|
SuzukiSavage.com
/cgi-bin/YaBB.pl General Category >> The Cafe >> AI phase-in rocks all computing worlds /cgi-bin/YaBB.pl?num=1520358046 Message started by Oldfeller on 03/06/18 at 09:40:46 |
|
Title: AI phase-in rocks all computing worlds Post by Oldfeller on 03/06/18 at 09:40:46 This is a new thread, following the explosion of AI as it puts MS and Intel off into the weeds technologically. One hopes that MS can manage to join in on to some other type of processor tech that can tap into these AI worlds, but one holds out much much fainter hopes for Intel as they got "the not invented here" disease really really bad & very very big time lately. ARM is leading this thrust with DynamIQ and all the primo chipsets out now are already DynamIQ compatible. Now here comes the lower end DynamIQ compatible graphics sets that will bring full AI to mid range and low range phone SoCs. https://liliputing.com/2018/03/arm-launches-new-mali-gpu-display-and-video-designs-for-mainstream-and-mid-range-devices.html ARM launches new Mali GPU, display, and video designs for mainstream and mid-range devices http://https://liliputing-wpengine.netdna-ssl.com/wp-content/uploads/2018/03/mali.jpg ARM Mali-G52 and G31 GPUs First up are ARM’s new graphics processors. The Mali-G52 is said to offer up to a 30 percent performance improvement over the Mali-G51, while also bringing a 15 percent enhancement in energy efficiency. Perhaps the biggest change is that the company is promising up to 3.6x the performance for machine learning tasks. The Mali-G52 is still one of ARM’s “high area efficiency” designs, rather than a “high performance” design, which means it won’t offer the same kind of performance as a Mali-G71 or G72 GPU. Meanwhile, the new Mali-G31 is the company’s latest “ultra-low power” GPU. It’s basically designed to fill the slot that’s long been occupied by the ARM Mali-400 line of GPUs. The new chip ARM says it’s 20 percent smaller, offers 20 percent better performance density, and it’s ARM’s “smallest processor to support OpenGL ES 3.2 and the latest generation Vulkan API.” Mali-D51 display processor This new mainstream display processor borrows some features that had previously only been available in the company’s premium version, including HDR support and ARM’s Komeda architecture. It also supports 4096 x 2048 content at up to 60 frames per second, twice the scene complexity available from a Mali-DP650 display processor, 50 percent better memory latency, and 30 percent lower power consumption. Mali-V52 video processor The new video processor is available with between 1 and 4 cores, supports 4k60 video decoding, 4k30 encoding, and a 20 percent boost in encode quality (which means that you can get the same quality of video with a 20 percent lower bitrate, helping save storage space and/or bandwidth). One use case that might seem a bit odd in the west is the ability to decode and play 16 HD videos at once… something that ARM says is common with Chinese TV boxes, where users want to see live previews of available programs in a 4×4 grid. Notice the last paragraph -- your Chinese "everyman" isn't buying any primo stuff at all right now, he is buying lots of low end to midrange stuff. This is an AI set built for him, specifically. Mediatek has just released the first of these graphics cores teamed up with their same old Cortex A73/Cortex A53 combo they have had in play for several years now. It makes up an ass-kicking very low cost to implement "winner for right now combo" for the China marketplace that will grow lots of "me too" competitors out the butt very very very soon. Even Qualcomm has recognized this threat and has identified their low middle "me too" SoC with a chipset designation that has of yet no real design or production reality behind it at all. Also note that ARM will never say anything about a new tech until one of their vendors actually ships a product containing it. Mediatek popped this cherry a week ago, so here is the formal announcement of the new technology coming out from ARM. |
|
Title: Re: AI explosion rocks computing/phone worlds Post by Oldfeller on 03/06/18 at 17:47:39 Global Foundries has now fully announced its 7nm EUV lithography process, making the third major foundry to announce that 7nm production is imminent. Samsung and TSMC have already announced their stuff (and both are currently actually making memory chips at both 7nm and at 5nm at this time). Furthermore, TSMC is now supposedly in full "private Apple production" now on this year's newest Apple SoC using 7nm lithography. Intel is still trying to get their 10nm process to work right ...... sad, but they still have not gotten acceptable yield rates yet off their 5th try at bat with their proprietary 10nm process. =================================================== DynamIQ was always originally planned as a 7nm lithography stage, although several major players "upsized" it one generation so as to get out the gate a year earlier than would be possible at the delayed 7nm lithography. This was Samsung and Qualcomm doing that (and to a lesser degree Huawei at a yet still larger lithography stage while still using A57 bigs and A53 littles and a slightly more modern ARM GPU set). Mediatek has jumped on this particular A57 bigs and A53 littles swing cheap bandwagon but Mediatek differs by using a very modern ARM G52 GPU set, one that is complete with an good sized 2.6 TOPS AI block integrated right into the GPU function set. Now that the right 7nm lithography node is finally ready, DynamIQ will run in primo, middle and low end ARM designed chipsets, with each level carrying an AI integrated GPU set that best matches the main cores functionality levels. This is 2.6 TOPS of AI speed that can be used by freely by all apps just by making a call to it. Since it saves so much battery (not having to radio the cell tower all the time, the calculation delays and getting the results back from the tower) it will save a lot of battery run time. The AI functions will get used first, for both speed and battery reasons ..... I look to see the next big thing to be lots of full Octacore DynamIQ A55 SoCs with either an A52 GPU set or perhaps an A62 GPU set depending on which GPU set has the better mix of AI and graphics power and the lowest cost for the actual uses the Octacore chipset will be put to. Using 8 littles of the newest generation has always been the first stage of a lithography generation's general adoption pattern, as it is the cheapest / quickest way to get into full production with something that will clearly beat all previous generations. Plus, you only have to buy one new license and optimize for the one A-55 core type so you can move quicker. |
|
Title: Re: AI explosion rocks computing/phone worlds Post by Oldfeller on 03/07/18 at 06:50:22 https://liliputing.com/2018/03/microsoft-will-offer-s-mode-of-various-windows-10-editions.html http://https://liliputing-wpengine.netdna-ssl.com/wp-content/uploads/2018/03/windows-10-s.jpg ...... right Microsoft begins to unveil their counter plan to Chrome AI, and it is a fairly weak re-naming of what they are already doing (but the implications for the existing older Windows users are quite large). Windows 10 S as it exists today is basically a streamlined version of Windows 10 Pro that offers features such as Bitlocker disk encryption and shared PC configuration, but what it lacks is the ability to run apps that aren’t downloaded from the Microsoft Store. Details are scarce at the moment, but Microsoft’s Joe Belfiore confirmed on Twitter that “next year 10s will be a ‘mode’ of existing versions, not a distinct version” of Windows 10. In other words, you might see Windows 10 Home S (or maybe Windows 10 S Home), Windows 10 Pro S, or even Windows 10 for Education S. Belfiore’s comment basically confirms a report from Thurrott.com in February that suggested “Windows 10 S is dead, it’s now Window 10 S mode,” and that “each version will have an S Mode.” That report suggested the change could take effect starting in April or May, while Belfiore says “next year.” But maybe he means next school year or next fiscal year, in which case we might not have to wait until 2019 to see Windows 10 S become a “mode” rather than a distinct operating system. While the primary difference between Windows 10 S and other versions of Windows 10 is that the S version restricts users from downloading apps that aren’t in the Microsoft Store, there are a few other differences as well. Among other things, it doesn’t support the Windows Subsystem for Linux, doesn’t have the same level of compatibility with third-party peripherals, and doesn’t allow users to change the default search engine or web browser (you are stuck with Bing and Edge). In other words, it’s a more restrictive experience. On the one hand, that could be frustrating for long-time Windows users that want more freedom over the software running on their PCs. On the other hand, it could help provide better security, stability, and simpler management for a group of computers, which is why Windows 10 S was initially targeted toward the education market, where it would compete with Google’s Chrome OS. But now the S thing is going to be across the board for everybody -- you will start out with a default STREAMLINED or START-UP minimal Windows installed on the machine "as built" and then YOU PERSONALLY will opt into buying the upgrades that you need from the Windows store. Note the rename, the S stands for streamlined or start-up -- it does not stand for school any more as school is a paid for upgrade now (and is only free on a Chromekiller machine). This way Mickey gets a net lower initial sales price on the machine itself (to encourage you to go buy it) while locking you mentally into an upgrade path that never ends -- $49 here, $49 there, $89 for a major year's upgrade -- you get the picture. ..... can you say "Maasi blood bowl" ???? Fact, Mickey has gone after my wife's machine now, she is being told her license isn't from the Microsoft store and Mickey has begun deleting "unauthorized softwares" off of her machine and that is pissing her off weekly as calling her a pirate and taking stuff off her machine when she knows better makes her a confirmed MS hater. The first time Mickey deletes something the IT guys from work put on her machine and it stops working right as a "dial into work" machine -- well, then the fireworks will be fun to watch, but not from inside the room with them. MS's nightly twiddles are now slowly downgrading your machine to a pure Win S default status ..... to get a lost functionality back you will need to buy an upgrade package from the Windows Store. EXPECT to start losing an occasional old software functionality as MS "modernizes" your machine. Like I said, MS has created the environment for a perfect storm (complete with lots of pent up demand) for whenever a replacement technology becomes available. ==================================================== https://liliputing.com/2018/03/switching-from-windows-10-s-to-windows-10-will-remain-free-of-charge.html Microsoft does a quick 180 on plans to land S version on all items and then to stick you with the costs of upgrading to a non-crippled version. Windows 10 S mode should work a lot like the current version of Windows 10 S, offering a user experience that looks a lot like the full version of Windows, but with a few restrictions that add simplicity, tighten security, and make it easier for businesses, schools, or other institutions to manage a set of computers. For one thing, Windows 10 S mode doesn’t let you install third-party apps unless they come from the Windows Store. For another, it doesn’t let you change the default web browser or search engine. It doesn’t support the Windows Subsystem for Linux, may not support third-party antivirus software, and may have limited support for hardware peripherals. But all of those restrictions are a bit easier to accept if it turns out you can buy a Windows 10 S device and know that you’ll be able to switch to a full version of Windows 10 for free. (and then have to rebuy all your various softwares all over again .... hee hee ....) It’s unclear if you’ll be able to switch back though. Belfiore says Windows 10 S mode will be available “starting with the next update to Windows 10, which should be the Redstone 4/Spring update release expected to launch in the next month or two. However, please note that ever since Microsoft first released Windows 10 S in 2017, the company has allowed users to upgrade to Windows 10 Pro for free… but Microsoft had also always suggested that this was a limited time offer, and that eventually users would have to pay a fee to switch to a full version. This has been a consistent official message from the very beginning. What Joe Belfiore says today isn't really official, he is just a talking head that says what the teleprompter reads today. Read the ULA to see what is really real ..... |
|
Title: Re: AI explosion rocks computing/phone worlds Post by LANCER on 03/07/18 at 10:43:47 So they are eating their own tail, and all in an effort to save their tail. Greedy greedy greedy, as they go swirling down the tube. |
|
Title: Re: AI explosion rocks computing/phone worlds Post by Oldfeller on 03/07/18 at 14:17:51 https://liliputing.com/2018/03/microsoft-is-building-support-for-ai-apps-into-windows-10-with-windows-ml.html ..... it is a powerpoint slide, that is all it is, just a slide ..... A vague promise with no reality behind it AT ALL --- MS requires you to have the free Google Tensor FOSS apps loaded in the background to actually do what little it does do http://https://liliputing-wpengine.netdna-ssl.com/wp-content/uploads/2018/03/windows-ml_01.jpg But this spring Microsoft will include a brand new API that lets third-party apps leverage your computer’s hardware for machine learning-related tasks. It’s called Windows ML, and it’ll ship with the next major Windows 10 Update, which is expected to be the Redstone 4 release coming this spring. In a nutshell, Windows ML is a set of tools that lets developers to add pre-trained machine learning models to their apps using visual studio and allow them to run locally on a PC by leveraging a computer’s CPU and GPU resources. Microsoft may also eventually add support for specialized chips such as Intel’s Movidius hardware, which is designed specifically for machine learning. Among other things, this could enable apps to use machine learning for things like computer vision. In fact, all of Microsoft’s sample apps for Windows ML are related to computer vision, with examples that use a camera feed to: * interpret a number drawn by a user * classify an image Right now your cell phone can tell you exactly where you are if you take an image of a building front, or it can auto-translate a foreign language sign if you take an image of it. Your existing Cell Phone AI can identify random images at a rate of 1,000 a second using built in AI blocks that are part of modern ARM SoC designs. It can recognize your face behind sunglasses in less than a second ....... Now, Microsoft proudly announces it can "interpret a number drawn by a user" and "classify an image" as in a square or a triangle or a rectangle. ::) Wow ..... Mickey has such a commanding lead in AI technology using all the power of them massive Intel processors ........ :P I don't know how those thumbnail sized ARM SoCs with built in 4.6 trillion operations per second while using less than 2 watts of power AI blocks are ever gonna compete against the almighty Mickeysoft and their fearsome Powerpoints ........ ::) |
|
Title: AI phase-in rocks the computing/phone worlds Post by Oldfeller on 03/08/18 at 12:44:31 https://www.pcgamesn.com/amd-7nm-tsmc-globalfoundries AMD are splitting 7nm Zen 2 CPU and Vega GPU manufacturing between TSMC and GloFo http://https://www.pcgamesn.com/sites/default/files/TSMC%20Fab%202.jpg AMD once incorporated the manufacturing arm of what is now known as GlobalFoundries - which was split from the company back in 2009. As a result of this special relationship, AMD tend to rely on the foundry for most of their semiconductor needs, including their Polaris GPU architecture, Vega architecture, and Ryzen processors. For 7nm, however, AMD will move some of their product lines over to TSMC. With the upcoming Ryzen 2000-series processors remaining with GlobalFoundries for their 12nm LP node, it seems more than likely that AMD’s 7nm Vega card - expected to launch later this year - and upcoming Navi architecture will both be manufactured by TSMC. Zen 2 may then ship from GlobalFoundries 7nm fab once it launches. “So in 7nm, we will use both TSMC and GlobalFoundries,” Su says to Anandtech. “We are working closely with both foundry partners, and will have different product lines for each. I am very confident that the process technology will be stable and capable for what we’re trying to do.” GlobalFoundries expect the move from 14nm to 7nm to offer around 40% performance boost, or a 60% total power reduction. TSMC are already breaking ground on a 5nm process fab, which has firmly cemented 7nm as ‘old news’ in their books. That’s probably all the better for gamers, as a stable process likely means we will get higher quantities of better performing graphics cards. “We have so many new features in Vega, such as adding some of the compute centric features - but the beauty of 7nm is density and power,” Su says. “When you think just about how much horsepower you can put in the new technology, it made sense… the GPU is usually for us the first product in a technology. Graphics does have the capability for a lot of redundancy on it and so we feel like it’s a great utilisation of the technology.” TSMC also produce Nvidia’s 12nm-based Volta and 16nm-based Pascal GPU chips, although so far there has been no indication of Nvidia’s future development on the 7nm node. Nvidia are sure to make use of the denser tech at some point, and if Nvidia’s Ampere truly is the next generation after Volta, then this could be when the 7nm lithography comes into play for the green team. It seems the 7nm process node will be a stable technology ahead of AMD’s new Zen 2 and Navi launches, which is good news for AMD as their roadmaps depend on this process for both arms of their business. The machine-learning focused 7nm Vega shrink will be the first 7nm product out the door, which should offer some indication of what benefits this new process will bring to AMD’s next big architectural changes. What does this all mean? Samsung, Global and TSMC are all cranking up production on their versions of 7nm with Samsung and TSMC having a full quarter's lead currently on Global Foundries. So, all the other major foundries made it to 7nm production before Intel ever got their 10nm to run at full production rates ....... go Intel go --- circle that drain one more time --- swirl baby swirl All of the hockey stick boys are making their best SoCs at 10-12nm lithography or slightly better now, Intel is the sole remaining laggard still stuck at 14nm and is still struggling to get up on 10nm. Apple has their dedicated 7nm lines already up and running at full rates at TSMC and there is LOTS and LOTS of spare 7nm capacity available for anyone with the pocketbook to license the full on 7nm tech designs from ARM Holdings or to develop / tune their own custom 7nm cores in house. AMD is looking at an immediate 50% better throughput compared to Intel with more yet to come as 7nm settles in to the second EUV generation sometimes later on this year. Intel will lose yet another ~ 20% in additional comparative throughput losses due to the bigger productivity bite that fixes to Spectre and Meltdown are taking out of Intel's overall productivity numbers compared to AMD's better post fix numbers ..... Intel's figures and advertising are seen as becoming increasingly SUSPECT as "misleading BS" is being used to bridge the performance gap between Intel and the others right now. AMD figures that it will take some significant market share from Intel using 7nm in 2018, so much so that extra foundry support will be required to meet the additional demand. |
|
Title: Re: AI phase-in rocks all computing worlds Post by Steve H on 03/10/18 at 19:51:55 So, why has nobody put out a decent desktop using one of these nice chipsets with expandable memory, disk drives, optical drives, etc instead of these dinky little web surfing tablets? The capabilities for use of external memory, disk drives, etc. are all built into these chipsets. A very nice desktop could be made without the $800 intel processor at a good price. It has to have expandability, useability, capability to interface with existing devices. No problem for a system running linux which will easily utilize these new processors and their capabilities. |
|
Title: Re: AI phase-in rocks all computing worlds Post by Oldfeller on 03/11/18 at 07:45:09 The largest commercial ARM units made so far are TVs and TV boxes running Android and Linux internally. These sell well in the Asian markets. Kodi Linux TV boxes do carry the current state of the art ARM processors and quite a lot of memory. Still, they only carry half the resources that are required by the Wintel units because they only need that many resources to do the same job. Commercial ARM based Linux computers have been made (and are still available) but do not sell very well outside their niche market as people in general are willing to pay 2x more for a equally capable and 2x more expensive Intel processor that runs a restrictive OS system that they are totally embedded inside (Windows). The used PC side of things really does intrude into the Linux marketplace, as tons of "dead" off lease Intel PCs can be bought for peanuts and repurposed for Linux use (I am typing on one of these right now). This is still the cheapest/best way to implement a nice Linux machine. For $69 I bought an old 3.5 ghz Core Duo full sized engineering desktop that runs Linux Mint very very well. It was very encouraging last year when MS said that they were going to be fully supporting ARM chipsets on Win 10, but this now turns out to be more myth than reality -- large numbers of software restrictions have been rung into that Win on ARM support at the last minute, enough to make it far less desirable. =================================================== https://www.staples.com/Shuttle-XPC-nano-NS02E-Desktop-Computer-ARM-Cortex-A53-RK3368-1-5GHz-2GB-DDR3L-SDRAM-16GB-Flash-Mem-Android-5-1-1-Lollipop/product_IM18G9290?cid=PS:GooglePLAs:IM18G9290&ci_src=17588969&ci_sku=IM18G9290&KPID=IM18G9290&cvosrc=ppc%20pla.dotcom-sales-google.Computers&cvo_crid=225643316625&cvo_campaign=950206958&cvo_crid=225643316625&cvo_campaign=950206958&gclid=CjwKCAiAxJPVBRB4EiwAsCA4aUfsk4IVaIElkNmTfpo62rkaFIEA_NTgMlP6BZ9XGKw_w2uaD3HumBoC3HIQAvD_BwE&gclsrc=aw.ds http://https://www.staples-3p.com/s7/is/image/Staples/m006139773_sc7?$splssku$ $159 from Staples Shuttle ships a ARM desktop running Android that swings specs very much in line with a standard Chromebook. You can buy a bevy of Chromeboxes at right about this same price point that run ChromeOS (and Linux a la Crouton) and likely Win 10 ARM once things finally jell a bit. All of these still run an Intel low end processor and need about half the horsepower of the least of the current crop of Windows machines. HOWEVER, phones are still where ARM, Android and Linux shine the brightest -- and it is still on phones where all the notable newest AI technology is landing first. Could ARM Holdings make up a "PC Capable DynamIQ design"? Yes, they have already posted the needed A-75 Core design and the needed AI Block design and the needed Graphics set designs along with the interconnect that allows stacking in lots of 2 to 8 cores per group (up to 36 groups possible, 288 total cores). Someone who wants to do it simply needs to buy the licenses and use the designs. Qualcomm made up a 96 core Intel beating data center (rack center) core set using these exact tools, as has China's National Gov Data Center and at least one other rather secretive private company who has also done so at this point in time. Nobody is lined up to take on Intel publicly by building a PC sized set of cores because Intel fights dirty and they fight to win (any durn way they can). Intel owns PC space and jealously guards their property. No one wants that battle, actually. It is simpler and safer and easier to let phones naturally grow in power to overlap PC space, which is what they are doing right now. ================================================== Fuschia Friday This is now an on-going weekly news bulletin that shows up on my Google Cards feed every Friday. The Fuschia development group found it needed a weekly coordination update to keep all the players in tune and acting in sync with each other, so they made one up. Needless to say, it goes out to far more people than originally was intended ...... if you have shown an interest in Fuschia then your phone's Google Assistant will pluck up the card and put it in your Friday card stack automatically. About half the stuff in it is actually written in Fuschia now, but it all shows up fine on my oldest and on my newest cell phone ...... and on my Linux PC. I think the boys are doing something right if this is possible. Their avowed development goal is a "cross all devices OS" that keeps up with what it shows you no matter how you choose to get to it this time. So, by eating their own dogfood each week and serving it up to a slew people who will give them feedback if this weeks serving of Alpo don't work right, they are moving right along towards their goal. |
|
Title: Re: AI phase-in rocks all computing worlds Post by Oldfeller on 03/12/18 at 12:59:11 https://www.google.com/search?q=intel+plans+to+buy+broadcom&oq=intel+plans+to+buy+broadcom&aqs=chrome..69i57.9167j0j8&sourceid=chrome&ie=UTF-8 Intel announces plans to buy Broadcom outright if they suceed in buying Qualcomm (if Qualcomm buys NXP-Freescale) Why would Intel do this? FEAR of complete irrelevance Not only would Apple immediately stop buying radio chipsets from Intel (which would hurt Intel a lot) but all avenues going back into phones and many current avenues into automotive world would dry up as well if Broadcom/Qualcomm/NXP-Freescale became one. Intel is also petitioning the US gov to block the Broadcom sale because of "security issues". Intel's house of cards is trembling again ...... |
|
Title: Re: AI phase-in rocks all computing worlds Post by justin_o_guy2 on 03/12/18 at 16:24:33 I don't even know who to root for. I want the consumer to win. |
|
Title: Re: AI phase-in rocks all computing worlds Post by Oldfeller on 03/12/18 at 19:39:10 In this case we want the rich Asian company dumbfooks to keep their sticky fingered hands off the American innovator company, you know the technically active and proficient one that creates all that newest golden goose egg stuff. https://liliputing.com/2018/03/trump-blocks-broadcoms-effort-to-takeover-qualcomm-for-national-security-concerns.html Trump wrote an executive decree saying "Stop all actions" to Broadcom -- let's see if Hoc Tan is stupid enough to go try out the Trumpster. If so, Broadcom will become less important than it is now. |
|
Title: Re: AI phase-in rocks all computing worlds Post by justin_o_guy2 on 03/12/18 at 19:42:50 The innovator , small and investing in New technology, is a target for the lumbering giant that is not able to compete. |
|
Title: Re: AI phase-in rocks all computing worlds Post by Oldfeller on 03/12/18 at 19:46:01 Yep, both to steal Qualcomm's tech and to play "Let's make a deal" with Apple on the side. This thinking plays for Broadcom right now and for Intel should they be stupid enough to continue with their own acquisition plans. ================================================== Broadcom took heed of Trump's executive decree and has dropped all efforts to buy Qualcomm as of 2/14/2018. |
|
Title: Re: AI phase-in rocks all computing worlds Post by Oldfeller on 03/13/18 at 22:12:29 http://www.tomshardware.com/news/design-flaws-backdoors-amd-ryzen,36657.html Speaking of games being played a relatively unknown start-up, an Israeli virus research firm has suddenly and loudly (as in with less than 24 hours between first notice and full on web publication of "suspected vulnerabilities") slammed AMD Ryzen and EPYC Chips chipsets for 5 to 15 different discrete "manners that the chipsets could possibly be made vulnerable". All of which are predicated on breaking all of the industry standard protection protocols that both AMD and Intel currently rely upon. And accusing the AMD platform security module of being a total "backdoor" failure while NOT being able to crack that door open at all in their very limited to non-existing testing. The Israeli company is being lambasted for premature and erroneous assumptions, however, even AMD states that the potential threats found ARE PERHAPS POTENTIALLY REAL and if they are ever exploited they could render the entire affair into an open book. As such, AMD is investigating the claims. The fact they cannot be exploited now is a relief, but it points out that both AMD and Intel indeed rely too much on these "platform security modules" and "management engines" and both have blended them into the boot cycle so thoroughly that it they if they were cracked, then the machine instantly and permanently becomes a defunct paperweight that cannot be trusted without a complete motherboard swap out. This "off the cuff" form of unverified slander attack smells just about like an old-style paid third party FUD attack as would have been orchestrated in the past by Intel, and it is just what the old Intel would have done 10 years ago to any upcoming competitor that was actually threatening their PC business. Paint them thoroughly with thick coating of stinky lumpy brown mud ...... so what if 3 months from now it is all disproven, once it is painted on it cannot be completely removed --- and by 3 months it has done the job it was intended to do -- create FEAR, UNCERTAINTY and DISMAY which derails the competitor's plans. Stink is stink is stink ...... AMD was squaring up to take 25-30% of Intel's market share away from them this upcoming year based in part because AMD lost 20% less of a performance hit due to Meltdown and Spectre --- the highly suspicious timing and general shakiness of this FUD attack is fairly obvious to all experienced industry review persons. However, the responsible review mangers at Tom's Hardware and the ones at AMD all say the various attack methodologies mentioned need to be properly evaluated by SOMEONE OTHER THAN CTS-LABS, THE ISRAELI FIRM THAT OUTED ALL THE VARIOUS PREMATURE VAGUE ACCUSATIONS. Also note: CTS-LABS has a disclaimer about their claims down at the bottom of things, intended to cover their butts because they didn't TEST their allegations at all, really. They simply ran with allegation info taken directly off of the web in creating some of their scenarios. The report and all statements contained herein are opinions of CTS and are not statements of fact. To the best of our ability and belief, all information contained herein is accurate and reliable, and has been obtained from public sources we believe to be accurate and reliable. Our opinions are held in good faith, and we have based them upon publicly available facts and evidence collected and analyzed, which we set out in our research report to support our opinions. We conducted research and analysis based on public information in a manner that any person could have done if they had been interested in doing so. You can publicly access any piece of evidence cited in this report or that we relied on to write this report. Although we have a good faith belief in our analysis and believe it to be objective and unbiased, you are advised that we may have, either directly or indirectly, an economic interest in the performance of the securities of the companies whose products are the subject of our reports. Any other organizations named in this website have not confirmed the accuracy or determined the adequacy of its contents. What to take away about it all --- PCs on the web are always going to be vulnerable to some degree, but right now they likely suck a lot more than they should suck and a completely new boot system is really needed to tighten up this whole constant bleeding sore "security module / management engine" vulnerability thing. This can only take place over time and will be announced as "improvements" and "new features". Allowing security firms license to "out" completely unproven and theoretical attack modes functionally without any peer review should be the subject of normal lawsuits under the libel and slander "count up the damages" legal code we have today. That "cover your butt disclaimer" from CTS makes up a fine smoking gun for a libel court lawsuit to use, actually. The standard Google used was to get initial verification, then to advise Intel and AMD privately up front, give them 200 days to get their act together, ample time to get additional confirmation from MULTIPLE world class industry security peers (while moving on fixing the issues) that the flaws are both real and are able to be accessed and THEN GO PUBLIC with a calm, factual announcement that laid out data and facts and the actual corrections done to date. Kudos to AMD for keeping calm and responding very correctly and responsibly to this overnight hack job. Intel did not do nearly as well given 200 days to get ready. After loudly denying the issues completely for several weeks, the first Intel patches were full of non-corrective BS junk and the original set of Intel fixes actually caused re-booting loops and had to be pulled back by Microsoft the very next week. Also note that if any of these items are judged by impartial review persons to be "worth re-coding to avoid their effects" that all Intel units are vulnerable to each one of these specious scenarios in an equivalent fashion. My take on all of it is that Linux (Torvalds and crew) had fixes in place inside the Linux Kernel for all of Meltdown and most of Spectre before the waiting period was half over as Torvald's Linux boys were part of the impartial review crew and they reviewed it by FIXING it so it couldn't happen (and by doing so said that Industry Wide that all chipsets had some real issues that really did require fixing). Linus will do likewise with these allegations. =================================================== http://www.tomshardware.com/news/cts-labs-amd-ryzenfall-ryzen-epyc,36660.html To What End And For What Purpose? Yet it's important to note that the circumstances surrounding the vulnerabilities' disclosure, and the fact that this is a brand new company, have raised questions about CTS Labs' intentions. It feels like a hit job on AMD, aimed at torpedoing its stock price. That may be unfair to CTS Labs, but optics and decorum are important to perception, and perception is reality for many. People are beginning to investigate CTS Labs and the hit job they just did on AMD. I would expect a lawsuit or two to follow along in due time as both AMD and ASMedia (the company that makes the secure boot chipset system that AMD uses) will suffer both real and clearly accountable financial damages due to CTS Lab's irresponsible and unsupported accusatory actions. CTS Labs is now under the gun to do a proper peer review and they are DENYING they have any need or responsability to do this. CTS's stock holdings, financial transactions and any direct communications with various other entities are also being investigated at this time. ==================================================== https://www.google.com/search?q=torvalds+slams+cts&oq=torvalds+slams+cts&aqs=chrome..69i57.7380j0j4&sourceid=chrome&ie=UTF-8 http://www.zdnet.com/article/bogus-linux-vulnerability-gets-publicity/ Torvalds, in a Google+ discussion, wrote: "When was the last time you saw a security advisory that was basically 'if you replace the BIOS or the CPU microcode with an evil version, you might have a security problem?' Yeah." Or, as a commenter put it on the same thread, "I just found a flaw in all of the hardware space. No device is secure: if you have physical access to a device, you can just pick it up and walk away. Am I a security expert yet?" They've got a point. Linus and Zdnet have actually articulated the root issue ---- small Security Companies feel that if there is not a killer CRISIS taking place right now then they got no new customers. So they are making urgencies up out of nothing now so they can be paid to "fix" them. CTS should be sued for slander and libel and clearly put out of business as a warning to other little crap security companies to be A LOT MORE CAREFUL IN WHAT THEY SAY. If CTS was paid to do this "research", then DOJ resources need to do up a case of conspiracy to commit fraud and nail all the parties involved. We already see some of the elements of conspiracy to commit fraud on the part of the "confirming" little security company that has just suddenly popped into existence and entered the discussion with the intent to bail out CTS when CTS got their tit into the wringer for not doing any form of confirmation. If CTS leadership had any stock purchases / sales involved in these various precipitous accusation actions that they made, then their CEO or whomever should go to jail ...... Ditto if CTS was paid to do this "research". Jail for both of the parties involved is indicated. =================================================== AMD flaws independently verified by two credible sources A few hours after CTS Labs took a beating on social media and some infosec blogs, Dan Guido, the CEO of Trail of Bits —another security company— came forward to confirm that the CTS Labs report was real and contained actual vulnerabilities. Yaron Luk, the CFO of CTS Labs, confirmed to Bleeping Computer yesterday via email that the company had asked the Trail of Bits team to run an independent review of their findings, a fact that Guido confirmed on Twitter. 13 Mar Dan Guido @dguido So this http://AMDflaws.com business... CTS Labs asked us to review their research last week, and sent us a full technical report with PoC exploit code for each set of bugs. Dan Guido @dguido Regardless of the hype around the release, the bugs are real, accurately described in their technical report (which is not public afaik), and their exploit code works. 2:36 PM - Mar 13, 2018 First wave reports from technical experts that had nothing to do with it originally are coming in now. Furthermore, earlier today, Alex Ionescu, one of the most respected figures in the security research community, also confirmed that, he too, had seen the technical report that CTS Labs sent AMD, and that it contained "legit design & implementation issues." This last guy is well known and he had nothing to do with any of it, he like Torvalds is a responsible early mover on bug stuff like this. Alex Ionescu @aionescu On the #AMDflaws — I have seen the technical details and there are legit design & implementation issues worth discussing as part of a coordinated disclosure effort. The media storm and handling around that is sadly distracting from a real conversation around security boundaries. Oh well, AMD has got some fixing to do. |
|
Title: Re: AI phase-in rocks all computing worlds Post by Oldfeller on 03/15/18 at 23:14:04 https://www.anandtech.com/show/12536/our-interesting-call-with-cts-labs This is a long interview review that tracked through all of CTS's reasons and excuses for the torpedo job they did on AMD. look at the kid's pictures -- millennials ??? It carefully documents the CTS use of a PR firm to do up all the web pages, etc. etc. etc. It also covers the lagging but quick ringing in of Trail of Tears, an Israeli security testing firm that was PAID $16,000 for the confirmation that they instantly provided for CTS's shaky judgements (all done within hours apparently). What Anand Tech discovered was that CTS had a customer that actually ordered this work done, specifically instructing CTS to key in on the ASMedia security module and its functions. It was also uncovered that CTS pre-primed several news media sources to break the story hard, and that several investment world folks got prior knowledge of the reveal in time to perform illegal stock trades. This could constitute conspiracy and insider trading, which both mean go to jail time. CTS realized they had said too much, and cut off the interview at the point when they were asked point blank who the original requesting customer was. CTS is just a handful of greedy 20 something "millennial style" kids who are playing loose and fast in areas they really don't understand, while claiming lots of experience in these areas "due to their training as Israeli intelligence security analysts". And now they are claiming protections behind that covert background, something that Israel isn't going to give them as the Trump admin purely hates that sort of action. The kids are going to jail. They will flip on their secret customer who will go down with them. Who picked the PR firm that did the work up for CTS? Who paid that PR firm? Who provided the money to pay Trail of Tears for instant within an hour verification? Who bankrolled this whole fiasco? Killer tech question ..... can you get into the ASMedia security module without using the actual ASMedia language and access codes? Where did the kids get the language and the access codes from? CAN YOU REALLY PARK A MALWARE INSIDE THE ASMEDIA SECURITY MODULE where it remains totally hidden and potentially active for years and years? Somebody is sounding a lot like they KNOW this for a fact. Sounds like NSA/Israeli Intellegence shite, doesn't it? We all know that NSA can get inside all of Intel processors "everything" at will -- this has been a constant all along. Ditto for AMD. Does the Israeli Intelligence Agency share these skills and did they teach these dumb millennial kids how to do this during their military enlistment? If so, these kids really will likely catch a bullet in the head before any confirmation can come out that would taint Israeli intelligence. READ THE COMMENTS AT THE BOTTOM -- SIX PAGES OF REALLY INTERESTING STUFF THERE THAT WAS CHURNED UP BY ANAND TECH'S INTERVIEW. |
|
Title: Re: AI phase-in rocks all computing worlds Post by Oldfeller on 03/16/18 at 00:11:18 https://www.extremetech.com/computing/265695-cts-labs-responds-allegations-bad-faith-amd-security-disclosures-digs-deeper-hole OK, this one is fairly direct and damming -- ASMedia chipsets are not just used in AMD processors sets, they are used in far far far more INTEL processor sets than AMD and this high Intel use level is over six years old at this point. This makes Intel a suspiciously likely source for the giving out of the CTS assignment, the instructions (including language and codes) and the giving out of a tightly coded test coding sample to show the issue is real, a test coding sample that works on AMD processor sets (and perhaps the same carefully written test coding sample will NOT work on Intel processor sets using the same compromised ASMedia chipsets ????) ::) oooooh, this is a fun one, huh? Spy vs Spy all the way ..... The stock market manipulation has become a lot clearer too. A simultaneously released report from a firm trying to short AMD’s stock made the entire affair look particularly shady, especially since the firm in question, Viceroy Research, carried out a nearly identical attack on a German company just a week ago. In that case, Viceroy took a large short position on the German company ProSieben, then accused it of questionable accounting practices. Also the pure biased intentional "finding of the issue with AMD" had become well framed by CTS's own people giving out statements like this: CTS's Zilberman tacitly acknowledges this when he writes: We have started researching ASMedia chips about a year ago. After researching for some time, we have found manufacturer backdoors inside the chip which give you full control over the chips (ASM1042, ASM1142, ASM1143). We wanted to go public with the findings, but then saw that AMD have outsourced their chipset to ASMedia. So we decided to check the state of AMD, we bought a Ryzen computer, and whimsically ran our exploit PoC, and it just worked out of the box. Zilberman, who stopped you from announcing the Intel vulnerability that you found about a year ago by your own words? You, who are so gung ho about instantly exposing dirty laundry? Who gave you the language and the access codes etc to get inside THREE GENERATIONS of ASMedia chipsets? Who gave you the methodology to PARK A PERMANENT, HIDDEN MALWARE INSIDE THE ASMEDIA CHIPSETS ???? |
|
Title: Re: AI phase-in rocks all computing worlds Post by Oldfeller on 03/17/18 at 04:15:40 OK, now what does this all shake out to mean? All PCs are fairly insecure spaghetti strainers, they all leak badly because of construction choices made way back at the dawn of the Wintel/IBM era that are all still around to haunt us. Yes, there are multiple backdoors and various ways inside the physical construction of a modern PC machine's boot cycle, most egregious are the secret pathways that Microsoft had built into it to be used to update your machine at night even though the machine was turned off (and yes, if they want to they can cause your machine to boot up secretly showing no activity lights just as long as it is connected to the web through a internet service and a router and both have power to them). You are now being set up by the Wintel boys to make you want to go buy an expensive new PC just so you can get some small sense of security and some small part of your old operating speed back. Do not go buy another PC !!! --- spend your money on the next wave of tech which is defining itself as you watch. |
|
Title: Re: AI phase-in rocks all computing worlds Post by Oldfeller on 03/19/18 at 05:10:13 Fuchsia Friday https://9to5google.com/2018/01/23/what-is-google-fuchsia-os/ Fuchsia Friday has morphed a bit from a programmer's discussion site to a programmer training sort of structure. Each post has a discussion below it, look there for the programmer folks seeking advice on whichever/whatever. And yes, your device (Android, Chrome, Linux, Windows, ARM, Intel) can read and execute Fuchsia code items if you have the resources loaded. And they tend to go get themselves as needed if you don't have them on your machine already. This is one of Fuchsia's neat tricks --- seamless resources. Go get it, use it, and then to put it away so as to not jam up your hard drive. bing, bada boom, done Is it like Windows? No -- nor is it overmuch like Linux either. More like Android/Chrome and the web pages having a love child together. Different, but phone/Chrome familiar (sorta)..... For those who are curious (copy it up to your search bar) Fuchsia Friday: Ledger picks up where you left off Fuchsia Friday: Everything is an Entity Fuchsia Friday: A system built for ‘Instant Apps’ on steroids Fuchsia Friday: The structure of Google’s Lego-like modular OS, explained Fuchsia Friday: How Flutter is paving the way for Fuchsia (and our first Fuchsia app!) |
|
Title: Re: AI phase-in rocks all computing worlds Post by Oldfeller on 03/20/18 at 05:39:29 https://www.anandtech.com/show/12509/xilinx-announces-project-everest-fpga-soc-hybrid http://https://images.anandtech.com/doci/12509/everest_tilted_678x452.png OK, so what is this ??? It is a signal that the buy-out wounding of Qualcomm and the complete non-starter recent performance of Intel has prompted some of the AUTOMOTIVE people to step up and offer their automotive standards based product to both the rackspace data center people and to the PC people as a replacement tech for Intel. Xilinx is offering a set of automotive based stuff that does all the modern items, AI, etc. etc. etc. and this new Everest system supposedly can make up a PC competitor right out of the gate. Or a rack processor. Or whatever the heck you need it to be ...... (it can be reprogrammed on the fly after all). Somewhat similar to ARM DynamIQ, actually, except it has a larger host of support chipsets available including heavy use of large FPGAs and larger AI blocks in addition to CPU cores and GPU sets. But different enough not to be prone to Spectre, Meltdown and the current list of specious stuff from CTS ...... Well, so now you got multiple people actually offering new 7nm hardware setups now and Google offering a cross system OS under development now too --- both that do all the modern things (including concurrent self-maintenance) so I suspect that before too awful long somebody will put out an actual product that will likely fail to make Joe and Rita Six Pack happy for the first couple of times simply because it isn't Windows but may indeed eventually mature into a real replacement technology. Joe and Rita's kids and grandkids will understand it and "get with it" right away. More likely ARM Holdings will look at the Xilinx offering and thoughtfully tune their DynamIQ stuff up a bit -- and vice versa as any new standard replacement technology needs agreed upon cross industry standards and a cross-pollinated system is stronger than the ideas of just one company. Your modern operating system must work across systems seamlessly if it wants to play in the phone/PC/Automotive/IoT world we are headed towards. Google is actively seeking participating companies as they go about building Fuchsia, with the intent that it be a cooperative group-built standard from the very beginning. Change she comes ...... |
|
Title: Re: AI phase-in rocks all computing worlds Post by Oldfeller on 03/20/18 at 14:51:27 https://liliputing.com/2018/03/windows-10-arm-reviews-hint-at-good-things-to-come-but-mixed-bag-at-present.html http://https://liliputing-wpengine.netdna-ssl.com/wp-content/uploads/2018/03/novago_02.jpg Long story short ---- all the intentional Win 10 S restrictions mentioned by Microsoft in the last months are real and the Snapdragon 835 simply isn't nearly strong enough to make an appealing Windows 10 system out of those limitations. Looks like there will never be a third machine, a Lenovo machine and the other two out of the promised three Snapdragon 835 machines are definite disappointments at this point in time. https://www.laptopmag.com/reviews/laptops/asus-novago-tp370ql https://www.theverge.com/2018/3/20/17143554/microsoft-windows-snapdragon-always-connected-qualcomm-pc-review-asus-novago Theoretically Windows 10 on ARM is full-fledged Windows 10. But compatibility and performance issues mean that there are some things you can easily do on an Intel-powered computer that you wouldn’t do on a device like the Asus NovaGo. That may change in the future. Upcoming ARM-based chips are likely to be faster, and Microsoft is likely to continue improving Windows on ARM performance… assuming enough people buy the first-generation devices to justify building second-generation models. And right now it’s a little hard to imagine that this’ll happen. :P Snapdragon 845 doesn't look to be able to lift this load either. Intel has cheaper chipsets that can do Win 10 (full) and Intel can slap a modem on these units easily (they build modems now for Apple after all). Battery life is over 10 hours and that is good enough for Joe and Rita Sixpack as it does not require them to learn or do ANYTHING NEW ..... Win 10 on Snapdragon 835 is a Dead Horse, in other words ....... |
|
Title: Re: AI phase-in rocks all computing worlds Post by Oldfeller on 03/22/18 at 09:09:40 https://liliputing.com/2018/03/samsung-unveils-exynos-9610-chip-for-high-end-phones.html http://https://liliputing-wpengine.netdna-ssl.com/wp-content/uploads/2018/03/9610.jpg The Samsung Exynos 9610 features: Four 2.3 GHz ARM Cortex-A73 CPU cores Four 1.6 GHz ARM Cortex-A53 CPU cores ARM Mali-G72 graphics High Speed 4G LTE (support for speeds up to 600Mbps down and 150 Mbps up) 802.11ac WiFi Bluetooth 5.0 The chip also features an embedded ARM Cortex-M4F sensor hub to enable gesture recognition, context awareness, and always-on sensor features without waking the more power-hungry primary CPU cores. Other features include support for 480 frames per second slow-motion video at 1080p resolution, 4K video encoding and decoding at up to 120 frames per second, and a neural network engine for face detection and other camera effects. So, Samsung also lands on the "bump up the old cores with a modern AI boosted Mali G72 GPU plus various other adder blocks to do whatever else you want" bandwagon. This makes both Qualcomm and Samsung designing a "backed down" last generation big / little set up swinging a Mali G72 or other modern AI block graphics processor plus adder blocks to satisfy those customers who are simply not going to buy new 7nm lithography based cores for another year or more, but who still want to participate in the power increases that are coming from the new AI usage. This "retooling the old stuff methodology" apparently is indeed a speedy, low cost fairly efficient way to make up a pretty good phone SoC these days. The background idea is that last year's winner chipset is STILL A LOT MORE POWERFUL than what you really need, and when you bump it up with AI functions it really gets on up the power curve up towards the current crop of full on primo chipsets. Understanding the cost/performance magic this trick can do, then understand that this trick is also going to be functionally slowing down the adoption of the latest styles of ARM DynamIQ processors (the Cortex A-75 bigs and the Cortex A-55 littles) to some degree. Until TSMC's 7nm lithography production is really truly up to full cost effective full on production rates the newest ARM cores will really struggle to price justify themselves against the current trick of adding AI blocks from the most modern Mali G72 set on to the two year older 10nm natural core designs that chip producers already own the licenses for. Look to see ARM Holdings to do an interim re-design and a license cost downshift of the A75 & A55 cores if the Asian folks simply won't buy these newest designs simply because of lithography adoption issues and the relative high license cost issues that are involved ..... ARM has had to do this before when lithography availability issues stalled them out for a year or so. And this stall out is always aggravated by Apple's habit of locking up all the initial production lines on that critical first year out of the gate of each big lithography shift ...... |
|
SuzukiSavage.com » Powered by YaBB 2.2! YaBB © 2000-2007. All Rights Reserved. |