In the early morning of September 28, Beijing time: Intel On Technology Innovation Summit was held, this year is the second. The word "On" is taken from the last two subtitles of Innovation, which is intended to highlight the meaning of innovation. At the
Summit, Intel announced that it will launch the 13th generation Core and a gaming graphics card for consumer users; updated the Intel developer cloud platform and Intel Geti computer vision platform for developers; and also introduced more updates to the data center market about the fourth generation of Xeon scalable processors and data center graphics cards.
1, 13th generation Core, the price is really good. The release of Core has always been very high. Since the release of 12th generation Core with large and small core design, Intel has steadily won a game in the consumer processor market.
Not long ago, AMD released the new generation of Ryzen 7000 series processor. In the past two days, there have been a lot of actual test comparison data on the Internet with the 12th generation Core. However, AMD's new products do not have obvious advantages, so Intel's 13th generation Core is much easier. The
press conference came too quickly. I wonder if I didn’t have time to change PPT. At a glance, the 13th generation Core has also improved a lot. Single-threaded and multi-threaded performance has been improved by up to 15% and 41% respectively, and there are quite a lot of multi-threaded improvements.
In addition, I found that the specification parameters of i9-13700k are the same as the cores of i9-12900k, and the cache and main frequency have also been slightly increased, but the price is still the price of i7-12700k. If nothing unexpected happens, the 13th generation Core allows users to buy i7-12900k with the money of i7-12700k, which is really delicious.
In addition, Intel said it will launch some Core processors that can reach 6GHz earlier next year, and AMD may be a backup plan due to the absence of 7800X in the Ryzen 7000 series.
2, the graphics card part: gaming graphics card
Intel Ruixuan A770, priced from US$329, can be used for content creation and 1440p gaming scenarios, and will start on October 12.
From the picture displayed, it adopts a dual-slot design, 8pin+6pin power supply design, and is expected to have a maximum power consumption of more than 200 watts, with a maximum ray tracing performance of 65% higher than competitors.
From the specifications, it is initially expected that the performance level will reach the level of mid- and low-end gaming graphics cards. This level of gaming graphics cards can meet the gaming needs of most users.
Finally, if you really need a graphics card at this stage, but you are really afraid of buying a mining card, you can pay attention to this graphics card.
3, part of the graphics card: Data center graphics card
In August this year, Intel released the data center GPU Flex series, mainly meeting the capabilities of video cloud encoding. In this update, Intel said that the Flex series will also support mainstream machine learning frameworks such as TensorFlow and PyTorch, and of course, it also includes its own OpenVINO.
is not clear at present whether it is used for training or reasoning. It is likely that it is just the reasoning part. Intel said that some users use the Flex series GPU to do some reasoning, and their performance is excellent.
At present, machine learning training pays great attention. If you want to talk about inference performance, you have to look at Ponte Vecchio (in fact, the training load is mainly done by Gaudi) . Intel previously stated that the Ponte Vecchio is more powerful than Nvidia's A100 in some scenarios, and if everything goes well, the Ponte Vecchio is quite worth paying attention to.
At this summit, Intel stated that the blade server for Intel data center GPU codenamed Ponte Vecchio has been shipped to Argonne National Laboratory and will provide computing power for Aurora supercomputers.
4, the fourth generation of Xeon scalable processor with "DLC" codenamed Sapphire Rapids will use Intel 7 technology and support PCIe 5.0, CXL and DDR5 memory. The hardware alone is worth looking forward to.
The fourth generation of Xeon scalable processor provides a new Intel on-demand activation mode, where customers can turn on additional accelerator combinations in addition to the basic configuration of the original SKU, gaining greater flexibility and more choices when business needs it.
Since Intel Xeon scalable processors now have more and more functions, normal users may not use all functions, so some functions are set to be paid separately to use, similar to the game's DLC. Therefore, some people jokingly call this mode the Xeon DLC.
From a pessimistic perspective, this is Intel's approach to overcharge money, which will increase user costs. From an optimistic perspective, it is possible to reduce the overall price, and users only need to pay for the extra functions they need to use. In this way, it is actually reducing costs.
In addition, Intel has also announced the latest technology roadmap for FPGAs. Intel Agilex FPGAs and Intel Agilex D series and Intel Agilex FPGAs (R-Tile version) all use Intel 7 process technology.
Among them, the performance of the Agilex D series can support mid-range FPGA applications, which is smaller in appearance and lower in power consumption than the current Agilex F/I/M series. Compared with other 7nm FPGA products in the industry, its logic structure performance per watt will be 2 times higher. It is expected to be launched in the second half of 2023 and put into production in the first half of 2024.
Intel Agilex FPGA is suitable for edge, embedded and other application scenarios. It consumes less power and has a smaller appearance than the Agilex D series. Compared with other 16 nm FPGA products in the industry, Sundance Mesa has 1.6 times per watt performance. It is expected to be launched in the second half of 2023 and put into production in the first half of 2024.
In addition, Intel Agilex FPGA (R-Tile version) adopts a Chiplet (core particle) design, and will also support CXL and PCIe5.0, with stronger IO capabilities. Currently, the first batch of equipment for this product is being shipped to early customers.
5, open more to developers
Intel Developer Cloud attempts to enable developers and partners to test and verify Intel's latest platforms earlier:
includes the fourth-generation Intel Xeon Scalable Processor (Sapphire Rapids), the fourth-generation Intel Xeon processor with built-in high bandwidth memory (HBM), Intel Xeon D processor, Habana Gaudi 2 deep learning accelerator, Intel® Data Center GPU (codenamed Ponte Vecchio) and the Intel Data Center GPU Flex family.
Another update is the collaborative Intel Geti Computer Vision Platform (formerly codenamed Sonoma Creek), a software service that helps data scientists and experts in various fields to quickly develop AI models. With optimizations for OpenVINO, development teams can also deploy high-quality computer vision AI solutions in their enterprises.
