HW News – 2.5nm Transistor, AMD Graphics Brain Drain, RTX 2060 | Nexus players


Engineers build 2.5nm transistor

Researchers at MIT, in collaboration with the University of Colorado, have succeeded in building a 3D transistor at just 2.5nm, using a manufacturing process that works at the atomic level. Researchers believe that to follow Moore’s Law, which some have declared dead, semiconductors will have to be manipulated atom by atom.

To do this, the researchers use 3D transistors (FinFETs), which stand up vertically like fins, and are thousands of times thinner than a human hair. To create the smaller FinFETs, they use a modified thermal ALE (Atomic Level Etching) technique. The result is FinFETs as narrow as 2.5nm and theoretically more efficient than their commercial counterparts.

The MIT website describes the process very well, writing:

“Microfabrication involves deposition (growth of a film on a substrate) and etching (etching of patterns on the surface). To form transistors, the surface of the substrate is exposed to light through photomasks having the shape and structure of the transistor. Any material exposed to light can be etched with chemicals, while material hidden behind the photomask remains. The advanced techniques for microfabrication are known as atomic layer deposition (ALD) and atomic layer etching (ALE). In ALD, two chemicals are deposited on the surface of the substrate and react with each other in a vacuum reactor to form a film of desired thickness, one atomic layer at a time. Traditional ALE techniques use a plasma with highly energetic ions that remove individual atoms on the surface of the material. But these cause surface damage. These methods also expose the material to air, where oxidation causes additional defects that adversely affect performance.

The researchers believe this method could be easily integrated into existing manufacturing processes, with real impact in the form of denser chips, higher performance, and increased yields.

For more on the intricacies of the process, check out the MIT report linked below.

Source: http://news.mit.edu/2018/smallest-3-d-transistor-1207

Intel: Sunny Cove, B365, Gen11 and Xe Graphics

Intel hosted its Architecture Day 2018, where they made several announcements, including the new Sunny Cove microarchitecture for Xeon and Core chips. Additionally, they made announcements for their brand new Xe Graphics (no new Arctic Sound yet) and x86 hybrid processors. While unrelated to Intel’s architectural day, Intel has also quietly released the B365 chipset.

Intel unveiled Sunny Cove, a sort of clean slate architecture for the company. For years, Intel maintained a ticking cadence, where a new architecture was tied to a new, smaller process node. Intel encountered a huge roadblock with the 10nm shrinkage, being stuck with 14nm and Skylakeesque architecture since 2014. This, coupled with AMD‘s resurging competitiveness, meant that Intel had to rethink its approach to designing the ‘architecture.

As such, Sunny Cove is the first of three new architectures on Intel’s roadmap that are designed to be portable between nodes, and will be Intel’s first gross increase in IPC since Skylake. Sunny Cove will debut on the 10nm node and is scheduled for the first half of 2019, although it is not clear which chips the architecture will accompany. Interestingly, Anandtech speculates that Sunny Cove cores paired with Gen 11 graphics will debut as Ice Lake.

The new Intel Gen11 graphics engine accompanies Sunny Cove. With Gen11, Intel increased the number of threads from 24 to 64, with compute performance increasing by more than 1 TFLOPS. The new Gen11 design will also support tile-based rendering and offer an Intel version of adaptive sync. Intel’s adaptive sync technology will not require any external hardware and will be compatible with FreeSync monitors.

Beyond Gen11 is Intel’s new Xe Graphics brand, which will take them into discrete GPU territory. Details are scarce and Intel more teased than anything else. However, it seems that with Xe, Intel is aiming for scalability, hinting that they intend to design GPUs ranging from integrated solutions to mid-range and enthusiastic discrete GPUs, even targeting data centers. . These designs will likely all be supported by the same architecture. Intel has also reaffirmed its intention to release discrete GPUs by 2020.

Finally, Intel released a new “x86 hybrid” processor, built with its new 3D chip stacking process, Foveros. Foveros will allow different matrices and silicon to be stacked on the same case, in conjunction with EMIB (Embedded Multi-Die Interconnect Bridge). This will allow chips to be built with a mix and match approach combining CPUs, GPUs, and different I / O elements with high speed interconnect.

Intel presented a working sample for an anonymous customer, in which they combined a Sunny Cove core and four Atom cores in a 10nm design, mated to a separate 22nm I / O chip connected via TVS (Through Silicon Via). Intel has announced plans to roll out a full line of products with chips designed by Foveros.

By the way, Intel also quietly released the B365 chipset. Like the H310C chipset, the B365 is a 22nm variant of the 14nm B360 and uses a Kaby Lake PCH. This is undoubtedly another move to free up fabulous space for the production of 14nm silicon, as Intel continues to struggle to meet the demand for 14nm processors.


LARP Deep Dive with David Kanter: https://www.youtube.com/watch?v=629r1Ud4Cro


Fact Sheet: New Intel Architectures and Technologies Target Expanded Market Opportunities

Rumor: AMD Vega II and New Vega 10 and 20 GPU

New reports that AMD have filed a trademark for a new Vega logo suggest the company is not done with Vega just yet. Looking at the logo, we could interpret it in two ways: either as Vega II (2), or as the Roman numeral VII, denoting a shrinkage of 7 nm.

The discovery of new device IDs in a Linux driver further fuels Vega’s refresh rumor mill. AMD has released fixes for the RadeonSI Mesa and AMDKFD / AMDGPU kernel drivers, where a total of 7 new Device IDs have been spotted by Phoronix – one new Vega 20 ID and six Vega 10 IDs.

While logos and device IDs can only lead us to speculation, it is clear that AMD is planning Something with Vega, presumably to exist alongside Navi. Hopefully we’ll find out more around CES 2019.

Source: https://trademarks.justia.com/882/10/n-88210086.html


Rumor: RTX 2060 Alleged 1920 Core Host, TU106

Videocardz would have “confirmed” the existence of a Gigabyte RTX 2060 card, launched early next year, according to the source. Leak claims RTX 2060 will boast 1920s Cuda kernels and silicon GPU TU106. The card is also said to have 6GB of GDDR6 memory, unsurprisingly.

Comparatively, the current crop of GTX 1060 runs 1280 Cuda cores, 6GB of GDDR5 or GDDR5X after a recent refresh. Still no word on Tensor cores or real-time raytracing for the 2060 cards. We, along with many others in the tech space, have speculated that the RTX 2060 will not offer this technology. Also, it’s entirely possible that none of this is accurate to begin with.

Again, we’re just around the corner from CES 2019, where a lot of these rumors will be confirmed or dismissed. Probably, anyway.

Source: https://hothardware.com/news/rumored-gigabyte-geforce-rtx-2060-oc-leaks-with-1920-cuda-cores

AMD Brain Drain to Intel

Intel has poached yet another source of talent from the AMD group, this time in the form of a certain Damien Triolet. Triolet, some may recall, was the former editor of the French publication Hardware.fr. Triolet specialized in GPU testing, which ultimately attracted AMD to hire him as a technical marketing manager.

In November 2018, it appears that Triolet joined Intel and now occupies a technical marketing position within Intel’s Gaming and Graphics department. Intel recently bought Ryan Shrout, formerly of PC Perspective.

Source: https://www.techpowerup.com/250509/ex-hardware-fr-gpu-editor-damien-triolet-jumps-ship-from-amd-rtg-to-intel

NVIDIA shares slide to half their value, investor wants out

We recently reported on Nvidia’s earnings and how they’re suffering from a self-induced ‘crypto hangover’. It looks like that hangover isn’t going away anytime soon. Nvidia put a lot of eggs in the cryptocurrency basket, unable to predict when that bubble would burst (as AMD did, for that matter) and ended up with excessive amounts of GPUs in the channel, in especially the GTX 1060.

As such, they had to temper their revenue expectations for the fourth quarter with a dose of that reality, cutting it down by over $ 1 billion. Subsequently, their stocks began to plunge. Well, that dive looks like it hasn’t found a low end yet, as Nvidia’s shares recently hit about half of their peak value – from $ 289.36 to $ 149.

Worse yet, one of Nvidia’s biggest shareholders, Softbank Group, wants to step down. According to a Bloomberg report, Softbank plans to sell next year, after Nvidia’s less than stellar stock performance. Softbank acquired a $ 3 billion stake in Nvidia in 2017, and if it chooses to sell, the investor could make a profit of $ 3 billion.

Source: https://www.bloomberg.com/news/articles/2018-12-11/softbank-is-said-to-plan-sale-of-stake-in-nvidia-next-year

That’s it for our HW News this time around. To catch up on past episodes of HW News, head over to our material section. Be sure to check out our recent deep diving videos with David Kanter and Scott Wasson. Also take a look at our brand new mousepad in the GN Shop or grab our newest t-shirt.

Editorial: Eric Hamilton
Host: Steve Burke
Video: Andrew Coleman, Josh Svoboda

Source link


Leave A Reply