SK Hynix Inc. today announced that the Company will construct a new semiconductor fabrication plant (or 'the FAB') at its headquarters in Icheon, Gyeonggi-do, to respond to growing demand for memory chips and to secure a future growth engine.

Construction on the 53,000 m² site in Icheon will begin late 2018 and is slated to be completed in October 2020. SK Hynix will invest 3.5 trillion won in the FAB. The production portfolio of the FAB shall be decided considering future market conditions as well as the Company's technology capability.

It's not just Micron, but also Korean DRAM giants Samsung and SK Hynix, that are the latest victims of large-scale industrial espionage by Chinese DRAM makers to steal vital DRAM intellectual property (IP), according to Korea Times. Today's DRAM makers build their products on IP acquired over decades, and that is time Chinese companies do not have, and aren't willing to license from established DRAM makers, either.

"Samsung Electronics and SK Hynix have become the target of industrial espionage by Chinese memory chip manufacturers. In semiconductors, patents are critical to the cost structure. The companies have to protect what they have spent decades building. The result is Chinese companies are attempting to infringe on Samsung and SK patents," said a Korean official involved in the investigation of IP theft.

Digitimes reports that Korean memory manufacturers Hynix and Samsung have both been hit by unstable yield rates for their 18 nm server DRAM production.

While the yields are claimed to be sufficient for notebook and desktop PC production, they are not good enough for server memory, which has higher quality requirements. Due to the shortage, Chinese enterprises like Alibaba, Huawei, Lenovo and Tencent are now switching to use 20 nanometer DRAM for their servers, which is in better supply. Other vendors have even requested that no more 18 nm chips are shipped by these Korean suppliers, in a bid to improve quality, which might take several months, but shouldn't have a significant impact on overall DRAM prices.

NVIDIA's upcoming GeForce GTX graphics cards based on the "Volta" architecture, could feature GDDR6 memory, according to a supply deal SK Hynix struck with NVIDIA, resulting in the Korean memory manufacturer's stock price surging by 6 percent. It's not known if GDDR6 will be deployed on all SKUs, or if like GDDR5X, it will be exclusive to a handful high-end SKUs. The latest version of SK Hynix memory catalogue points to an 8 Gb (1 GB) GDDR6 memory chip supporting speeds of up to 14 Gbps at 1.35V, and up to 12 Gbps at 1.25V.

Considering NVIDIA already got GDDR5X to run at 11 Gbps, it could choose the faster option. Memory remains a cause for concern. If 8 Gb is the densest chip from SK Hynix, then the fabled "GV104" (GP104-successor), which could likely feature a 256-bit wide memory interface, will only feature up to 8 GB of memory, precluding the unlikely (and costly) option of piggy-backing chips to achieve 16 GB.

In a bid to ensure sufficient supply of NAND flash memory to meet the growing demands of not just PC, but also smartphone markets, China's Tsinghua Unigroup and Intel are in talks to license-manufacture 64-layer 3D NAND flash, based on existing IMFlash Technologies designs. IMFlash is a joint-venture between Intel and Micron Technology. Tsinghua Unigroup is one of the biggest beneficiaries of the Chinese Government's ambitious plan to invest RMB 1 trillion (USD $158 billion) over the next five years, to increase China's semiconductor self-sufficiency to 70 percent, by 2025.

The move will significantly increase supply of NAND flash memory, and is seen as a market threat to Korean NAND flash giants Samsung and SK Hynix, and Japanese Toshiba. IMFlash Technology released its first 64-layer 3D NAND flash to the market in 2017, and is currently developing a 96-layer 3D NAND flash chip, which, along with newer 10 nm-class silicon fabrication process, could double densities over the current 64-layer chips.

SK Hynix Inc. today announced that the Company recently completed developing an enterprise SATA Solid State Drive (or 'eSSD'). With its 72-Layer 512Gb (Gigabits) 3D NAND Flash chips, the Company is paving the way for its full-fledged entrance to the high value-added eSSD market. SK Hynix combined the 72-Layer 512Gb 3D NAND Flash with its in-house firmware and controller to provide the maximum density of 4TB (Terabytes). SK Hynix makes the most of its 72-Layer 512Gb 3D NAND chips to double the biggest density of the SSD of the same size with 256Gb NAND chips.

A single 4TB SSD could contain 200 UHD (Ultra-HD) movies, each of which is generally as large as approximately 20GB (Gigabytes). The new eSSD supports sequential read and write speed of up to 560MB/s (Megabytes per second) and 515MB/s, respectively, and it can perform 98,000 random read IOPS (Input/Output operations per second) and 32,000 random write IOPS. SK Hynix also improved the read latency, which is of the utmost importance in eSSD performance. The Company is sampling the product to server and data center clients in the United States.

Sk Hynix has added to its product catalog single-die 16 Gb DDR4 memory chips, which should enable a two-fold increase in maximum memory capacity per single DIMM. This allows SK Hynix to sell same-capacity chips with fewer memory semiconductor dies, due to the increase in storage density, and to increase maximum memory capacity at the same memory die populations as before. The benefits are lower power consumption (due to the reduced number of memory dies to power), and the possibility of putting together either dual-ranked 64 GB modules, quad-ranked 128 GB LRDIMMs and octal-ranked 256 GB LRDIMMs. That last part is the most important: theoretically, the maximum amount of memory on top Intel or AMD server platforms could double, which could enable up to 4 TB RAM in EPYC systems, for example. And as memory-hungry as big data applications have become, there's ever need for higher memory capacity.

SK Hynix's 16 Gb DDR4 chips are organized as 1Gx16 and 2Gx8 and supplied in FBGA96 and FBGA78 packages, respectively. Current 16 Gb density speeds stand at DDR4-2133 CL15 or DDR4-2400 CL17 modes at 1.2 V. SK Hynix plans increase the available frequencies in the third quarter of this year, adding DDR4-2666 CL19 to the lineup.

SK Hynix, the other Korean DRAM and NAND flash giant than Samsung, updated its DRAM product catalog to reflect immediate availability of its 8-gigabit (1 GB) GDDR6 memory chips. The company is selling four SKUs, part "H56C8H24MJR-S2C" in 14 Gbps and 12 Gbps variants; and part "H56C8H24MJR-S0C" in 12 Gbps and 10 Gbps variants. The -S2C chips are more energy efficient, in achieving 14 Gbps at 1.35V and 12 Gbps at 1.25V; while the -S0C achieves 12 Gbps at 1.35V, and 10 Gbps at 1.25V. Unless NVIDIA decides that the GTX 1080-successor should feature 16 GB of memory, the company could be in the market for 8 Gb GDDR6 chips.

Meanwhile, rival Samsung announced that it began mass-production of 16 Gb (2 GB) GDDR6 memory chips, which should enable 16 GB of memory across a 256-bit memory interface, or 8 GB over a 128-bit interface. Samsung's latest chips not only have double the density as SK Hynix's, but also tick faster, at 18 Gbps, with a voltage of 1.35V. It is widely expected that GDDR6 will be the dominant memory standard for entry, mainstream, and even high-end graphics cards, which launch through 2018-19. NVIDIA is expected this year to launch its new "Volta" graphics architecture across various consumer-graphics market-segments.

Reuters reports that a sudden (if ridiculous) 5% drop in memory chip prices in Q4 2017 has brought revenue expectations and investors' profit measurements to a teetering halt. 5% may not look like much - it certainly isn't much when we look at the historic price increases that almost doubled the cost of DDR4 memory kits, as you can see in the PC Part Picker chart below. This memory module price chart doesn't include the 5% drop yet, probably because it takes time for memory chip pricing to materialize in end-user module pricing. But for investors, it's like a spark in a paper archive - it could signal an impending price decrease that would push all profit estimates out the window.

This 5% drop in pricing has prompted industry analysts to review their profit estimates for 2018, and expect that the memory industry's growth rate will fall by more than half this year to 30 percent. You read that right - investors are scared because growth rates will be 30 percent instead of 60 percent. Oh the joys of inflated pricing, and slower-than-usual ramp-up to keep demand higher than supply. The joys of economic capitalism, where prices for consumers go up, and an industries' value skyrockets by more than 70$ in a single year (2017).

It's been a couple years now that we've seen continuously increasing pricing of DRAM and NAND semiconductors. The price increase, which has been hailed and documented over, over, andover again (and there are way more articles on this subject here on TPU), follows reported increased demand which has failed to be accompanied by its respective manufacturing and supply ability.

However, reports that companies were planning on increasing production of DRAM and NAND below the expected increases in supply demand may have turned at least some regulatory eyes towards the issue. China's National Development and Reform Commission's Pricing Supervision Department (NDRC) said they are aware of the situation, how it could point towards price-fixing from the four major NAND production players (Samsung, Hynix, Micron and Toshiba), and are looking into the matter. "We have noticed the price surge and will pay more attention to future problems that may be caused by 'price fixing' in the sector," the official Xu Xinyu was quoted as saying in an interview to Chinese newspaper Daily China.

This news may really not come as such; it's more of a statement in logical, albeit unconfirmed facts rather than something unexpected. AMD is working (naturally) on a GDDR6 memory controller, which it's looking to leverage in its next generations of graphics cards. This is an expected move: AMD is expected to continue using more exotic HBM memory implementations on its top tier products, but that leaves a lot of GPU space in their product stack that needs to be fed by high-speed memory solutions. With GDDR6 nearing widespread production and availability, it's only natural that AMD is looking to upgrade its controllers for the less expensive, easier to implement memory solution on its future products.

The confirmation is still worth mention, though, as it comes straight from a principal engineer on AMD's technical team, Daehyun Jun. A Linked In entry (since removed) stated that he was/is working on a DRAM controller for GDDR6 memory since September 2016. GDDR6 memory brings advantages of higher operating frequencies and lower power consumption against GDDR5 memory, and should deliver higher potential top frequencies than GDDR5X, which is already employed in top tier NVIDIA cards. GDDR6, when released, will start by delivering today's GDDR5X top speeds of roughly 14 Gbps, with a current maximum of 16 Gbps being achievable on the technology. This means more bandwidth (up-to double over current 8 Gbps GDDR5) and higher clock frequency memory. GDDR6 will be rated at 1.35 v, the same as GDDR5X.

DRAMeXchange, a division of TrendForce, has come forward with the expected announcement that DRAM output in 2018 likely won't be enough to fully satisfy supply. This has been the case for some time now. However, what started with simple insufficient output that could contain the explosion of DRAM capacity in smartphones seems to now be turning into a conscious decision by the three top memory manufacturers. Samsung, Micron, and SK Hynix are seemingly setting output at a lower than required level so as to artificially inflate pricing due to low supply. TrendForce themselves say so, in that these suppliers "(...) have opted to slow down their capacity expansions and technology migrations so that they can keep next year's prices at the same high level as during this year's second half. Doing so will also help them to sustain a strong profit margin."

DRAM production is expected to increase by 19,6% in 2018; however, this ratio is lower than the expected growth in demand, which is being pegged at 20,6%. This means 2018 is likely to see increased constraint in the supply channels (whereas 2018 was actually expected to see a slight relief in supply issues). This means that pricing will either stabilize or tend to increase from current levels. To be fair, semiconductor production isn't as simple as hitting a "increase production by 10x" button; reports say that all three players are contending with insufficient room to expand output on their production lines, and getting a new production facility online isn't a trivial effort - neither in funds, nor on time. However... All involved companies would much rather keep prices as they are than see them being brought down by oversupply.

Toshiba today announced that it has signed an agreement with a consortium led by Bain Capital to sell its memory business for JPY 2 trillion (approximately USD $18 billion). The consortium includes Korean memory giant SK Hynix, which contributed JPY 395 billion (nearly 20 percent) of the consortium's investment toward acquisition of Toshiba Memory Corp. Apple and Dell are the other members of the consortium.

Toshiba Corporation (erstwhile parent of Toshiba Memory Corp.) continues to hold a 40.2 percent stake, which along with another Japanese company, Hoya Corporation, holding a 9.9 percent stake, ensure that Japanese firms hold 50.1 percent of the business, keeping Japanese regulators happy. The Bain Capital-led consortium will hold a 49.9 percent stake, ensuring that Toshiba Memory Corp. stays afloat, and away from rival Western Digital, which has dragged Toshiba to a multitude of international courts and arbitrators. Apart from SK Hynix, the Bain Capital-led consortium includes Apple and Dell, who feared they would lose DRAM and NAND flash price bargaining power if Toshiba Memory fell in the hands of Western Digital, which owns SanDisk.

Western Digital Corp. (NASDAQ: WDC) today commented on Toshiba Corporation's ("Toshiba") statement regarding the transfer of its interests in its NAND flash-memory joint ventures operated with Western Digital's SanDisk subsidiaries ("JVs"):

"We are disappointed that Toshiba would take this action despite Western Digital's tireless efforts to reach a resolution that is in the best interests of all stakeholders. Throughout our ongoing dialogue with Toshiba, we have been flexible, constructive and have submitted numerous proposals to specifically address Toshiba's stated concerns. Our goal has been - and remains - to reach a mutually beneficial outcome that satisfies the needs of Toshiba and its stakeholders, and most importantly, ensures the longevity and continued success of the JVs.

Furthermore, it is surprising that Toshiba would continue to pursue a transaction with a consortium led by Korea-based SK Hynix Inc. and Bain Capital Japan without SanDisk's consent, as the language in the relevant JV agreements is unambiguous, and multiple courts have ruled in favor of protecting SanDisk's contractual rights. We remain confident in our ability to protect our JV interests and consent rights."

DigiTimes is reporting, through "sources from the upstream supply chain", that AMD's current shortage of RX Vega cards to distribute to the retail market will continue at least until October. The tech reporting site says that sources are pointing towards the package integration of HBM2 memory (from SK Hynix or Samsung Electronics) and the Vega GPU (manufactured on Global Foundries' 14 nm FinFet) as being at fault here, due to low yield rates for this packaging effort. However, some other sources point towards the issue being with the packaging process itself, done by Advanced Semiconductor Engineering (ASE) through use of SiP technology. Whichever one of these cases may be, it seems the problem lies with AMD's choice to use HBM2 on their Vega graphics architecture.

As a footnote to its story, DigiTimes is also reporting that according to industry sources, NVIDIA has, in light of RX Vega's performance, decided to postpone the launch of Volta-based GPUs towards the first quarter of 2018.

The DRAM industry is experiencing an acute shortage of various classes of GDDR memory (graphics DDR), which could affect graphics card prices come Holiday. Supplier quotes for various graphics memory components have risen by as much as 30.8% in August, from an average of USD $6.50 in July, to $8.50. Top graphics memory suppliers Samsung and SK Hynix have committed a bulk of their inventories to manufacturers of servers and mobile handsets, which triggered the price rally. Samsung is the largest supplier of graphics memory, with a 55 percent market-share, followed by SK Hynix at 35 percent, and Micron Technology at 10 percent.

If there's one thing enthusiasts usually complain regarding SSDs is that here doesn't seem to exist any viable alternatives to a good old high-capacity platter-based drives. Where are my 2 TB SSDs for 200$? How can I possibly save my entire music library in this puny 512 GB SSD that has already cost me my arms and legs?

Well, those answers might not be coming anytime soon (even though the advent of QLC NAND might change that.) In the meantime, we can put our eyes on Viking Technology, so as to see that a halo SSD product can achieve amazing storage capacity in a standard form factor. The company has just started selling their UHC (Ultra High Capacity) Silo Series, which leverage SK Hynix's MLC NAND (if the capacity doesn't put your $ klaxons running, the use of MLC should.) The SSDs leverage a 6 GB/s SAS interface, and deliver 500 MB/s read and 350 MB/s write speeds. The manufacturer says these can sustain up to 60,000 IOPS on random reads, and 15,000 IOPS on random writes. These may sound low in high-performance terms (and they sort of are), but remember these are products geared for the enterprise market. Pricing wasn't (maybe smartly) disclosed. I believe I'd laugh maniacally should I know how much they cost. However, if you must have a high-capacity SSD, you know where to look for.

You must remember the ongoing house of horrors that is Toshiba's financial situation. Granted, it isn't that bad - the company is still managing to push the envelope on its semiconductor production business. Still, I'm sure the company would have liked to not be on the verge of selling out 20% of its memory business production stake - which is one of the company's most profitable divisions to begin with.

After announcing their intention to begin mass production of their latest 72-Layer 3D NAND Flash back in April, SK Hynix has now confirmed that it has entered mass production of the high density NAND modules. Apparently, SK Hynix has already achieved the much sought-after "golden yield" ratios, where the semiconductor yield is now at such a level that it is advantageous to finally enter mass production. Apparently, SK Hynix's leadership was fearful of not being able to achieve the golden yield in a timely manner after their announcement of the technology only three months ago; however, after its "management team and engineers repeatedly spent nights doing research, yield went up vertically and has become comparable to Samsung Electronics'" own yield - and as you know, Samsung is kind of the golden standard when it comes to NAND technology.

According to industry sources, SK Hynix is already mass-producing SSDs (Solid State Drives) with the company's own controllers and firmware which leverage this new 72-layer 256Gb NAND flash memory. This is a welcome change for the company which should allow it to increase revenue, since this is the first time controllers are developed in-house. The company is also said to be already producing eMMC (embedded Multimedia Card) for mobile devices based on this technology, with supply already arriving to its customers.

Ah mining. The revival of an old craze. Who doesn't want to make their room's temperature increase to insane levels over the summer in order to cash in on the mining wagon? Who doesn't want to pull their hardware by the ankles and wrists, stretching it in utilization so as to maintain the PoW (proof of Work) cryptographic security in cryptocurrencies? Apparently, a not insignificant number of users and would-be miners does want that. That has, in turn, placed a whole lot of pressure on the graphics card market from both AMD and NVIDIA, with prices climbing and skyrocketing for graphics cards in the $200-$400 price ranges, as you know. It remains to be seen whether the flow of new miners decreases somewhat now, considering the recent market correction (read: dip) in the cryptocurrency market value (down around 42% from the all-time high of 357€ [~$400] of June 12th.)

After ASUS, it would seem like it's Sapphire's time to try and sway miners from their consumer-oriented, gaming graphics cards, through the launch of five different graphics cards models especially geared for mining. These are currently available for pre-order on Overclockers UK, and there are five different products in total, one based of RX 560 silicon, and four different takes on the RX 470 silicon (no, that's not a typo; it really is the 400 series.)

DRAM prices have been high for quite some time now, due to a general increased demand over a slowly improving supply capability from manufacturers. Pricing of DRAM has been increasing (to the tune that if I wanted to double my memory capacity, I would have to pay double of what I paid a mere 11 months ago.) NAND pricing has been affected as well, with newer technologies such as 3D NAND not having a relevant impact on end user pricing as was expected, since tight supply and growing demand means process-level savings are dwarfed by the increasing prices on the balance of supply and demand.

Most of our woes can be traced back to high-end smartphones, which make use of up to 6 GB of RAM and have copious amounts of NAND memory. Now, reports are coming in that due to the iPhone 8's impending launch, supply is even tighter, with several firms being either unable to secure the amount of Ram they are looking for, or having to order in significant advance (futures speculation anyone?) Reuters is reporting that some clients have moved to 6-month supply agreements for their DRAM and NAND purchases, accepting higher prices than the customary quarterly or monthly deals, to make sure they get enough memory chips for their products.

Plextor today introduced the S3 series value SATA SSDs. The series has two main variants based on form-factor, the S3C built in the 7 mm-thick 2.5-inch form-factor with SATA 6 Gbps interface; and the S3G series, built in the M.2-2280 form-factor, featuring SATA 6 Gbps wiring. Both drives combine Silicon Motion SMI2254 controllers with SK Hynix planar TLC NAND flash memory. The S3C comes in 128 GB, 256 GB, and 512 GB capacities, while the S3G comes in only 128 GB and 256 GB.

All capacities of the S3C and S3G series offer sequential reads of up to 550 MB/s, while sequential writes are rated at up to 500 MB/s, 510 MB/s, and 520 MB/s, for the 128 GB, 256 GB, and 512 GB variants, respectively. 4K random read performance is rated at up to 72,000 IOPS, 90,000 IOPS, and 92,000 IOPS, respectively; and 4K random write performance at up to 57,000 IOPS, 71,000 IOPS, and 72,000 IOPS, respectively. Endurance is rated at 35 TBW for the 128 GB variant, and 70 TBW for the 256 GB and 512 GB variants. Available now in the EU, the 128 GB, 256 GB, and 512 GB variants are priced at 62€, 106€, and 213€, respectively (including taxes).

It turns out that the EVO X AMD Edition isn't GeIL's only AMD Ryzen-optimized DDR4 memory, with the company unveiling AMD Edition variants of pretty much all its DDR4 memory brands. This includes the EVO Potenza, EVO Spear, EVO Forza, Super Luce, Pristine, and DragonRAM series. Each of these "AMD Edition" branded memory kits has been stability-tested with AMD Ryzen processors, at their advertised clock speeds and timings. They come in a variety of speeds, including DDR4-2400, DDR4-2666, DDR4-2933, DDR4-3200, and DDR4-3466; in densities of 8 GB and 16 GB, making up 16 GB and 32 GB dual-channel kits, respectively.

It's not known if all of these are based on Samsung b-die DRAM chips. AMD spokespersons have publicly stated that Samsung b-die isn't the only DRAM chip needed for high memory clocks on Ryzen processors, and that even with older versions of AGESA, certain memory kits with SK Hynix and Micron DRAM chips are having some success in achieving high memory clocks. AMD is working to improve support for faster DDR4 memory kits through updates to AGESA, which will be dispensed through motherboard vendors to end-users, as motherboard BIOS updates.

South Korean DRAM and NAND flash giant SK Hynix updated its product catalog to feature its latest GDDR6 memory, besides HBM2. The company had April announced its first GDDR6 memory products. The first GDDR6 memory chips by SK Hynix come in 8 Gb (1 gigabyte) densities, and data-rates of 14 Gbps and 12 Gbps, with DRAM voltages of 1.35V. The company is giving away small quantities of these chips for product development, mass production will commence soon, and bulk availability is slated for Q4-2017. This would mean actual products implementing these chips could be available only by very-late Q4 2017, or Q1-2018.

A graphics card with 14 Gbps GDDR6 memory across a 256-bit memory bus (8 chips) features 448 GB/s of memory bandwidth. A card with 384-bit (12 chips), should have 672 GB/s at its disposal. Likewise, the 12 Gbps memory chips offer 384 GB/s in 256-bit (8-chip) setups, and 576 GB/s in 384-bit (12-chip) setups. Meanwhile, SK Hynix also updated its HBM2 catalog to feature a 32 Gb (4 gigabyte) HBM2 stack, with a clock speed of 1.60 Gbps. The 2.00 Gbps stack which featured in the Q4-2016 version of this catalog is no longer available. At 1.60 Gbps, a GPU with four stacks has 819.2 GB/s of memory bandwidth. A chip with two stacks, such as the purported "Vega 10" prototype that has made several media appearances, hence has 409.6 GB/s.

SK Hynix Inc. today introduced the world's fastest 2Znm 8Gb (Gigabit) GDDR6 (Graphics DDR6) DRAM. The product operates with an I/O data rate of 16 Gbps (Gigabits per second) per pin, which is the industry's fastest. With a forthcoming high-end graphics card of 384-bit I/Os, this DRAM processes up to 768 GB/s (Gigabytes of graphics data per second). SK Hynix has been planning to mass produce the product for a client to release high-end graphics card by early 2018 equipped with high performance GDDR6 DRAMs.

GDDR is specialized DRAM for processing an extensive amount of graphics data quickly according to what graphics cards command in PCs, workstations, video players and high performance gaming machines. Especially, GDDR6 is a next generation graphics solution under development of standards at JEDEC, which runs twice as fast as GDDR5 having 10% lower operation voltage. As a result, it is expected to speedily substitute for GDDR5 and GDDR5X. SK Hynix has been collaborating with a core graphics chipset client to timely mass produce the GDDR6 for the upcoming market demands.