AMD Introduces Epyc Server Chip, 16-Core Desktop Chip

At this week'due south financial analyst meeting, AMD unveiled a 16-core, 32-thread desktop processor called Ryzen Threadripper—its new Epyc brand for server chips—and introduced its starting time graphics board aimed at the machine learning market.

But I was also glad to see the company unveil a roadmap of successive generations in its CPU, GPU, and server lines, with migration to 7nm and 7nm+ process nodes through 2022, which is crucial for the business firm to regain brownie with business concern buyers. At this bespeak, competitors Intel and Nvidia boss their markets, particularly on the server side, and business buyers demand to be convinced that AMD volition be a long-term actor in order for it to land on the consideration list.

"Immersive and instinctive computing will transform all of our daily lives," AMD CEO Lisa Su said at the conference when defining the visitor's vision for the future. Immersive computing with high-end graphics is all around u.s., but instinctive computing—which involves using huge amounts of information and auto learning algorithms—is just outset to evolve. All of this requires "high-functioning computing," she said, a term she used to describe all sorts of high-stop computing and graphics, not just the HPC or supercomputing market place.

Su talked at length well-nigh investing for "multi-generational leadership" in x86 CPUs, graphics for both PCs and integrated, and software—a large alter for a company whose primary products accept targeted the mainstream or low-end markets. AMD will at present focus on premium products, she said, and noted that while the mainstream accounts for well-nigh of its units, the premium part of the market accounts for most of its revenue and profits.

Her biggest announcement was probably Epyc, the branding for the new line of server fries, which had been codenamed Naples. Su said that the Zen compages, which debuted in the Ryzen desktop chips was "created with the new information center in listen." And while she is enthusiastic for what Zen tin can bring in the desktop and laptop markets, she is even more than excited about what Zen can do in the data centre market. She talked about how AMD has also targeted the data eye market place through Radeon Instinct, a version its next-generation graphics architecture known as Vega—which will provide 25 teraflops of performance—and how these would work together in a vision of heterogeneous computing.

"Today'south data eye really requires heterogenous computing to be successful," she said, describing AMD as the only provider of both high-performance computing and graphics.

CTO Mark Papermaster gave more detail about the mucilage that volition agree the new chips together, also as the firm's process for designing new chips in a way that will "provide sustained innovation going forward." Papermaster described the major features of the Zen and Vega architectures for CPUs and GPUs, most of which had been described previously, and said the firm has to design "not only for performance, just likewise for efficiency."

Tying the fries together is the firm'due south new Infinity Cloth, which connects CPUs, GPUs, memory controllers, and other features within a chip and between fleck sockets. Papermaster chosen Infinity Fabric a "hidden precious stone," and explained how it includes a control fabric that manages sensors on the scrap; tin can regulate operation and security; and tin also work as a information cloth, moving data between the various parts of the system.

He said this allows for "well-nigh perfect scalability" through 64 CPU cores. AMD is likewise supporting some new industry standard interconnects between systems, known as Gen-Z & CCIX, as it pushes for open up standards (since it doesn't have the heft of Intel in the market).

Papermaster said that 1 big challenge for the firm over the next few years volition exist "defying the slowing of Moore'due south Law." He said that through integration, software, and system pattern it would be able to stay at the pace of generational operation improvement, even without frequency improvements.

To that cease, he showed roadmaps for both the graphics and CPU lines through 2022, saying that the team was non only rolling out the current generation of products, merely working on the adjacent two. The plans show moving to 7nm and 7nm+ production modes, with continuing improvements in both raw performance and performance per watt. While these weren't detailed, they were groovy to see.

A 16-core, 32-Thread CPU and a GPU Aimed at Auto Learning

Jim Anderson, Full general Manager of the Calculating and Graphics grouping, said that in addition to the Ryzen 7 and Ryzen 5 products already announced, a lower-cease Ryzen three will ship in the third quarter. More importantly, all five of the top PC OEMs (Acer, Asus, Dell, HP, and Lenovo) volition have Ryzen desktops available for consumers in the market place by the end of the quarter. Commercial desktops should follow in the 2d half of this yr, he said.

For those interested in the "absolute highest performance," Anderson announced a new version called Ryzen Threadripper with 16 cores and 32 threads coming "this summer."

Anderson said that a mobile version of the Ryzen processor, with on-die Vega graphics, volition be available for consumer systems in the 2d one-half of 2022. A commercial version should follow in the kickoff half of 2022. He said the mobile chip will offer l percent more CPU performance and 40 per centum more GPU performance while using 50 percent less ability compared to the company's electric current seventh-generation APU.

Meanwhile, in the graphics world, Raja Koduri, Main Architect for the Radeon Technologies Group, discussed the company'south plans for a new serial of graphics boards based on the new Vega architecture. Koduri noted that the visitor's electric current Polaris architecture GPUs generally address the mainstream and mid-market segments of the graphics market (with graphics boards under $300), and acknowledged that the visitor didn't play at the elevation end.

This will change with the new Vega architecture. Among the features Koduri described were a new loftier-bandwidth cache controller (which can double or quadruple the available retentivity), a new programmable geometry pipeline, Rapid Packed Math (for 16-bit floating bespeak), and an advanced pixel engine. He showed demos featuring much smoother motion in high-end gaming and said Vega will support 4K 60 Hz gaming.

For the professional market, Koduri talked about getting more certifications and supporting SSG, which will allow for a 16GB high-bandwidth cache and up to 2TB of on-board SSG NVME wink memory. He showed demonstrations of this working in real-time ray-tracing and on producing 8K video clips in Adobe Premiere.

Koduri then turned his attention to motorcar learning, where competitor Nvidia has made huge inroads in deep learning with its GPU-based products and its CUDA compages. He acknowledged that "AMD is not even in the conversation today," merely emphasized the company'due south heterogenous computing approach and what information technology calls the Radeon Open Compute Platform (ROCm), which supports running machine learning applications on all of the leading frameworks, including TensorFlow and Caffe.

Koduri demonstrated the new product, which will be called Radeon Vega Frontier Edition, and showed that information technology is very competitive on Baidu'due south DeepBench machine learning training benchmark. He said this volition offer thirteen teraflops of performance at 32-flake, 25 teraflops at sixteen-bit, as well equally upward to 16GB of high-bandwidth memory (HBM2), which is four times that of AMD's current Fury X lath. This is due out in late June. One would assume other Vega-based boards will follow soon.

Koduri said that he believes Threadripper and Naples volition exist disruptive, in part because the CPU bottlenecks will go away, leaving more than room for GPU performance.

AMD Promises 'A New Day for the Datacenter'

Forrest Norrod, SVP and General Director for the Enterprise, Embedded and Semi-Custom Business Group, gave details on the Epyc server transport. He said "datacenter leadership" is the company's No. 1 priority, despite having what he described as "a rounding to zero percentage market share."

Norrod said the visitor'due south previous achievements in server engineering—which include having created the offset 64-bit x86 cores, high-speed coherent interconnects, and integrated memory controllers—are what brand it "plausible… predictable… inevitable" for information technology to once again compete in this market place.

Norrod said that Epyc will include 32 Zen cores, eight retentivity channels, 128 lanes of high bandwidth I/O, and a defended security engine. But while that sounds like a huge chip, Norrod explained that the Infinity Fabric Papermaster described earlier enables the chip to actually be built from four 8-core die, which makes it much easier and less expensive to produce.

Norrod said that in a 2-socket configuration, the scrap could offer 64 cores, 4TB retentiveness, and 128 PCI Express lanes, thus giving it 45 percent more cores, 122 pct more memory bandwidth, and 60 percent more I/O than the Xeon E5-2699A v4 (Broadwell). And he presented a couple of demos that showed it outperforming the Xeon in tasks such equally compiling Linux.

I was more impressed with a demo of how a single-socket version of the chip could outperform a middle of the market dual-socket Intel configuration, which he said accounts for virtually of the market. (Equally ever, I accept vendor benchmarks with a grain of salt, and propose yous practice as well.)

Epyc volition offer the "best value for cease users," he said. AMD is looking for leadership in specific segments of the market; Norrod noted it's probable to get traction at get-go from the bigger datacenter customers who write their own software. 1 big difference between AMD and Intel's arroyo, he said, is that every Epyc will exist "unrestrained, with all of the I/0, memory channel, high-seed retentiveness, security stack, and integrated chipset features supported on all models. (Intel offers certain features but on higher-end models).

Norrod said Epyc offers a simpler architecture compared to Intel, and fabricated a big deal about combining Epyc and the Radeon Instinct platform for motorcar learning. Norrod said Epyc is scheduled to be launched in late June, with more than 30 server models expected to be shipped this year.

Norrod also stressed that this is the starting time in a series of chips, and presented a roadmap with versions called "Rome" and "Milan" (to follow "Naples") between now and 2022, using 7nm and 7nm+ processes. He emphasized that information technology isn't all about core performance, merely rather continuous innovation.

Following a financial presentation, CEO Lisa Su returned to close the issue, and told the financial analysts that while the numbers are of import, "this company is all about the products." It'south certainly skillful for the manufacture to take more contest in CPUs, graphics, and especially in the data center market. Later all, the periods when we've seen the nearly competition have also been the periods when nosotros've seen the most innovation.

Source: https://sea.pcmag.com/feature/15640/amd-introduces-epyc-server-chip-16-core-desktop-chip

Posted by: doyonfrimilt.blogspot.com

Share this post