Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×

Comment Re:geolocation redundancy (Score 1) 99

raising the possibility of fallout for the technology industry and perhaps the global economy.

I don't know but I do some very small scale hosting and I have presence in multiple datacenters in multiple continents. Wouldn't be doing the same with chip production be a good idea?

You are 100% correct. PRoblem is that doing geographical redundancy with datacenters is oreder of magnitude more difficult to do with chip manufacturing than it is to do with datacenters.

The knowledge needed to do chips with good yields is highly specialized, and highly specific to each manufacturer, so, due to lack of specialized workers is easier for Foundries to expand existing fabs, or keep fabs in the same country, than it is for them to put fabs everyhere.

Comment Re:East coast though... not so scary (Score 3, Interesting) 99

While the earthquake is clearly powerful and devastating and I hope that there is only limited loss of life, nearly all if not entirely all of the chip fabs are on the west (Hsinchu) and south (Tainan) areas which are distant from the epicenter. Of course the equipment is very sensitive so it makes sense to be cautious but this particular earthquake may not have much impact on the semi industry.

As for loss of life, last time I checked it was only 1 death (probably more by now), so, thanks the lord in the hights tha the mortality was low.

As you aslo said, the equipment is very sensitive, but I think you are underestimatingt the disruption. You see, the engineers at TSMC, UMC and Nanya (to name three), will have to completely purge all production lines, then recalibrate the equipment, and then re-start production, causing a disruption (a bubble + delays) in the supply chain...

But guess what?

If you are an engineer (or an Egineress) charged with recalibrating the epuipment and re-starting production, and you were hurt because a piece of concrete fell on your back during the eartquake, you will be out of commision for a long while, and even if you are 100% Ok, if your parents, offspring or SO were hurt, you will be taking care of them, and also unavailable for a while... or, if they "force" youto be there, your mind will not be 100% there...

So, fabs face a shortage of workers in the aftermath of the quacke, precisely when they need them more....

Comment Re:"Silicon Shield" prevents operational redundanc (Score 1) 99

I highly doubt the Taiwanese government (TSMC's largest shareholder) would be willing to bring more operational redundancy outside of Taiwan for the sole purpose that Taiwan's semiconductor fabrication dominance has single-handedly prevented an invasion by China. There is a reason the semiconductor industry has been dubbed Taiwan's "Silicon Shield".

Unfortunately for the rest of the world, Taiwan happens to be right smack in the middle of the Pacific Ring of Fire and is also in the same region where many of the world's most powerful typhoons occur. A powerful earthquake, tsunami, super typhoon, or the invasion of a vengeful neighbor (which keeps calling Taiwan a "rogue province") might well be what causes the next global economic catastrophe.

Actually, is the other way around. China wants taiwan back as a National pride issue, not for economic reasons. Is too long to explain, and dates back to the Opium wars (you can google around if interested). Actually, it almost happened in the mid of the XX century, when there was no semiconductor industry to speak of.

Actually, if china invaded, the taiwan semiconductor industry would be obliterated, either by planed destruction by the taiwanese govt, or as non-intended collateral damage of the invasion per-se, or because a samiconductor outfit without the personnel is not worth it, as that is now adays so specialized and different from one another that restarting it without the people, even if it went 100% unharmed, will be a long process, and once re-started, it will be at much lower nodes and yields.

So, one of the (many) reasons china does not invade is that a big chunk of the rest of the world do not want Taiwan's semiconductor industry disrupted, and therefor, have implemnented disuasive meassures, not because the chinese govt wants to capture Taiwans semiconductor economic benefit (because it will evaporate in case of a invation for a loooong time).

Comment And less advanced chips are important too (Score 3, Insightful) 99

While everyone and their dog is fixated on the "most advanced chips", less advanced chips are important too.

UMC, and Nanya (to name 2) are Fabs that operate in Taiwan too, and guess what, many of the chips they manufacture end up in your car, your home theatre or your microwave. And even if your chips were "Fabbed" elsewhere, chances are they will be packaged by ASE group, also located in Taiwan...

Tell Toyota that a $30.000.oo corolla is stuck in the assembly line because a $30 non-advanced chip for the ECU is missing, either because a Fab in Taiwan could no manufacture it, or because the silicon is ready, but it could not be packaged.

Or tell tesla that a $60,000.oo car is also stuck on the assembly line for lack of an ABS/ECU non-advanced chip

Or tell a manufacturer of an appliance (say, a non-smart fridge, that still needs semiconductors) that a U$D appliance is stuck because they could not get a $10,oo non-advanced semiconductors in time.

And contrary to popular beliefs, many of those fab and/or packaging services, and sometimes, even the finakl chips themselves are not easily fungible.

Is not like a manufacturer of a non-advanced ECU chip can shift the VHDL design from the UCM or Nanya rules to the Glofo or ON Semicondutor at the drop of a hat. And is not like a chip designed from the get-go to be packaged by ASE group using ASE's propiertary test/packaging process can be packaged elsewhere by the drop of a hat...

So, do not fixate fetishistically on the bleeding edge, the non-bleeding edge will be equally damaging.

Comment Copyright infringement (Score 4, Informative) 395

Ms'. Fosen latest stance of the matter is that she does not want the image used anymore. And the copyright holders (which is Playboy magazine, not Ms. Fosen) are siding with her.

So, the "morally correct" * thing to do is to stop using the images. Lest you as a researcher (and you employer) are ready to get a bunch of cease and desist letters and copyright infringement lawsuits. I do not know if the IEEE (and Nature before them) is doing it to be "morally correct" * or to cover their asses, but well done anyway!

May I sugest using relevant non-provocative pictures (high contrast, fluff, varied detail, colour, etc ) of REAL sicentists? Claude Shannon, Harry Niquist, Alan Turing, Hedi Lamarr (nee Hedwig Eva Maria Kiesler), grace hopper, steve sansonn, etc... I think the archives should have some relevant pictures, even negatives, to scan

Full disclosure: I am a memebrr of IEEE since 1994

* Not moraly correct in terms of "do not use adult magazine pictures for research papaers in unrelated fields" but rather morally correct in terms of "respect the wishes of the picture subject and the copyright holder"

Comment Re:What the *bleep* is AMD doing? (Score 1) 44

seriously, I get that CUDA is a tough nut to crack, but it's a $2 trillion dollar market.

AMD should be getting flooded with capital right now. But I suspect that the investors aren't all that interested since they already own Nvidia stock, and why would you compete with yourself?

AMD has been bussy changing their "CUDA Killer" architecture (and the Accelerated Graphics one too) every few years. Close To Metal, Mantle, Stream, GPUOpen, HIP, ROCm... Does any one of those ring a bell?

AMD has been bussy having knee-jerk reactions, developing new CUDA-Killer software frameworks, without reason or Ryme, and then discarding them, without stoping to take a long hard look and reflecting at what need to be done, and then doing it.

Hope intel has better luck with their unifyinf FOSS effort, and AMD hops into intel FOSS bandwagon.

Comment It will fumble it, similar efforts fumbled before (Score 3, Insightful) 44

Just for example, Think about the shitshow that is OpenCL v3. OpenCL was SUPER-Slow but at least overall rationaly developed until OpenCL 2.2.. and then it went apeshit with OpenCL 3.0, there, 1.2 is mandatory (a regression), and ABSOLUTELY EVERTHING beyond that is optional.

Who could develop interoperable SW on top of OpenCl like that?!

Similar crap with other efforts like SYCL...

So, history does not repeat itself, but it surely rymes, this effort wil fumble too

And the worst part is that I wish there was a viable alternative to CUDA...

Submission + - ASK Slashdo: What the MINIMUM Windows AI PC will be like? (slashdot.org)

williamyf writes: On March 21, 2024, Microsoft introduced their "First AI PCs" , the Surface Pro 10 and The Surface Laptop 6. Jokes about the "Copilot Key" aside, if the specs of these machines were the bare minimum for a Windows 12 "AI PC", again a sizeable swath of current machines would not be able to upgrade. I guess tht's an scenario Microsot will try to avoid. So, what a MINIMUM Spec Windows 12 AI machine would look like?

Currently, DirectML can work with a machine with a DX12 Graphics card (Feature Level 12), so even a lowly intel 9th gen iGPU or an nVidia GTX9xx can do it (sorry, I am not very knowledgeable about AMD APUs). But I am not sure if DirectML can be used for other AI tasks beyond DirectX itself. Could that be the cutoff?

Maybe OpenCL 1.2 with a dash of selected 2.x and 3.x features will do? Again, a 9th gen intel iGPU or a nVidia 9xx can do it, but the devil is in the details, and it will depend on the 2.x and 3.x features required.

Perhaps require those to at the same time, with a dash of AVX2 to supplement anything missing?

With XeSS intel has demonstarted that DP4a is viable for AI tasks, while that can be done with an nVidia RTX10xx, only 12th gen Intel iGPUs can do that, leaving in the dust a lot of iGPU only machines...

OR perhaps, a processor supporting VNNI (or more exactly, AVX10.x VNNI) can do the deed? But that would leave even more machines in the cold.

Or perhaps Microsoft will do a clean sweep, and require in NPU/MXM/whathaveyou either in the CPU or GPU, in which case, only RTX20xx or intel 14th gen need apply... Truly an ewaste aocalypse (or another big opportunity for Linux to give new life to those machines, who knows?)

Also, ne has to wander? How much memory available to run (or train) the models Microsoft will require. While there are "smallish" models out there, anything above 6GB will kick to the curb many a graphics card which otherwise could have handeled the load compute-wise.

Also, will Microsoft run/train many models device side (like voice and writing recognition) for added privacy and battery life? Or will they do most AI work in the cloud, with the added latency and battery penalty (after all, those 5G modems chew battery like crazy).

So, my Slashdot brethen: What will the MINIMUM Win12 AI PC will look like?

Comment Re:Unusable (Score 1) 32

I tried OCLP with the latest Sonoma, which worked okay for getting Xcode 15.3 on a 2014 mini. It performed somewhat acceptably after disabling just about every service imaginable. I was going to just deal with it. After I realized Apple started forcing the use of their hypervisor, I wiped Sonoma for Ubuntu. Aside from losing an $80 box that can properly build Apple stuff, Iâ(TM)m quite pleased. If I had to, I guess I could dial boot, but Iâ(TM)d rather avoid the hassle.

In principle, you can run MacOS on a virtual machine on ubuntu, as long as it run on Apple Hardware. the ToS allow it 100%. Now, if a 2014 mini can run sonoma in a VM, is another matter entirely, I think it can be done, but will depend on what your VMM is capable of simulating or not.

If you make that work, you could build. and more importantly, notarize, Apple Apps, be it MacOS or iOS.

Comment Re:Pay Up, Or Else (Score 1) 33

Well, is not really pay up or else. Is more:

"We suspected we could sue you for a long time, but you had no money, so it would cost more in lawyers+executive focus&distraction+Bad publicity than what we could get trying to squeeze a stone. Now that you finally have money, suing you is worth our while"

If that was the play, I would have expected Nokia to spring this after a successful IPO.

Do you realize that most analysts think that this IPO will be a flsh on the pan, have a one or to day frenzy for the share price then to crash?... right?

Comment Re:Nokia (Score 1) 33

Do they actually sell anything or have they pivoted to patent troll full time?

They are the 3rd largest telco vendor Worldwide. The largest ones being Huawei and Ericsson (which one if first and which one is second varies quarter to quarter). ZTE is a distant fourth.

Nokia is the Amalgamation of the telco part of nokia + the telco part of Siemens + The telco part of motorola + Alcatel-Lucent (which in turn was Alcatel + Lucent).

You may recall that at some point Nokia, Siemens and Alcatel manufactured their own phones, with their own OS and their Own Software. I can not say anything about Siemens and Alcatel, but nokia had some Forward-thinking stuff between '95 to '10. All those implementations had patents associated to them, some implementations were successfull, some not, but the patents remained.

Nokia still has a consumer electronics division, doing R&D, licencisng IP and manufacturing a little bit of stuff.

In another note, the cellphone division of nokia was sold to Microsoft, along with a limited time licensing of the Brand. Human Mobile Devices (HMD, a company staffed mostly by former nokia cellphone business executives let go duriing the turmoil years and whose main shareholder is Foxconn) bought the cellphone operation from Microsoft, including some factories, and got a licensing agreemnet with nokia for the nokia brand, and are manufacturing Smart and feature phones to this day.

Comment Re:Pay Up, Or Else (Score 1) 33

This strikes me as a bit of a shakedown, settle with out patent claims or we'll screw up your IPO by creating a new potential liability.

I'm assuming this is a somewhat standard practice, both with patents and other legal liabilities, and probably isn't a big issue for investors either way.

Well, is not really pay up or else. Is more:

"We suspected we could sue you for a long time, but you had no money, so it would cost more in lawyers+executive focus&distraction+Bad publicity than what we could get trying to squeeze a stone. Now that you finally have money, suing you is worth our while"

Comment Re:Thoughts and Ideas. (Score 1) 33

Thoughts and ideas should not be patentable, only actual implementations. It's like saying 5 times 5 equals 25, then suing anyone who comes up with a formula that equals 25. Copy write your code, but fuck off if you think you can bottle up the idea that said code performs and/or results in.

Nokia did/implemented (and patented) some pretty forward thinking stuff in the Consumer cellphone area in the ~'95 ~ ~'10 timeframe (one could say things that were ahead of their time), that includes but is not limited to mobile chat, bulleting boards moderation (including but not limited to keep refreshing the feed, keeping a coherent view, push notifications et al). Some of those implementations were sucessfull on a regional basis, but forgotten on the midst of internet time, some implementations were not successfull at all, but the patents stayed...

I guess things are similar with Alcatel and Siemens in their consumer divisions (remember when both Siemens and Alcatel made their own phones, with their own SW?), but I am not that well versed in that respect.

Also, even now, Nokia has aconsumer electronics division, granted, no more than an appendage, but still doing R&D AND selling products and/or IP.

TL;DR Nokia (and Alcatel and Siemens, that are now part of Nokia as well) did not patent "ideas", many of those patents were also implemented.

Nokia hardly qualifies as a "non practicing entity" (i.e. patent troll).

PS: How do I know? Well, I worked at a predomiantly Nokia telco customer from '98 until '04. While we were small, they treated us very good because we were their "showcase client" in LatAm (and their firts customer in LatAm too). Also, during my tenure, we flirted with Siemens from 2000 'til 2003.

PS2: With the 5x5=25 you choose a poor example, as patent law around the world has provisions to explicitly forbid that kind of patent.

Slashdot Top Deals

If a thing's worth doing, it is worth doing badly. -- G.K. Chesterton

Working...