Donate!
Welcome, Guest. Please Login or Register :: View Members
Pages: 1 2 3 
Send Topic Print
Browsers and AI PCs (Read 601 times)
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12638
Fayetteville, NC
Gender: male
Re: Browsers and AI PCs
Reply #30 - 04/05/24 at 06:48:40
 

Google intends to charge $$$ for AI PC services.  

So, go mash that CoPilot key on your new keyboard and then say some sort of verbal gibberish as your data request and find out after the fact that Google has charged you a hundred dollars of processing time to try to figure out what the heck it was that you wanted .......

Roll Eyes
Back to top
 
« Last Edit: 04/06/24 at 15:02:20 by Oldfeller--FSO »  

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

Hobby is now
"concentrated
neuropany"

Posts: 12638
Fayetteville, NC
Gender: male
Re: Browsers and AI PCs
Reply #31 - 04/06/24 at 14:07:12
 

Intel has started losing money this year in a rather large, clear fashion.   If it wasn't for the latest Biden Bucks latest infusions Intel would really be a completely crippled wounded duck.  

(but it is not a duck, it is a monstrously large reptile named Chipzilla)

Companies that are betting on Intel Foundry Services for their future success are simply deluded right now --- because Intel now lacks the funds to rebuild itself in any large meaningful RAPID fashion.

Intel is failing to build new processors that are competitive with AMD's new processors.   Intel is failing to take market share back away from Apple AMD and the various ARM boys.  

Intel products are starting to suck more and more as the phone boys all roll into motion swinging new ARM chip sets that can pull off full low end PC functionality, competing especially well when using the AI PC stuff (which is being focused on more in ARM phone type chipsets) is being considered as a moderate to mid level performance booster.

Remember, Intel and all the rest are getting significantly stronger compared to their historical past.   The rub is that Intel is just not getting better as fast as the rest are getting better and as such it falls further and further behind the pack.

Intel is still strongly linked to MS Windows and the historical Wintel conglomeration is faltering as Microsoft hits "excess complexity issues" of their very own.

It is bad when MS chooses to use Linux internally for all their troops.  It is bad country wide when China and Germany dump MS for Linux so very very publicly and officially.   It is bad when MS puts Linux in their company upgrade slides as being the upgrade path for MS old Windows 10 versions .....

RISC is still  coming on very strong in the commodity electronic sub-processor chipset markets.  

Raspberry Pi is struggling to find some upgraded Broadcomm processors to use as Broadcomm no longer makes a line of stronger ARM processors as they did in years past.  

Broadcomm has been bought out and the new owner/management does not support hobbyists like they did in years past.   Eric Upton is out in the cold, so to speak.

Raspberry Pi is now looking at various pathways that involve other sources ...... including building their own processors of their own design.
Back to top
 
« Last Edit: 04/17/24 at 23:44:59 by Oldfeller--FSO »  

Former Savage Owner
  IP Logged
Eegore
Serious Thumper
*****
Offline

SuzukiSavage.com
Rocks!

Posts: 8011

Re: Browsers and AI PCs
Reply #32 - 04/12/24 at 06:14:29
 
So, go mash that CoPilot key on your new keyboard and then say some sort of verbal gibberish as your data request and find out after the fact that Google has charged you a hundred dollars of processing time to try to figure out what the heck it was that you wanted .......


 I think it is much more likely Google will clearly and repeatedly inform people, most likely by means of a Premium Account, instead of facing the tons of lawsuits for charging "after the fact", a blatant and easily provable violation of law.  If people don't read their ToS, and they choose to be charged without looking, that's on them.

 I think they should charge for AI services.  When someone uses an AI generative search through Google they are drawing power and usage from literally one of the most powerful chips ever created.  Maybe 12% of AI proceeds go towards future innovation, these things cost a ton to keep operating.  Spreading that cost to non-AI users is like charging property taxes on humans without homes.  

 I want to use a chip that literally uses the entirety of documented humanity, in seconds, that filters out all the p edo ph elia, sexual assaults, FB live animal torture etc. for me, yeah I will pay for it, and it will be very unlikely that I won't be made aware of this before it happens.
Back to top
 
 
  IP Logged
Oldfeller--FSO
Serious Thumper
ModSquad
*****
Offline

Hobby is now
"concentrated
neuropany"

Posts: 12638
Fayetteville, NC
Gender: male
Re: Browsers and AI PCs
Reply #33 - 04/12/24 at 14:40:11
 

https://www.tweaktown.com/news/97493/amd-ryzen-9-9950x-cpu-tested-zen-5-chip-...

AMD processors just got roughly a 26-46% improvement in processor throughput with the advent of Zen 5 high performance chiplets and (more specifically) the new Zen 5c power efficient chiplets.    Zen 5 is a 4nm product with improved fabric and throughputs but the star of this new show is a mixed lithography 3nm and 4nm Zen 5c chiplet that has better throughput than the old Zen 3 and 4 main cores used to have.

This was done for the larger many cored mainframe style processors, but as in all things AMD the bigger AMD improvements get played across the board when they show up anywhere.

So most new AMD processors will carry a whole bunch of Zen 5c processors to keep the efficiency numbers as high as possible,  but they also have a smaller crew of very high performance Zen 5 normal big chiplets will allow raw performance levels that means Intel can't even sniff the ground behind the new AMD products.

Wink

Remember, Zen 4 was where Intel lost their competitive position, now we have the new Zen 5 and Zen 6 stuff coming in that moves the goal posts roughly ~ 46% ~ further down the field in top end performance .......

Yes, this moves AMD that much further away from Intel's historically lower processing levels.

So. Please don't run out and buy any of Intel's current stuff, instead buy you some AMD Zen 5 and Zen 6 based products if you want to get you some state of the art performance .......


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


If you want the old traditional "Intel Inside" levels of performance, you can buy a much much cheaper level of AMD Zen 3 or Zen 4 processor from two plus years ago.   AMD had built up a bunch of these processors that are still "sitting in inventory as available" and they are currently for sale right now for much much less money than the current stuff.

Us old retired folks should mebbe consider this as a pathway to the future as we can buy 2 year old "as good as Intel's current best" AMD stuff for dirt cheap right now that will handle our needs for as long as we will likely be around to use it.


This assumes you are not going to chase the as yet undefined MS and Intel BS into "the AI PC world" that is currently churning up such huge fog banks of icky stinky brown vapor all over at the moment ......
Back to top
 
« Last Edit: 04/15/24 at 07:17:20 by Oldfeller--FSO »  

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

Hobby is now
"concentrated
neuropany"

Posts: 12638
Fayetteville, NC
Gender: male
Re: Browsers and AI PCs
Reply #34 - 04/16/24 at 09:50:05
 

https://liliputing.com/amd-launches-ryzen-pro-8040-mobile-and-ryzen-pro-8000-...

AMD is bringing its Ryzen PRO chips for laptop and desktop computers into the AI age. The company’s new Ryzen PRO 8040 Series processors are basically business-class versions of the Ryzen 8040 Mobile chips that launched last year, meaning they have the same Zen 4 CPU cores, RDNA 3 integrated graphics, and Ryzen AI NPUs. But they now get AMD’s business-class security and management features.

Meanwhile the new Ryzen PRO 8000 Series desktop chips are the first desktop processors to feature neural processing units for hardware-accelerated AI.

So what are you actually supposed to use those NPUs for? AMD highlights a couple of different applications including:

Summarizing meetings with AI-generated notes
Providing live captioning for video calls (or real-time language translation)
Managing device performance and maintenance with battery life and WiFi management, and predictions of impending equipment failures or performance degradation
AI-enhanced image and video editing and processing, or text-to-image generation without relying on cloud servers
Enhanced security features including threat detection and security issue diagnosis

AMD’s new mobile chips include Ryzen PRO versions of every existing 8040U and 8040HS processor from the Ryzen 5 PRO 8540U chip to the Ryzen 9 PRO 8945HS. All of these chips feature Ryzen AI NPUs with up to 16 TOPS of AI performance.


Intel announces new stuff 2-3 years out and it never never comes to pass exactly as promised.  
AMD announces stuff a month or two out and it comes to pass exactly as announced.

AMD always meets its chipset performance goals, while Intel always redefines what they say they are going to be making so they don't have to admit they had missed their original goals completely.

Intel simply sucks, in other words.

AMD is currently sitting at 16-24 TOPS (instead of the 40 TOPS that MS says it has to have to "meet AI PC requirements").

So, this new stuff still isn't MS AI PC grade capable yet ......



I wonder whose standard is this 16-24 TOPS level and when they are going to kick their stuff off big-time ????      Roll Eyes    Or, if at all .......
Back to top
 
« Last Edit: 04/17/24 at 11:48:53 by Oldfeller--FSO »  

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

Hobby is now
"concentrated
neuropany"

Posts: 12638
Fayetteville, NC
Gender: male
Re: Browsers and AI PCs
Reply #35 - 04/17/24 at 07:11:45
 

Intel's effort to build a foundry biz is costing far more – and taking far longer – than expected

If you wanna make money you've gotta spend money. And against Samsung it's gonna cost a lot

ANALYSIS Three years after CEO Pat Gelsinger announced Intel would create a foundry business that took on contract manufacturing gigs, Chipzilla has committed to more than $185 billion in spending across new and existing fab, packaging, and test sites.

Now with $8.5 billion in US CHIPS Act subsidies in hand and up to $11 billion of loans secured, Intel Foundry – an independently managed unit under the Intel umbrella – is pushing ahead with its plan to become the number two foundry operator behind TSMC by 2030.

In a report issued on Monday, Intel offered an update on how its ten largest projects are going – and it looks like it's going to be a little while longer before it's ready to leapfrog Samsung and give TSMC a run for its money.

US expansion gains momentum as costs rack up
US-based Intel has made domestic manufacturing a priority. So far that intention has seen it announce four fresh fab sites stateside – two in Arizona and a pair in Ohio.

The two facilities in Chandler, Arizona – Intel's Fab 52 and Fab 62 – have made the most progress since Gelsinger announced them in early 2021. As of December 2023, Intel revealed work on the fab's concrete superstructure had been completed. Construction crews are now working to install the automated material handling system, which Intel describes as an "automated highway" used to transport wafers.

The fabs are expected to come online later this year or early in 2025, and are slated to produce chips based on Intel's next-gen Angstrom era process tech – including its mass market 18A node.

Meanwhile, in Ohio – Intel's first new US fab site in more than 40 years – construction crews are busy digging. The foundry biz claims that in 2023 crews shifted more than four million cubic yards (3.53 million metric tons) of earth – the equivalent of 248,000 dump truck loads – and installed more than 32 miles (51.4 km) of conduit.

Much of the work this year will center on building out the fabs' utility level and bringing in "superloads" of manufacturing equipment necessary for the next phase of construction.

Intel's Ohio fabs are slated to come online in 2025. However, as we reported in February, it appears that delays to US CHIPS Act funding and changing market dynamics may have pushed completion dates into late 2026.

Beyond manufacturing delays, Intel faces several other challenges bringing these fabs online. First is rising construction costs – which have ballooned since Gelsinger first announced the Arizona and Ohio projects.

Intel's Arizona fabs were initially projected to cost about $10 billion apiece. But a little over a year later, the chip shop revealed the actual cost would be closer to $30 billion – and that it had brought in Brookfield Asset Management, a private equity firm, to help pay for the plants.

More recent reporting indicates the Arizona two fabs, along with upgrades to its existing Ocotillo manufacturing plant, will run roughly $32 billion. It's a similar story with the Ohio fabs, which were also slated to cost roughly $10 billion a pop. However as of early 2024, the estimate is closer to $28 billion.

To be clear, Intel isn't the only foundry operator dealing with rising costs. The expected scale – and by extension cost – of both TSMC and Samsung's Arizona and Texas fab sites have also exploded since their announcement.

Another challenge facing Intel and others building fabs in the USA is staffing. Shortages of skilled staff have forced delays for TSMC, while Intel is actively working to train personnel to run these facilities when they come online.

Last summer, the Xeon-slinger told us that the "US semiconductor industry could face a shortage of 70,000 to 90,000 workers over the next few years."

Wafer production is only part of a complex supply chain, which increasingly relies on advanced packaging technologies.

Along these lines, Intel has previously announced upgrades to its Rio Rancho, New Mexico facility –known as Fab 9 and Fab 11x. These upgrades – originally expected to cost $3.5 billion but now estimated to cost closer to $4 billion – aim to enable the facility to support advanced packaging in high volumes.


Headlines to check into to get a fuller picture
Intel's foundry business bled $7B in 2023 with more to come
Samsung snags $6.4B in CHIPS Act funds for Texas fabs
Intel scores $8.5B in government cheddar to supercharge fab builds
TSMC scores $11.6B funding infusion for Arizona fabs, now plans for third plant
Over the past few years, advanced packaging has emerged as a key technology for scaling compute beyond the reticle limit of a single silicon die. Technologies like Intel's EMIB 2.5D packaging and Foveros 3D packaging tech enable multiple dies to be stitched edge-to-edge or even stacked atop one another. We saw both technologies on display with Intel's GPU Max product family, better known as Ponte Vecchio.

With more chip shops (including AMD and Nvidia) embracing multi-die designs, the foundry challenger appears to be leaning on its tech to woo prospective customers away from TSMC – which is currently responsible for most advanced packaging not performed by Intel.



Intel celebrated the reopening of Fab 9 in January, and plans to install and qualify additional tools at it and Fab 11x later this year.

On the topic of advanced technologies, last northern Fall (Autumn) Intel broke ground on a new support building near its D1X fab in Hillsboro, Oregon. The R&D site will reportedly support the development of next-gen process tech.

When complete, the building will add an additional 35,000 square feet (3,251 sqm) of clean room space, and feature six docks to allow for faster installation of the latest tools at the fab. Over the next few years, Intel plans to invest $36 billion into its Hillsboro site.

Foreign investment abounds
Intel's foundry investments aren't limited to the US. Over the past few years, the chipmaker has invested heavily in upgrades to its Irish chip plants, announced new fab sites in Germany and Israel, and detailed assembly, test, and packaging facilities under development in Malaysia and planned for Poland.

Among the largest of these is the fab site in Magdeburg, Germany. It was announced in early 2022 as part of a €33 billion ($35.1 billion) investment in manufacturing across Europe, of which the German plant would account for approximately €17 billion.

The plant was expected to break ground in early-to-mid 2023 and begin producing components as soon as 2027. However, much like Intel's other fab projects to date, it quickly ran into trouble.

For one, the total cost of the facility continued to rise. In mid 2023, when the plant was expected to break ground, Intel had finally reached an agreement with the German government to subsidize costs by €10 billion ($10.6 billion) of the by then larger €30 billion ($31.9 billion) project.

The latest Intel update on construction progress contains little detail as to the status of the German build, but does mention an apprenticeship program to train local workers to run the facility when complete.

Intel's European expansion has, like its US developments, been plagued by staffing shortages. Intel needed about 3,000 staff to run the plant, but its three-year apprenticeship program reportedly had just two candidates enrolled for 2023, and 20 slated to start in 2024.

The reports raised concerns that unless Intel resolves this issue – either through additional training and incentives or through increased levels of automation – the plant could face additional delays before production commences.

As well as the Magdeburg plant, Intel is also working on a new fab at its existing campus in Haifa, Israel. Intel has operated out of Israel for 50 years and currently employs 11,000 people across four locations in the country.

The $25 billion Haifa project, detailed late last year, is slated to produce chips using extreme ultraviolet lithography (EUV) and is set to receive $3.2 billion in support from the Israeli government.

According to Intel, construction of Fab 38 – which is expected to come online sometime in 2028 – has commenced. However the report offers little detail. Considering the timeline of Intel's other projects, we'd wager that construction crews are probably still moving earth.

In addition to fabs, Intel has also made upgrades to existing facilities – including Fab 34 in Ireland, where the story is much the same as the others. Announced alongside Intel's investment in Germany, the Leixlip, Ireland-based fab was due to receive a €12 billion ($12.7 billion) investment to upgrade the facility to support EUV lithography necessary to produce chips using the Intel 4 process.

However, by the time the plant began volume production of the new node in September, costs had increased to €17 billion ($18.1 billion).

Intel also detailed progress on its advanced packaging facility in Penang and a fifth assembly test manufacturing plant in Kulim, Malaysia.

The former dates back to late 2021, and saw $7 billion set aside for a 710,000 square foot facility designed to manufacture products built using Intel's 3D Foveros packaging tech.

Intel also provided something of an update on its planned assembly and test facility in Wrocław West, Poland. The $4.6 billion project was first teased in mid 2023 and is designed to support Intel's European build out – including the facility in Magdeburg.

Beyond this, Intel didn't provide much of an update – other than to highlight joint research and development projects with Wrocław University of Science and Technology and to cultivate local talent to fill the 2,000 positions the site is expected to employ.

A long costly game
With its foundry build out well under way, Intel has begun taking steps to separate its Product and Foundry businesses to avoid the kinds of conflicts of interest that can arise when building chips for competitors.

Intel detailed the split at its Foundry Direct event in February. The two organizations will be separate legal entities with independent sales forces and ERP systems, Intel Foundry Services head Stu Pann explained at the time.

Intel Product and Foundry financials will be reported separately. We caught a glimpse of this earlier in the month, when Chipzilla released revised financial disclosures reflecting the change.

The new reporting structure effectively makes Intel Products look more like a fabless semiconductor manufacturer. Meanwhile, Foundry looks rather unhealthy – owing in no small part to the fact that Intel is its only major customer at this stage.

In 2023, Intel's revised records showed that its Foundry business had an operating loss of $7 billion. And with so much money wrapped up in capital expenditures, losses are not expected to disappear in the short term.

Speaking with analysts, Gelsinger warned that it would likely be 2027 before Intel turned the corner, with a not inconsiderable part of that being a reduced reliance by its Product group on rival foundries.

Many Intel products – like its Gaudi and GPU Max accelerators – are manufactured in part or in whole by TSMC. By 2027, Intel aims to reduce the amount of kit it outsources to other manufacturers from 30 to 20 percent of its total output.

Intel's mass market 18A process node – on which much of Gelsinger's Foundry bet hinges – won't see broad adoption until at least 2026.



Cheap and easy prediction time:

If China does not invade Taiwan, Intel will forfeit a bunch of big loans and be forced to divest itself of some of the failed Intel Foundry program.

In this progression, Intel will beg more $$
Back to top
 
« Last Edit: 04/22/24 at 13:46:41 by Oldfeller--FSO »  

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

Hobby is now
"concentrated
neuropany"

Posts: 12638
Fayetteville, NC
Gender: male
Re: Browsers and AI PCs
Reply #36 - 04/21/24 at 18:35:15
 
Intel
..... estimates it will take them until 2027 to begin to break even on their existing commitments.   So Intel will lose money (a lot of money) for the next 3-5 years.

Intel is actually betting that Trump will continue to dump US GOV money down the Intel rat hole during his entire administration.


Hmmmmmmmm ?????


If Intel does not get the huge amounts of money it has been getting from the governments---- what happens?   Intel has to slow down a good bit, but as long as it has a contractual lock on all the PC box makers, it has a built in cash flow big enough to keep all its vital plans afloat, but these plans will not grow at the promised rates without .gov financing.

Please remember, Intel has a massively larger market share than AMD does because of the contractual lock it has on the PC box builders.   Not that its processors are better, in recent  history Intel products have been inferior but the Intel products have been more readily available in retail channels.

AMD has to fight Apple and all the others for TSMC allocation time and now that Intel is being built by TSMC there are signs that AMD is being choked out intentionally by Intel.  Intel builds very large lots and warehouses them, AMD finds at critical points of their product introduction cycle that Intel  has bought up all the TSMC space that Apple has left open.

All AMD has actually done is force Intel to get better at a faster rate that it would have chosen to do otherwise.

Back to top
 
« Last Edit: Yesterday at 04:52:45 by Oldfeller--FSO »  

Former Savage Owner
  IP Logged
Pages: 1 2 3 
Send Topic Print


« Home

 
« Home
SuzukiSavage.com
05/09/24 at 15:06:09



General CategoryThe Cafe › Browsers and AI PCs


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