Listings in two benchmark databases reveal that Intel, or perhaps a company close to the chip giant, seems to be testing a rather mysterious unannounced graphics processor that has slightly more compute horsepower than the company’s Iris Xe Max discrete GPU for notebooks.
Someone has submitted benchmark results of what is called the Intel Gen12 Desktop Graphics Controller featuring 128 execution units (EUs), clocked at 1.40 GHz, and equipped with 3.03GB of memory. Given the ‘Gen12’ designation, the GPU is based on Intel’s Xe-LP microarchitecture as other Xe microarchitectures are referred to as Gen12.x in the drivers. Since Intel’s Xe-LP EUs are 8-wide, 128 EUs may be compared to 1024 stream processors of AMD or Nvidia GPUs.
So far, benchmark results of the Intel Gen12 Desktop Graphics Controller with 128 EUs have been submitted to Geekbench 5 and SiSoftware databases, as found by TUM_APISAK/Twitter. The part scored 9,311 points in Geekbench 5 and 82 points in Sandra.
To date, Intel has released three products featuring its Xe-LP graphics: the 11th Generation Core ‘Tiger Lake’ processor with up to 96 EUs at up to 1,350 MHz, the Iris Xe Max discrete GPU for notebooks with 96 EUs at up 1,650 MHz, and the Server GPU with 96 EUs at up to 1.1 GHz.
The Iris Xe Max with 96 EUs is advertised as a standalone graphics processor for laptops and Intel has never disclosed plans to make it available on the desktop market. Meanwhile, when Intel started its ‘Odyssey’ discrete GPU marketing campaign back in 2018, it clearly announced plans to offer graphics cards for desktops in 2020 and even demonstrated a small Xe Graphics DG1 development board at CES 2020. Keeping these facts in mind, we can speculate that the Intel Gen12 Desktop Graphics Controller with 128 EUs could be an entry-level desktop GPU.
(Image credit: Future)
Assuming that the Intel Gen12 Desktop Graphics Controller with 128 EUs runs at up to 1.4 GHz, its compute performance will be around 2.86 TFLOPS, about 13% higher when compared to that of the Iris Xe Max running at 1650 MHz (~2.53 TFLOPS). That said, it is hard to expect such a discrete GPU to beat gaming graphics cards running chips from AMD or Nvidia. Meanwhile, Intel could market it as a co-GPU for certain tasks in some niche markets.
Intel has not officially announced an Iris Xe-LP GPU with 128 EUs. In fact, it is not even guaranteed that the product will actually be released. Nonetheless, it looks like Intel has something in the works.
The global server market is growing while Intel is losing market share. This is the result of estimates by various market observers. International Data Corporation (IDC) estimates third quarter sales of server processors 2020 at 22, 6 billion US dollars, the equivalent of 18, 5 billion euros. Compared to the previous year, this corresponds to an increase of 2.2 percent.
Server with x 86 – processors from Intel (mainly Xeon) and AMD (Epyc) made around 21 billion US dollars 93 percent of the market. The small “Non-x 86” – remainder is mainly attributable to IBM systems with power processors and the expensive Z mainframes. But ARM servers are slowly making themselves felt.
According to IDC, AMD has reduced sales of its own Epyc processors within a year by 112, 4 percent increased. This fits in with AMD’s last quarterly report, according to which the Enterprise, Embedded and Semi-Custom division increased its sales to more than 1, 13 Has doubled billion US dollars – this also includes the console processors for the Playstation 5 and Xbox Series X as well as Instinct GPU accelerators.
Current figures on market allocation between IDC does not name AMD, Intel and other chip manufacturers. Mercury Research stated in November that AMD’s share in the pure server comparison Epyc vs. Intel Xeon SP last at 12, 1 percent was – with a strong trend upwards.
ARM grows Besides AMD’s Epyc, ARM server processors are also starting to nibble on Intel’s market share. IDC estimates that sales of ARM processors increased by more than 430 percent within a year, but starting from a small base per year 2019.
Amazon in particular is promoting the use of the ARM processors Graviton and Graviton2 developed in-house in the AWS cloud business, because the EC2 instances running on them are simply cheaper than comparable x 86 – instances. For example, according to a tweet from its CEO Don MacAskill, the photo service Flickr plans to put all cloud services that do not require GPU instances on Graviton machines.
An analysis by financial advisor Jefferies showed that in August 2020 around a tenth of all AWS instances were running on Gravitons. AMD’s share was accordingly above average at Amazon with 20 percent: Among the four large cloud operators Alibaba, Amazon, Google and Microsoft had of the chipmaker an average market share of , 8 percent. Intel came up with “only” 84, 3 percent.
Magnetic tape was invented back in 1928, 92 years ago, but it took years before it got commercialized for audio applications and then as a storage medium for digital data. By today’s consumer standards, magnetic tapes are slow, not convenient to use, and relatively expensive. But as magnetic tape prepares to celebrate its 100th birthday, it can offer capacities not available on modern hard disk drives. At least, IBM and Fujifilm believe that their newly developed Strontium Ferrite (SrFe) magnetic layer will enable LTO-8 tapes to store up to 580TB of data.
Data Storage Needs Increasing
Nowadays, data storage needs are increasing exponentially as numerous companies and organizations tend to collect big data for their services, devices generate zettabytes of data themselves, ultra-high-definition videos are gaining resolution and color depth (18K cameras are already here), and end-users are generating more data than ever. About 20% – 30% of the world’s datasphere — which is predicted to grow to 175ZB by 2025 — needs real-time processing, so it has to be stored on hard drives or solid-state drives. But anywhere from 70% to 80% of data is the so-called ‘cold data’ that is rarely accessed but still has value for businesses.
(Image credit: Rackspace)
For cold data, 3.5-inch hard drives featuring shingled magnetic recording (SMR) technology are good enough. Still, they can only store up to 20TB of uncompressed data, and at $500+ they are pretty expensive. In contrast, modern LTO-8 magnetic tape cartridges can store up to 12TB native or 30TB of compressed data and cost less than $90. With an attractive combination of capacity and price, magnetic tapes have outlived three types of diskettes, CDs, and DVDs.
Fujifilm, IBM Research, and Sony are still developing magnetic tapes and manage to hit new milestones rather regularly. Since 2006, IBM and Fujifilm have increased the capacity of an LTO-8 tape from 8TB to potentially up to 580TB in the coming years.
Strontium Ferrite Particulate Tape: 317Gb/inch2 Recording Density
In general, magnetic tape evolves like hard drives. Developers tend to reduce track width to increase recording density, reduce tape thickness to increase tape length within a cartridge, and introduce new methods to reliably record and then read the data. This time around, IBM and Fujifilm are talking about a 317Gb/in2 recording density, a 1255 meters tape length, and 580TB of capacity.
(Image credit: Fujifilm)
One way to increase recording density — and therefore capacity — is to switch to a new tape material. Today, most tapes use Barium Ferrite (BaFe) magnetic layer, but to move on to higher densities, they need to switch to something else. IBM Research and Fujifilm propose to use a magnetic layer consisting of Strontium Ferrite (SrFe) magnetic particles. The companies say that SrFe particles have higher magnetic characteristics and 60% less particle volume compared to BaFe particles, which enables 56.2-nm wide tracks as well as a linear density of 702Kb/inch. Furthermore, SrFe features a good recording/readback performance with low noise, and since it is an oxide, it is very stable chemically, which makes it a viable candidate for long-term data storage.
(Image credit: IBM)
Just like with hard drives, the right magnetic layer is only a part of the equation. Positioning read/write heads over 56.2-nm wide tracks on a tape that is streamed at a speed of approximately 15 km/h is challenging. To do so, IBM and Fujifilm developed a low friction tape head technology that can position its ultra-narrow 29nm-wide TMR read sensor at an accuracy of 3.2nm. The technology is enabled by a series of servo-mechanical innovations, including new servo controllers that support a new servo pattern that is pre-recorded in the servo tracks to assist the controller in maintaining a precise positioning of the heads relative to the tape as well as a new head actuator. Also, the two companies had to implement a new tape-handling method to control tape vibrations.
Tape in the Era of Exascale Datacenters
Neither Fujifilm nor IBM have revealed when to expect the commercial arrival of SrFe particulate-based LTO-8 magnetic tapes that can store up to 580TB of data, but the two companies are confident that the market for storage tapes will only grow in the coming years and new types of tape will be needed.
(Image credit: IBM)
Analysts from Synergy Research estimate that in Q3 2019, there were more than 500 hyperscale data centers in the world, with more than 150 under construction. These massive server farms need to store their cold data somewhere, and tape has numerous advantages over both HDDs and SSDs.
Firstly, at $0.002 ~ $0.007 per gigabyte (native/compressed), tape is cheap. Secondly, it does not require power when not in use. Thirdly, data recorded on tape will still be readable in 30 years if the tape is stored properly. Fourthly, tape can be air-gapped to protect sensitive data against unauthorized access or manipulation. Finally, an LTO-8 tape provides superior storage density over modern HDDs and SSDs.
Researchers from Fujifilm and IBM note that even at a 317Gb/in2 recording density, which will be commercialized years from now, tape is still way behind today’s hard drives (which are on track to hit 2Tb/in2 in the coming years with HAMR) as far as this metric is concerned. But they argue that this means that tape has a lot of headroom for growth and will therefore stay with us for decades to come.
In the interactive iX – online course DNSSEC in practice, the Unix, mail and DNS expert Patrick Ben Koetter teaches the basics of DNSSEC, TLS / SSL transport encryption and DANE. In practical exercises, admins and security officers learn how to implement DNSSEC and DANE with the name server bind9 and the mail server postfix.
DANE (DNS-based Authentication of Named Entities) is a protocol standardized by the IETF that uses the DNS security extension DNSSEC to ensure the identity of mail servers. This enables secure TLS / SSL-encrypted mail communication that cannot be attacked by inserting a manipulated X. 509 certificate.
The online course will take place on February 5th 2021 using video conferencing software. A browser and an SSH client to access the training systems are sufficient to participate. It is possible at any time during the course to ask the speaker questions and to exchange ideas with the other participants. The speaker Carsten Strotmann specializes in Unix systems, DNS, DNSSEC and IPv6 security. If you book by 8 January 2021 you will receive 10 Percent early bird discount. The course is limited to 20 participants.
Further information and registration :
DNSSEC in practice, online course, 5.2. 2020 (odi)
“Google has repeatedly and shamelessly broken antitrust and consumer protection laws. Its modus operandi is to monopolize and twist the facts,” says the Antitrust lawsuit launched Wednesday, “Google is using its position of power on all sides of the online advertising market to unlawfully exclude competitors. It also boldly claims that ‘we will never sell your personal information to anyone’ despite the entire business model of targeted advertising is. ”
With the on Wednesday In a US federal district court, ten US states are suing Google for abuse of its advertising monopoly, including insider trading with advertising space and the formation of a kind of cartel with Facebook. Google denies this, a statement from Facebook is not yet available.
Countless allegations The description begins 2008 shortly after the competition authority FTC gave Google the green light to take over the online advertising service provider DoubleClick. Google is said to have broken promises made to the FTC and the US Parliament soon. Since then, Google has been trading on all sides of the market: with advertisers who have to pay fees, with operators of websites and apps who also have to pay fees, and when processing advertising contracts via an exchange, where even higher fees are incurred.
On such advertising exchanges, advertisers bid in real time for the display of their advertisements in end-user browsers and apps. Initially, Google prohibited website publishers from working with more than one exchange. Because this reduced the income of the websites, they found another way: Header Bidding.
With header bidding, not an advertising server but a JavaScript in the user’s browser opens up the possibility for several advertising exchanges for to offer the advertising space. The website operator’s advertising server (usually the Google Ad Manager) then takes precedence. That was lucrative for the website operator, but Google saw its skins swim away
Since then, Google has been fighting header bidding with great effort and success. At first it allowed the use of several advertising exchanges – but is said to have secretly won his own exchange, even if an external bid was higher.
In order to be able to better use the advantages of the Java alternative Kotlin such as expressiveness, security and support for writing asynchronous code with coroutines for its internal gRPC infrastructure especially for Android and cloud development, Google implemented the implementation gRPC Kotlin built – and released as open source in April 2020. Now the RPC framework has reached the first major version and is being launched with revised documentation, numerous bug fixes and expanded platform support.
Version 1.0 of gRPC Kotlin not only benefits from Kotlin’s coroutines for reactive input / output when creating data transmission objects and the services surrounding them, but also uses the Flow API to process streams with reactive input / output. In order to optimize the generated code for Android clients in particular, gRPC Kotlin uses the lite mode of gRPC Java: It enables Protobuf compilers to generate smaller, more powerful classes. The artifact grpc-kotlin-stub-lite depends on the associated grpc-protobuf-lite from. The generated Kotlin stub client can be used in the same way as on the JVM.
It also works without a JVM The Lite mode also allows the use of GraalVM Native Image, so that developers can create applications that compiled from Ahead-of-time (AOT) without a JVM native images run out. This enables smaller applications to be built that require less memory and are therefore also suitable for environments with limited resources. An example project shows how a native client application can be created with an only 14 MByte executable that, according to Google, starts in a split second, Kontakt records to the server, makes a request and processes the response.
The gRPC (google / general-purpose Remote Procedure Call) developed by Google is an open source framework for calling up functions in distributed computer systems and is primarily used by developers for communication between microservices. It uses the HTTP / 2 protocol for bidirectional streaming and protocol buffers as Interface Definition Language (IDL) or as the underlying message exchange format.
gRPC is primarily aimed at web and Android applications, but works in principle cross-platform and is available for a number of programming languages - including Java, C ++, Go, Python and Ruby. The gRPC released as Open Source by Google 2015 has since proven itself as an “incubating project” at the Cloud Native Computing Foundation (CNCF).
Google sees the advantages of gRPC Kotlin not only on the client side, but also uses the RPC Framework also for backend services. With gRPC Kotlin, these applications should be particularly easy to package and make available in containerized form wherever Docker containers or JVM apps can be run – for example in Kubernetes clusters. However, Google is also targeting its own serverless computing service called Cloud Run, which was officially launched a year ago and promises a fully managed infrastructure for application operations.
Further information on version 1.0 of gRPC Kotlin can be found in the blog post about the publication of the major release as well as on the GitHub pages of the project.
Over the last few years we have witnessed the postponement of successive premieres at Intel many times – first, lithography was implemented with delay 14 nm, then the times of reheating 4-core cutlets came. In the meantime, the 10 nm process proved to be a complete failure, as evidenced by the Cannonlake series, which was introduced quietly and just as quickly removed from memory. Last year’s Ice Lake processors were far from perfect. Only this year’s Tiger Lake systems turned out to be noticeably better, but also in this case it is difficult to talk about the success of the lithography 10 nm SuperFin. There are also problems with the implementation of the Sunny Cove architecture to servers and more and more problems with the Ponte Vecchio accelerator. According to the latest reports, further delays of further premieres and a change of the CEO at Intel are planned.
Information has appeared on the network suggesting further Intel problems with the implementation of 7 nm lithography and fine-tuning the process 10 nm. There are further delays in both processors and graphics cards.
There are disturbing information on the network about Intel’s problems, which concern both the refinement of 10 nm lithography as well as the implementation of more advanced, 7 nm technological process. According to informants, 10 nm lithography still does not fully meet expectations, despite the fact that 10 nm SuperFin is much more polished than the base process. OEMs are not getting the required number of Tiger Lake generation processors 11, which delays the introduction of new laptop models to the market. In addition, Intel is expected to continue to face problems in the server sector, where Ice Lake-SP chips still fall short of expectations in terms of performance and energy efficiency. Officially, the new Xeons are expected to debut in the first quarter of the year, but unofficially there is talk of another delay between the middle of the second quarter and the beginning of Q3 2021. So we are talking about the interval between May and July. The next generation of Sapphire Rapids server chips is to have an even greater delay. Instead of the second quarter 2021, there is talk of an annual slip – until Q2 2022 of the year.
If the information about further delays Ice Lake-SP and Sapphire Rapids find confirmation in fact, it will be an extremely difficult situation for Intel. A competing company – AMD – attacks this market very strongly. Originally, Ice Lake-SP chips were meant to compete with AMD EPYC Rome. Meanwhile, the next generation of EPYC Milan will be released in a few months, and Intel still does not have any competitive processors. The annual delay of the Sapphire Rapids generation will mean that these systems will have to compete with AMD EPYC Genoa, already based on the rebuilt Zen 4 architecture. .
Yet worse news comes in the topic of the 7 nm technological process. The implementation of this lithography was officially delayed a few months ago. According to unofficial information, 7 nm will be finally delayed for another 6 – 12 months, until 2023 year. This will affect, for example, the Intel Ponte Vecchio accelerator, based on the Xe-HP architecture. Originally, most of the GPU elements were to be produced in Intel factories in 7 nm lithography. Most likely, the production of the entire accelerator will be transferred to TSMC. Intel also plans to transfer the production of other chips – in this case it is said that part of the orders will be allocated to TSMC and some to Samsung. All these problems are expected to lead to another CEO change at Intel – a decision on this subject is expected in the first quarter 2021 of the year. However, we emphasize once again that all the above information is unofficial. Intel has not addressed them in any way at present.
Lots of news from Intel’s storage division , both for the Optane range and the classic SSDs based on 3D NAND memory. The new Optane SSD P 5800 X stands out, a very fast drive based on second generation 3D XPoint memory aimed at datacenter.
by Manolo De Agostini published 16 December 2020 , at 15: 53 in the Storage channel Optane Intel
During the event Memory and Storage 2020 , Intel unveiled six new storage products . We are facing the first big announcements since the US company announced the sale of the storage division to SK hynix. We must not forget that it will be a not rapid “posting”, but it will take several years: for this reason the division will continue to work “as if nothing had happened” even during the transition period.
Optane SSD P 5800 X : a very fast SSD for datacenter
To stand out among the news is undoubtedly Optane SSD P 5800 X , defined as “the fastest data center SSD in the world” : with interface PCI Express 4.0 x4 , a new controller and the second generation 3D XPoint memory , offers three times higher performance compared to the previous generation Optane SSD DC P 4800 X.
More precisely, P 5800 X reaches 1.5 million IOPS with cas loads which 4K and sequential performance of peak equal to 7.2 GB / s . The unit is also capable of achieving performance with random loads 400 B up to 4.6 million IOPS and ensure a maximum of 1.8 million IOPS with mixed loads. The new P 5800 X is also capable of a write resistance up to 67% higher than P 4800 X, consequently touching a peak of 100 drive writes per day (DWPD) .
However, these are just some of the capabilities of this very interesting product for datacenters. Intel has improved the SSD’s quality of service (QoS) over the previous model and worked to ensure consistent performance even under very heavy loads. The P 5800 X is capable of sustaining a very high load throughput of 8GB / s. Furthermore, the new SSD ensures a latency of only 66 us.
Optane SSD P 5800 X is also able to take advantage of the bi-directionality of PCI Express, which is very useful in mixed workloads. In a scenario with reads and writes respectively at 70% is 19%, the SSD is capable of touching 2 million IOPS, 3.7 times higher than the previous generation model.
Intel has not yet released information on pricing or availability , but has announced that the new Optane will be available in the form factor U.2 with capacity of 400 GB, 800 GB, 1.6TB and 3.2TB. It is plausible that the new Optane SS DP 5800 X arrives on the market together with the new Intel Ice Lake server platform, the first to support PCIe 4.0 , expected in the first quarter 2021. Obviously the SSD can be installed without any problem even on systems based on AMD EPYC CPU.
Optane Memory H 20 for reactive notebooks
Still within the Optane family, but very different than the P 5800 X, it is the new Optane Memory H 20: it is a SSD in M.2 format that matches 3D NAND QLC to 144 3D XPoint memory layer managed by a renewed controller. Evolution of model H 10, fuses the storage capacity of NAND with the speed and low latencies of 3D XPoint, so as to allow notebook manufacturers (will not be sold in stores) to make responsive, fast and light products.
The debut of Optane Memory H 20 is expected for Q2 2021 in two versions with 512 GB or 1 TB of 3D NAND flanked by 32 GB of 3D XPoint memory . Optane Memory 20 is accompanied by a PCI Expres 3.0 interface and we will find it on board products with Intel Core CPU of 11 th generation of the U series and series chipset 350.
SSD D7-P 5510 and D5-P 5316, 3D NAND a 144 layer for datacenters
Intel has renewed the patrol of SSD for datacenter with new solutions based on 3D NAND memory a 144 layer , of the TLC type in the case of the D7 and in the QLC meaning in that of the D5. However, the two share the same controller and support PCI Express 4.0. The D7-P 5510 is presented in U.2 format with capacity of 3, 84 and 7, 68 TB , but also a series of specific news for the world of the cloud and virtualized environments. It will be available by the end of the year, obviously only for companies in the sector.
Model D5-P 5316 will arrive instead in the first half of next year with the aim of taking the place of hard drives. Available in capacities of 15, 36 is 30, 72 TB, the new SSD with 3D NAND QLC memory will be available in U.2 and E1.L formats (in quest ‘last case will allow you to insert 1 petabyte of storage in a 1U rack).
Intel SSD 670 p, QLC memory to 144 layer the consumer area
In the first quarter of next year Intel will introduce the SSD 670 p, a consumer model characterized by a new controller compared to the previous solutions and from the memory 3D NAND QLC to 144 layer . Also in this case it is a SSD M.2 PCIe 3.0 , which will come in capacities of 512 GB, 1 TB and 2 TB . Price and performance will be shared by Intel near the debut, but the company has made it known that it has made some changes regarding the functioning of the dynamic SLC cache.
In practice, while the maximum and minimum sizes will not change from the past, Intel has succeeded in improve available cache size when drive is partially full : a 670 p with half the capacity occupied will have almost the maximum size of the SLC cache available. Furthermore, the minimum size will not be reached until the drive is full at the 85%. A change that should be useful to sustain performance in some situations where short and fast writes are required.
Optane Persistent Memory, “Crow Pass” is coming
Finally, Intel also talked about the future of Optane Persistent Memory , i.e. the memory modules that fit into the classic DIMM slots of the DRAMs but, unlike the latter, are able to store much more data and keep them even in the absence of energy. Intel has made it known that after the proposals of the series 200, codenamed Barlow Pass, it will be up to the “Crow Pass” generation to support future Xeon “Sapphire Rapids” processors . There will therefore most likely be a major change in terms of the interface, as future CPUs will support DDR5 memory.
As leading makers increase their 3D NAND bit output, prices of flash memory drop due to oversupply. This leads to sinking prices for solid-state drives as other types of NAND-based storage devices. Analysts from TrendForce predict that next quarter prices of client and server SSDs will drop by another 10% to 15% compared to Q4 2020.
3D NAND ASPs Set to Drop
There are six major makers of 3D NAND and a number of niche/local flash memory suppliers on the planet today, many more than there are DRAM manufacturers. All of these producers compete fiercely against each other both for market share and for technology leadership.
(Image credit: TrendForce)
These companies regularly introduce new 3D NAND process technologies that increase the number of layers (i.e., bit density) and/or make new memory architectures (e.g., QLC, PLC) more viable for certain applications. In addition, leading makers methodically add 3D NAND capacity as they see increasing demand for their products. As a result, 3D NAND bit supply grows both as a result of technological advancements and because of new capacities.
Analysts from TrendForce estimate that client SSDs currently account for 31% of 3D NAND bit supply, enterprise drives consume 20%, eMMC/UFS storage devices use 41% of 3D NAND bits, whereas only 8% of 3D NAND bit output is sold as raw wafers.
TrendForce researchers claim that in Q1 2021 four 3D NAND makers — Intel, Samsung, SK Hynix, and YMTC — plan to expand their 3D NAND bit output, which will spark competition and will cause a 10% – 15% quarter-over-quarter decline of 3D NAND average selling price (ASP) in Q1 2021.
(Image credit: TrendForce)
Client SSDs to Get Cheaper
Traditionally, PC makers decrease production of computers in the first quarter due to a seasonal downturn. As a result, demand for components, such as memory and SSDs, drops and their prices drop too. But there are two other factors that will drive them down in the first quarter of next year, according to TrendForce.
(Image credit: Western Digital)
Firstly, PC makers have plenty of SSDs in their stocks already. Secondly, 3D NAND suppliers are actively sending samples of their 1xx-layer 3D NAND to SSD makers thus energetically supporting launch of new drives.
Demand for computers is high and PC makers have backorders to supply, so their current inventories will likely be consumed to satisfy current demand. Nonetheless, new drives based on 1xx-layer 3D NAND will still be cheaper on per-bit basis than existing SSDs. In any case, TrendForce predicts that client SSD prices will drop by 10% to 15% in Q1 2021.
Prices of Enterprise SSDs to Drop Too
As the ramp of Intel’s Xeon Scalable ‘Ice Lake-SP’ platform (also known as Whitley Gen 2) comes closer, server makers are beginning their transition from existing PCIe 3.0-based drives and intensify testing as well as certification of PCIe 4.0 SSDs from Intel, Samsung, and Kioxia, reports TrendForce.
(Image credit: Micron)
The analysts expect that as competition on the enterprise SSD market intensifies in Q1 2021, prices of appropriate products will decrease by 10% to 15%.
eMMC/UFS Prices to Decline
Smartphones, tablets, and other types of portable electronics (such as low-end PCs) are usually among the first to start using new types of 3D NAND. Producers of eMMC and UFS storage devices usually use new types of memory to increase capacity of their devices, but this naturally boosts bit supply and reduces per bit prices.
(Image credit: Samsung)
Suppliers of eMMC and UFS plan to start transiting their products to 1xx-layer 3D NAND only after Q2 2021 in a bid not to oversupply. 64GB and higher capacity devices already use 92/96-layer 3D NAND, so without a new transition there will not be an oversupply in terms of bits.
As a result, TrendForce believes that due to dropping demand for everything towards the end of Q1, prices of mobile storage devices will drop by 5% to 10%.
Prices of 3D NAND Wafers Set to Drop
Many types of 3D NAND based products, including SSDs, flash drives, and memory cards, are in a state of oversupply, according to TrendForce. However, since it is difficult to get components, which are made by foundries, 3D NAND flash suppliers are diverting their production to wafer market.
(Image credit: Micron)
According to the report, Samsung, the company that has historically focused on actual products and not on selling raw NAND, has begun shipping its 92-layer 3D NAND wafers to module houses. Western Digital, another company that tends to use its 3D NAND internally, has also started sampling of its 112-layer 3D NAND with select customers. Meanwhile, Kioxia continues to supply its 96-layer 3D NAND to clients, whereas SK Hynix is sampling its 128-layer memory.
Aggressive selling of high-density 3D NAND wafers to third-party SSD makers creates bit oversupply on the wafer market, which will cause a 15% quarter-over-quarter price drop in Q1 2021, the analysts predict.
Flash memory is fast, but Intel’s Optane memory is faster – Intel was already able to do that with the Datacenter SSD P 4800 X show. Now comes the much faster successor, according to Intel at three times the speed: The P 5800 X uses Optane of the second generation and PCIe 4.0. In terms of sequential transfer rates, other SSDs are a bit further ahead, but Intel sets new standards when it comes to accessing random addresses: The company promises 1.5 million IOPS for exclusive reading or writing. In mixed operation, i.e. when reading and writing at the same time, it should even be 1.8 million IOPS.
At the same time, Intel promises a durability of 100 Drive Writes Per Day, the SSDs may therefore be written with their own capacity one hundred times a day within the framework of the five-year guarantee. The SSDs will come onto the market in the coming year, Intel has not yet commented on prices.
More data center SSDs Intel plans to introduce the data center SSD D7-P 5510 launch. It is equipped with TLC flash in 144 layers and should be up to 7, 68 TByte capacity. For larger space requirements, Intel will bring the P 72 in the next quarter: The SSD equipped with QLC flash should be equipped with up to 30, 72 TByte available.
If you buy 30 of it in the E1.L Format, a 1U server rack can be equipped with up to 1 PByte of memory – that is exactly what Intelv promised when introducing this “Ruler” format.
QLC has prevailed QLC Flash is also used in Intel’s desktop SSDs. The 670 p is the successor of the 665 p and is only offered as M.2 cards. It also uses the current 144 – layer version of the memory; Information on capacities and prices is not yet available.
Intel already had a combination of QLC flash and Optane memory with the H 10 in program. Here Optane serves as a fast cache for the rather slow Flash. With the H Intel used two PCIe lanes for the Optane part and the other two as access to the QLC Flash, retrofitting in existing notebooks was not possible. Whether it will be different with the H 10, which is equipped with the newer versions of the memory is yet unknown. (ll)
A few days ago, attacks on various US authorities became public. The government-sponsored group, which is also said to be responsible for the successful break-in at the IT security company Fireye, used infected updates for the Orion network management software from SolarWinds as an attack tool. She smuggled this, provided with a valid digital SolarWinds signature, to the company’s update server and from there to the target systems.
This report summarizes new information, including on possible points of attack and protective measures taken. It complements the reporting of the past few days:
Cyber attacks against US ministries – Moscow under suspicion Trojans in SolarWinds updates allow cyber attacks Period of attack and number of potentially affected persons In a so-called “FORM 8-K”, a standardized report, investors and the public about important events informed, SolarWinds has provided information on the period in which manipulated Orion updates were delivered, as well as the number of potentially affected customers. The SolarWinds report cites the period from March to June as the “relevant period” 2020; This largely coincides with the findings from a detailed analysis by FireEye.
According to SolarWinds, products that were downloaded or implemented before or after this period did not contain any malicious code. Observations that security researchers shared via Twitter contradict this representation, however: Shared screenshots and links are intended to document that malware-infected “updates” were still available via the company’s download portal last Monday, i.e. after the incidents became known.
According to SolarWinds, Orion’s software build system – and not the source code of Orion products – has been compromised and tampered with. The company has taken countermeasures and continues to investigate the incidents. There are currently no indications that other products could be affected. During and after the “relevant period” would have approximately 33. 000 Customers using the Orion platform; However, the company assumes that less than 18. 000 customers would have used a manipulated Orion installation.
In the meantime, the company has published two hotfixes for the Orion platform.
Older access data leak as an initial point of attack? How the attackers initially gained access to the SolarWinds network infrastructure and ultimately to the build system is still unclear. SolarWinds is investigating, among other things, whether an undiscovered vulnerability in Orion products could be responsible, according to the report.
Interesting in this context are the observations that security researcher Vinoth Kumar shared via Twitter : He pointed to a public GitHub repository that apparently contained plain text FTP access data for a SolarWinds software update server. Kumar told The Register that SolarWinds was already on 19. November 2019 to have notified the data leak by email. Using the (extremely insecure) password “solarwinds 33” he said that it was vulnerable uploaded a file to the server. In addition, he even warned the company in his email that criminal hackers could use this method to upload a malicious .exe file as a fake update for SolarWinds products. The Register article does not give details on the GitHub repository.
(Image: @vinodsparrow via Twitter)
The company has its E -Mail on 22. November 2019 answered; at the time, the access data had been available for at least two to three weeks. The email that Kumar shared as a screenshot via Twitter states that the repository is no longer publicly accessible. In addition, the leaked credentials were “taken care of”.
It is still unclear whether “take care” is to be equated with “change” or even with “use stronger passwords in the future and avoid leaks” . A statement from SolarWinds on whether there is any connection between the leak from last year and the current incidents is not yet available. In any case, Kumar’s description does not cast a positive light on SolarWind’s previous handling of the topic of network security. And it raises the question of whether the success of the attack, however sophisticated it may actually have been, is not also due to some extent to negligence.
As with the valid digital signature of the manipulated Updates came, is still unclear.
Microsoft takes over C&C domain IT security blogger Brian Krebs has pointed out that the domain avsvmcloud com , which the attackers used to communicate with compromised systems, was apparently taken over by Microsoft. The company has already taken control of malware domains in coordination with (US) security authorities, for example to help break up the Necurs botnet or to attack the infrastructure of the Trickbot gang.
Microsoft only confirmed the domain takeover indirectly: Senior Director Jeff Jones replied to Krebs’ tweet that everyone should do their part in the area of cybersecurity. (ovw)
The open source project Nexus has reached version 1.0. The declarative approach to creating GraphQL APIs should be stable for productive use with the new release, and there are some significant changes associated with the version jump. Nexus is now available under a new package name, and fields may by default have the value null .
Everything is code The Nexus Project relies on a code-first approach for the creation of GraphQL interfaces: The definition of the APIs takes place entirely via JavaScript, instead of describing the schemas and associated resolvers for using the endpoints, as is otherwise usual with GraphQL.
The approach requires a rethink at the beginning, especially for the definition of the schemes, but should be easier to maintain in the long run and for larger projects. One advantage is that schemes and resolvers are not separate from each other, but are in one file.
SDL and TypeScript from tape Nexus can automatically create suitable SDL files (Schema Definition Language) by calling the function makeSchema , which can be used, among other things, to provide the appropriate schemes for applications such as development environments. In the same block, TypeScript types can be created that ensure type safety when the interface is called from Microsoft’s programming language, as the following example from the blog post on the release of Nexus 1.0 shows:
import path from ‘path’ const schema = makeSchema ({types: , outputs: {schema: path.join (__ dirname, ‘generated / schema.gen.graphql’), typegen: path.join (__ dirname, ‘generated / nexusTypes.gen.ts’),},}) A web-based SDL converter helps with the reverse Way: converting SDL definitions to Nexus.
Zero is an option Version 1.0 brings a few innovations that can lead to incompatibilities. Nexus 1.0 can now be found in the uniform package called nexus and no longer in @ nexus / schema . Developers must adapt the import accordingly:
import {makeSchema} from ‘nexus’ In addition, fields are now allowed by default have the value zero . While this nullability previously required an explicit declaration, there is now a function for fields that cannot be null :
const Post = objectType ({name: ‘Post’, definition (t) {t.nonNull.id (‘id’) t.nonNull. string (‘title’) t.nonNull.string (‘body’)},}) In the schema definition language, the correspondingly declared fields are given a final exclamation mark to indicate that the value is not null may be:
type Post {id: ID! title: String! body: String! } The nullability can be customized globally for any API created with Nexus. If you decide to deactivate it by default, you can set the value null for individual fields using the function nullable .
Lists and abstracts Nexus 1.0 also introduces the new function list for processing lists as input or output values, for example to [String] to be defined as list (‘String’) .
In addition, can Nexus describe the abstract types of GraphQL in the form of unions or interfaces. To do this, it uses one of the three strategies integrated in the official GraphQL JavaScript package, Centralized, Discriminant Model Field or Modular. A detailed description of the individual procedures with examples can be found in the Abstract Types Guide on the Nexus site.
Source: GraphQL.org
GraphQL is a specification for describing APIs. It defines how clients can describe and execute queries and how, on the other hand, the server has to process the queries. With GraphQL, data changes can be read and written in real time.
Facebook had the query language 2012 initially developed internally and 2015 published . The SDL has been part of the specification since the beginning 2018, and the end 2018 the language got its own GraphQL Foundation under the umbrella of the Linux Foundation, which has been taking care of further development ever since.
More details about Nexus 1.0 can be found in the blog post. The full list of new features can be found in the changelog on GitHub. The source code of Nexus is also available there under MIT license.
Assembling a PC yourself not only has the great advantage of tailoring the hardware to your own wishes and requirements, but is also a lot of tinkering fun. The c’t special issue PC-Selbstbau presents four construction proposals, which you can use to put together inexpensive budget gamers, a flexible all-rounder or a home server. Furthermore, the booklet provides purchase advice, numerous tests of PC components and tips for the optimal hardware for the Microsoft Flight Simulator 2020 .
Assembling the Ryzen-Allrounder Plus The construction proposals from the c’t laboratory are suitable for a wide range of applications: The Ryzen Allrounder Plus is equally suitable as a fast everyday computer as well as with the optional 16 – Core processor, large RAM and high-end graphics card for a potent luxury computer. PC gamers can choose between two inexpensive budget gamers with AMD Ryzen or Intel Core i. The flexible home server not only accepts 99 TByte data, but also offers it some professional features like ECC-RAM and remote maintenance.
Hardware tests and purchase advice In order to keep track of the huge selection of attractive hardware, you will find in the c’t special issue PC-Self-Assembly, in addition to extensive purchase advice on common PC components, additional in-depth tips for the right graphics card, SSD and the optimal one Processor. You will also find out which applications and 3D games benefit from how many CPU cores.
Special issue PC-DIY 2020 (5 pictures) The special issue contains detailed test reports on processors, solid-state disks with SATA and NVMe interfaces with up to 4 TByte capacity and graphics cards. Using the mainboard tests, you will find the right foundation for current Core-i and Ryzen CPUs. In addition, the booklet provides suggestions for quiet, efficient PC power supplies of the 500 – watt class.
Printed and digital c’t self-made PC is available on paper as well as also in digital form. Who the print edition for 14, 90 Euro up to and including 29. 12. 2020 in the heise shop ordered, pays no shipping costs.
The digital version as PDF is available immediately for 12, 99 Euro in the heise shop, in our Android and iOS apps or on Amazon.
The printed booklet and the digital edition are also available as a bundle for 19, 90 Euro. (chh)
The PICMG (PCI Industrial Computer Manufacturers Group) has published a preliminary version of its specification Computer-on-Modules for High-Performance Computing COM-HPC. COM-HPC defines Computer-on-Modules in five module sizes for powerful edge servers and clients in the industrial environment. It should take into account the growing demand for computing power in the edge and provide significantly more computing power and I / O performance for edge servers and edge clients than the previous x 86 – Computer-on-Module specification COM Express capable.
COMe was 2003 introduced by Intel and Kontron as ETXexpress and 2005 as PICMG specification COM Express Revision 1.0 released. It defines a bootable PC with CPU, graphics processor, RAM and standard interfaces on a module that is connected to an application-specific carrier board via a maximum of two connectors.
COM-HPC is intended to supplement COM Express, but is not limited to x 86 processors. Instead, it also allows the use of RISC processors, such as ARM CPUs, but also FPGAs (Field Programmable Gate Arrays) and GPGPUs (General Purpose Computation on Graphics Processing Units).
Significantly increased All modules are connected to the carrier board via two 400 – pin BGA connectors that transfer a maximum of GT / s (Gigatransfers) and have a system management interface. Up to eight full-size DIMMs or four SO-DIMMs and a power consumption of 150 watts by the CPUs are permitted. COMe modules are on the other hand on 10 GT / s, two SO-DIMMs and 80 Watts limited.
COM-HPC also significantly expands the limits for I / O: The COM-HPC client modules can have up to 32 PCIe Gen4 or Gen5 lanes four USB 3 or USB 4 ports, four video interfaces and two Ethernet interfaces with a maximum of 25 Record gigabit / s. The specification sees the module sizes as 95 or 120 or 160 mm x 120 mm in front. The COM-HPC server modules may have up to 49 PCIe Gen4 or Gen5 lanes, two USB 3 or 4 USB ports, four graphics interfaces and eight 25 Gigabit Ethernet interfaces , also two GBase-T interfaces. The server modules may 160 or 200 mm x 160 mm in size and therefore significantly larger than COMe modules, for which, for example, only 32 PCIe Gen3 lanes, one 1GBase-T and four 10 may have GE interfaces. Support for PCIe targets on the module, for example for FPGA or GPU modules, is new.
The PICMG develops patent-free, open embedded computing specifications . Now that its technical subcommittee has approved the COM-HPC specification, it goes into the membership review with the pre-release. This review phase acts as the final review by the PICMG members. They include IT companies such as AMI, Intel and Supermicro as well as industrial IT providers such as ADLINK, Advantech, Amphenol, congatec and Kontron, but also Bielefeld University. The ratification is planned for the beginning 2021, in addition a platform management interface specification, a COM-HPC EEEP and a carrier board design guide should supplement the basic specification.
The abridged preview specification is available for download.
This year, video games provided escapism for a lot of people. But for many, myself included, the games I remember most fondly in 2020 were the ones that became social spaces. The pandemic didn’t make me want to escape; it made me miss my friends and family.
Animal Crossing: New Horizons came out amid the shutdowns, and it quickly became my virtual island paradise, even when I wasn’t playing. I couldn’t log into Facebook or Twitter without seeing almost everyone on my feeds talking about Animal Crossing — people were sharing memes, asking for Dodo codes to visit their friends’ islands, begging for specific fruit that they wanted to grow.
Some turned to Nookazon, a third-party trading platform, to buy outfits they had a hard time finding. The latest Animal Crossing title ushered in a new community of people that were taking the game so seriously that someone without context could easily think all these people purchased some island real estate after reading a travel blog. It was the early months of 2020 and we were stuck inside. As it got warmer, Animal Crossing provided us with the tropical getaway we could actually visit. I even know people who purchased a Switch just so they could play this game.
One of my best friends lives in another state, but that distance has never stopped her from being the person I go to when I need to vent about life. So I begged her to buy a Switch so we could play Animal Crossing. She came to the game a few months after I did, and since I was so far ahead, we would FaceTime and I would guide her through the early tasks of the game, occasionally dropping off extra resources at her island if she needed to build the museum or wanted to decorate her house. We would constantly “hang out” on each other’s islands.
Since I had just started working at The Verge, I was constantly telling my friend my experiences of my new job and some of the weirdest news I saw that day. It felt like we were actually hanging out like old times, despite the fact we were states apart and unable to see each other because of the shutdowns and travel advisories.
Fall Guys.
Even at work, my colleagues were also playing. Our workplace is organized by Slack channels, so naturally, two were dedicated to Animal Crossing. One was a social space centered around the game, sharing advice, pictures of bugs, and organizing happy hour visits to someone’s island. (The Verge’s news writer, Jay Peters, once planned an extremely sophisticated scavenger hunt.) The other channel was all business, though: namely, colluding to manipulate the game’s “stalk market” so we would all get rich shorting turnip prices.
One week, I had an incredibly high turnip price on my island. I shared the number in the Slack channel and it blew up. I opened my gates to visitors during my lunch break and the demand of people far exceeded the time it took me to make and eat a peanut butter and jelly sandwich. I left my Switch on idle, went back to work, and returned later to collect two million bells’ worth of tips for my generosity. It felt like my birthday in a way, which was in late April and I, unfortunately, could not celebrate it with many people. So, I soaked in all the love and praise and endless thanks for being kind enough to open my gates and suffering through the hellish nightmare that is Animal Crossing’s multiplayer loading screens as travelers came in and out of my island.
For a minute, it seemed like everyone was talking about Animal Crossing: discussing favorite villagers, sharing tips on capturing bugs and fish. I was exchanging Dodo codes with people I went to high school with but would not have called friends — until they arrived on my island. We could all share the common ground of playing a low-stakes adventure. If the game is a satisfying loop, reconnecting with old friends and making new ones is the real reward.
Fall Guys was a game that came out during the dog days of summer, and word got out on how addicting its gameplay was thanks to streaming platforms like Twitch, which is the way games go viral these days. But most importantly for me, I have two younger cousins I hang out with frequently and they were looking for some new games to play since Fortnite was removed on iPad. So together, we let our asses get kicked by Fall Guys.
It was fun sitting with them on the couch at my aunt and uncle’s house. We would take turns seeing how far we would get in each match before we were smacked hard, knocked off the platform into a never-ending sea of Pepto Bismol, and subsequently eliminated. I am not a huge fan of battle royale games like Fortnite and PUBG, but Fall Guys is easily accessible for all ages and doesn’t have convoluted gameplay mechanics. Also, it’s hard to not constantly laugh at all the sounds these bean-shaped characters make as they waddle to the end of the course.
The most surprising game of 2020, and arguably one with a lot of staying power in the long term, is Among Us, a cartoony riff on Werewolf, wherein an imposter or two attempts to murder all their friends. Like Animal Crossing and Fall Guys, it’s a game the internet won’t shut up about. While it was originally released in 2018, Among Us’ popularity grew exponentially this year. My friends from college also started playing the game this year and a bunch of us ended up creating a Discord server dedicated to Among Us. We have channels dedicated to memes or finding new matches.
Among Us.
I think fondly of all the memories we have of hanging out in the college dorm rooms playing Mario Kart or compromising our friendships with an intense game of Mario Party. With college now over and most of us living pretty far away from each other, I missed the days we could huddle in the dorms and play the GameCube classics.
Now we’re doing the equivalent of that in Among Us, getting nostalgic over stories from college while debating who the imposter is. Over the last few months, I would often play this game after work, for hours on end, until I realized it was midnight. It’s in those moments that reality kicked in and I realized I was not a 21-year-old who would stay out with my friends partying until 4AM; I am a 25-year-old adult with responsibilities and a full-time job. My priorities have changed, but making time for my friends will never get old no matter the circumstances.
But the common theme here for all these games is the sense of normality they have provided in my life. While I enjoy happy hours with my colleagues and weekly family Zoom meetings, these activities grew tiresome for me. These things serve as a painful reminder of what life was like before the pandemic. I miss going to rooftop bars after work and unwinding from work mode; I miss going to my grandparents’ house on the weekends and having cookouts and big Sunday dinners. I love my family, and I love my colleagues, but I craved the more interactive and creative conversation that games offer.
I was playing games with friends and family before the pandemic began, and I play a lot of games solo too. But Animal Crossing, Fall Guys, and Among Us are good-ass games that continued to get better as I enjoyed them with familiar company. They allowed me to make up for the lost time that was part of my old life.
The point of going out to bars was never the drinking. It was a common social space, and in a year where those places disappeared, we found new ones in fake islands and spaceships, and whatever the setting of Fall Guys is supposed to be. Especially in 2020, it’s not about where you are, but who you’re with.
We use cookies on our website to give you the most relevant experience. By clicking “Accept”, you consent to the use of ALL the cookies.
This website uses cookies to improve your experience while you navigate through the website. Out of these, the cookies that are categorized as necessary are stored on your browser as they are essential for the working of basic functionalities of the website. We also use third-party cookies that help us analyze and understand how you use this website. These cookies will be stored in your browser only with your consent. You also have the option to opt-out of these cookies. But opting out of some of these cookies may affect your browsing experience.
Necessary cookies are absolutely essential for the website to function properly. This category only includes cookies that ensures basic functionalities and security features of the website. These cookies do not store any personal information.
Any cookies that may not be particularly necessary for the website to function and is used specifically to collect user personal data via analytics, ads, other embedded contents are termed as non-necessary cookies. It is mandatory to procure user consent prior to running these cookies on your website.
