TechInsights Publishes Preliminary Analysis of 3D XPoint Memory

Now that the Intel Optane Memory M.2 SSDs are readily available on the open market, anyone with an electron microscope and the skills to use it can begin to probe the secrets of 3D XPoint memory that Intel and Micron have been keeping tightly under wraps since announcing the new technology in August 2015. The reverse engineering experts at TechInsights have been doing just that, and they recently published their initial findings.

Die Size

With some of the first high-resolution die photographs of 3D XPoint, TechInsights has provided precise measurements of the die size and memory density. The 128Gb 3D XPoint die is 206.5 mm2, much larger than is typical for modern NAND flash or DRAM but comparable to Intel’s 128Gb 20nm planar MLC NAND. A large total die size is typical for Intel and Micron, as they have historically not catered to the mobile market with their NAND flash while competitors like Samsung and Toshiba have strived to ensure their flash will physically fit in devices like smartphones. (That trend is changing with the introduction this year of 64-layer 3D NAND where Intel and Micron are producing both a larger 512Gb TLC part and a smaller 256Gb TLC part.)

Bit Density

The Intel Optane SSD DC P4800X is using memory of similar density to the Intel SSD DC P3700 that it is displacing as the flagship of Intel’s SSD product line. When comparing similar chips, die size is a strong predictor of manufacturing cost, but 3D XPoint memory is quite different from NAND flash memory, both older planar NAND or newer 3D NAND. Still, there’s some value in noting that the P4800X is arriving with a price tag about 25% higher than the P3700 initially carried. This suggests that the manufacturing process for 3D XPoint is either more expensive than planar NAND or that 3D XPoint yields are not mature enough, but a lot of the markup can also be explained by the lack of high-performance competition for Optane SSDs.

TechInsights calculates that 91.4% of the 3D XPoint die area is occupied by the memory array itself. This is a much higher figure than for NAND flash, where the record is 84.9% for Intel/Micron 3D NAND with its “CMOS under the array” design that puts a large portion of the peripheral circuitry underneath the memory array instead of alongside. Samsung’s current 48-layer 3D V-NAND manages an array efficiency of just 70%, and 3D NAND from Toshiba and SK Hynix has been comparable. This means that once Intel gets around to increasing they layer count in future generations of 3D XPoint memory, they should be able to get much closer to the ideal capacity scaling than 3D NAND memory can currently achieve.

The analysis from TechInsights confirms that 3D XPoint memory is manufactured using a 20nm process, with the same pitch in both the bitline and wordline directions of the memory array. The DRAM market is only just moving beyond this milestone, so comparing the density of 3D XPoint to current DRAM highlights the fundamental capacity advantage 3D XPoint enjoys: around 4.5 times higher density compared to typical 20nm DRAM, and about 3.3 times higher than the most advanced 1Xnm DDR4 on the market. This gap is likely to widen with future generations of 3D XPoint.

The materials and construction of an individual 3D XPoint memory cell have not been fully analyzed, but it appears to be a phase change memory element with a doped chalcogenide selector switch. The 3D XPoint memory array is constructed between the fourth and fifth metal interconnect layers above the silicon die.

Autore: Billy Tallis AnandTech

Memo to the Prez: Germans Already Build Cars Here

It wouldn’t be a normal week, or day, if President Trump didn’t ruffle feathers. Here’s one affecting the auto world: In a meeting with top European Union leaders, Trump is said to have criticized Germany for allegedly dumping cars on the US. According to Der Spiegel, Trump said, “The Germans are bad, very bad. See the millions of cars they sell in the US, terrible. We will stop this.”

Der Spiegel said the comments were provided by people in the meeting, which included European Commission President Jean-Claude Juncker and the European Council President Donald Tusk. Later, however, Juncker publicly said Trump was “not aggressive.”

BMWs headed for export from South Carolina. 70 percent of BMW’s US production is sold elsewhere.

Poster child for good corporate citizen: BMW

To the extent the comments reported what Trump said, the President has an incomplete view of automaker manufacturing in the US and North America. German are already building cars in the US and in some cases exporting cars to other countries.

BMW’s plant in Spartanburg (Greenville), South Carolina, is the company’s largest in the world. Opened in 2004, it has produced 3.7 million vehicles and exported more than 2 million. It currently builds all BMW SUVs — X3, X4, X5, X6 — except the X1. It will also begin building the full-size BMW X7 in 2018.

In 2016, BMW built 411,171 vehicles in Spartanburg and exported 70 percent of them, or 287,700, worth about $ 10 billion. No US-flag automaker has that kind of export numbers. Capacity in Spartanburg has been increased several times, most recently to 450,000. Between BMW and the dozens of suppliers located nearby or along the I-85 corridor, upstate South Carolina has become comparatively affluent.

BMW just began work building a factory in San Luis Potosi, Mexico, to open in 2019. It will build the BMW 3 Series for the global market, not just the US, the company said in January. It was this facility that got Trump riled up in January, along with the factory Ford planned for Mexico and then scrapped under pressure from the then-President-elect. Some analysts say Ford CEO Mark Fields’ decision to back off from the $ 1.6 billion plan is one of the reasons Fields was dismissed this week, along with Ford stock price dropping 36 percent under Fields.

Trump in January said he wanted to slap a 35-percent tariff on BMWs imported from abroad. So far, he hasn’t followed up.

The Mercedes-Benz Alabama plant passed 300,000 production in 2016.

Other German automakers in the US

Mercedes-Benz has a factory in Vance, Alabama, near Tuscaloosa. It builds the GLE SUV (sibling to the E-Class sedan), the GLS SUV (S-Class), the C-Class sedan, and the GLE coupe using, Mercedes says, “a combination of American, Japanese, and German automotive best practices.” 2016 production amounted to more than 300,000 units, including $ 1 billion in exported vehicles. In comparison, MB’s US sales (including Sprinter trucks and Smart cars) was 381,000.

VW has had a plant in Chattanooga, Tennessee, since 2011, with a capacity of 150,000 cars a year, a figure it hit 2-3 years after opening. In the wake of dieselgate, VW sales have been softer. Audi has a new plant in the Americas, if not the US: In San Jose Chiapa, Mexico, that country’s first luxury-car factory. Its capacity is 150,000 vehicles a year, primarily for export.

In addition, every German automaker, along with most all Asian makers, have R&D facilities and design studios in Michigan, Silicon Valley, and/or Southern California.

Autore: Bill Howard ExtremeTechExtremeTech

Apple Neural Engine, un chip per intelligenza artificiale e realtà aumentata

Apple Neural Engine, un chip per intelligenza artificiale e realtà aumentata

Apple sarebbe al lavoro su un chip che sgraverà il lavoro del processore e della GPU nelle applicazioni di intelligenza artificiale e realtà aumentata. Sarà integrato nei futuri iPhone e forse in un prodotto nuovo.

Si chiamerebbe Apple Neural Engine il chip per l’intelligenza artificiale che la Mela starebbe sviluppando in proprio e che verrebbe inserito nei futuri dispositivi, iPhone compresi.

La fonte è Bloomberg che spiega come l’idea di Apple sia quello di spostare sul nuovo chip Neural Engine i carichi di lavoro più impegnativi che sono comunemente richiesti dai meccanismi di riconoscimento facciale e vocale, dalle applicazioni per la realtà aumentata e dai sistemi di visione artificiale (computer vision) in genere. Processore e GPU verrebbero notevolmente sgravati.

Apple Neural Engine, un chip per intelligenza artificiale e realtà aumentata

Grazie all’Apple Neural Engine, device come gli iPhone di nuova generazione potranno compiere un ulteriore balzo in avanti nell’elaborazione automatica migliorando allo stesso tempo le prestazioni della batteria e le performance complessive.

Non è chiaro se il Neural Engine possa fare il suo debutto già quest’anno ma sembra certo che i tecnici a Cupertino siano già al lavoro per i primi test sugli iPhone.

Tim Cook ha più volte osservato come la realtà aumentata sia al centro dell’attenzione per Apple e come l’azienda stia lavorando alacremente su questo campo. Non è quindi escluso che il Neural Engine possa costituire il pilastro portante di un prodotto innovativo che sarà presentato più avanti.


Thunderbolt diventerà uno standard aperto nel 2018

Intel ha annunciato che renderà disponibile il suo protocollo Thunderbolt 3 come standard aperto per l’industria il prossimo anno e lo implementerà direttamente all’interno dei suoi prossimi processori. La mossa del gigante di Santa Clara arriva in un momento di transizione verso lo standard USB-C, che è però anche il connettore scelto per ospitare il protocollo Thunderbolt 3.

Nel comunicato diffuso dall’azienda si afferma che “immaginiamo un futuro dove dock ad alte prestazioni che fanno uso di un solo cavo, incredibili fotografie e video in 4K, realtà virtuale che sembra vera e periferiche di archiviazione ultrarapide sono la normalità – un mondo dove un connettore USB-C rende possibile tutto questo oggi e per molti anni a venire”. Già da qualche tempo si sta delineando un indirizzo generale dell’industria IT riguardo gli standard di connessione delle periferiche: dalla moltitudine di qualche anno fa si sta passando ad un solo connettore che è in grado di gestire i compiti più disparati.

Se l’integrazione tra Thunderbolt 3 e USB-C non è, però, una novità assoluta, altrettanto non si può dire per quanto riguarda il rilascio delle specifiche complete con una licenza gratuita. Intel cerca, di fatto, di imporre il suo standard, similmente a quanto fece con USB quasi vent’anni fa (in quel caso riuscendoci, pur non essendo l’unica autrice). Il colosso in blu è già da tempo al lavoro con Microsoft, la quale ha portato la compatibilità delle periferiche Thunderbolt 3 in plug-and-play su Windows 10 con il Creators Update, ma le due aziende affermano che ulteriori progressi saranno fatti nei prossimi mesi.

Tra i vantaggi che potrebbero essere resi disponibili dal protocollo Thunderbolt 3, Intel segnala la possibilità di gestire direttamente flussi audio e video (come già avviene oggi sulla MSI GeForce GTX 1080 Ti Gaming X), di permettere il collegamento di periferiche quali le schede video o la gestione dei flussi audio e video necessari per la realtà virtuale con un unico cavo. Questo annuncio potrebbe rappresentare finalmente un impulso ad un’adozione più ampia anche dello standard USB-C che, nonostante prometta maggiore semplicità e maggiori funzionalità, stenta a decollare nell’adozione da parte dei produttori di hardware.

Autore: Le news di Hardware Upgrade