After 4 or 5 years of development, the ARM-based server chip has gradually developed, but in the 4 or 5 years, it is a few joys. First, the new entrepreneur Calxeda exited the market, NVIDIA announced that its Tegra K1 chip application direction turned, and rumors that Samsung originally intended to develop ARM architecture server chips, but also changed the strategy due to the operating downturn, and then disbanded the team.
Some operators have withdrawn from the market, and some companies have entered. For example, Qualcomm announced the development of ARM-based server chips in November 2014, and then established Huaxintong with the Guizhou government. In addition, Guangzhou Feiteng Information Technology (PhyTIum) is also in development, Huawei (Huawei) HiSilicon, its own, even invested earlier.
After Calxeda, NVIDIA, and Samsung are unsustainable, there are also manufacturers who are weak. For example, Marvell's chip ARMADA has entered Baidu (Baidu) to become its personal cloud storage application server, or HP (HP) Moonshot server. Texas Instruments (TI) chip KeyStone II, but because they are 32-bit chips, data center operators are more inclined to 64-bit chips, so it is difficult to have more follow-up.
After the 64-bit market became the basic entry ticket to the market, the new technical threshold reappeared and must be as multi-core as possible. In this regard, the ultra-micro (AMD) Opteron A1100 reaches 8 cores, HiSilicon's Phosphor V660 reaches 16 cores, Qualcomm and Huaxintong cooperate with chips of 24 cores, APM's X-Gene 3 has 32 cores, and Kai is semiconductors ( Cavium's ThunderX has 48 cores and ThunderX 2 has 54 cores. Feiteng Information introduced its 64-core chip Mars (R&D code) at the HotChip conference.
Soon, AMD canceled the more follow-up ARM architecture chip plan. Originally, after the A1100, there was a new plan codenamed SkyBridge. However, AMD's operation was sluggish and financial difficulties. AMD still returned to its own position without seeing a clearer market direction. I am good at the x86 architecture chip route, and part of it should be due to the backwardness of the audit competition.
It can be seen that in order to establish a foothold in the ARM architecture server chip market, the funds should be sufficient to avoid falling into the similar situation with the new entrepreneur Calxeda. The original ontology operation should be stable, avoid the situation of Samsung and AMD, and also the market. Sturdy enough to avoid the trend of the chip development as the automotive electronics market is more attractive, like NVIDIA.
In addition, the technology must continue to improve, can not stay at 32-bit, must meet the requirements of data center customers to promote to 64-bit, and then can not fall behind in the audit of the number of audits.
Meeting these is still not enough, and the biggest chip that must further challenge the current server market, Intel's Xeon, has been challenged by APM and Cavium, claiming that its ARM architecture chip is comparable to the Xeon E5 series. Then the chip industry must also win the system software vendors and system manufacturers. Currently, various Linux operating systems support ARM architecture, such as Debian, Fedora, Ubuntu, SUSE, RedHat, etc. The hypervisor is also supported by Xen and KVM with open source code. . System vendors are actively wooing server makers, such as Wistron and Gigabyte.
Finally, the most important is the application. At present, ARM architecture server is still difficult to become a mission-critical application, but pure Web Cache web cache, Proxy proxy host, Storage storage and other applications are feasible, and can also be used for high performance computing (High Performance) CompuTIng, HPC), but in this application, the GPGPU chip is the corner, and the ARM chip retreats into an auxiliary supporting role. Others are also suitable for applications such as Memory Cache and Big Data.
Of course, Intel has also responded to the rise of the ARM camp. Therefore, there is a highly integrated Xeon D chip, and an Atom chip that specializes in servers and storage markets, in order to integrate circuit integration, cost, power consumption, etc. Can compete with ARM architecture chips.
At the same time, Intel also provides chip customization services for larger data center buyers (such as Amazon and Google), and retains customers to avoid changing their ARM architecture chips. Is this kind of supporting measures effective? Look at the data center's technical strategy, operational strategy and other attitudes.
Disposable elastic earloop face mask is made of non-woven fiber fabric, comfortable and breathable, soft and skin-friendly, and easy to use.
These Face masks are made to use in an easy and friendly way comfortable and breathable and will not be stuffy but breathe freely, extra soft elastic earloop will make it fit comfortable on your face.
Disposable face mask are effective in protecting against germs, dust and air pollution, safeguard your health and life.
Surgical Mask,Disposable Face Mask,Medical masks ,Disposable medical mask
Dongguan City Leya Electronic Technology Co. Ltd , https://www.dgleya.com