 Wysłany: Śro 14:51, 15 Gru 2010 Temat postu: processor for netbooks after |
|
| Bulldozer: specially designed for computer enthusiasts
carefully considered, AMD 7 months the company has not released a newly designed CPU products. But as far as we know, Phenom II product will soon put replaced by a new product, this new product will be designed for servers, desktops, notebooks and netbooks three major platforms. AMD new processors designed products will be divided into three series: Bulldozer, Bobcat, and Llano. After the three products listed expected, AMD and Intel will return to competition between the white-hot stage, as we categorize the following we will introduce three designs in advance of the product. Bulldozer core design of the processor-based products high scalability, each of the basic core module contains two processor cores, and the basic core modules can be combined as a unit . Each basic module which have adopted the cluster type named multi-threaded (Cluster-Based Multi-Threading: CMT) technology. CMT to understand the concept, we can first look at their fellow brothers and SMT ( symmetric multi-threading), Intel P4 processor with Hyper-Threading Hyper-Threading is based on this technology. in Intel's SMT program, using the method of copying the state of processor architecture to achieve SMT, the internal processor core that is not an additional extra hardware execution units to deal with SMT, but only increased the processor to store the number of thread-related data elements, and the hardware implementation of the data unit will be sent free to which treatment, in order to increase the utilization of processor execution units. image that, Intel's Hyper-Threading Technology is like a car factory in the assembly line, although only one assembly line, but the line at any time for each step of a work to receive automotive parts for assembly, Parts not in place will not stop the assembly line and other parts caused by the phenomenon. After years of promotion, many of the current operating system has been well coordinated with Intel's Hyper-Threading Technology, Intel Hyper-Threading processor fully guaranteed line availability. However, this design also has some shortcomings, it only uses a command window (instruction window: responsible for buffer scheduling instruction unit) is responsible for two thread scheduling, execution,Washington Redskins Jerseys, and retire. Returning to our example above, the assembly line, this looks like only one production line management and dispatching personnel, a person is obviously very difficult task of simultaneously two, so sometimes there will be production line failures, and processor encountered such a situation occurs when the apparent decline in performance. Bulldozer and each module is added an additional execution units, each module can have a big task subdivided into multiple parallel tasks the ability of these lines can be any integration need not the efficiency of the assembly line will be affected, this is called the CMT technology. CMT than the Hyper-Threading technology with greater efficiency and performance. When Bulldozer transferred between the processor results in the pipeline, these results are stored in the cache, the program can be read from a data cache the results of these calculations. The data is like a car factory in the installation has been completed, wait for the finished car to the customer. Intel's Hyper-Threading technology products like car factory car parked in a parking lot on the whole, there are various types of products parked cars, trucks, such as cars, trucks, so that when the only customers to provide delivery vehicles,Arizona Cardinals Jerseys, their delivery efficiency will be low. The Bulldozer in the establishment of two such parking. Bullzoer the two lines share a single floating point operation unit (shown in red circle to play the part), with two 128bit FMAC (converged multi-precision) lines, higher precision floating-point operations, so Bulldozer When carrying out floating-point processor performance and accuracy than existing Phenom II significantly improved. 128bit FMAC floating-point pipeline is also able to meet the new AMD Bulldozer added SSE5 and the demand for AVX instruction set extensions, which included a large number of instruction set extensions 128bit multimedia instructions. And after using FMAC technology can also be achieved in one cycle to be completed within Intel AVX instruction computing purposes. SSE5 AVX instruction set extension and upgrade the processor on multimedia, encryption, and significant performance scientific computing. Finally, Bulldzoer PhenomII also on the structure of the processor's cache has been improved, and now the processor of the secondary and tertiary structure shared cache are used, each module can be shared in the two core the cache. Zambezi: AMD introduced the first product will be the Bulldozer core processor, code-named Zambezi products, this product contains four basic modules (each module contains two processors core), it is understood Zambezi forms will still use the AM3 socket, so in theory, all current PhenomII DDR3 motherboard BIOS upgrades can be supported simply by the new processor. Unfortunately, part of the current technical details Bulldozer core AMD side is still not public, these questions include: -Bulldozer core of the \\ -Bulldozer in the number of pipeline stages in the end? -Bulldozer internal floating-point execution units specifically how to configure it? - inside the module is how the two lines share a single floating-point execution units? lack of information, so we are still difficult to determine the Bulldozer processor compared to the current rate of processor performance in terms of the number in the end. Bobcat: processor for netbooks after Bulldozer, AMD will introduce an additional product line, this is the Bobcat, this architecture is similar to the Athlon primarily for Neo and Intel Atom processors that access the product. According to AMD's roadmap shows that the first use of Bobcat will be a 32nm processor design process, code-named Ontario's APU processor, this processor internal integration of the two Bobcat processor module, and integrated set of DX11 GPU was. Bobcat processor modules each contain a single core processor, set up a single core within an integer scheduler and an integer execution unit, the integer execution units include I-line, Ld - pipelines and St pipeline three parts, each responsible for different parts of these three types of tasks, like assembly line workers to install electrical and mechanical installation work as the division of labor. The core of the integer Bulldzoer scheduler also manages four integer pipeline, but so far we do not know these four integer pipeline is responsible for what their specific job. In addition to the integer part addition,my old PC, Bobcat also has a dual-core internal structure of floating-point processing part of the pipeline,think about it, the two floating-point pipelines are named \M pipeline \ power consumption. Bobcat core standby heat less than 1W, 0.5W only about the actual use of the TDP will be around in 5-10W, is ideal for Internet use. performance, AMD announced Bobcat processor can achieve \If they are talking about here \ But the most compelling place Bobcat processor is not the architecture design, but AMD processors in the design of this new design is the use of technology. They used something called HLS (high-level synthesis) design technology to design the processor, designers use C + + program to design the processor's behavior patterns, and then enter the code manufacturing equipment, will be able to manufacture to meet the requirements. HLS technology with the design speed, the advantages of conversion product is easy to adjust, but using this technology in products in terms of operating frequency is not as good as manual design of the product, according to AMD sources, using technology designed by HLS Bobcat processor running frequency of hand-designed product than about 20% lower. comprehensive look, Bobcat processor on the market than the existing ultra-low voltage processors such as the VIA Nano, Intel Atom processor and AMD Athlon Neo so much faster, but also have the backing of HLS technology, products Improved response to market demands faster,let us look at its configuration information bar, which is designed using the old methods of competing products can not. Fusion: for low-end laptop and desktop 2006 年 10 月 26 日, AMD acquisition of ATI, the company completed the work, when they had solemnly made the following a statement: \\AMD processors were not much happened. But taking into account the set of Intel processor products was GPU performance is with the ATi / Nvidia's existing products can not match, so people still have great expectations of Fusion. AMD finally announced the recent launch next year of the Fusion processor, code-named Llano. Llano: Llano AMD Fusion processor will be available the first model. AMD earlier this year, according to Llano released photos show the core, the processor core and Athlon X4 (codenamed Propus) is very similar (such as the following comparison chart shows it), but the process evolved from the 45nm 32nm, and joined the GPU core modules only. Llano and Propus same products for the mainstream market, so that the results of the comparison is not surprising, after all, designed to improve the existing process after the introduction of new products, the core compared to redesign the entire processor spent much less time. is worth mentioning that the core scan Llano figure above shows all of the core design is not yet complete, the figure shows the bottom of the GPU chip and some other foreign release has been made fuzzy processing, which means that the actual launch of the Llano future will be shown in the figure there is a big difference. However, according to available information and pictures, we can see the core of the Llano within each volume is equipped with 512KB-1MB secondary cache,football uniforms, three-level cache is not set, the other part is to configure 6 GPU-based HD5000 Series 480 stream processors designed units. Simply put, Llano in fact, like the core Athlon II X4 with open fifths (66%) of the Radeon HD 5750 and the skill composition of a product, from this point of view, at least in the graphics performance better than existing Intel processors Clarkdale and Arrandle a lot stronger. summarized in the text before us, let's sum up so far mentioned in this article several architecture and product: Summary: AMD has long been talking about the beginning, \was designed to set the demand. Perhaps the readers will read this opinion article questioning the \\Although AMD's first Fusion significant processor sets alone can not catch up with significant performance, but after all, they have taken the first step, and to Llano for a low-voltage laptop, netbook and mainstream units type machine market, the performance of such products has enough water. Not only that, AMD and more simmering in the Sha Zhao. Although Bulldozer processor was not designed to set, but the floating-point unit design it appears relatively simple, we see that they probably will in the next-generation architecture design completely by the external part of the floating-point operations GPU or integrated graphics card to handle, and GPU's performance in floating point advantage is the general CPU can not match. Under the guidance of this idea in future applications will be able to take the initiative to identify the most suitable for dealing with the current operations of the CPU or GPU core tasks, and automatically give them these tasks were handled. Development strategy which is obviously a poor relation of mastery of GPU technology, the Intel feared. CNBeta compile Original: icrontic |
|