Thanks OF, all banking is done on Linux.

https://www.thefpsreview.com/2020/03/05/amds-latest-cpu-roadmap-hints-at-zen-...



AMD is rolling their plan forward by a year since TSMC is ready to go early on the 5mn lithography reductions.

It feels like it was just yesterday that AMD released its Zen 2-based Ryzen 3000 CPUs, but the next generation of performance is right around the corner.

AMD CTO & EVP of Technology & Engineering Mark Papermaster took the stage during today’s Financial Analyst Day 2020 event and teased some of AMD’s future CPU plans. Among those were a roadmap that confirmed the progress of its Zen 3 and Zen 4 processors.

The former looks like it’ll be out before the end of the year, but there’s a surprising quirk: what happened to Zen 3 being a 7 nm+ part? Well, AMD has explained that it could no longer use “7 nm+” to describe a successive iteration of 7 nm after TSMC formally defined its EUV-based N7+ process – which Zen 3 isn’t actually using.

TSMC has three versions of its 7 nm process:

N7: the basic initial version using “DUV”-only tools (no EUV)
N7P: the second generation version of N7, which is also only DUV
N7+: EUV version of N7 for a number of layers in the metal stack

The roadmap also confirmed that Zen 4 would release by 2022. This generation is especially exciting because it marks Ryzen’s debut on the 5 nm node, which should bring significant performance and power improvements.

As for the more distant future, Papermaster teased a stacking technology called “X3D” packaging. AMD believes that this will increase bandwidth density by as much as ten times.


What is new knowledge ..... AMD is not going with each separate step of the TSMC lithograpy shrinks, but is going to skip the 7nm+ lithography stage (while still using the extra lithography burn levels that the 7nm+ process equipment can give them as the extra layers are a real advantage when building a processor)

AMD is instead going to be doing a lot of the connection leg work on a processor stacking technology, spending the "skip a weak TSMC generation" time working out all the stacking details in advance of the 5nm lithography wave which takes place the very next year.



Interestingly, AMD will still remain several generations ahead of Intel by doing this, adding a massive amount of new core counts to their chipsets (making them all threadrippers in essence).

Cooling these stacked up CPUs will likely require a new cooling technology since they are stacked up on top of each other with the bottom ones located too far away from a soldered in place heat sink for it to work well.

Commentors in the comments below theorize about the use of an oil based cooler that seals to the chipset using an "O" ring gasket, a solution which sounds potentially complicated to me.  

5nm lithography is twice as nice and the associated improved liquid cooling may mean even higher performance speeds, which is always a good thing.  

Doubling or trebling or quadrupling the number of cores doesn't hurt either.  

Getting 10 times the bandwidth density on communicating with each core means lots & lots more throughput which makes your PC act a lot more like a mainframe chipset.

AMD will have better than 5 mhz processor speeds on top of it all, too.

So, Intel has a series of very real challenges headed their way

AMD is simply going to continue to improve by leaps and bounds while Intel has to make up more & more dubious and dodgy 14nm responses that never seem to reach physical reality ......


Reminder .....  5nm is about 1/3 the size of 14nm

https://liliputing.com/2020/04/lilbits-391-better-wifi-and-arm-based-macs.html

https://www.bloomberg.com/news/articles/2020-04-23/apple-aims-to-sell-macs-wi...

5nm next year marks the start of ARM coming up strongly against Intel (and coming out against AMD too, incidentally).  

Remember, AMD will still be in there with a 5nm competitor chipset, but so far Intel will not have a 5nm competitor (but Intel claims to have some new "future plans" to make up one late next summer ......)

ARM itself has is now licensing out that 5nm Big-Little tech that is now available to everybody so you can make up whatever you want, be it a mainframe or laptop or desktop or whatever processor ---- so since Intel has a full ARM design license already Intel really has NO EXCUSE other than "dinosaur bean picker thinking" to blame for missing this particular boat yet again.

We also read that Apple is going to go with the same big-LITTLE approach that ARM recommends, basically using the mostly stock "dual sized cores" ARM approach that is being used by ARM in the new mainframe chipsets and elsewhere.

The first Mac processors will have eight high-performance cores, codenamed Firestorm, and at least four energy-efficient cores, known internally as Icestorm.

Apple is exploring Mac processors with more than 12 cores for further in the future, the people said.

As such, with the cores produced on TSMC's 5nm lithography lines customers can expect some really great processing power to happen for them at very minimal power usages.

This is standard ARM licensed stuff, so every cell phone maker can get into this action and help to cremate Intel's moribund obsolete bloated corpse very very quickly.

Note that AMD will be in there from the very start, competing in this 5nm market against Apple and all of the larger cell phone guys.

All this tends to explain Apple's latest push for new benchmarking standards that treat Apple's A-14 chipsets fairly and it also explains Intel's current active abandonment of the existing benchmarking standards bodies and all forms of marketing honesty.

https://www.theverge.com/2020/1/6/21054007/amd-7nm-ryzen-4000-cpu-ces-2020-in...

https://www.techradar.com/news/amd-ryzen-4000

There are rumors that AMD's Mobile 4000 will be based on TSMC's new 7nm EUV (extreme ultraviolet) process, similar to what's rumored to be seen with Nvidia Ampere. If this is true, the processors could be much more power efficient, which could see clock speeds see a sizable bump - which could seriously threaten Intel's chips in the gaming scene.

Another thing that could make Intel start sweating is the rumor that with Ryzen 4000, AMD may introduce more powerful hyperthreading, with each physical core having four simultaneous processing threads, as opposed to the two found on today's silicon. This is a rumor we'd definitely take with a grain of salt, but if it's true it could even further widen the gap between AMD and Intel when it comes to multi-threaded workloads.


So, what does Covid19 summer bring to us for AMD vs Intel action?

Intel's 10nm (7nm ???) is out now, swinging lots of very strange Passmark benchmarking shenanigans and some "Intel optimization" changes that don't really change anything much once you get away from the monkey'd up benchmarks Intel was pushing off on everybody.

Even the very latest Passmark version 9c  now shows existing AMD chipsets running faster than Intel's newest and best using a just released "impartial version" of the benchmark which was created to satisfy the outrage from the consumer performance fan base.

Come October of this year AMD will release the very last of the 7nm+ Ryzen 4000 laptop chipsets with a very quick sell through planned for the generation as the 5nm wave will sweep it all away next year ---

5nm will be really really big coming out from Apple, Huwawei and Mediatek all at the same time, followed by AMD's chiplets being run on the same process as soon as it clears off enough room for the protracted AMD runs.

Once built, AMD's 5nm chiplets can go into mainframe rack chipsets, Threadripper high end consumer chipsets and the sorted out slower ones will go into the first shipped Consumer Ryzen desktop versions.

Watch out for two main things at 5nm AMD consumer release next year in 2021 .......  thread counts may jump up to double to what you are used to seeing per core, and the raw CORE COUNTS themselves may quadruple as well.   On chip memory buffers will grow by 6x in size as more layers for burned on silicon memory will happen with the latest 5nm production systems.     Integrator / board builders can take the size and efficiency changes as 30+% energy efficiency increases and the 20% CPU thermal improvements as a mixture of either/or much better battery life or as a large processing through-speed  boost, depending on the intended use of the product.

This huge upsurge in core count and throughput will all be worked out in mainframe and rack space uses first, where it gives the very greatest benefit for the dollar spent, while the eventual Consumer Ryzen stuff may actually see a shifting changing consumer market where the new 5nm ARM chipsets from the phone boys have altered the landscape a good bit simply by being MUCH MUCH MUCH cheaper while still being quite powerful compared to what we have now today.  

Consumers aren't really using the recent throughput increases really as the software base hasn't changed enough to utilize it much at all yet.   So expect "cheap and good" to be what sells next year as everything will be powerful enough compared to the uses to which consumers are putting it.

AMD may find themselves just out there with everybody else, dancing on the hot coals right along with the phone boys, jest a hopping and skipping on top of the glowing coals that is all that is left from the same industry wide wild fire that has just finally cremated Intel.




===================================================




Harsh Truth Time

Intel isn't going anywhere.  Intel is still selling most of what gets sold due to their market binding legal agreements and until country by country these agreements are abrogated by the local courts then Intel isn't going anywhere no matter how sorry their products actually become.

Next fact ---- Intel always has a lot of money on hand and Intel could spend their way out of trouble at will by dropping multiple 10s of billions of dollars on it.    I am reminded of the 20 billion dollars Intel dropped on price supports for the 20nm and 28nm "phone chips" they created while attempting to take over the phone market with ATOM chipsets ......  Intel could patch their lack of technology problems with multiple layers of greenbacks laid on top of non-competitive CPU designs (same old Intel garbage in other worlds).

Intel has re-tooled their 14 nm and 10 nm and 7nm production lines several times each now .......    just not making as much progress as you would think by spending all that money.

Intel's bean pickers simply dance to an internal music we cannot understand.  But Intel isn't out of business and Intel isn't massively behind on making money, which is what Intel is all about after all.  

So Intel is fine, just peachy fine just the way it is according to the people running it.

https://liliputing.com/2020/05/tsmc-to-cease-business-with-huawei-following-u...

TSMC to cease business with Huawei following US order

Chinese electronics company Huawei’s future looks bleaker than ever. The company is a major player in the telecommunications industry and has risen to prominence in the consumer space recently thanks to a series of well-regarded smartphones.

But a series of actions from the US government in recent years have lefts countries around the world wary of using Huawei networking equipment, and recent Huawei smartphones have shipped without the Google Play Store and Play Services, since Google isn’t working directly with the company anymore either.

Now Huawei may also have a hard time manufacturing chips for its own devices, as Taiwan Semiconductor Manufacturing Co (TSMC) has announced it will no longer be filling new orders for Huawei.



The move comes a few days after the US government issued an order blocking companies outside of the United States from doing business with Huawei (if they’re also using US technology).

As Nikkei points out, the move could hurt TSMC — Huawei is the the company’s second biggest client. But violating the US order could hurt the company even more — the company’s biggest customers is the US-based Apple.

Meanwhile Huawei is likely going to be looking for other manufacturing partners, but there aren’t a lot of good options. Samsung would likely be subject to the same US government regulations, and China’s SMIC (Semiconductor Manufacturing International Corporation) is far behind the TSMC when it comes to manufacturing capacity and technology.

It’s certainly possible that the Trump administration could ease restrictions on Huawei and/or other Chinese companies such as ZTE as part of ongoing trade negotiations with China. And it’s also possible that a change in US leadership could also lead to a change in US/China relations when it comes to technology. But US skepticism of Huawei pre-dates the Trump Administration.

Engadget has a partial timeline, including the time in 2008 when Huawei decided not to buy 3Com because of US concerns, and the time in 2012 when a Congressional report suggested tight ties between Huawei and the Chinese government (claims which Huawei disputes).

But with Huawei virtually cut off from Google, TSMC, ARM, and other key partners at this point, things definitely don’t look good for the company right now.


Anybody want to place some "fill in the blank" orders for 5nm chipsets to be produced at TSMC starting next week?

You see, TSMC has this sudden 50% opening in their year 2020 5nm production schedule ........

Naw, this is  thought is wrong because most of what Hwawei sells is in China and India, etc.

hwa....way

speeded up it comes out wa..way

https://wccftech.com/intel-2021-2029-process-roadmap-10nm-7nm-5nm-3nm-2nm-1nm...



Note:  this is not a roadmap nor a product plan --- it is a design philosophy at best.  

So, Intel fires back at Apple's announcements, basically saying that Intel has NO PLANS TO EVER COMPETE with anybody ever again.

Intel is now defining themselves by their new "slow reaction" plans to only adopt new lithography and chip processing technology 2-3 YEARS behind the lithography leaders.

So, Intel isn't spending a penny on their own processor lines until any new technologies become completely mature .....  i.e. until they have been actually replaced by a "new current technology" and the selected techs have become totally mature and fixed by passage of time.  

Somebody else will spend risk capital to develop these techs.   Intel is being completely run by bean pickers that will play no part in expanding new technologies anywhere.

14nm to 10nm to 7nm to 5nm to 3nm in The Next 10 Years

Starting off with the process roadmap, Intel will be following a 2-year cadence for each major node update. We got a soft launch of 10nm (10nm+) in 2019 which will be followed by 7nm in 2021, 5nm in 2023, 3nm in 2025, 2nm in 2027 and 1.4nm in 2029. What's interesting here is that this 2-year cadence is referred to as the optimal cost-performance path by Intel themselves. So it would be Intel's priority to follow this path, but there's also a yearly cadence for the + / ++ nodes that offer more performance leverage and scalability opportunities on an existing node.

Before we talk about the optimized nodes for each process, we should focus on the key features that each major node update has to offer. For 7nm, Intel is saying the biggest feature is that it is made using EUV (Extreme ultraviolet lithography) technology. Similarly, all other major nodes will come with new features, but Intel hasn't explicitly stated what new features we could expect. At the same time as Intel introduces their 10nm++ products, they will also have production and launch planned for their next-gen 7nm process node. The 10nm and 7nm nodes were already detailed by Intel during their 2019 Investors Meeting.

Me, I think Intel is banking on all those very restrictive legal agreements that folks signed up for in order to get Intel chips at all, forgetting the standard boiler plate in those agreements that says the customer can jump ship selectively if Intel does not have a current technology item to ship to them in a timely fashion.

When ARM based chipsets grow up a bit more and become more prevalent and competitive in the marketplace, at that point in time not being able to buy
Intel Inside  will not be the showstopper that it was in the past.

https://www.tomshardware.com/features/amd-vs-intel-cpus

This is the exhaustive, once per year processor comparison that Tom's Hardware does to rank all production processors.   This 20 page article (with many exhaustive references) is what most pundits use to rank what is available for real in any yearly comparisons that they do.

This ranking excludes any of the promised Intel products that didn't make it into reality (as one in four Intel processors are wont to do nowadays) so do not be surprised to see entire families of Intel chipsets fail to make an appearance.

It does not apply any of the current mitigations for various security issues (as I wish it did) as that would take Intel completely out of the running.  Any Intel wins are always by less than 10 percentage points, while applying Intel mitigations would remove ~20%~ of Intel's "reported performance" across the board (and that would knock Intel's wins down to perhaps only 1 win as "drivers and software" are the only thing that would not be affected).

Lastly, Tom's ranks this year by not-warmed-up totally cold processor performance (a trick which Intel relies upon) as any test results beyond the magic minute have to reflect BIOS thermal throttling which again would knock Intel down by about ~15-20%~.   NOTE PLEASE  ---  AMD uses stock in the box fan and heat sink based cooling in all tests and posts all processor results after 5 minutes of warm up and with any resulting fully warmed up processor throttling being firmly in place.
We wish Intel was as honest as AMD in all their testing and reporting.


AMD vs Intel CPUs

                                                Intel                  AMD

CPU Pricing and Value                                          ✗
Gaming Performance                 ✗      
Content Creation/Productivity                               ✗
Specifications                                                       ✗
Overclocking                               ✗        (requires specialty motherboard and liquid cooling which Intel does not supply)
Power Consumption                                             ✗
Drivers and Software                  ✗      
Process Node                                                      ✗
Architecture                                                          ✗
Security                                                                ✗

Winner: AMD

Totals                                    3                  7


Final points, Tom's Hardware giving Intel a Win for having "drivers and software" is completely bogus for two reasons.  AMD uses industry standard drivers  in all cases because they PUT their drivers into those industry standards while designing and building the first prototype processors.   By the time the processors reach the full production stage, several trial lots and whatever driver tweeking that was needed has already been done and is already fully incorporated into the applicable standards.  

Next, unlike Intel, AMD doesn't have to tweek their drivers 3 times a year just to be able to claim some sort of bogus short term "advantage" over AMD's competition like Intel does --- AMD does not need bogus short term "created advantages", they (AMD) have sizable real advantages coming on all the time.

AMD really does beat Intel on all points using stock motherboards, stock coolers, using fully warmed up processors and using industry standard drivers for whatever OS standard you choose to use.

,,,,,,       that is the simple truth about the whole thing as it currently sits at 7nm.



Next year at 5nm Intel will start losing out to the phone boys too, due to very large 5nm process advantages that will ring in with ARM 5nm very very shortly.