RUMOR TIME

Intel foresees TSMC being sidelined by a Chinese military take over of Taiwan.  

When this happens, Intel will ask for and will quickly get 100 billion dollars from Biden to build an AMERICAN BASED MODERN chip making capability of some size.   This will be a military sort of program and it will move quickly.  ASML has already been told by their own Dutch government to expand their scanner building capability to support such a rapid construction of new fab sites, as the Dutch government also sees this potentially happening soon.  

NATO as a whole is concerned over chip supply ......

Folks theorize that the Ohio and the Arizona complexes will be expanded to their second stage size and that their original plans included this expansion scenario.   Stage 1 was for well known high yield processes, stage 2 will run towards current state of the art processes.

Buckle up boys, it is going to be a wild and crazy ride IF THIS HAPPENS ......

DEEP BREATH TIME ........


Intel is going to bet the farm on TSMC 3nm super fin very soon.   If it works out, Intel becomes undisputed Industry Leader again.

Intel's backup plan is to become its own TSMC here in America if China invades Taiwan.   This will be very very very messy if it happens.  

Eventually, when it all works itself out ...... once again Intel becomes undisputed Industry Leader again.

If Intel stubs their toe trying to build up Intel FAB as an industry leader and China does not invade, then AMD/TSMC will continue the competition to be the industry leader.  



Odd isn't it, that AMD only wins if Intel screws the pooch again big time which is exactly what happened in the last 10 years with Intel.

https://www.tomshardware.com/news/amd-cpus-see-less-than-10-performance-drop-...

Spectre-v2 isn't going anywhere soon. VUSec, a group of researchers from Vrije Universiteit Amsterdam, recently discovered Branch History Injection (BHI), a byproduct of Spectre-v2 that affects Intel and Arm processors. According to Phoronix's latest report, while Intel has taken a performance hit up to 35%, AMD's processors got off easy at less than a 10% hit.

Yep, we have had 4 days of frantic back and forth over Spectre version 2 issues.

Here is my nutshell on this mess.

THERE IS A SPECTRE VERSION 2 ATTACK possible on both Intel and AMD that simply steps around all the existing hardware mitigations developed in the last 2 years.   Taking it all back to ground zero on fixing this mess, it is a brand new mess that is even worse than before.

Two very recent Software Mitigations have been written by the Linux boys for the new Spectre v2 attacks.   Two main mitigation variations have been written, with only one being seen as effective enough to go forward into widespread distribution.

Intel is simply eat up with the stuff and is seeing potential throughput hits of 20-35% just to do a software based mitigation that Intel couldn't even be bothered to write.   Intel can't fix their own problems and is acting all denial paralyzed at the moment.

Intel is currently spending all their time in smear campaigns against AMD products instead of trying to fix their own stuff.

Intel is looking at 20-35% hits in throughput on common processor tasks.  

AMD is looking at 5-10% throughput hits because of AMD architecture differences.   AMD has options to potentially cut down their mitigation hits to less than 5% by using on chip hardware mitigations that is built into the newest AMD processors.

This Spectre stuff totally wipes away Intel's high watts draw "processing advantage" and puts Intel Alder Lake functionally firmly back in #2 position across the board.   Intel must build their own hardware mitigations into their new silicon to be competitive to AMD.

Intel has now dropped back into their Intel default lying mode as that is all they have right now ........ BS, bad data and bigger lies.

AMD is going to try to address this mess in hardware with their newest processors as AMD has a re-programmable on board security chip built into their new hardware that can actually do this job --- while Intel has NOTHING effective that they can do at the moment other than the software mitigations that have been written by somebody else.

All of this very bad situation will change and evolve very quickly.   I look to see Intel try to ignore it all since they cannot fix it without a huge throughput hit.    I look to see MS adopt the best of the Linux boys software mitigations and put it out internal to Win 11 very soon.   I look to see the processor ranking shift some when this happens as Intel cannot ignore the performance hit to their processors any longer.

 
===================================================    two days later


Is your old machine affected by:

1) the Spectre / Meltdown version 2 new illnesses and their software mitigations --- yes it is.

2) AMD is relatively less affected by the two illnesses, losing less than 10% functionally by existing software mitigations and less than 5% by using the new proposed hardware mitigations that AMD is pursuing.     (only new machines need apply for this hardware fix)

3) Intel is being hit much harder, 30-45% throughput declines reading through on some of the older processor based equipment.


Should machine through puts be ranked in the charts using existing best mitigations to give consumers the very best most practical ranking system?   Yes, because this Spectre/Meltdown v2 stuff is part of your current reality going forward.

Microsoft will blend this mitigation stuff into Win 11 and Win 12, and the Tom's Hardware processor speed re-rankings will flow from that.  

Intel will try to ignore the whole thing until MS forces mitigations down upon them.

Apple M2 Ultra chipsets are:

Performance superior

Cooling superior  (lower sized lithography spread over a 4x larger surface means measured temps are actually lower than AMD's temps anywhere you take the Ultra's readings)

Graphics superior

Still as vulnerable to Specre V2 and Meltdown V2 but with far fewer exploits that have been written against the Mac softwares


........ when compared to AMD or to Intel.


Structurally, the Apple M1 Max and M2 Ultra look very much alike EXCEPT the M2 Ultra is physically doubled.   Yes, it is two identical M1 MAXs laid out top to bottom on the same substrate.  

Think Threadripper big or Exynos big if you need something to compare M1 Ultra to.



The chipset is 4 to 6 times larger than a standard A4  socket AMD chipset, it can pull power like 4-6  of those A4 AMD chipsets and it  has the throughput of 4-5 AMD chipsets.

Apple has built a series of HUGE BEASTS for their largest processors.   Intel no longer has the biggest power draw nor the largest heat produced ---- Apple is now the king of Huge and Excessive.

Apple has a working design for a workstation and a mainframe style chipset now .......



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


Intel is a reactive company .......   Intel reacts to what its competitors do (generally by copying them).

Intel was all set to compete with AMD.  

Now Intel has Apple and NVIDIA to deal with.

Poor Intel.   More and bigger challenges.

Intel is now suddenly #4 instead of #2 with a much much much bigger gap in performance to try to overcome.




 

Intel has just released a new power supply spec that is somewhat revealing.

Intel has spec'd their newest class of power supplies has to provide a minimum of 1,000 watts of power.   1,500 watts is included in the spec range on the upper end.

Half of this will be likely needed for the Intel graphics, the rest is for the processor and cooling package.

Intel is signalling that they will want to play in the new "super huge" brackets right along with Apple and NVIDIA.

https://hothardware.com/news/amd-ryzen-zen-4-cpus-integrate-radeon-rdna-2-gpus

Let's talk some more about those AMD chiplets that each carry their own set of graphics cores and spread over 12-16 chiplets can aggregate together to make up a full gaming level graphics solution.





Leaked documents arising from the recent Gigabyte security breach is the gift that keeps on giving. Not that we condone such actions in the slightest, but now that there are interesting tidbits about unreleased products out in the wild, we can't bury our heads in the sand, either. And so with that said, the leaked documents seemingly confirm a previous rumor suggesting all of AMD's next-generation Zen 4 CPUs will have integrated graphics.

This is outlined in yet another published document from when hackers pilfered Gigabyte's servers, who then tried extort a ransom from the company. Those efforts were apparently unsuccessful, as the hackers proceeded to publish a trove of stolen data online, containing documents outlining various details about unreleased products from multiple companies, including AMD and Intel.

In this case, the folks at Chips and Cheese got their mitts on the data dump and posted a handful of diagrams and other information related to Zen 4. One of the slides details the upcoming AM5 socket that will replace AM4, and what features it will be compatible with, like DDR5 memory and so forth. This is also the chart that suggests all Zen 4 processors will have integrated graphics.





As you can see above, the chart highlights three types of "Family 19h" processors, which are Zen 4 chips. In each case, it points out on-chip graphics support. So in other words, integrated graphics will not be the exclusive domain of dedicated APUs as part of the Ryzen G-series, but available on all product lineups, to some extent.

We say "some extent" because the document also notes, "Some OPNs...may not support GFX." We read this to mean that on some processor SKUs, the integrated graphics will still be there, but will be disabled, just as Intel does with its F and KF processor models.

Assuming the information is accurate, this will almost definitely entail the pairing of RDNA 2 graphics with Zen 4. This is something that has been previously rumored, with a supposedly leaked roadmap showing AMD's next-gen CPUs pairing Zen 4 with "Navi2" (RDNA 2).

There is still plenty of time speculate. Just a few weeks ago, AMD CEO Dr. Lisa Su reiterated that Zen 4, which is being built on a 5-nanometer manufacturing process, is on track to launch in 2022. Same goes for AMD's next-generation RDNA 3 graphics products.

Zen 4 will also bring support for DDR5 memory, just as Intel is doing with its upcoming Alder Lake processors. However, the same set of leaked documents seemingly confirm that AMD will stick with PCI Express 4.0 on the consumer side, and only support PCI Express 5.0 on its next-gen EPYC processor. That really shouldn't matter a whole lot in terms of real-world usage (for consumers), but if that's the case, it does give Intel bragging rights.

It's also expected that Zen 4 will hold the line with 12-16 cores and 24-32 threads among the mainstream lineup. Combined with a rumored 20 percent lift in IPC performance, Zen 4 could be a huge upgrade over Zen 3.


 AMD Zen 4 processors built on TSMC 5nm engineering samples of the chiplet based distributed graphics core Ryzen 7 and the Ryzen 9 are known to be in the hands of board makers and systems builders at this time.

3nm Zen 5 will follow along later in a year or two with its planned core count increases to make the graphics aggregation tricks work even better.  12 -16 core equivalents (or threads if you prefer) is what it takes now  (Ryzen 7 at a bare minimum)  but this will likely change as AMD rolls down to 3nm lithography late next year.  Rumors say the core counts on all levels will roll up to at least the counts used by the next bump up as there is massive amounts of room to do this core count increase inside the same AM-5 socket because of the much smaller more dense 3nm lithography.

Both Intel and AMD are going to provide strong gaming graphics across the board going forward.   This cannot help but to hurt NVIDIA as it means as long as Intel and AMD stay current on their graphics standards and continue to provide some really good gaming graphics then NVIDIA will lose out on supplying the video cards to these machines.  

Remember, video cards are one of NVIDIA's main market segments and video cards are the cash cow that makes most of NVIDIA's money for them ......

https://www.tomshardware.com/news/amd-future-cpus-could-feature-direct-attach...



AMD has patented a processor featuring a machine learning (ML) accelerator that is stacked on top of its I/O die (IOD). The patent indicates that AMD may be planning to build special-purpose or datacenter system-on-chips (SoCs) with integrated FPGA or GPU-based machine learning accelerators.

Just like AMD can now add cache to its CPUs, it might add an FPGA or GPU layer on top of its processor I/O die. But, more importantly, the technology allows the company to add other types of accelerators on top of future CPU SoCs. As with any patented work, the patent doesn't guarantee that we'll see designs with the tech come to market. However, it gives us a view into what direction the company is moving with its R&D, and there is a chance we could see products based on this tech, or a close derivative, come to market.

Stacking AI/ML Accelerators on Top of Memory on top of a standard socket I/O die

AMD's patent titled 'Direct-connected machine learning accelerator' rather openly describes how AMD might add an ML-accelerator to its CPUs with an IOD using its stacking technologies. Apparently, AMD's technology allows it to add a field-programmable processing array (FPGA) or a compute GPU for machine learning workloads on top of an I/O die with a special accelerator port.

AMD describes several means of adding an accelerator: one involves an accelerator with its own local memory, another implies that such an accelerator uses memory that's connected to an IOD, while in the third scenario, an accelerator could possibly use system memory, and in this case, it does not even have to be stacked on top an IOD.



The rumors progressed from rumors about how AMD could do it, to patents saying AMD has exclusive rights to the AMD based technology to go do it, and now the post patent release rumors extend to proposed release times and the exact lines of AMD product to be affected by the new technology now that the patents have been granted.

Intel runs off secrecy and Intel BS while AMD runs off logical chains of conjoined technologies and solid partnerships with other companies like TSMC and XILINX ......





ONCE AGAIN, THE AMD-XILINX-TSMC PARTNERSHIP  WITH CONJOINED DISTRIBUTED CHIPLET BASED PROCESSING THAT IS BEING BUILT AND PACKAGED AT THE NEW TSMC FACILITIES USING PROCESSING STACKS THAT ARE TO INCLUDE THINGS LIKE TSMC IN-CHIP LIQUID COOLING.  THIS OFFERS SOME REALLY REALLY MASSIVE ADVANTAGES TO AMD,  GIVING AMD SOME WORLD CLASS PROCESSOR COOLING TECH THAT INTEL ALSO REALLY DESPERATELY NEEDS (but doesn't have) FOR THEIR OWN SUPER HIGH WATT DRAW INTEL ROOM HEATERS.