though is that in Intel and AMD, much like in most other companies, the team responsible for developing the product is relatively small.
We're talking about maybe a few dozen engineers designing the CPUs.
They are surrounded by thousands that test the product and run the business (administration, HR, sales, IT, marketing, analytics, client support etc.).
AMD is expected to have relatively low number of people in sales and support/consultancy. The rest is likely quite limited as well. But the engineering team may be just as large as Intel's.

I would say that AMD's team's are bigger than Intel's since from all reports, Intel only has a team for 1 cpu design, whereas AMD has multiple teams leap frogging eachother in development, but again, that giant pool of people around the team is very important, and AMD would need to hire people to fill these spots, which takes money, and time.

And keep in mind that a lot of Intel's staff is responsible for semiconductor process and fabrication (anything from material science to physical work at the plant).
By comparison: TSMC employs 48,000 people (Intel: 110,000; AMD: 10,000).

Which means in the case of Intel, they are able to do everything in house is faster, cheaper and more reliable, unlike AMD who has to rely on TSMC or other foundries.

No offense, but why do you believe in (let alone: cite) a guess from a bloke put on an icky website?
He explicitly says that a "reputable firm" estimates this to be $14B, but he thinks it's $20B (because why not?).

That's even worse than youtube vloggers. What's wrong with the world?

I chose that source because it was one of the few sources I found that actually had something beyond "Intel lost billions", and I agree with the additional loss of reputation factor. Intel isn't looking good, and hasn't for the last few years with all the security flaws, and non-existent 10nm success. Combine this with AMD's massive return into the industry and ARM encroaching into the server business Intel is looking like a dated relic that is going to be left behind.
This also reflects very poorly on their GPU business (which is already in trouble) because it casts a very poor light on their situation, is the Intel GPU going to be the next Intel modem?

Really the only way I could see AMD successfully entering the Modem market would be with the help of a company like Samsung since they are partnered already with RDNA, but even then there is a ton of risk for little reward.

somethinggenericThey don't and they frankly shouldn't. AMD is competing VERY favorably against Intel and has had good success against Nvidia and entering into a market with juggernauts like Qualcomm and Huawei where they have zero experience is an extremely bad idea.

AMD is in a comfortable position of being smaller than Intel and Nvidia in markets they compete. So yes, they can focus on niches and gain market share.
This will not last forever. The reason why Intel and Nvidia invest so much time in new activities (interconnects, AI, autonomous driving etc) is that every company tries to grow and they couldn't do that in their core business.

We'll see how it goes for AMD, but - obviously - they'll have to change strategy at some point - becoming more like Intel and less like a sexy, agile, pro-enthusiast competitor.

Finding these people, the tools and the IP will not be cheap, easy, or fast coming. This is why I brought up the company size issue because a company like Intel either has the people, tools and IP's on hand, or can throw money at this problem much more effectively than AMD and steal established talent away from companies, buy IPs and buy tools.

You say this like if Intel was a giant company in Silicon Valley and AMD was a tiny family business in wasteland.
They're both big. They're placed in the same place. They offer similar jobs and pay similar money. And workforce moves between them. Always has.

For someone who wants to design chips or write low-lever libraries, it's still a dream job. Maybe Intel is more prestigious, but that doesn't mean they hire all talented candidates available. They're not THAT big. :)

I would say that AMD's team's are bigger than Intel's since from all reports, Intel only has a team for 1 cpu design, whereas AMD has multiple teams leap frogging eachother in development

I'd lover to see some of these reports.

that giant pool of people around the team is very important, and AMD would need to hire people to fill these spots, which takes money, and time.

Once again: where does this opinion about AMD come from? They're not a small startup. They're not a newcomer. They've been in this business for decades. They have the people they need.

Which means in the case of Intel, they are able to do everything in house is faster, cheaper and more reliable, unlike AMD who has to rely on TSMC or other foundries.

So first of all: a larger company is not doing things faster. It's the opposite most of the time.

And no, AMD has the advantage here. Outsourcing means they can deliver some things faster and cheaper. And with much lower risk.
Because if you're developing both the CPU and the production node, you have the risk that they won't sync well. You could have a good CPU design, but no way to make it. Or a plant that does nothing because the CPU isn't ready or the orders are low right now.

The reason why Intel didn't decide to sell the fabrication part is simple: they're big. They must have a guarantee that a plant will be available.
Once again: it's a disadvantage of being a large company and a clear leader.

I chose that source because it was one of the few sources I found that actually had something beyond "Intel lost billions", and I agree with the additional loss of reputation factor.

I understand it fits right into your opinion, but it's just a random site. You can't believe in something written online just because it's original or you like it.

SteevoIntel doesn't have the foundry to support what smartphone manufactures want

Foundry had little to do with it, Intel modems were trash as simple as that. When Apple went with Intel (& QC) 2 gens back, IIRC they limited the QC handsets to match Intel's performance of course Intel modems also ran hotter so :shadedshu: