See also:
- The results of the 2015 drives.
- The end of 2015 — the video card.
- The end of 2015 — PC platform.
- The end of 2015 — best iron.
10. Dispersal of processors of Intel for the poor on a smaller chip sets
The LGA1151 platform and a line of 14-nanometer processors the item has been added will be considered the most advanced Intel products over the next two to three years. As always, chipmaker seriously “screwed the nut”, but because among all this abundance only released two overclocking models feature an unlocked multiplier: Core i5-and Core i7 6600K-6700K. They dispersed only on the boards, built on top (and most expensive) Express chipset Z170. All the younger variations of processors Core i3/i5/i7 and logic H170/H110/B150 Express overclocking functions are deprived. These realities no one was really surprised. Intel has long sought a business model, according to which the acceleration is only for the upper caste users — rich. I want more MHz and a faster your computer? Buy the most expensive chip and a corresponding motherboard! Personally, I never cease to be amazed by this business rule, because by itself, the overclocking appeared just as a means to increase capacity of budget decisions.
Intel Core i5-6600K
Well, time has shown that the yield of the new processors did not affect sales of personal computers continued to decline. Which is logical, serious losses began to suffer motherboard manufacturers. ASUS and GIGABYTE are the leaders in its segment in 2015 shipped only 17-17. 5 million units. This is 10% less than what was implemented last year. Other manufacturers of things was still worse, so for the past year, the market has fallen by 20% in volume. It is obvious that “matlatzinca” you must look for options to increase sales. One of them was the removal of the prohibition on the acceleration of Intel the item has been added without the unlocked multiplier by the old method of raising the clock generator frequency. But only for cards on the Z170 chipset Express. Today new firmware released BIOS ASUS, ASRock, GIGABYTE, MSI, Biostar and Supermicro.
Yes, such a limitation is solely software. The Whim Of Intel. As I said, at the moment, the relief only applies to motherboards chipset Z170 Express, but it’s clear that this half-measures. I want to believe that in 2016 Chinese manufacturers will go further and “push” the fundamental Intel. And then overclocking CPUs without an unlocked multiplier will be implemented on all motherboards for LGA1151 platform.
Overclocking Intel Core i3-6100 to 6100 MHz
9. HBM-memory of the second generation
The first graphics card equipped with HBM memory, appeared last year. Distinguished by AMD, presented a line of gaming accelerators Radeon R9 Fury. The use of new technology has thus improved some technical aspects of the flagship cards of “red”. So, HBM consumes significantly less power. Secondly, this facilitates the layout of components on the Board that allows you to create very compact devices. The most obvious example — video card Radeon R9 NANO. Its length is 152 mm. It’s the most powerful unit among peers of similar form factor. Thirdly, the first generation of HBM has an impressive capacity of 512 GB/s.
Comparison of HBM memory with GDDR5
Unfortunately, first generation, and disadvantages. The main limitation on volume. All the Radeon R9 Fury/NANO has only 4 GB VRAM which is enough only for games in lower resolutions. With the introduction of the second generation of HBM, this option will be doubled, but capacity will be unprecedented 1 TB/s. we Already know that HBM memory will continue to be used not only in the decisions of AMD, but in NVIDIA GPUs. A similar development has Intel.
AMD Radeon R9 NANO
8. The first 10-nanometer processor
The transition to finer process technology becomes chipmakers increasing headache. There are already prototypes, released on a 10-, 7 – or even 5-nanometer standards, but they will remain prototypes until the factory will be able to establish their commercial production. In 2015, Intel has faced serious challenges in the transition to the new process. The first generation 14-nanometer processors Broadwell was supposed to appear in 2014, but the release occurred in a truncated form only last summer. Behind him (with a difference in a couple of months) there was another ruler — the item has been added. The production concept of “tick-tock” for the first time failed, but Intel still believe that she will recover. In this case, in 2016 chipmaker architecture should translate the item has been added to 10-nanometer process technology. But will it succeed?
I personally don’t really believe, therefore, that this alleged event took only the eighth place in the ranking. But wait and hope. If the 10-nanometer Intel chips won’t appear in 2016, that generation tick-processor (an old architecture on a new process) will again be defective, or it does not at all. In any case, the release cycle for new chips can significantly stretch out over time. Modern CPUs are very slowly progressing in terms of performance.
The concept of “tick-tock” clearly
7. The first hard disk HAMR
In fact, for the same reason, the event concerning the appearance of the first hard drives, made on the technology HAMR (Heat-Assisted Magnetic Recording) was only in seventh place. The probability of the announcement of such devices is extremely small, although the talk of thermomagnetic recording has been going on for three years, if not more.
Let me remind you that the use of HAMR will produce hard drives with a capacity of 20 TB and more. Currently developing plates with a capacity of 2 Tbit/IN2, but eventually plans to release a elements density of 5 Tbit/IN2, which will result in a 100-terabyte hard drive.
Schedule of emergence of new technologies for hard disks
Not the appearance of HAMR devices this year will not be a reason for sadness. Mechanical storage still continue to evolve. To date have not exhausted the resource of the classical perpendicular recording technology. In parallel with her developing technology tiled entry. The year 2015 is remembered for the release of the first hard drive capacity of 10 TB. This year HGST plans to produce a model with 12 TB.
The first commercial hard disk capacity of 10 TB
6. Games on DirectX 12
The main technical feature released last year operating system Windows 10 was the support for software interface DirectX 12. Low-level interface with the API from Microsoft has never been. In a nutshell, DirectX 12 will allow developer to enable all the computer resources: CPU, graphics card and memory. In the end the game supporting the new API should be better optimization, which, in turn, will allow to significantly reduce the system requirements, or, conversely, to increase the number of objects and textures that can handle modern hardware.
Compare DirectX 12 with DirectX 11 3DMark
But all of this theory. Windows 10 was released in July of last year, but to date not on paper there are only two games that support DirectX 12. One — Fable Legends — is still in the status of open beta testing. There’s another benchmark from Futuremark. And that’s it. It is strange that Microsoft to announce Windows 10 is not rolled out something exclusive as an illustrative example of the possibilities of the new “headmistress”. In the end, gamers will have to wait and hope that in 2016 we expect a lot of games on DirectX 12. And then it will become absolutely clear in which direction the industry moves.
In Fable Legends new API works so-so
5. The first 10-core processor
The top line of desktop processors Broadwell-E platform LGA2011-v3 was supposed to appear last year, but for obvious reasons described above, this never happened. According to rumors, the announcement is postponed to summer 2016. At this time usually passes Computex 2016. If confirmed, then the flagship will be a 10-core model (probably, Core i7-6950X) — for the first time in the segment table.
Since the architecture is more efficient Broadwell Haswell is only 5%, then the only thing the chips Broadwell-E can outperform Haswell-E is exactly the number of cores. With the long-awaited flagship all clear — a quick 10 cores Intel will appreciate at 999 USD. I wonder more whether there will be in the line of “budget” 8-core chip? Otherwise, the series processors Broadwell-E is unlikely to get any benefits in comparison with Haswell-E.
Intel Core i7-5960X
4. New home development in microelectronics
This year we trace the development of the two domestic products. The first 8-core processor “Elbrus-8S”, developed by MCST. Promise that in 2016 it from the development stage will come to the stage of production and sales. The chip is built on the same architecture of the third generation. Using thinner process technology (28 nm) has enabled to significantly increase the clock speed and overall performance in comparison with the scheme of the last generation “Elbrus-4C”. If we talk about peak capacity, the Russian CPU is superior to some modern 8-core models of AMD. It is planned that the “Elbrus-8C” will be used in server systems, since the chip supports DDR3 memory in 4-channel mode.
“Elbrus-8S”
Hardly “Elbrus-8C” becomes a mass product, cleverly competing chips from AMD and Intel. But the developers of the national system-on-chip “Baikal-T1” seriously intend to make a statement. It is planned that by 2020 will be implemented over five million copies of this chip. With the high production cost per chip will be small by today’s standards 60 USD.
Let me remind you, “Baikal-T1” is the first Russian system-on-chip created on 28-nanometer process technology. Chip got functional and powerful enough stuffing. It is based on two core Warrior P5600 32 R5, operating at a frequency of 1.2 GHz, and the units with Ethernet ports, PCI Express 3.0, SATA 3.0, and USB 2.0. To see Baikal in conventional computers and mobile devices will not work. The MIPS architecture is designed for use in various embedded systems and networking equipment.
I want to believe that the ambitious plans of the Russian company fulfilled in full. I hope that in 2016 domestic chipmaker will please us with new developments.
“Baikal-T1”
3. First drive XPoint 3D
In my humble opinion, Intel and Micron have made a real revolution in the segment of storage systems. The technology was introduced XPoint 3D, but in 2016 based on it will be the first SSD Optane. A full line of implements on the basis of modern form factors M. 2, U. 2, PCI Express x4 and NVDIMM (non-volatile DDR4 modules for the not yet released Intel Xeon E5 v5 generation Broadwell-EP).
The evolution of types of memory
3D XPoint for storing information does not use transistors. Memory cell changes the resistance to create either a logical zero or one. This principle allows to increase the recording density at least 10 times. Access to the memory cell is carried out by combining certain of the voltages on lines intersecting conductors. Hence the name of the technology. In addition, these lines are superposed on each other in several layers. In the end, the new development promises a performance with 1,000 times the performance of modern flash memory. Plus 3D XPoint somewhere 1000 times more reliable.
Perhaps the only drawback of SSD Optane will be their very high cost. On average, the XPoint 3D chip is worth five times more expensive than the most common today MLC NAND. And yet in the server segment of the new development will quickly pay for itself. Intel also plans to release a line of Optane for drives and consumer market.
Variations Intel Optane
2. NVIDIA Pascal
In 2015 AMD and NVIDIA have released their flagship graphics card based on the largest 28-nm chips. To increase the usable area of the crystal and on the factory TSMC can’t, and it means that with this process it’s time to say goodbye. Of course, both manufacturers can still release GPUs based on time-tested technological limits, but a truly new 3D accelerator cards will have 16-nanometer chips. Finally, since the first graphics processors that are manufactured on 28-nanometer process technology, appeared more than five years ago!
The positioning of Pascal in comparison with previous architectural solutions
New NVIDIA graphics cards will be built on the architecture of Pascal. This is the main innovation. Information about it is not very much. We only know that 3D accelerators will support different precision computations: FP16, FP32 and FP64. In calculations with double precision peak performance flagship Pascal chips will be about 4 teraflops. Indicator space, as the most powerful to date, dual-processor accelerator Tesla K80 in turbo mode shows only 2.91 teraflops.
The second constructive feature is the use of the latest tire NVLink bandwidth 80 GB/s. To further “green” plan to increase this parameter to 200 GB/s. New tire significantly speed up the work of Tesla accelerators installed in the supercomputers, as well as improve the relationship between multiple graphics cards combined in SLI.
Thirdly, 3D-accelerators Pascal will receive the chips HBM memory of the second generation. The volume of the array will range from 8 GB to 16 GB. Bandwidth, as we have seen, will amount to 1 TB/s.
The prototype NVIDIA Pascal
1. CPUs AMD Zen
CPUs on the AMD Zen architecture — the most anticipated products in 2016, according to Ferra. “Red” two years of not releasing anything new, but if you dig deeper, with its modular architecture Bulldozer, the company hasn’t gone anywhere since 2011. Something needs to change.
More precisely, it is necessary to change all. Platform AM3+ and FM2+ are seriously outdated. Instead they will have one, but modern — AM4, developed in conjunction with ASMedia. But the most important thing is the new chips. According to guiding the tops of AMD, the core of Zen will execute 40% more instructions per clock than Bulldozer core. This is not an absolute indicator of performance, but the transition to 14-nanometer process technology allows for a crystal much more than that. As a result, we may expect these to 8-10 cores. In addition, the crystals Zen will support Simultaneous Multithreading technology is analog Hyper-Threading technology from Intel. In the AMD report that they are satisfied with the performance and yield of the new chips.
So AMD at the time praised, modular architecture Bulldozer
Zen — the exact opposite of Bulldozer. The new kernel can be considered complete. No modularity! All these facts allow us to hope that the new chips “red” will be 40-50% faster than their predecessors. This means that the competition between AMD and Intel will be reborn. If not Zen, then what?
Performance AMD Zen