AMD urgently recruits light and pursues talents: Optimizes the game for RDNA2 architecture
Although in the eyes of some people, ray tracing is just a "gimmick" for NVIDIA to promote RTX 20 series graphics card , with the launch of Xbox Series X and PS5 consoles at the end of the year, ray tracing will gain more recognition, especially the number of supported games will increase greatly. For this reason, AMD's game R&D team has to urgently recruits light and pursues talents, thinking that game integration is a supporting package. Compared with RDNA1, the RDNA2 architecture of
AMD has increased its performance by up to 50%, and has added variable-rate shading VRS technology to further improve the operation efficiency of the GPU and realize GPU hardware-level ray tracing support.
's new generation of gaming graphics cards using RDNA2 architecture will be released at the end of September and officially launched in October. We might as well wait and see what surprises AMD will bring to us this time.
SK Hynix mass-produces HBM2E video memory
Sk Hynix announced in early July that it has begun mass-produced a new generation of HBM2E DRAM memory/video memory chips, which is only 10 months after the first HBM2E specification was proposed in August last year.
SK Hynix's HBM2E has the characteristics of large capacity, high speed and low power consumption. It uses TSV silicon perforation technology to vertically stack eight 2GB chips, so as to achieve a single capacity of 16GB, twice that of Shangdian HBM2. In theory, it can achieve a stack of twelve and a single capacity of 24GB. The data rate of each stack of HBM2E is as high as 3.6Gbps (which can be understood as the equivalent frequency of 3.6GHz). With 1024bit I/O, the bandwidth can be easily achieved at 460GB/s. If four combinations are used, the total capacity is 64GB, the total bit width is 4096bit, the total bandwidth exceeds 1.8TB/s, and the voltage is maintained at the 1.2V level of HBM2. Such specifications are beyond the reach of GDDR6.
SK Hynix said that HBM2E can be widely used in cutting-edge fields such as Industry 4.0, high-end GPU graphics cards, high-performance supercomputers, machine learning, AI artificial intelligence, etc. Samsung is also actively preparing for HBM2E, while Micron vaguely stated that it will launch a new generation of HBM before the end of the year.
NVIDIA RTX 3070 parameters and specifications are exposed
Recently, foreign media revealed that the RTX 3070 graphics card was reported.
According to the leaked news, the RTX 3070 graphics card will be made of GA104 core, and NVIDIA has designed three sets of solutions around the GA104 core: the full version of the GA104 core has 3072 CUDAs, but it is divided into two PCB versions of PG141 core PG142. The former uses 8GB GDDR6X video memory, and the latter is 8GB GDDR6 video memory; there is also an incomplete version of the GA104, with 2944 CUDAs, the same number as the current RTX 2080, and is equipped with 8GB GDDR6 video memory.
According to speculation, the 2944 CUDA versions of GA104 should be RTX 3070 graphics cards, which means that the performance of the new generation of RTX 3070 graphics cards is at least RTX 2080 level. As for the two sets of solutions for the full version of GA104, it is speculated that NVIDIA will choose one to pass retail verification as the RTX 3070Ti to launch. However, it cannot be ruled out that one of the sets will be released as the "super" series. In terms of price, the RTX 3070 is expected to be $399 and the RTX Ti is expected to be $499.
NVIDIA Ampere Core Downgradation: Switch to Samsung 8nm process
According to the whistleblower kopite7kimi, the upcoming GeForce Ampere graphics card GPU is not TSMC 7nm, but is created by Samsung 8nm LPP.
Samsung 8nm is an improved version of 10nm. As an improved process for transitioning from 10nm to 7nm, it was developed in 2017. Previously, there were more of Qualcomm. According to Samsung, 8nm LPP reduces power consumption by 10% and reduces area by 10% compared to 10nm LPP.
Considering the comprehensive advantages of 8nm production capacity and cost, NVIDIA is likely to choose Samsung 8nm for higher returns.
NVIDIA internal notice: It is recommended that graphics cards start to increase their prices in July
According to the news learned by Juxunmen, NVIDIA has internally notified all AIC brand graphics card manufacturers, and it is recommended that some models can be increased from July. One of the factors that raises the price of
is to make up for the space left by the 618 special offer - has China's 618 e-commerce promotions have had such a big impact on the global graphics card market? Could it be that there will be a wave of increase after Double 11?
. Another reason is that the demand for "mining" of virtual currencies has increased significantly in the recent period. The market demand for popular mid-to-end mainstream models such as GTX 1660Super and RTX 2060 is very strong. NVIDIA recommends that prices be raised appropriately.Not long ago, some investors accused NVIDIA of using mining card sales to mislead investors and filed a class action lawsuit against NVIDIA. In this regard, NVIDIA stated in its latest statement that some investors deliberately pointed out in the company's investor information, and emphasized that it is impossible for both the company and the management to fully understand what the sold GPU graphics cards are used for.
. According to the latest news, NVIDIA's high-end chips are currently insufficient and stocking is very tight, so there is a high probability that there will be serious shortages in stock, and it is even possible that there will be a window period. It is not clear whether NVIDIA will make up for production capacity in the future. These factors combined make N card price increase inevitable. Currently, RTX 2060 is already rising.
AMD is not spared. Affected by the growth in mining demand, AIB brand graphics cards have begun to be out of stock, and RX 580 2048SP graphics cards have begun to rise in prices.
Samsung discovered new semiconductor materials: memory, flash memory benefit
Recently Samsung Electronics announced that Samsung Institute of Advanced Technology (SAIT) and Ulsan National Academy of Science and Technology (UNIST) and the University of Cambridge in the United Kingdom have discovered a new semiconductor material "amorphous boron nitride" (amorphous boron nitride) abbreviated as a-BN, which is expected to promote the accelerated development of the next generation of semiconductor chips.
Amorphous boron nitride originates from white graphene and is composed of boron atoms and nitrogen atoms. However, unlike white graphene's hexagonal structure, amorphous boron nitride has a molecular structure without a fixed shape.
Samsung said that the dielectric constant of amorphous boron carbide is very low, only 1.78, and has strong electrical and mechanical properties. As an interconnect isolation material, it can greatly reduce electronic interference, and it can grow to wafer-level size with an ultra-low temperature of 400℃.
Samsung believes that amorphous boron nitride is expected to be widely used in the manufacturing of DRAM memory and NAND flash memory, which is particularly suitable for the next generation of large-scale server storage solutions.
AMD Ryzen 3000XT officially launched
AMD officially released the Ryzen 3000XT series processor on June 16, including Ryzen 9 3900XT, Ryzen 7 3800XT and Ryzen 5 3600XT, and officially released on July 7.
Ryzen 9 3900XT: 12 cores 24 thread , 6MB secondary cache , 64MB 3 cache, base frequency 3.8GHz, maximum acceleration 4.7GHz, thermal design power consumption 105W, priced at 3899 yuan, no original radiator.
Ryzen 7 3800XT: 8 cores, 16 threads, 4MB L2 cache, 32MB L3 cache, base frequency 3.9GHz, maximum acceleration 4.7GHz, thermal design power consumption 105W, priced at 3,049 yuan, no original radiator.
Ryzen 5 3600XT: 6 cores and 12 threads, 3MB L2 cache, 32MB L3 cache, base frequency 3.8GHz, maximum acceleration 4.5GHz, thermal design power consumption 95W, priced at 1,859 yuan, and comes with a ghost radiator.
compared with the previous Ryzen 9 3900X, Ryzen 7 3800X and Ryzen 5 3600X, the three Ryzen 3000XT series processors have increased the acceleration frequency by 100MHz and will replace the former's current market position. The former will be priced appropriately. Players who want to enter the third generation of Ryzen may wish to take the opportunity to buy it.
According to AMD's official data, the single-threading performance of the Ryzen 3000XT series is 4~5% higher than the old model, and the multi-threading performance is 1~3%. Among them, the benchmark performance of the Ryzen 9 3900XT is increased by up to 10%, the gaming performance is improved by up to 7%, the benchmark performance of the Ryzen 7 3800XT is improved by up to 3%, the gaming performance is improved by up to 13%, the benchmark performance of the Ryzen 5 3600XT is improved by up to 3%, and the gaming performance is improved by up to 15%.
In addition, AMD also released the newly upgraded StoreMI2.0 hard disk acceleration technology, which is suitable for accelerating the application and game startup on large-capacity HDD , and can automatically transfer the most commonly used data to the SSD cache. According to the data provided by AMD, when StoreMI2.0 is mixed with HDD+SSD, the loading speed of game loading can be up to 13% faster than using HDD solely, the application startup speed can be up to 6 times faster, and the Windows system startup speed can be up to 31%. The first release supports X570 motherboards, and will gradually support X470, B550, B450, X399, and TRX40 motherboards within this quarter.
Intel Core i9-10850K exposed
When AMD released the Ryzen 3000XT series processor, Intel also quietly prepared a new product Core i9-10850K, but in contrast to AMD's frequency extraction, this new U is the product after the i9-10900K down frequency, and has now appeared in the GeekBench4 test database.
Core i9-10850K belongs to the 10th generation Core Comet Lake-S series, with 10 cores and 20 threads, 2.5MB Level 2 cache, 20MB Level 3 cache, 3.6GHz reference frequency, and the maximum acceleration frequency is 5.2GHz. Compared with the existing flagship i9-10900K, the reference and acceleration frequency are reduced by 100MHz, and the thermal design power consumption is still 125W. In terms of running scores, the i9-10850K and i9-10900K are very close, and the multi-core running score is only about 5% lower than the latter.
I9-10900K has multi-core, high-frequency, and overclockable characteristics. Although the price is not cheap, it is still very popular after it is launched. Some spot stocks are even hyped up to about 5,000 yuan. If the i9-10850K can reduce the price by just a little lowering the frequency, and highlighting the cost-effectiveness, it will undoubtedly attract many high-end players, which will be a powerful blow to the Ryzen 3000XT series processor.
Intel7nm process is scheduled for 2023
ARM Company's China executive @After thinking, I still Winnie exposed an internal report of Linli Group not long ago, which included the circuit diagram of the Intel processor. Although it is in the server field, it also revealed enough information to us.
According to this Linli Group's roadmap, the Intel10nm process will be extended until 2022, and 7nm will not be launched in 2023, which is one year later than the Intel official roadmap, suggesting that Intel may be delayed in the future.
