Donate!
Welcome, Guest. Please Login or Register :: View Members
Pages: 1 ... 15 16 17 18 19 ... 21
Send Topic Print
AMD & others --- Intel dominance in 2022 (Read 9739 times)
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12629
Fayetteville, NC
Gender: male
Re: AMD & others --- post AMD dominance in 202
Reply #240 - 10/13/21 at 11:47:40
 

We have us an historical moment in the past 3 years that is occurring right now ----- Intel has three chipsets out of their entire current crop of chipsets that can claim "competitiveness" against the currently released AMD products ...... if you hold your mouth just right, that is ......   and compare them to a slightly non-standard match up that is giving the advantage to Intel.

Intel does have some issues that MS just rang the combined Wintel effort 15-25% back down into the hole with a very strongly flawed Win 11 initial release  but hey, when you are Intel you  have to take whatever crumbs you can get when you can get them.

This is a lot closer than Intel has been in 3 years ......  

But sorry, Intel, there are no awards for close second place.

Hey Lookie Over There, Intel --- AMD has got two entirely new generations of their entire AMD line up sitting over in the wings waiting to ship out very early first quarter 2022 and then again later third quarter next year  (processors and packages in the first wave have already been produced in TSMC's new facilities no less).   They both swing HUGE 1.5 gig caches on the top of the processor stack, yep some right huge brand new TSMC tech 15% product improvement L3 caches ..... and these chip assemblies are also carrying some 25-35% throughput TSMC architectural improvements inside the processor itself all across the board.

 ...... AND A BRAND NEW TSMC LITHOGRAPHY SHRINK OR TWO are scheduled for 2022, let's not forget those .......

Intel will go back in the hole rather deeply yet again as each new AMD wave is released and the products go out for benchmarking.

And TSMC has also come up with some additional AMD 6nm and 5nm in the first quarter of 2022 and some additional 4nm and 3nm production volume allocations for AMD to use in the 3rd quarter of 2022, so AMD can actually go peel off some more fresh Intel Market-share each month in 2022.

Roll Eyes



Fair and balanced coverage

Intel has stronger commercial agreements with their box makers, to say the box makers trust Intel to be the better option in the long term is an industry truism.   Intel can screw up royally in the short term and the box makers just keep on using their current stuff on the theory that Intel will fix it next time.   They also trust Intel to spread enough brown vapor around such that consumers will buy the stuff when the box makers deliver it.

Intel does ink some very restrictive distributor agreements that are in restraint of free trade and they are allowed to get away with it by US regulators.   EU regulators fine Intel every year or so and every year Intel stalls paying the penalties, endlessly stringing the legal process along until the judgments become moribund.

The box makers will always support Intel BECAUSE THEY REALLY HAVE NO CHOICE BUT TO DO SO.

Watch the budding RISC-V roll out in the EU break up a few of these rigged games, but also watch Intel and MS take legal actions to protect their duopoly in the long term.

Dirty Tricks Department

The new AMD L-3 cache (based on TSMC packaging tech) isn't recognized by Win 11.   Why?   Because it wasn't invented by Wintel and Wintel has no CURRENT answer to the 15% power boost it gives to AMD processors.    

So, jest go break Windows so the new AMD stuff doesn't work right, that seems to be the answer the Wintel boys always seem to have for innovations coming from somebody else.

Please note that older versions of Win 10 still works fine (this week anyway) and Linux's increased performance is unchanged as well.




UPDATE:     It turns out that Microsoft had already been given the correct drivers by AMD, but MS simply hadn't rolled them out in time for the Win 11 introduction.    oops     Embarrassed

MS has caught a whole bunch of shite over this "unintentional omission" and has corrected it as of this week's software roll out.

Intel can no longer claim any form of equivalent performance as the AMD 1.5 gig L-3 cache yields a clean 15% performance boost for AMD's current production processors.

Intel got to see the bright sunlight for a day or so on 3 processors, but now it it's getting stuffed back down underneath the big flat rock again --- poor old Intel ......


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


Intel has finally broken ground on their two new fabs in their Chandler Arizona complex this past week.   Samsung and TSMC have their buildings up already and are busy installing production lines.    ASML is building extreme EUV machines at the rate of one a week to put into all these new plants.

ASML is the working standard for all this stuff (5nm and 3nm are fully developed standards that all computers will be built to).    Intel will be late on getting equipment as they didn't order it back in a timely fashion like Samsung and TSMC did.

This is good for the industry as a whole as Intel vs Apple vs ARM have shown us that grossly un-mixable standards really don't help things to move along together harmoniously.

If Intel is going to be a fab they need to GET ALONG with everybody and work within the current most modern standards instead of always "doing it the Intel way ......."  and going off on their own as they always did before isn't going to work any more.

Back to top
 
« Last Edit: 10/22/21 at 18:25:55 by Oldfeller--FSO »  

Former Savage Owner
  IP Logged
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12629
Fayetteville, NC
Gender: male
Re: AMD & others --- post AMD dominance in 202
Reply #241 - 10/19/21 at 20:23:22
 

https://liliputing.com/2021/10/lilbits-raspberry-pi-legos-alibaba-open-source...

Alibaba open sources its RISC-V chips, Mac Pro could have 40-core CPU next year and 128-core graphics too



This is a big deal as it shows both RISC-V and ARM servers picking up real muscle power to take on  the x86 boys and to simply slay them with power efficiency and superior processing power.

Alibaba Cloud (China) Unveils New Server Chips to Optimize Cloud Computing Services

Alibaba introduces Yitian 710 server chips with up to 128 Armv9 CPU cores. The company also plans to open source the code for its XuanTie RISC-V processors cores as well a Linux, Android, RTOS, and AliOS software stacks for those chips. It looks like source code and documentation for E902, C906, and C910 are already on GitHub
.

And being open source on BOTH hardware and software, anybody can use it .......    

Intel and AMD both took a lick on this one.   Open Source RISC-V processors are quickly becoming more competitive to ARM mobile chipsets and both are edging closer to being competitive to Intel/AMD.

Alibaba and Google and Amazon Web Services are showing up now as major customers/competitors for mainframe processor designs.

I will communicate the closing race between the processor types as they each develop down their relative pathways.

Back to top
 
« Last Edit: 10/22/21 at 18:14:59 by Oldfeller--FSO »  

Former Savage Owner
  IP Logged
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12629
Fayetteville, NC
Gender: male
Re: AMD & others --- post AMD dominance in 202
Reply #242 - 10/22/21 at 17:55:06
 

https://liliputing.com/2021/10/sifive-says-its-next-risc-v-processor-outperfo...

When RISC-V chip designer SiFive introduced its Performance P550 processor in June, the company said it was the highest-performance processor based on RISC-V architecture to date.

Now the company is previewing a new processor that SiFive says is 50% faster than their previous efforts, allowing the new chip to theoretically outperform an ARM Cortex-A78 processor.



SiFive’s announcement is a bit light on details at the moment, the company hasn’t even given the new chip a name yet. But the company says more details will be revealed at the RISC-V Summit in December.

So far we know that SiFive’s next chip:

* Supports up to 16 CPU cores
* Runs at frequencies up to 3.5 GHz
* Features 16MB of L3 cache, up to 2MB L2, and
* SiFive says the new chips are 64-bit processors with quad-issue out of order processing, which can be scaled to support a wide range of devices including PCs, servers, or mobile and embedded devices.

SiFive says the new chips are 64-bit processors with quad-issue out of order processing, which can be scaled to support a wide range of devices including PCs, servers, or mobile and embedded devices.

High-performance versions would likely combine a cluster of chips together, offering up to 128 CPU cores.

This chip and some other planned new ones was why Intel was so hot to buy up SiFive a few months back.   SiFive has backed away from the deal totally now and no longer plans to use Intel as their exclusive one and only foundry as Intel's attitude and recent actions had left a bad taste in their mouth.

Intel has a whole lot to learn about being a fab and about remaining "properly neutral" about it ........



===================================================    one week later



Intel has posted their stock holder guidance for the next quarter and Intel IMMEDIATELY lost 10% of their stock value in ONE SINGLE DAY as Wall Street downgraded Intel according to their own guidance information.

Why?   Reading between the lines on the attached fiscal information it was clear that Intel didn't have the ready funds needed to execute all the various things they had already committed to do early this upcoming year.

It also became clear why Intel chose to drop their SiFive buyout bid and their Global Foundry purchase bid so abruptly.   All of Intel's pin money has to go towards the equipment to put in the two new fabs in Arizona.

Sad

Intel isn't broke yet by any means, but Intel lacks the sufficient very very large positive cash flow "free money" they used to have in years past.

Intel is also getting downgraded by the futures traders because relative to their current COMPETITORS  (and they have a lot more competitors than just AMD)  the performance of what Intel builds now at 14nm is beginning to smell more like sardines left out on an aluminum boat seat in the hot sunshine.  

Yep, it stinks like very very dead bait fish ........

Intel has no winners out there at the moment.    Intel has no credible plan to have a up and coming winner in the short term either.

Intel has no plans for any long term planned winners laid out in any credible fashion going out for next year or so (apart for getting TSMC to build some selected 3nm stuff for them, actually).    

TSMC building some stuff for Intel at TSMC 3nm is Intel's next best window to achieve anything close to a "winner" product.

Intel will have to continue to lower their 14nm general processor prices even more to be able to move their "non-preferred" 14nm product line going out into the foreseeable future.

So, Intel is facing a short term "crisis of confidence" both on Wall Street and in the actual market place.   This will all get worse over time until Intel starts building something at a modern competitive lithography level and comes up with some winning new designs.

Undecided

(poor Intel, so sad)



Back to top
 
« Last Edit: 11/03/21 at 03:01:48 by Oldfeller--FSO »  

Former Savage Owner
  IP Logged
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12629
Fayetteville, NC
Gender: male
Re: AMD & others --- post AMD dominance in 202
Reply #243 - 10/25/21 at 05:13:15
 

https://liliputing.com/2021/10/intel-launches-12th-gen-alder-lake-chips-for-d...

Intel has begun their next brown vapor campaign, this one is based on using their Big, followed by their "middle" and then the smaller cores set ups.   Yep, up to 16 cores per unit being tested is what you are seeing the results for.



Intel has a software that puts the loads on the currently coolest Big processor core, switching them out as they heat up.   This is a neat trick, and when it finally works right it should help out with better real world throughput results .......

Please realize any intentionally leaked results you see now a days from Intel are based on the best core being measured for the single core results and then being run out to the same size core count of the entire chipset to make up the many proposed many core results.   This is not how it actually works in the real world ........   Intel marketing is systemically lying and cheating again.

First, only one core out of the entire set actually does work that good, not 4, not 6, not 8 and certainly not 16.   The little cores are not able to do the work of the middle sized cores much less the amount of work that can be done by the big cores.  

We are now seeing that the smaller cores are actually all the same thing, simply ranked and split up by the Intel processor management software, they were originally intended to all be middle cores, but some didn't make the grade and the Intel processor management software keeps them sorted.

Forget about running all of this at once in a mobile product.   300 watt combined power draws running the cores all at once is NOT A REALISTIC SCENARIO on any mobile product.

The new LGA1700 socket to hold this 10nm monstrosity is simply HUGE for a mobile processor equipped product.   It is bigger than the old Intel desktop socket used to be.

Intel sees this as a plus for cooling, since their new processor management software shuttle bumps the processing load around all over the large chipset to keep things as cool as possible.

For both the new Intel and the new AMD products, look to your cooling methods as all of these new chipsets are cooling thermal limit bound at the top end of their performance range, even if they could theoretically churn out these huge potebtuak outputs with no regard to their internal thermal limits.

I would say we are getting into the realm where inside the chip liquid cooling is going to be needed by both Intel and AMD, so look for results which are made using real processors in real laptops using a fully identified stock cooling system (whatever it is).

(use of freon based cooling systems by some test houses has been both suspected and has been occasionally verified by catching them red handed, yes, Intel we are talking about you at the big sales show a few years ago.   Intel had a freon rig the size of a small dorm refrigerator tucked away under their display table and their demo shown on their display screen was actually being run on a dual core mainframe motherboard with a dual set of Freon chilled mainframe processors ...... )

Tongue

Next, place no confidence in the results of any small custom test houses.   One reviewer called them "hired liars" and he wasn't far wrong in that estimation.

Lastly, you don't really need the most gee whiz best thing out there ---- Linux actually runs faster on lesser processor sets, as does older versions of Windows.



Back to top
 
« Last Edit: 11/05/21 at 12:35:12 by Oldfeller--FSO »  

Former Savage Owner
  IP Logged
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12629
Fayetteville, NC
Gender: male
Re: AMD & others --- post AMD dominance in 202
Reply #244 - 10/29/21 at 07:30:54
 
       
Folks, due to a hard hitting Covid booster shot my wife and I are ailing a bit.   Not dying, but not feeling well.  

Expect fewer comments from me for a while .......



AMD just posted some very good market share and financial return numbers.
Intel did not, and Intel has lost 11-12% of their total stock value per share recently as a result.

Pundits are reviewing the Intel presentation of Alder Lake and the public is asking questions like "Is it safe to swim in Alder Lake at this point?"

So far their overall answer is " NO '" as Alder Lake is actually pretty green and scummy.    Seems appropriate that the lake and the processor both share a similar review, huh?

Intel is suffering strongly from their own Marketing BS and AMD is busy still taking more Intel market share as TSMC has found quite a few extra wafers to supply AMD's market share takeover efforts.

Intel also released fiscal information that leads some to think they are getting cash strapped by the two Arizona plants they are building ----- some are even saying internal Intel documents refer to intentionally pausing building up the new lines inside the buildings until Intel demand picks up enough to keep the new lines 100% busy.

Joe Biden had promised Intel a lot of USA gov money that is not readily apparent at this time.   Instead the USA gov is asking Intel for contractual default money payments off the Frontier supercomputer project, the one from two years ago that Intel totally dropped the ball on.

Intel's plans to build a German or Italian plant are now seen as quite dubious accordingly .......
All of these plans are contingent on the country in question providing massive tax breaks and contributions.  TSMC, Samsung and Intel all require lots of up front money from the state were the stuff is going to be built in.

They aren't getting it right now from USA gov as the Democrats have spent it all elsewhere on stupid stuff.   Expect all these Fab guys to hit pause until the funds flow again .....


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


BIG PICTURE TIME

Intel has created their own crisis recently by renaming everything, suddenly saying that their old 10nm is now 7nm and 4nm and by "creating" lithography modes that have dubious benefits.   This creates distrust and confusion with chipset buyers.

Wall Street is crushing Intel for their flaws, taking away stock value by the week.   Meanwhile, AMD and TSMC are making real progress by carefully following their published road maps.   Intel is having to use lots of brown vapor BS to make up for their lack of progress over the last 3 years,  creating new roadmaps to disguise the lack of completing the old ones.

Intel is likely going to bankrupt these split off company divisions, ie. have to stop building on the internals of their new plants due to being late to the party and NOT HAVING ANY ORDERS TO RUN IN THOSE PLANTS TO KEEP THE PLACES BUSY ENOUGH TO PAY OFF THE LOANS.

Both TSMC and Samsung are deeply committed to building out what they have started, but you can understand their reluctance to keep putting their money into anything if the USA gov money (local and federal funds) do not flow as promised.

Next opportunity for Intel to create some form of "real winner" comes in 1 year from now when Intel has TSMC to build a line of processors for them at one of TSMC's 3mn advanced nodes  (which Intel will call something else that is complete BS of course).

Cooling is now critical for everybody as all state of the art processors are growing in power and complexity.   300-600 watts to drive a chipset means a significant chipset cooling system is needed as well.

I am losing all interest as I cannot see me ever buying any of this state of the art stuff as the cost of it has gone all vertical on me and my computer discretionary funds are getting slimmer and slimmer.
 
Actually, my real computer needs are getting less and less as well .......

If you are buying used, buy a cheap used Dell workstation machine as they are all well understood and they have all the software bits and pieces available in a stock Linux OS system.

If you are buying Windows OS machines from Wintel, good luck and Godspeed to you, brave pilgrim, as the Wintel boys are forcing you to buy a very expensive brand new machine to meet their new security requirements ......

Both Intel and AMD are going to chiplet based Big Little designs with larger core counts than you are used to seeing.  There will be some AI boosters stuck in the heterogeneous core mix and on-board graphics good enough for a worker bee's needs will be built into all of them.

All of these will be quite expensive and far far more powerful than anything I need.


Roll Eyes  


The world economy is being crippled right now due to a lack of shipping throughput when the container ships hit ports and cannot be unloaded in a timely fashion.

Covid is being blamed for all of this but Covid was not all of it as other root causes and a general recession are playing out as well.   Governments will eventually lack the money printing capacity due to dilution and will not be able to support all of this with more and more large gov checks to people who don't want to work.

Back to top
 
« Last Edit: 11/02/21 at 16:22:07 by Oldfeller--FSO »  

Former Savage Owner
  IP Logged
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12629
Fayetteville, NC
Gender: male
Re: AMD & others --- post AMD dominance in 202
Reply #245 - 11/02/21 at 12:24:37
 
 
Welcome to the Angstrom Era
PS: Moore’s Law is Still Dead
by Steven Leibson



Intel took a lot of flak when it announced new names for its future semiconductor process nodes during the Intel Accelerated event in Late July. The new nodes are called Intel 7, 4, 3, and 20A. Industry pundits have knocked the company for calling its 10nm enhanced SuperFin process node “Intel 7.” (Intel announced and demonstrated 12th Generation i5, i7, and i9 Core processors built using the Intel 7 process node just last week at its Intel InnovatiON developer event, so this process technology is clearly well in hand.) The company now uses the name “Intel 4” for the node previously called 7nm. Intel 3 and Intel 20A are all-new node names. Part of this node renaming has roots in marketing, and part is simply facing reality.

In the past, Intel had said that its 10nm enhanced SuperFin node (now renamed Intel 7) was roughly equivalent in power and performance to TSMC’s 7nm node. Confusing, no? The company has also stated that industry analysts have asked Intel to update its process node nomenclature to reflect the company’s true competitive position in the semiconductor process arena. All of that is the marketing part of the renamed nodes.

Years ago, process node naming was based on a transistor’s smallest feature size. That naming convention was specifically developed for a planar MOS transistor, and the smallest transistor feature was invariably the transistor’s gate length. That naming convention went out the window when FinFETs took over from planar MOS transistors, and all vendors’ process node names became nothing more than estimated equivalents for the power and speed you’d get from an equivalent planar MOS transistor. Except that they couldn’t actually make those planar MOS transistors any more. Planar transistor technology gave out. MOS transistors just don’t work well at current lithographic levels. That’s why we turned to FinFETs. Intel introduced its first FinFET process at the 22nm node – way back in 2011.

RibbonFETs and angstroms

Ten years later, the fundamental structure of transistors is about to change yet again.

FinFET gates are driven from three sides. That’s two more sides than are driven in planar MOS transistors, and which results in much better transistor performance at the expense of more complex manufacturing techniques. However, driving three sides of the FinFET transistor’s gate no longer achieves the desired speeds and leakage currents. We must now drive all four sides of the gate. These 4-sided gate-drive structures are universally called “gate all around” or GAA.

Universally, except for Intel, which calls its GAA transistors “RibbonFETs,” which are currently scheduled to appear in the Intel 20A process that’s due to appear in the first half of 2024 (barring delays). Notice the “A” in the process name. “A” is for “angstrom.” Instead of calling this process node “Intel 2” as with the previous process nodes, Intel switched units from nanometers to angstroms. An angstrom is 10-10 m, or one tenth of a nanometer. I think we’re supposed to simply disregard the fact that Intel dropped the “nm” from the bigger process nodes.

Presumably, this nomenclature change to angstroms allows for fractional nanometer node naming. That’s similar to the situation we had with the previously used unit, microns. Back in the dark ages, we had 3, 2.5, 2, 1.5, 1.3, 1.2, and then 1 micron process steps, followed by 0.8, 0.75, 0.7, 0.5, and all the way down to around 0.25 microns. Somewhere around that point, we jumped from microns to nanometers. Based on my best recollection, that nomenclature change occurred at 0.18 microns, which was popularly called 180 nm. When was that? Around 1998. More than two decades ago.

That means the nanometer process node ruler will have been in use for almost a quarter of a century when the angstrom era starts. The Intel angstrom nomenclature allows the company to have process nodes named Intel 18A, Intel 17A, Intel 16A, and so on. That sounds so much better than Intel 1.8, Intel 1.7, Intel 1.6, and so on, doesn’t it? Sounds like they’re making more progress this way, right? However, the progress from one node to the next isn’t nearly as big as it once was. From the image above, you can see performance/watt improvements of between 10 and 20 percent from one Intel process node to the next, and the image doesn’t even discuss density improvements.

Moore’s Law lies a moldering in the grave

And that gets us to the meat of the matter for this article. Based on all that’s written above, I have very sad news for you. Moore’s Law is dead, despite Intel CEO Pat Gelsinger’s bold and enthusiastic promise to meet or beat Moore’s Law every year for the next 10 years, made last week during the Intel InnovatiON event. Scaling a process node by shaving a few angstroms from one generation to the next does not achieve a doubling of transistor density, and that is the true essence of Moore’s Law.

Moore’s Law did not refer to transistor power or speed. Moore’s Law says that the number of transistors on a chip doubles approximately every two years. That’s all it says. You can look it up if you like. Gordon Moore’s original article in Electronics magazine from April 19, 1965 was titled “Cramming more components onto integrated circuits.” It’s conveniently posted on Intel’s Web site for you to read. This article appeared three years before Moore and Bob Noyce founded Intel, while Moore and Noyce were both still working for Fairchild Semiconductor.

Moore performed an unbelievable feat of prognostication in his groundbreaking article. The semiconductor industry eventually embraced and transformed his prediction into a self-fulfilling prophecy, which was based on a very few data points. The first data point was a single transistor. The second data point was one of the first commercial integrated circuits, a 3-input NOR gate called a Fairchild μLogic Type G RTL chip.

According to Dr. David Patterson, Moore’s Law lasted through 2015 when it ran out of gas. (See “Fifty (or Sixty) Years of Processor Development…for This?”) How could Patterson say that Moore’s Law died in 2015 when Intel’s executives would uniformly have you believe that Moore’s Law is alive and well today? It’s because of a phrase on page 2 of Moore’s 1965 paper. That phrase is: “…the production of larger and larger circuit functions on a single semiconductor substrate.” (Emphasis is mine.) Moore’s Law is about monolithic integrated circuits, and that’s just not where the semiconductor industry is heading.

In fact, on the 50th anniversary of the International Solid State Circuits Conference, in 2003, during a presentation titled “No Exponential is Forever,” Gordon Moore himself stated categorically that “…no physical quantity can continue to change exponentially forever.” Moore’s Law was already dying nearly two decades ago, and Moore saw it himself.

My current poster child for signaling the end of Moore’s Law is Intel’s own Ponte Vecchio GPU. Intel is assembling this integrated device using 47 active “tiles” – Intel’s name for the chiplets or die in a multi-chip package – manufactured by multiple semiconductor vendors from five different semiconductor process nodes, all combined in one package using 2.5D and 3D assembly techniques to produce an integrated product with more than 100 billion transistors. (Please be sure to read that last sentence with a Carl Sagan accent and emphasis on “billion.”)

Some people have claimed that Moore foresaw multi-chip packaging in his article. They cite this phrase:

“It may prove to be more economical to build large systems out of smaller functions…”

But they seem to omit the second half of this sentence:

“…which are separately packaged and interconnected.”

Here, Moore was clearly discussing the use of multiple individually packaged chips on one board, a staple of board-level design since integrated circuits first appeared in the 1960s. From my perspective, Moore was clearly not predicting today’s multichip packaging with this sentence. In fact, his article discusses the possibility that monolithic ICs with 65,000 components per IC could appear in 10 years, by 1975, which was far more discrete components than any single printed circuit board could accommodate back in 1965 when Moore’s article appeared. Who could possibly need more than 65,000 components? If Moore thought about it at all back then, and he probably did, he must have seen that multichip packaging technology would not be needed until far, far in the future. Well, that future has arrived.

Intel’s Ponte Vecchio GPU incorporates 47 tiles from five different process nodes to cram 100 billion transistors into one package.

Multichip packaging only makes sense only because different process nodes deliver different cost/performance/capabilities tradeoffs, because we’re at the reticle limits of current chip-making equipment, and because 2.5D and 3D packaging techniques are now sufficiently practical and economical to make this approach work commercially. Why shouldn’t blocks and subsystems be made from the most efficient semiconductor process node possible, assuming you have the manufacturing processes needed to assemble all of these tiles or chiplets reliably and economically? Ponte Vecchio is admittedly an engineering marvel, but it’s most definitely not a monolithic chip, and so it’s not an example of the original Moore’s Law in action.

Except for the massively mythological underpinnings of Moore’s Law, it’s really not important to most of us how Intel crams 100 billion transistors into the Ponte Vecchio package. It’s not important to systems engineers using Ponte Vecchio GPUs in their designs. It’s not important to people using graphics software or computer games that run on Ponte Vecchio GPUs. The device’s performance, power, and price (the three fundamental “P”s of all engineering design) are what’s important to those of us who live outside of the package.

Intel plans to build Sapphire Rapids, the code name for the company’s next-generation Xeon processor, in a similar manner. Intel will manufacture Sapphire Rapids using four CPU tiles and 2.5D assembly based on the company’s EMIB (Embedded Multi-die Interconnect Bridge) technology. Versions of the Sapphire Rapids processor will also incorporate multiple HBM2 (High Bandwidth Memory 2) DRAM stacks in the same package.

Multi-chip packaging is not unique to Intel. Far from it.

AMD, Nvidia, and Xilinx all manufacture integrated devices that look like monolithic integrated circuits on the outside but are multichip devices – interconnected collections of tiles or chiplets – on the inside. (Actually, these fabless semiconductor vendors all have their devices manufactured and packaged for them by 3rd-party foundries – often that’s TSMC.) This has been the situation for more than a decade.

For example, Xilinx introduced the Virtex-7 2000T FPGA in 2011. It’s based on a multi-chip package that puts four FPGA dies on top of one silicon interposer. TSMC fabricates that device for Xilinx. Around the same time, multi-chip packaging allowed Xilinx to incorporate 28 Gbps transceivers into its Virtex-7 580T FPGA before it was possible to build these transceivers directly into the CMOS FPGA die. Xilinx has expanded its use of multi-chip packaging with each new FPGA generation. That’s a good indication that multi-chip packaging works well, at least for the high end of the IC market.

So, welcome to the angstrom era. It’s a “Beyond Moore” era.
Back to top
 
« Last Edit: 11/02/21 at 15:18:19 by Oldfeller--FSO »  

Former Savage Owner
  IP Logged
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12629
Fayetteville, NC
Gender: male
Re: AMD & others --- post AMD dominance in 202
Reply #246 - 11/04/21 at 19:46:02
 

https://liliputing.com/2021/11/alder-lake-review-roundup-intel-retakes-the-cr...

This review roll up contains all 17 benchmarks by the best in the industry, so click on it and read up if you are interested.



Alder Lake had a messy introduction.   It was all mashed in with the more than slightly screwed up Windows 11 roll out ........  and was also involved in AMD not getting the correct drivers issued by MS with the initial Windows 11 release.

These have been corrected by MS in Win 11 nightly updates and retests have resulted in the following roll up of results which is a best a very slight win for Intel contingent on the testers using ONLY THE VERY BEST AND MOST MODERN LIQUID COOLING SYSTEMS.



Alder Lake indeed sucks in massive amounts of power and makes for massive levels of heat at new levels that had never been seen before .......  this resulted in the COOLING SOLUTIONS used on the first released Intel Alder Lake products that simply were not adequate to allow the first samples to even equal the performance of existing AMD processors using stock AMD fin and fan coolers.

BUT (and it is a big but)  if  you give Intel a huge power supply and an even bigger liquid based cooling system Intel does indeed eek out an slight edge over win over AMD --- temporarily --- as Intel runs through their limited thermal headroom and BIOS slowdown then occurs.

Give both companies the same level of cooling,  AMD always wins.

AMD has two new generations of processors already sampled that will outshine Alder Lake by double digits and then later on next year by triple digits (assuming both get the same level of cooling).  

Going on out into the future ......

TSMC and AMD have cooperated on new packaging advances for future generations that actually have inside the chipset liquid cooling, a trick that will open up processor performance to levels never even dreamed of before.

Intel will be there, too, nicking at whatever lead AMD achieves with products that do almost as well.



Tit for tat,   AMD and Intel will crab walk on up the performance stairway to heaven while engaging in an endless back and forth sword fight ........



FOR THOSE WHO PREFER TO WATCH RATHER THAN READ, THIS IS A GOOD VIDEO FOR YOU TO WATCH THAT ROLLS ALL OF THIS UP FAIRLY TIGHTLY.

https://www.youtube.com/watch?v=dE9N95uSqHA


And yes, he is talking about 36 CORES and 72 threads inside an AMD mobile processor which is potentially possible using those new TSMC lithographies and packaging techniques.  When you drop down to 4nm and less, there is a whole lot of room on that AM5 socket set to put lots of chiplets (at 8  processors per chiplet no less).




Back to top
 
« Last Edit: 11/06/21 at 09:59:26 by Oldfeller--FSO »  

Former Savage Owner
  IP Logged
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12629
Fayetteville, NC
Gender: male
Re: AMD & others --- post AMD dominance in 202
Reply #247 - 11/07/21 at 22:33:06
 

We now have 3 new Intel Alder Lake processors announced.   A pattern is showing clearly now, Intel intends to beat selected AMD processors and furthermore it intends to sell these AMD beaters at 25% less than AMD's current pricing.

Intel can do this.    For a while, anyway.

Intel can spec out a chipset that is physically twice as large, pulls up to twice the power draw (but not for long as it overheats very badly) and even then it only marginally outperforms the AMD unit until the system BIOS shuts the processor down due to overheating.  

Cooling is King now for both sides ......

Intel will be happy with a win right now, no matter how marginal or energy expensive or short term in duration it might be.

It is clear Intel intends to refine these competition points, expand on them and try to take some market share back from AMD no matter what it costs Intel to do this in the short term.

Intel knows it has to do this now, quickly, before the next generations of AMD lithography and structure progress roll out to smash them flat again ......

Wall Street is simply killing Intel's stock value right now because Intel is not making any progress so Intel will now try to brute force some progress at any amazingly high cost levels, then Intel will have to sell the resulting new processors for less money than the AMD competition for long enough to cement their win in the eyes of Wall Street.

Intel is faced with multiple "house of cards" fall down potentials right now and Intel is doing whatever they can to keep all the cards upright and still standing .......

Back to top
 
« Last Edit: 11/09/21 at 17:09:51 by Oldfeller--FSO »  

Former Savage Owner
  IP Logged
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12629
Fayetteville, NC
Gender: male
Re: AMD & others --- post AMD dominance in 2020
Reply #248 - 11/08/21 at 09:58:24
 

https://youtu.be/ECHhuvuiNzs

Lisa Su presents the next 2 years of AMD progress, with next year being only 4 weeks from now.

Now you see why Intel is forcing their hand so early and abruptly, just to get it counted before it gets overcome by AMD events.
Back to top
 
 

Former Savage Owner
  IP Logged
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12629
Fayetteville, NC
Gender: male
Re: AMD & others --- post AMD dominance in 202
Reply #249 - 11/09/21 at 03:18:57
 
grey bar means discontinued



Tom's Hardware is says  "10nm" Alder Lake is being treated as a serious Intel technical advancement by Tom's analysis team.

It can slightly out-game the current AMD units by using a selection of the fastest of all the Alder Lake big cores at once and overclocking the hell out of them until they overheat.

It can "out compute" AMD by using a mix of Intel efficiency and performance cores and overclocking the hell out of them until they overheat.

The Intel processor management software works, but needs some updating to fix a few minor bugs that showed up after launch.   Repeat, the Intel processor management software is working fairly OK now and the nightly updates will fix the current minor bugs.

Alder Lake is designed as non-homogeneous and the Intel Processor Monitoring software is key to using the complex non-homogeneous processor.   As a core gets tired and the performance drops, it becomes a next level down core automatically.

This software management means there are no substandard Alder Lake chipsets, as there is room at the bottom for the downgraded cores to be useful.

As each processor cores get overly warm, they are deselected by the processor management software and the load is moved over to lesser performing cores.  The huge chipset has enough backup cores that are needed to do this trick ongoing, permitting overheating to dissipate while maintaining overall performance.

The only downside Tom's sees is Intel's very large power consumption and the BIG BIG HEAT LEVELS which inexorably leads to across the board BIOS throttling as the chipset cooling cannot keep up with the heat flux.

Tom's recommends spending some extra Intel system money on a first rate liquid cooler system.

Intel will be rated #1 on Tom's processor listing until the next AMD generation ships in a few weeks.

This head to head bash up takes place at only 3 processors, with Intel lowering their price to 25% less than AMD while edging past AMD's performance, with AMD moving their selling price down to match the Intel price point.  Lots of gamesmanship taking place here, with not a whole lot of across the board clear wins that will last more than a few weeks.

Warehouse supply of these contested chips has instantly evaporated causing some pundits to accuse both of the competitors of doing a paper launch and a paper cost reduction with no inventory to back either up.

My central point is both processor makers are ACTUALLY REACTING AND COMPETING AGAINST EACH OTHER and that is a good thing as they will cause each other to make real improvements at a greater speed .......

Intel's speed improvements and throughput improvements (at a stiff cost in power consumption and high processor heat, requiring a new motherboard loaded with state of the art very expensive DDR5 new memory and a needed water cooler) will remain, and Intel's package will be here all of next year to keep AMD honest.

AMD offers the same level of performance using your existing memory and motherboard socket for just one more upgrade, then next fall AMD will go to a new socket and a new motherboard as they drop down to 5nm.

Chiplet counts from both players are DOUBLING late next year, with better architectures and better throughput from each.   Intel will triple their power draw from what they are selling right now and will require expensive liquid cooling on all desktop chipsets to keep heavy throttling at bay.   Intel sees no issues with cooking things if the cooling system hits a hiccup, this has actually been the Intel way for a while now.  

So, if  you don't have the recommended cooling solutions, it's on you, baby.

Apple will double their processor counts also, while dropping down to 4nm and 3nm lithography levels.   Rumored core counts of 32 and 40 cores are being floated by folks who claim to have seen them in TSMC factory packaged Apple chipsets.




Back to top
 
« Last Edit: 11/17/21 at 21:30:34 by Oldfeller--FSO »  

Former Savage Owner
  IP Logged
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12629
Fayetteville, NC
Gender: male
Re: AMD & others --- post AMD dominance in 202
Reply #250 - 11/16/21 at 19:15:35
 

https://liliputing.com/2021/11/lilbits-windows-10-21h2-is-here-and-windows-11...

Apple, MS, Intel and AMD are squaring up to take some bigger risks.   The Blue Screen Of Death will reappear more and more on Wintel machines because MS is using all of installed PC's as their test group.

MS knows there will be issues as what they are putting out "cannot be thoroughly tested" items against all PC constructions all at the same time.   It will go out in a nightly update, fail, report back home to MS what went wrong and MS will trickle back out some fixes that will likely make some other issues happen.  

Can you say "cluster fook"?   Just do a rewind of 3 years ago to see what it will be like to live through this again ......

Linux isn't happy because Intel and MS are pushing out some of this strange crap to fail in their Linux world, too.   Mint is acting plumb defensively at the moment, holding back MS and Intel "advancements" until they become stable in the Windows 11 Linux Implementations.

Linux Mint values stability, so it is staying back an entire release generation from all this whacked out stuff.   Mint will implement the Win 11 improvements very selectively and very slowly after a separate run of Mint testing.  

Most of what Wintel is doing is trying to fix the current issues with Intel's latest processors and it involves 10-24% processor speed delays which frankly don't do much of anything for the older Linux crowd of hand-me-down machines and really aren't needed by them at all.

It looks like all of Intel's small performance gains are now being absorbed by the slowdowns put in by MS as the required Intel security meditations.

It is a whole lot of stuff to do just to prompt AMD to put in the next stair step to heaven.


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


Intel is moving off their dime finally.    Intel has placed orders and taken shipment of some ASML 3-5nm Scanners.    Intel is making waves and shooting for some significant processing advancements.   Intel is back, and that is a good thing for the industry in general.

Apple will soon be Intel's competitive focus because Intel has arguably equaled AMD at this particular point in time and both AMD and Intel will nudge ahead of the other in small baby steps going forward.  

But it is Apple that can break out into the commanding position with their 3nm 32 and 40 core ARM based chipsets currently in testing.   Apple gets all of TSMC's full support (they pay handsomely for this service, BTW) so expect to see  TSMC 2nm & 3nm trotted out in Apple products for the first year starting this upcoming year.

AMD is making big moves in mainframe chipset construction right now and AMD is putting the Xylinx accelerator stuff in there as well.  

AMD is now considering Samsung 3nm as a possible interim step as TSMC has some serious 3nm yield issues that are being very troublesome at the moment.  

If AMD is going to compete in two years, they need a large volume of 3nm and 2nm chiplet wafers.


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


Confirmation comes on AMD moving several products over to Samsung's 3nm process as TSMC is not having "readily expandable success with 3nm" off their current lines.   Plus, all existing 3nm capacity belongs to Apple for the next several years.   Apple's needs come first, and Apple isn't getting all they want right now either.

AMD going over to Samsung leaves Intel pretty much out in the cold for next year since AMD had priority over Intel going into this planning period.   AMD is not giving up any TSMC allocations, they simply are tacking on Samsung as chiplet volume expansion as it is covered by their TSMC agreements as the documented "first alternate supplier".

Samsung 3nm is a fairly firm and reliable source for a one-two year planning horizon.   Count on TSMC to fix their order issues and to come back to AMD with a good proposal when they have the free capacity to build all the AMD chiplets that AMD needs.

AMD is currently taking allocations for all the 5nm and 4nm AMD chiplets that TSMC can make, beyond the ever growing totals of what Apple needs with their new M1 Pro‌ and ‌M1 Max‌ chipsets.  

It remains to be seen if AMD can mix and match chiplets from TSMC and Samsung on the same product socket board.   If AMD can do this, then their current pain and suffering will diminish a lot accordingly.

All manufacturers using chiplets rank them in 2-3 different performance groups, allowing for a core to degrade some and simply be utilized as the next step down (no scrap in this ranking system).


Back to top
 
« Last Edit: 11/29/21 at 04:32:09 by Oldfeller--FSO »  

Former Savage Owner
  IP Logged
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12629
Fayetteville, NC
Gender: male
Re: AMD & others --- post AMD dominance in 202
Reply #251 - 11/22/21 at 22:40:35
 

https://www.tomshardware.com/news/phytium-d2000-eight-core-for-desktops



This new company mini PC has an ARMv8 processor and AMD Radeon RX 550 graphics

People wonder why Intel keeps losing 1-2% market share every time anyone reports market share numbers while everyone else gains 5-8% during the same period.  How is this possible?

The answer is that massive world-wide compute industry expansion is going on in a steady fashion.

The world market share overall pie is growing constantly with new users popping up all over China and India and Indonesia.   AMD is growing a little better than the Industry  Growth Average while new ARM and RISC-V processors are providing the bulk of the newly built processors making up the background growth numbers in a thin spread all over the place.  

Intel is shrinking steadily (relative to the market as a whole) as the brand new market players eat Intel's lunch for them.   Look to see 8 cores at least double with the next downshift in lithography ........

Intel always only compares its own numbers to itself on purpose, so with a straight face Intel can saying they have market growth while Intel's significance and relevance continues to shrink along with Intel's real world market share numbers.   It is a constant on-going lie that Intel has to provide to reassure the Intel stockholders ........

ARM v9 is coming next year, and with it come the first smells of a tipping point for x86 desktop dominance.


Watch what Apple does with ARM processors to see the pathway of the future unfold before you.   32 and 40  ARM v9 and X1 cores are coming this spring from Apple ---- some significant processing power here that is overshadowing both Intel and AMD on the low end consumer side.

Mediatek is doing something similar with some very strong ARM chipsets.   AMD is providing their state of the art graphics to both Mediatek and to Samsung.   Samsung is known to be doing runs of 3nm graphics chiplet wafers for AMD in exchange as nobody can find any loose TSMC  wafer capacity for next year.



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



https://liliputing.com/2021/11/this-mini-pc-has-an-armv8-processor-and-amd-ra...

Wow, here is another one popping up ---- possibly the same main board in different cases.





Back to top
 
« Last Edit: 12/10/21 at 16:50:34 by Oldfeller--FSO »  

Former Savage Owner
  IP Logged
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12629
Fayetteville, NC
Gender: male
Re: AMD & others --- post AMD dominance in 202
Reply #252 - 11/27/21 at 06:28:29
 

Rumor Time

These are still rumors until they actually start shipping.

The ARM buyout by Nvidia is now very unlikely as 5 different regulating bodies have given a long list of stuff to Huang to answer and he is balking at providing any answers to any of them.

So, he gets "no" votes by Britain, Brussels, China and the USA until such time he gives up satisfactory answers to each regulator body.    Huang is now content to let the buyout clock run out and not spend any more time, effort and money on his aborted ARM Holdings takeover efforts.

ARM is currently rolling out newly designed 5nm processors with even newer 4nm design specs being provided to the world as tryout parts for the next 4nm generation and the following 3nm generations, with Mediatek acting as the prime "first implementer" of the newly implemented ARM 4nm standard designs right now.

ARM/Mediatek 4nm is now looking at a new doubling on the core counts with a fresh 30% performance upper along with a memory capacity doubling bumper due to use of the new TSMC "vertical stacking" and dense packaging techniques.   This ships starting first quarter next year.

Ditto again for a 3nm speed bump in 2022 ......

For example, 8" Fire Tablets might pick up 30% faster processing with DOUBLE the systems memory on all the base units within the next 2 years.   Amazon contracts with Mediatek far in advance and Amazon only upgrades once every 2 years.  

This upcoming Amazon Fire processor bump (or mebbe the next one) will possibly memory and speed lap my old desktop PCs ...... in a silly cheap tablet no less.

                                              Roll Eyes

Graphics processing from ARM is getting better too, but those who want really good graphics (Samsung for example) are doing deals with AMD to get AMD Radeon ray tracing graphics for their devices.



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



Their toes are being nibbled at by odd looking Chinese ducks .......

For $17, RISC-V is selling an equivalent to the original Raspberry Pi device.    But who would want a 4 years out of date copy of the original Pi?
Back to top
 
« Last Edit: 12/01/21 at 21:36:02 by Oldfeller--FSO »  

Former Savage Owner
  IP Logged
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12629
Fayetteville, NC
Gender: male
Re: AMD & others --- post AMD dominance in 202
Reply #253 - 12/03/21 at 01:57:31
 
 
Now for some rumor becoming cold facts.

https://liliputing.com/2021/12/sifive-performance-p650-risc-v-processor-revealed
-as-a-challenger-to-arm-cortex-a77.html

This says that RISC-V is now only one-two generations back compared to ARM.



https://www.ftc.gov/news-events/press-releases/2021/12/ftc-sues-block-40-bill...

This says that the Federal Trade Commission has sued NVIDIA to STOP all their log jamming bullshite,  stating that NVIDIA will NEVER be approved to purchase ARM.  

NVIDIA will have to pay their cancellation fees to ARM and IMMEDIATELY STOP stalling the development of all of ARM'S new processors for all of ARM's other customers.  

Any NVIDIA tech that got crammed into ARM's designs unwillingly by NVIDIA during this grey period will get a perpetual license from NVIDIA and NVIDIA will pay ARM full license fees for any tech items NVIDIA lifted from ARM.

ARM and the phone industry in general were damaged by NVIDIA, and the FTC will try to see that NVIDIA picks up the tab for that damage in as much as possible.

What the FTC can't fix is all the yardage that RISC-V made up while ARM was all log jammed up by NVIDIA.    Many companies moved away from ARM in reaction to Jensen "Benny" Huang and instead have firmly committed to RISC-V and that RISC-V commitment remains and all the new RISC-V processors that are under design by multiple companies will be completed and shipped next year.

ARM Holdings may never really recover from this convoluted mess in the current generations of processors ----- look to see ARM expedite some new generations of processor designs to pick their market share back up again.

NVIDIA responds saying there is no basis for the FTC suit --- take us to court if you want to.    NVIDIA specializes in legal quagmires, so FTC would be foolish to do so.

Some would say that blowing off the Federal Trade Commission is a very poor idea, as payback is a right biatch when it come for you.

Next, new regulations can become pointed weapons if you piss the government agencies off badly enough.
Back to top
 
« Last Edit: 12/19/21 at 12:41:49 by Oldfeller--FSO »  

Former Savage Owner
  IP Logged
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12629
Fayetteville, NC
Gender: male
Re: AMD & others --- post AMD dominance in 202
Reply #254 - 12/12/21 at 23:28:01
 

Intel vs AMD

AMD is year on year moving on down the lithography chain at TSMC.   AMD is hitting each announced step in their roadmap like clockwork.

Intel is also hitting their roadmap steps, but is having a far bumpier ride in doing so.

Intel is still sucking down twice as much power as AMD and is still requiring a brand new motherboard per each step of the way.

AMD is becoming preferred in the data center market as AMD's progress steps work across the board making large improvements in data center that Intel cannot match.

Intel now has their Intel 10nm working in volume production (they call it Intel 7, but it is really the old Intel 10nm super fin).

Next year everybody goes down a lithography step.   Intel still tries to claim parity with AMD using some relatively extreme power sucking tricks, but hey, the tricks are working and Intel is moving on up in the processor comparison charts at Tom's Hardware.

As of Alder Lake, "Intel is back" as a competitor in the desktop and laptop space.   However, AMD has improved its supply of chiplet wafers at TSMC by roughly 20%  and is taking more market share from Intel accordingly in all nodes that are not Intel Alder Lake based (2.4% against the total x86 market just this last month).

AMD is now using 3nm, 4nm, 5nm and 7nm TSMC wafer allocations to make this market share progress.   Apple and Intel are beginning to squeeze in on the AMD's chiplet allocations by their own increased demands, so any new allocations will likely go to whomever is willing to pay more for them.

Intel has hatched up an impromptu plan to buy up all of AMD's wafer allocations right out from under AMD and to put AMD out of business.   AMD has countered this trick by paying for their allocations years in advance by putting in fully pre-paid production orders.  

TSMC is beginning to insist on everyone doing these firmly scheduled and fully pre-paid allocation production slots from everybody.   TSMC likes the AMD plan for production scheduling, so everyone will do that starting now.

If you don't like it, Mediatek or AMD will buy up your now unscheduled production slots and that is the harsh new reality in today's fabless chip production.

We will see if the new USA plants being built in Arizona by TSMC and Samsung can alleviate the "production wafer war" which is currently building up as we speak.


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


I have said several times that TSMC is treating Intel like a competitor fab not as a customer.   TSMC verifies this "competitor fab" status by telling Intel's CEO on his last trip just this past week that Intel has to pay up front for 100% of the materials and all run charges up front whenever Intel schedules a run, not when Intel takes delivery ---- and that all scrap produced on an Intel design run belongs to Intel.  

Take it or leave it, these are the terms.   Intel is not  a happy camper right now ........     Intel had often messed up initial runs of new designs as they kept designing things for their own Intel processes instead of designing things from the get-go according to the TSMC ASML standards.

Obviously, TSMC does not really want to make chips for Intel having seen first hand how Intel repeatedly screws up and Intel then stiffs their suppliers and their business partners over the years.   TSMC is simply acting to protect themselves by having these Intel restrictions put in place up front.

AMD will buy up any slack that Intel puts into the supply chain, followed by Apple, Mediatek and several others.    Unused production allocations will get resold promptly and Intel goes to the back of the line to schedule some new TSMC runs again.

TSMC lists Intel all the way down their list of preferred customers, all the way down at position  #11 on their list of preferred customers.    TSMC has only committed (by getting paid their full price up front) to building some selected Intel graphics processors for Intel at this point in time.   These graphics processors were the ones developed by TSMC on TSMC's production processes, so they should work without causing big scrap charges to anybody's bottom line.

This may change, of course, as Intel starts buying and using only industry standard ASML process lines as their bread and butter production equipment.



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




AND YES, THE INTEL ALDER LAKE FAMILY REALLY IS KICKING AMD'S BUTT FOR BOTH PERFORMANCE AND PRICE AT THE MOMENT ......      
And yes, this is intentional and active price fixing on Intel's part as these new chipsets are larger and more expensive to make but are being sold for less.

Back to top
 
« Last Edit: 12/20/21 at 00:21:55 by Oldfeller--FSO »  

Former Savage Owner
  IP Logged
Pages: 1 ... 15 16 17 18 19 ... 21
Send Topic Print


« Home

 
« Home
SuzukiSavage.com
03/28/24 at 19:09:36



General CategoryThe Cafe › AMD & others --- Intel dominance in 2022


SuzukiSavage.com » Powered by YaBB 2.2!
YaBB © 2000-2007. All Rights Reserved.