Current location - Quotes Website - Personality signature - On cpu Problems of General Edition n95
On cpu Problems of General Edition n95
To Landlord:

It has been a long time since Machine Emperor N95 and its brother N95 8G went on the market, and their powerful performance has been well known. Among them, we have to mention a great hero-OMAP 242, which is the strong "heart" of N95.

Omap242 is made by Texas instruments (TI) and is suitable for high-end mobile phone applications based on Linux, Windows and Symbian operating system (OS). It is the first product in Omap 2 series, and Omap2 series will eventually turn to the mixed field of "modem and application processor". Perhaps the most attractive part of this chip is the multi-processor core, which includes 33MHz ARM 11 RISC, 22 MHz TI C55 DSP, imaging and video processor with ARM7, and 3-D graphics processor of Imagination Technologies which supports 166 MHz mobile DDR SDRAM. The chip also integrates display and camera controller, SDRAM and flash controller, and more than 6 peripheral controllers are attached. Omap 242 can provide strong support for high-end multimedia applications, including 3fps CIF video conference, 3fps VGA codec, VGA and TV display, and cameras with more than 3 million pixels.

the biggest difference between p>242 and previous generaTIons of Omap application processors of ti is that the design process is reduced from 13nm to 9nm. In addition, the performance of the ARM core used in the 242 has also improved: the previous Omap chip can only support 22MHz at the highest, while the 242 has increased the speed to 33MHz. The cache capacity and total memory bandwidth of 242 have increased, and TI also replaced ARM9 with ARM11 RISC.

the 3-D graphics performance of p>Omap242 is improved by 4 to 5 times, and the video performance is also improved by an order of magnitude. And through parallel processing, its multi-task processing ability is more powerful.

242 is a multi-core design chip, and TI has made it clear that it will support as many cores as possible according to the application requirements in the future. In this way, parallel processing or multi-core processing has become the future development direction of TI in the field of mobile phone applications.

242' s unique version layer accommodates TI's multi-engine processing and power management functions. Here, TI turned to a standardized interconnection method called Open Kernel Protocol (OCP), which belongs to an independent non-profit standards organization OCP-IP.

OCP facilitates kernel reuse and enables the kernel to be independent of the integrated subsystem. OCP also implements a set of timing rules that can be used for the whole chip and a separate verification tool set. If necessary in the design process, it also allows the IP module to move around in the system, which makes the system analysis and chip debugging easier and more direct.

the power management technology of p>Omap242 allows each processor to optimize power consumption for different applications. At the die level, TI combines multiple power domains, and each domain can turn off the power to zero leakage, thus greatly extending the battery life. Many power management technologies used in Omap242 are part of TI's new SmartReflex technology.

at the transistor and software level, TI pays special attention to how to adapt to the zero leakage level. By using multiple power switches preset on the die, the architecture can power off different domains independently, thus helping TI to obtain the required granularity. In addition, the chip also contains a special embedded "diode-footed SRAM" switch to reduce the power consumption of embedded memory.

due to the different power requirements of end users, TI has realized a software programmable power mode, which can adjust the voltage and frequency according to different applications. For example, enter the lowest power consumption state by defining the "off" mode.

at present, Semiconductor Insights(SI) company is analyzing Omap242. We have determined TI's two-pass power switch control circuit, and are analyzing the influence of this control circuit on the whole power management system in the automatic wiring circuit. By analyzing the bare die, SI also reveals the specific positions of the ARM core, DSP core and graphics accelerator. The locations of the storage controller, TMS32C55x DSP, power management module and IVA (Imaging Video Accelerator) that once had doubts have also been confirmed.

there are five memory types in p>242, and all the embedded memories together account for about 1/4 of the die area, while the four different memory types in Omap171 account for about 17% of the die. However, the die size of Omap242 is about 1.9 times that of Omap171, so it seems that some sacrifices need to be made in die area to achieve higher integration.

from the software point of view, TI allows the code to manage the power state from two aspects: first, by monitoring the workload, the software knows which application is being used; Secondly, the software collects statistical information to determine the workload. This enables the chip to manage power based on prediction according to the actual usage mode.

TI is trying to transfer the Omap process to 65nm, and some additional process problems in the process have affected the leakage of the chip. As we all know, TI has a number of 9nm process technologies: some use high leakage transistors, while others use low leakage transistors; Therefore, it can be inferred that TI will continue this strategy in the 65nm process. In addition, TI can also compensate the extra leakage at the system level, so we expect TI to propose some unique circuit implementation methods to solve the 65nm leakage problem.

TWL9223 located at the periphery of power management design is an auxiliary power management device. It adopts 25nm analog BiCMOS process design, and can provide some special functions, such as on-board voltage regulation and DC/DC conversion, to supplement the functions of 242.

in fact, TWL9223 is specially designed for use with 242, and it contains some special protection mechanisms suitable for the latest Omap, such as bare die heating thermometer and overheating and power failure protection. Among them, the thermal power-off protection allows TWL9223 to turn off its DC/DC converter and then send an interrupt signal to 242. As the shape of handheld devices is shrinking, similar functions are very important.

TWL9223 also replaces many peripheral chips and discrete components in a single chip.

in the whole process of launching 242, TI has overcome numerous difficulties, and the complexity of the layout is the biggest challenge of 242. However, 24 finally got out of the difficulties and used the latest test version of standard tools for verification.

the memory part

TI supports packaging and stacking (PoP) memory on Omap24. PoP refers to placing one package directly on top of another package, and the packaging materials of the two packages are directly connected. However, such a memory stack may cause some problems. Because leakage current will generate heat, placing another chip on Omap242 will increase the heat generation of the whole system.

it's SmartReflex technology's turn here. In order to ensure compatibility and stability, TI simulates the coupling between SRAM or mobile DDR devices introduced by various memory manufacturers and Omap242 by POP and standard die stacking respectively.

from the perspecTIve of ti, PoP is a business model. Because PoP urges customers to negotiate with memory manufacturers, TI need not be limited to supporting a specific memory.

Samsung's latest NAND flash memory device raises a question: Is the total amount of memory that can be stacked and contacted limited? Because NAND flash memory and NOR flash memory seem to be very popular, some people may wonder which memory configuration users will adopt. In fact, there is no need to worry here, because TI supports various flash configurations and types, including NAND or NOR up to 1GB and mobile DDR up to 1GB. Even if NAND is more popular in the mass storage application of high-end mobile phones, TI still regards NOR as its good competitor.

in a word, TI has been working very hard to promote the scope of design using Omap242 to meet some "crazy" applications. For example, run high-quality 3-D games while running high-quality audio applications, and use OMAP242 to control a standard TV to play full-screen games. According to TI, such applications will become very common. At the same time, TI also claims that OMAP242 can run graphics and video processing on two monitors at the same time, and can use the same set of storage devices to run audio, video and 3-D games in parallel on an advanced operating system.

although Omap242 is fierce, it is not without competitors in this field. Broadcom7 launched its BCM275 multimedia processor in 27, and Nvidia and ATI Technology continue to design similar products to compete with TI. In addition, Qualcomm is still concerned about how to integrate high-end multimedia functions with baseband functions.

however, with Omap 242, TI has advanced the overall integration and functionality of the chip. It is hoped that 242 can help TI maintain its leading position in the field of application processors.

the advantages of choosing OMAP242 instead of OMAP243 for p>N82 CPU!

let's talk about something beside the point first. Most CPU loaded on Nokia phones belong to TI's OMAP series (OMAP171, OMAP242), while HTC (dopod) generally uses OMAP85. The difference is that OMAP85 has a built-in communication chip, which is an ARM7 GSM/GPRS digital baseband processing system. OMAP85 is a cost-effective scheme. However, Nokia obviously doesn't trust TI's communication module when it doesn't choose it. It would rather complete the peripheral circuit design and chip selection by itself than choose this highly integrated CPU. It can be seen that Nokia is excellent in call quality and signal.

get to the point! Originally, everyone thought that the N82 used OMAP243. However, through the disassembly of netizens, it is confirmed that the N82 uses OMAP242+ another chip specially used for processing cameras. Some people are somewhat sorry after hearing this news. If they buy a high-end N82, they naturally hope that the CPU they use is also high-end. Most people think that OMAP243 is higher than OMAP242, but it is not.

from the perspective of a mobile phone, 242/243 is a CPU. But from a technical point of view, 242/243 is not called CPU, but a chip that integrates CPU, DSP, memory image processor, accelerator, interface and many digital processing circuits needed by a mobile phone. As far as the important components are concerned, OMAP242, which came out in 25, integrates: CPU-SRAM 1136 (33 MHz), DSP-TMS 32C5x (22 MHz), 2D/3D graphics processor (with an operation speed of 2 million polygons per second), image/video processor IVA (which can process still pictures larger than 4 million pixels) and internal integration of 5M. It is also the CPU of ARM1136, but the frequency is increased to 46Mhz;; The whole DSP is removed; Reduce the operation speed of 2D/3D graphics processor to 1 million polygons per second; Upgrade the image/video processor to IVA2, and it only takes one second to process the 5-megapixel still picture. The related circuits of the second camera are also integrated; The internal integrated 5MbSRAM was removed, but the USB interface was upgraded to 2..

seeing this, I believe that even if you are not familiar with this knowledge, you can understand that OMAP243 has a very obvious advantage over OMAP242, that is, it is cheap! At the same time, another type of chip, OMAP2431, castrated the DSP of OMAP242, and then castrated the 2D/3D graphics processor, which was cheaper. In addition to low cost, 243 can only be worse than 242 in other aspects, and the biggest change has made the image/video processor a cow. However, this is a component that N82 will not use, because N82 uses an independent image/video processor-OMAP 29.

In fact, even if the image/video part of OMAP243 is awesome, it can't meet the camera configuration of N82. According to the data, it takes 1 second to process 5W, not to mention that it must be more than 1 second in fact. Just count it as 1 second. You press the shutter and the photo is displayed on the display screen. This has a delay of 1 second, which should be acceptable to many people. But don't forget that N82 needs autofocus. The lens should be divided into several segments from the shortest to the longest, and the images of each segment should be captured and then processed to determine where the lens is. Suppose it is divided into 1 segments (actually far more than that), which means that with 243, it takes at least 1 seconds from when you press the shutter slightly to when you are prompted to focus. If the N82 looks like this, will you buy it? You certainly won't. And if an N82 equipped with a 5W pixel xenon flash can't focus automatically, the photos taken are blurred. Even if you are willing to pay for it, Nokia won't want to make such a rubbish. Therefore, N82 can only use a special camera processor, plus OMAP242 which is better than OMAP243 in other aspects. Without DSP, the processing speed of 2D/3D is reduced. Is N82 still the speed you are using now? Maybe slower than N73 based on OMAP171!

it turns out that the N82 parts were selected correctly, and everyone bought the N82 correctly. Originally, friends who felt sorry that N82 didn't use 243 should be thankful: fortunately, it didn't use 242

It is necessary to explain to you what this DSP TMS32C5x is for. TMS32C5x is a programmable digital processor in Texas. What it can do is set by the user's programming, instead of being fixed at the factory like CPU. This processor is widely used in the communication industry. online