News AMD Comes Roaring Back: Analyzing CPU Sales at Mindfactory Over the Last Five Years

[Groveling_Wyrm]

This is what I heard too. Intel was all like ...

"We have too much money, and our investors are complaining how rich they are getting from their Intel stocks, I think we should 'allow' AMD to catch up for a while, so while they launch their new Zen architecture, let's just sit on 10nm for four years pretending we're having some difficulties or something. This will be good for the public." -- in fact I think it was this dialogue that got Brian Krzanich fired ....

Yeaaahh .... No.

AMD has fought their way back, and while there certainly will be many more rounds to come, AMD is doing ALL the bruising in this one, I don't think its really debatable.

I don't think anyone is complaining about the competition. I mean ,Intel promising 2x+ better performance per dollar in HEDT than their last gen? I would be so happy as an Intel enthusiast that the HEDT gouging is finally abating. They wouldn't be doing that at all (Investors are NOT liking that news I promise you), unless they felt AMD is currently a very serious threat in all areas.

So let's all stop pretending that what Intel has been doing for the last 2.5 years has nothing to do with what AMD is forcing them to do.

I know I wouldn't be happy with a $500 four core processor in 2019 ... just sayin'.

Take a look at Intel and AMD's actual reported numbers. Intel has massive amounts of money, in laymans terms or literal terms. They are heavily insulated and quite diverse. They DON'T put all their "Eggs" in one basket. Using this diversity, they can handle setbacks. And they are NOT sitting on their butt and letting the money just go out the window. They are still developing new and revised tech. And they have also recognized this, and proven that by reorganizing their business to change for the future, in their own infrastructure and workforce. They started doing that years ago.

Intel's investors are actually probably pretty bullish on Intel. Look at Intel's stock. It has RISEN by over 20% in the last 5 years, even with AMD gaining market share. If Intel is doing so bad, then why is their stock rising? If I were to tell you that getting 2x+ better performance for anything, you wouldn't believe me. If you actually believed that from either Intel or AMD, I really hope you didn't buy stock from that info.

AMD has NOT fought for anything. They developed their processors, and offered them to the public. Just as the PUBLIC has done before, THE PUBLIC STILL decides to purchase any of AMD's or Intel's processors. THE PEOPLE that buy their product is the judge. The only thing you could actually say that AMD fought for was in their lawsuits.

Even the current trend states exactly that. If Intel is making crap processors, then why are they STILL selling more processors than even AMD?

So AMD has finally forced Intel to react. Congratulations to them. Not the first time it happened, though. Hopefully, it won't be the last. Doesn't mean anything though. That is BUSINESS, not war. AMD can't keep doing it indefinitely. Research and development costs money, and AMD has a finite amount of that.

 

[jimmysmitty]

Not sure that we know this as a fact ... We do know that Intel 10nm as very initially planned was slightly more dense than TSMC's planned 7nm ... not "way more" dense as you purported in an earlier post. It was quite close actually, maybe a few percent.

7nm+ should be 20% more dense than 7nm according to TSMC - this would leap frog Intel's 10nm. Not sure the difference between Intel's 10nm and 10nm+ or timeframes but we do know that they can rock those +'s pretty well

So, I am not 100% confident that Intel hasn't made some revisions to their very aggressive 10nm design in order to actually get things working as they are currently - and the only direction in that regard would be looser, not tighter.

If you have any source or docs on this , I am actually curious about this ...



Overall though, at the end of the day node size isn't really all that relevant - but its a good "feature" for fanboi arguments I guess (sorry AMD fans, but this one is true). Performance, cooling, cost/perf, etc. is what defines a CPUs strength (sorry Intel fans this is also true) ... while a node may roughly indicate which CPU might be hotter, perform better etc. its only the final numbers in those areas that count. Node size is marketing.

The original 10nm spec was quite a bit more dense than TSMCs 7nm. I also just did some digging and TSMC has a "6nm" so they may not even have 7nm+ anymore. Can't find much.

I am sure much has changed. Intel was vastly too aggressive. To put it in perspective, Intels 10nm was about 2x as dense as Samsung or TSMCs 10nm.

And you are correct. The number is mostly meaningless these days. Its more marketing than anything. And value is dependent on many factors not just one part. Its why I look at a platform as a whole nut just the CPU itself. You could have gotten 8 "cores" with AMDs FX series, be it slower than 8 real cores but faster than 8 SMT cores, yet as a platform it fell flat.

The 10nm process that tops out at 4 cores and clock speeds lower than the 14nm mobile CPU's they are replacing vs the 7nm 64 core enterprise CPU AMD just announced? Oh yea, Intel's 10nm is clearly neck and neck with TMSC's 7nm as a usable process...



There you go again about the market. AMD has already launched their 7nm Rome CPU's. Where is Intel? Well, according to you, Intel is smartly targeting only the Ultrabook market right now with 10nm, because that's where the money is. Why sell enterprise CPU's for thousands, when you can sell mobile CPU's for hundreds?

https://www.anandtech.com/show/12436/intel-10nm-dualcore-cannon-lake

"Back in 2013, the company planned to make CPUs produced using its 10nm technology available in 2015. "

If desktop is so simple where are the 10nm Cannonlake desktop CPU's that were supposed to be released in 2015? Oh right, they got cancelled because Intel decided to focus on the Ultrabook market in 2019 instead because that's where the money is. It all makes sense now...



Are you fishing for likes with this? Nobody in this thread complained about competition, what are you talking about? However, competition as a result of Intel falling on its face for 4 years, has not benefited the public as much as if they had stayed on schedule and continued kicking AMD's butt. What Intel has been doing has had no effect on AMD's development of Ryzen. If Intel had stayed on schedule with their roadmap, we would still have Zen 2 today with at worst the same prices, but instead of having to choose between Zen 2 and Coffee Lake, we would have the choice between Zen 2 and a much faster and more efficient Intel product stack. Sure the Intel stack would cost more, but you would be getting a clearly better product, and Intel would still have to price their CPU's sensibly vs AMD's much improved Zen 2 (ie, no $600 quad core from Intel). So how exactly do we the consumer benefit from having inferior options to choose from?

Also, Wyrm, you keep repeating points I have already disproved while asking question I already answered. And you seem to have a problem with basic terminology like what IPC is. Core count and clock speed having nothing to do with IPC. Despite your attempt to invent new things like single core IPC, AMD still leads Intel in the IPC that the rest of us are familiar with.

So you ignore everything else I see. I specified that Intel focuses on mobile and enterprise. Thats why one of their first 10nm designs and Forevos designs is their Agilex FPGA, which would kill Rome or Xeons in HPC tasks or any CPU.

I never said they canceled anything due to that. We all know 2015s 10nm was canceled due to issues. However they do focus a lot more in markets that sell more. Desktop is an easy market to flip around. Enterprise is not. Intel is putting a heavy focus on enterprise which ,again, is why 10nm will hit it in 2020, they have an FPGA designed specifically for it (also using 10nm) and are heavily invested in the technologies you see in enterprise grade products like 10Gbe, new interconnects, Optane DIMMS etc.

 

[refillable]

I would disagree. 1 Company, IN GERMANY, does NOT translate to representing the world, in ANY market. Every country is it's own market, and anyone who truly says that 1 company represents the world doesn't know the world market.

It does represent the trend for the DIY PC market, and if you disagree, you don't know how the world DIY PC market works.

And what Intel products "SUCK"? Intel may not be releasing newer, higher speed stuff, but they ARE staying competitive right now with the Ryzen. If you recall, it wasn't so long ago that AMD was TRYING to keep up with the Vishera Core, and failing at that. They were not staying competitive.

Their server portfolio is in the same situation as AMD was 5 years ago. New gen products can't be added, plagued with issues, and must be sold at a loss to beat the competition on the same price.

Intel is putting a heavy focus on enterprise which ,again, is why 10nm will hit it in 2020, they have an FPGA designed specifically for it (also using 10nm) and are heavily invested in the technologies you see in enterprise grade products like 10Gbe, new interconnects, Optane DIMMS etc.

Interesting, but Ice Lake still worries me. Remember how Broadwell-EP was the first server grade CPU on Intel's 14nm? Broadwell-EP was released in early 2016, a year after the first (decent but not top of the line) 14nm CPU was released. Even in late-ish 2019 we haven't seen a real Ice Lake on the market, and those that were paper launched are only limited to 25W. Are you sure server grade 10nm will hit in 2020?

 

[jimmysmitty]

It does represent the trend for the DIY PC market, and if you disagree, you don't know how the world DIY PC market works.



Their server portfolio is in the same situation as AMD was 5 years ago. New gen products can't be added, plagued with issues, and must be sold at a loss to beat the competition on the same price.



Interesting, but Ice Lake still worries me. Remember how Broadwell-EP was the first server grade CPU on Intel's 14nm? Broadwell-EP was released in early 2016, a year after the first (decent but not top of the line) 14nm CPU was released. Even in late-ish 2019 we haven't seen a real Ice Lake on the market, and those that were paper launched are only limited to 25W. Are you sure server grade 10nm will hit in 2020?

One site in Germany does not represent even the DIY PC market. Even if it did the DIY is a much smaller market than the mainstream consumer market is. I know plenty of people who will never buy an AMD product. AMDs short time with a bad product hasn't helped mainstream consumer minds on their products plus AMD doesn't exactly market themselves as heavily as Intel does.

Per Intels current roadmap, pending any changes or possible delays, they state they will have 10nm server parts in 2020. I typically go by roadmaps until they get changed, be it AMD, Intel, nVidia etc.

https://www.anandtech.com/show/1319...oadmap-cooper-lakesp-and-ice-lakesp-confirmed

They can always change. Their 10nm FAB could blow up or WW3 might start but until there is a change I trust their roadmaps to be accurate.

FYI this is not the first time Intel has debut a new process on the ULP mobile market. In fact the Core uArch started on mobile and was refined into Core 2 on desktop.

 

[kinggremlin]

So you ignore everything else I see. I specified that Intel focuses on mobile and enterprise. Thats why one of their first 10nm designs and Forevos designs is their Agilex FPGA, which would kill Rome or Xeons in HPC tasks or any CPU.

I never said they canceled anything due to that. We all know 2015s 10nm was canceled due to issues. However they do focus a lot more in markets that sell more. Desktop is an easy market to flip around. Enterprise is not. Intel is putting a heavy focus on enterprise which ,again, is why 10nm will hit it in 2020, they have an FPGA designed specifically for it (also using 10nm) and are heavily invested in the technologies you see in enterprise grade products like 10Gbe, new interconnects, Optane DIMMS etc.

Come on man. Don't do this to yourself. Intel didn't just start selling(?) FPGA's to 4 early access customers because they are "targetting the enterprise." They're selling FPGA's and ultrabook cpu's because small low power chips is all their broken 10nm process can produce right now. When Intel is shipping high margin 10nm Xeons, then they will be targetting the enterprise.

 

[jimmysmitty]

Come on man. Don't do this to yourself. Intel didn't just start selling(?) FPGA's to 4 early access customers because they are "targetting the enterprise." They're selling FPGA's and ultrabook cpu's because small low power chips is all their broken 10nm process can produce right now. When Intel is shipping high margin 10nm Xeons, then they will be targetting the enterprise.

The FPGAs are specifically targeting enterprise HPC tasks. Thats basically all they do.

You can talk all you want but I don't see you developing a cutting edge process tech.

 

[refillable]

One site in Germany does not represent even the DIY PC market. Even if it did the DIY is a much smaller market than the mainstream consumer market is. I know plenty of people who will never buy an AMD product. AMDs short time with a bad product hasn't helped mainstream consumer minds on their products plus AMD doesn't exactly market themselves as heavily as Intel does.

People who avoid brands are a tiny minority of the DIY PC market. DIY PC market as in "those who individually buy components from select hardware supplier". DIY PC market =/= mainstream consumer market. Brand loyalty is very low and things are really easy to turn around.

For instance Amazon bestselling CPUs' top ten are populated by AMD CPUs. 7 out of 10 are AMD CPUs. One of many examples.

https://www.anandtech.com/show/1319...oadmap-cooper-lakesp-and-ice-lakesp-confirmed

Cascade Lake SP was launched earlier this year. Reality has proven that road map to be wrong.

FYI this is not the first time Intel has debut a new process on the ULP mobile market. In fact the Core uArch started on mobile and was refined into Core 2 on desktop.

It still worries me. The 25W is yet to be launched in Q3 2019. The situation on Yonah and Conroe cannot be compared. That was in 65 nm where yields were not an issue and a 100W CPU wasn't going to have any meaningful production disadvantage vs. a 15W CPU. Broadwell is a much better comparison and you should be able to deduce where Intel is realistically going.

 

[jimmysmitty]

People who avoid brands are a tiny minority of the DIY PC market. DIY PC market as in "those who individually buy components from select hardware supplier". DIY PC market =/= mainstream consumer market. Brand loyalty is very low and things are really easy to turn around.

For instance Amazon bestselling CPUs' top ten are populated by AMD CPUs. 7 out of 10 are AMD CPUs. One of many examples.



Cascade Lake SP was launched earlier this year. Reality has proven that road map to be wrong.



It still worries me. The 25W is yet to be launched in Q3 2019. The situation on Yonah and Conroe cannot be compared. That was in 65 nm where yields were not an issue and a 100W CPU wasn't going to have any meaningful production disadvantage vs. a 15W CPU.

Even newer leaked roadmaps show 2020 for Ice Lake-SP. So nothing has changed.

My only point with Core was that Intel has been known to start in one place and move to the next. Intel launching 10nm in the 25W ULP mobile market does not mean they wont be able to launch higher watt parts.

We have little to no information on their current 10nm yields. I will say though with a heavy push to 7nm in 2021 I am sure its probably not as good as 14nm (considering how old it is nothing probably is) and Intel is probably trying harder with 7nm than anything else.

 

[realneil]

All in all, AMD is doing better now.
So much so that I just sold one of my Intel setups to pay for a Ryzen 3700X system.
I won't be selling my 9900K or my 8800K boxes, but I look forward to another Ryzen box. (my wife has a 1600X Ryzen that I put together for her office machine)
To me, it doesn't matter if AMD is beating Intel at total sales, it only matters that they're now making a lot more money than they have been. This guarantees that they're gonna be around, promoting competition for the foreseeable future.
Dr. Su has really turned their fortunes around and in doing so, she's amplified the AMD brand in the marketplace.

 

[refillable]

Even newer leaked roadmaps show 2020 for Ice Lake-SP. So nothing has changed.

Something must definitely changed; you can't put full trust in these road maps.

My only point with Core was that Intel has been known to start in one place and move to the next. Intel launching 10nm in the 25W ULP mobile market does not mean they wont be able to launch higher watt parts.

We have little to no information on their current 10nm yields. I will say though with a heavy push to 7nm in 2021 I am sure its probably not as good as 14nm (considering how old it is nothing probably is) and Intel is probably trying harder with 7nm than anything else.

Not having a fully working 25W to the mass is an indication of low yields. We'll see what we can do, but I highly doubt they're going anywhere sooner than 2021 with their 10nm based on what I saw.

 

[Olle P]

I find it really surprising there hasn't been more of a bloodbath at Intel in upper management based on the disastrous last 5 years. ...
... roadmaps indicate they're not going to go anywhere until 2021.

I beg to differ.
Intel has had one "disaster", that hasn't resulted in any significant revenue problems... yet.
That one disaster is the delay of the 10nm node.

• Intel has (this far) been able to refine their Skylake architecture to keep up with AMDs performance.
• The shift to 10nm plants has resulted in a reduced 14nm production capacity, which has allowed Intel to concentrate production to the most profitable SKUs.
• The found vulnerabilities (resulting in a performance decrease when implementing fixes) has forced companies to buy more Xeons, keeping revenues up.
• AMD hasn't really caught up on the laptop/tablet side of CPUs, and Intel will most probably keep their advantage there for yet a while. (This is what will keep Intel alive.)
• Intel is about to get into the discreet GPU market as well. How well that goes is anybody's guess right now.

Intel will most likely take a hit soon though:

• Most companies building a new server park will opt for AMD. Plummeting Xeon sales will be tough on Intel, and that can't be reversed on a dime.
• AMD will most likely rule desktop and HEDT for years to come. There may be a few applications where Intel can have a SKU that is "the best performer", but Ryzen and Threadripper will be the go-to options.

Intel's upper management are persons that also own a lot of Intel shares. Replacing them prematurely will likely make the stock value fall, but won't fix the 10nm node that should have been up and running in 2015.
Now they've got time to dump their (share) holding before prices go into a free fall...

I'm sorry but I don't understand, what's disastrous about having growth in a market that was prognosed to be stagnant at best?

That's an interesting point!

 

[jimmysmitty]

Something must definitely changed; you can't put full trust in these road maps.



Not having a fully working 25W to the mass is an indication of low yields. We'll see what we can do, but I highly doubt they're going anywhere sooner than 2021 with their 10nm based on what I saw.

Of course you can't completely trust a roadmap. Especially now when process tech has become increasingly harder to design and much more costly. Intel put $7 billion to fit out FAB 42 for 7nm. However Intel has the money and resources to spend so I wouldn't put it past them to be able to push 10nm and then 7nm faster than most people think.

So for now I go based on the roadmaps until it is proven wrong as we have nothing else to go by. Q3 2019 is still going but you can alredy buy Dells and others with their 10nm CPUs in it. It takes time for a CPU to saturate the market and replace the old ones. This year we bought a few new ones and finally got 8th gen laptop CPUs while before we were still only able to get 7th gen laptop CPUs.

I beg to differ.
Intel has had one "disaster", that hasn't resulted in any significant revenue problems... yet.
That one disaster is the delay of the 10nm node.

• Intel has (this far) been able to refine their Skylake architecture to keep up with AMDs performance.
• The shift to 10nm plants has resulted in a reduced 14nm production capacity, which has allowed Intel to concentrate production to the most profitable SKUs.
• The found vulnerabilities (resulting in a performance decrease when implementing fixes) has forced companies to buy more Xeons, keeping revenues up.
• AMD hasn't really caught up on the laptop/tablet side of CPUs, and Intel will most probably keep their advantage there for yet a while. (This is what will keep Intel alive.)
• Intel is about to get into the discreet GPU market as well. How well that goes is anybody's guess right now.

Intel will most likely take a hit soon though:

• Most companies building a new server park will opt for AMD. Plummeting Xeon sales will be tough on Intel, and that can't be reversed on a dime.
• AMD will most likely rule desktop and HEDT for years to come. There may be a few applications where Intel can have a SKU that is "the best performer", but Ryzen and Threadripper will be the go-to options.

Intel's upper management are persons that also own a lot of Intel shares. Replacing them prematurely will likely make the stock value fall, but won't fix the 10nm node that should have been up and running in 2015.
Now they've got time to dump their (share) holding before prices go into a free fall...

That's an interesting point!

I wouldn't assume any company would just jump to AMD easily. There is a lot to it. I know a lot of IT professionals who wont touch AMD due to them having dropped out of the server market. There is a lot of trust to rebuild on AMDs side. I think the server market will be very slow to adopt AMD and by the time the momentum gets there Intel will have more equally competitive products.

 

[TerryLaze]

because small low power chips is all their broken 10nm process can produce right now.

Yeah how is that different from AMD?
AMD went from 5Ghz to 4Ghz and from a bazillion W to 65W for the top chip.

The only difference is that intel is not yet forced to sell them to desktop because they are still plenty fast on 5 year old tech.

If there is a shortage of 10nm laptops it's only because intel is selling them on m.2 boards for enterprise,you can have almost limitless numbers of cores on a single machine.
https://www.tomshardware.com/news/intel-ice-lake-10nm-nervana-npp-i-accelerator,39731.html

 

[refillable]

Of course you can't completely trust a roadmap. Especially now when process tech has become increasingly harder to design and much more costly. Intel put $7 billion to fit out FAB 42 for 7nm. However Intel has the money and resources to spend so I wouldn't put it past them to be able to push 10nm and then 7nm faster than most people think.

They put billions of dollars to 10nm as well, but despite that they were 2 years behind schedule and 1 year behind the competition (and counting, if you're talking about desktop and server processors). Let's hope they can make a comeback, though. But realistically, that'd at least be on 2022 unless they're doing wonders.

You can alredy buy Dells and others with their 10nm CPUs in it. It takes time for a CPU to saturate the market and replace the old ones. This year we bought a few new ones and finally got 8th gen laptop CPUs while before we were still only able to get 7th gen laptop CPUs.

Laptop takes time to saturate the market, but right after a new model is launched, Reviewers usually would recieve samples and retailers would put it on their website. None of these two are happening with Dell's new XPS, so I said it was a "paper launch". I know it might be too harsh, but one model that's not mass marketed 2 month after the "launch"? There's definitely still a serious issue with the fabrication yields.

I wouldn't assume any company would just jump to AMD easily. There is a lot to it. I know a lot of IT professionals who wont touch AMD due to them having dropped out of the server market. There is a lot of trust to rebuild on AMDs side. I think the server market will be very slow to adopt AMD and by the time the momentum gets there Intel will have more equally competitive products.

I agree on the first part, but the question is how "Intel will have more equally competitive products.". By the time Intel release their 10nm parts, which I'm 90% sure Intel won't be doing that before 2021, AMD would have released Zen 4 on 5nm. That is VERY worrying, but we'll see.

 

[jimmysmitty]

They put billions of dollars to 10nm as well, but despite that they were 2 years behind schedule and 1 year behind the competition (and counting, if you're talking about desktop and server processors). Let's hope they can make a comeback, though. But realistically, that'd at least be on 2022 unless they're doing wonders.



Laptop takes time to saturate the market, but right after a new model is launched, Reviewers usually would recieve samples and retailers would put it on their website. None of these two are happening with Dell's new XPS, so I said it was a "paper launch". I know it might be too harsh, but one model that's not mass marketed 2 month after the "launch"? There's definitely still a serious issue with the fabrication yields.



I agree on the first part, but the question is how "Intel will have more equally competitive products.". By the time Intel release their 10nm parts, which I'm 90% sure Intel won't be doing that before 2021, AMD would have released Zen 4 on 5nm. That is VERY worrying, but we'll see.

I doubt 10nm cost as much as 7nm and the $7 billion is just to fit the FAB, not counting R&D they have been doing.

If I can go to Dells website and order the product its not a paper launch. Broadwell-S was a paper launch.

There was another leaked roadmap showing 10nm Ice Lake-SP in 2020 and early 2021 would be the launch of the Eagle Stream platform which would have Sapphire Rapids, the successor to Ice Lake, using the improved 10nm and Willow Cove uArch. Then in 2022 they will launch Granit Rapids on 7nm using the Golden Cove uArch.

That is their current plans. I trust it will stay that way, again pending any possible delays that may come up.

We will see how TSMCs 5nm node plays out. I always say that if the largest Process Technology company has issues others will likely run into some too. And from specs their 5nm should compete with Intels 7nm from a density standpoint, not that thats the only metric that matters.

 

[Olle P]

I wouldn't assume any company would just jump to AMD easily. There is a lot to it. ...

I agree on that! If it was easy AMD would have had more of a success with the first generation of Epyc.
All of the security issues with Xeon (combined with the reduced running cost of Epyc) has made quite a few companies going full steam ahead with a switch to AMD for quite a while now, and soon that change will be executed...

Here's a quote about the current state: More's Law is Dead

If I can go to Dells website and order the product its not a paper launch. ...

Pre-orders don't make a launch.

 

[joeblowsmynose]

The original 10nm spec was quite a bit more dense than TSMCs 7nm. I also just did some digging and TSMC has a "6nm" so they may not even have 7nm+ anymore. Can't find much.

I am sure much has changed. Intel was vastly too aggressive. To put it in perspective, Intels 10nm was about 2x as dense as Samsung or TSMCs 10nm.
...

TSMCs 7nm+ is the EUV version of their 7nm. This article claims 20% density increase. 10% performance increase and increase in power efficiency by 15%

https://www.techspot.com/news/80237-tsmc-7nm-production-improves-performance-10.html

Article is a few months old though, but they've been already producing on that 7nm+ for huawei for months now.

 

[jimmysmitty]

I agree on that! If it was easy AMD would have had more of a success with the first generation of Epyc.
All of the security issues with Xeon (combined with the reduced running cost of Epyc) has made quite a few companies going full steam ahead with a switch to AMD for quite a while now, and soon that change will be executed...

Here's a quote about the current state: More's Law is Dead

Pre-orders don't make a launch.

Depending on the company and how they have their servers setup the cost to move to AMD might be vastly too much. VMWare requires all the same servers in nodes and while having different servers across nodes its easier and works better if the nodes have similar hardware, even if a few generations apart.

Then you have software support, channel and vendor support etc. Its a lot of work. AMD dropped out of the server market and it will take time and hard work to win back some companies trust. I say that by the time AMD makes enough progress Intel will be much more competitive per core.

And lastly, per Dells site I can order one with a 10th gen CPU and get it October 6th. We order some business laptops from them and their shipping is normally pretty long, we typically wait 1-2 weeks.

TSMCs 7nm+ is the EUV version of their 7nm. This article claims 20% density increase. 10% performance increase and increase in power efficiency by 15%

https://www.techspot.com/news/80237-tsmc-7nm-production-improves-performance-10.html

Article is a few months old though, but they've been already producing on that 7nm+ for huawei for months now.

What I meant is we are not 100% on if AMD is going to use the 7nm+ or might move to TSMCs 6nm.

Just remember though performance increases on process nodes do not always translate directly to CPU performance. Power is a lot closer normally.

 

[kinggremlin]

I beg to differ.
Intel has had one "disaster", that hasn't resulted in any significant revenue problems... yet.
That one disaster is the delay of the 10nm node.

• Intel has (this far) been able to refine their Skylake architecture to keep up with AMDs performance.
• The shift to 10nm plants has resulted in a reduced 14nm production capacity, which has allowed Intel to concentrate production to the most profitable SKUs.
• The found vulnerabilities (resulting in a performance decrease when implementing fixes) has forced companies to buy more Xeons, keeping revenues up.
• AMD hasn't really caught up on the laptop/tablet side of CPUs, and Intel will most probably keep their advantage there for yet a while. (This is what will keep Intel alive.)
• Intel is about to get into the discreet GPU market as well. How well that goes is anybody's guess right now.

Intel will most likely take a hit soon though:

• Most companies building a new server park will opt for AMD. Plummeting Xeon sales will be tough on Intel, and that can't be reversed on a dime.
• AMD will most likely rule desktop and HEDT for years to come. There may be a few applications where Intel can have a SKU that is "the best performer", but Ryzen and Threadripper will be the go-to options.

Intel's upper management are persons that also own a lot of Intel shares. Replacing them prematurely will likely make the stock value fall, but won't fix the 10nm node that should have been up and running in 2015.
Now they've got time to dump their (share) holding before prices go into a free fall...

I agree with almost everything you said. Good post. I do, however, disagree on the revenue hit part. They have not suffered any decrease in revenue, yet. However, the delay in 10nm was a major (though not the only) contributor to the months' long supply shortages. Intel never intended to release an 8 core mainstream CPU on 14nm, nor did they plan to still be 14nm when they entered a core race in the HEDT and Xeon lines with AMD once Ryzen was released. So even though they didn't see a revenue decrease, they should have seen much higher revenue had they been able to meet demand, which would have been achievable had they been on 10nm. Also, as you pointed out, Intel's own security woes ironically drove an increase in demand for their products to make up for the performance drop from security patches further contributing to the shortage. It's not often a company benefits so much from their own screwup, and I have a hard time given the executives at Intel a pat on the back and vote of confidence for that unexpected source of revenue.

The pain is coming very soon though. As already discussed, the huge price cuts coming for Cascade Lake X are going to affect their bottom line. That pricing scheme is going to have to make its way into the mainstream and enterprise lines as well. We'll see how close Intel gets to it claim of mid 2020 for 10nm Xeon. What they have out now doesn't build much confidence they'll be shipping a full product stack in volume in that time frame.

Yeah how is that different from AMD?
AMD went from 5Ghz to 4Ghz and from a bazillion W to 65W for the top chip.

The only difference is that intel is not yet forced to sell them to desktop because they are still plenty fast on 5 year old tech.

What 14nm Ryzen chip clocked to 5Ghz? Clock speeds increased when AMD moved from 14nm Ryzen to 7nm Ryzen 2. AMD has multiple 225W TDP 7nm Epyc CPU's. The process clearly scales up well beyond the ultra low power chips. Last I checked, AMD wasn't having any issues producing 7nm chips with more than 4 cores either.

The FPGAs are specifically targeting enterprise HPC tasks. Thats basically all they do.

No, not really. There are 3 versions of the FPGA. Only one of them is intended for compute usage. These chips will see more usage in networking and connectivity applications.

 

[TerryLaze]

What 14nm Ryzen chip clocked to 5Ghz? Clock speeds increased when AMD moved from 14nm Ryzen to 7nm Ryzen 2. AMD has multiple 225W TDP 7nm Epyc CPU's. The process clearly scales up well beyond the ultra low power chips. Last I checked, AMD wasn't having any issues producing 7nm chips with more than 4 cores either.

Compared to their previous architecture which is constructor cores ryzen is a low powered chip failing to go above 4Ghz in many cases when their previous architecture reached 5Ghz relatively easily,it's low powered.
In zen 2 each chiplet has two ccxs and each ccx has 4 cores,AMDs ryzen can't hit above 4Ghz on a quad core,it's small.
So zen 2 is just as broken a process as intel's 10nm is,intel just isn't forced to sell them to desktop yet.

Don't mix in server CPUs, the thread is messy enough,this is about mainstream/consumer CPUs.

 

[joeblowsmynose]

...
What I meant is we are not 100% on if AMD is going to use the 7nm+ or might move to TSMCs 6nm.
...

Recent leaked roadmap still indicates 7nm+ ...
https://www.techpowerup.com/259139/...a2-and-zen-3-onto-7nm-with-2020-launch-window

I have heard no news otherwise.

 

[kinggremlin]

Compared to their previous architecture which is constructor cores ryzen is a low powered chip failing to go above 4Ghz in many cases when their previous architecture reached 5Ghz relatively easily,it's low powered.
In zen 2 each chiplet has two ccxs and each ccx has 4 cores,AMDs ryzen can't hit above 4Ghz on a quad core,it's small.
So zen 2 is just as broken a process as intel's 10nm is,intel just isn't forced to sell them to desktop yet.

Don't mix in server CPUs, the thread is messy enough,this is about mainstream/consumer CPUs.

Everyone here is talking about the process node transition from 14nm to 10/7nm.
Why are you talking about architecture?

 

[NightHawkRMX]

Generally when there is a node shrink the engineers also give the cpu other architectural changes.

 

[kinggremlin]

Generally when there is a node shrink the engineers also give the cpu other architectural changes.

That's a definite no on the Intel side. Tick-Tock. As Intel defined it, Tock was the architecture change while tick was the die shrink.

Intel hasn't had an architectural change that dropped core speeds since they moved from Netburst to Core, 13 years ago. They've basically been tweaking core every generation since. There was no expected or announced drop in core speed when Intel announced Sunny Cove.

 

[NightHawkRMX]

Well, Intel's lack of innovation is finally starting to catch up to em.