NZCity NewsLinks -

All Links

12 Oct 2024

Search

Southern Institute to change Canterbury operation
The Southern Institute of Technology, which has a small campus in Hornby, refuses to provide details about proposed changes in Canterbury.

Innovators and scientists recognised at Beef + Lamb NZ Awards
The best people, technology and innovations in the red meat sector have been recognised at the 2024 gala.

Intel’s new PC chips strike another blow against hyperthreading
Intel’s latest Arrow Lake processor for desktop PCs fuses its “Meteor Lake” and “Lunar Lake” architectures together, carrying over Meteor Lake’s NPU and Lunar Lake’s abandonment of hyperthreading. Yes, hyperthreading has been banned from Intel’s desktop chips, based on a similar rationale for excluding the feature from Lunar Lake. Intel launched Arrow Lake, also known as its Core Ultra 200S processor lineup, on Thursday. The chip is Intel’s first “disaggregated” desktop processor, built on tiles, meaning each part of the chip is individually fabricated on a different process. In a twist, Intel unveiled a deep dive into the architecture of Lunar Lake and the models, prices, and performance of the Core Ultra 200S processor. A key omission? Hyperthreading, which also was not part of Intel’s Lunar Lake mobile processor. The story of Arrow Lake is a simple one: More performance and yet substantially less power than the 14th-generation Core chips. And to get there, Intel executives said they applied the same thinking to both Lunar Lake and its next-gen desktop counterpart: Make its cores as efficient as possible, both for power and for space. What is hyperthreading? Hyperthreading (also known as simultaneous multi-threading) is a fairly simple concept: While each processor core is designed to execute one instruction thread, hyperthreading creates a second “virtual” processor inside the single processor core. With hyperthreading, the idea is that the individual processor core is always executing instructions on at least one of the two cores, keeping it in operation the whole time. The last thing enthusiasts want is a CPU core sitting idle when it could be performing useful work. Intel Intel Intel The problem is that the second core is a virtual core, and not a “true” second physical core. That can lead to some contention of resources and additional overhead, enough that the question of whether to leave hyperthreading on or off while gaming has been a source of debate for years. Intel, meanwhile, has gone back and forth on the feature: Some of Intel’s 9th-, 10th-, and 11th-gen Core processors have excluded hyperthreading, such as the Core i7-9700K, and Intel’s Atom chips never used it. Most of Intel’s Core chips do, however. AMD has pretty consistently used hyperthreading, however, and still does. The question has always been: Does hyperthreading deliver a performance increase that surpasses the toll it takes in terms of system latency, the controller die cost, and the power hyperthreading consumes? In Lunar Lake, the answer was “no,” and that has carried over to Intel’s latest desktop chips, too. In part, that’s because Arrow Lake cribs heavily from Lunar Lake, with the same Lion Cove performance cores and the same Skymont efficiency cores that appear in Lunar Lake. Robert Hallock, a vice president and general manager of client AI and technical marketing for Intel, said that Intel basically comes out ahead in terms of power and performance by not using hyperthreading. Arrow Lake includes both desktop and mobile processors, and Hallock was being asked about the desktop implementation of hyperthreading. But it sounds like Hallock’s response applies to both desktop and mobile chips. “It’s a combination of a couple things, actually,” Hallock told reporters. “First, we knew that we can actually save the wattage for hyperthreading by not including it on the product, and you see that we’re still coming out net ahead by roughly 15, 20 percent in [multicore performance] without it. So we’re able to bump up efficiency and still hit our goals in overall compute performance. “The other thing that I would say is, you know, these are the same designs as leveraged from Lunar Lake,” Hallock added. “We took those cores, those designs, and were able to immediately integrate them because of [Intel’s] Foveros [technology]. So that’s the kind of one-two punch that influenced our decision: speed to market and maximizing performance per watt.” Will hyperthreading ever return? It’s possible it could. But it would have to justify itself in terms of performance, power, and die space, and it appears right now that it isn’t making the cut.

HPE unveils ProLiant XD685 for advanced AI model training
Hewlett Packard Enterprise has launched the HPE ProLiant Compute XD685, enhancing large AI model training with cutting-edge AMD technology and flexible configurations.

AMD pushes the power of Ryzen AI 300 CPUs to business laptops
AMD announced its third-generation Ryzen Pro processors for business laptops on Thursday, taking the AI power of its Ryzen AI 300 consumer processors and applying them to the business world. AMD launched three very similar members of its Ryzen AI Pro 300 family, with core counts ranging from eight to 12 cores. They’ll launch later in October. AMD launched its Ryzen AI 300 laptop processor for consumers earlier this year, and the Ryzen AI Pro is basically that chip with some additional security technologies layered on top. The AI 300 ushered AMD’s new Zen 5 architecture into the market, with more cores and more powerful RDNA 3.5 graphics. The chip also includes AMD’s XDNA 2 NPU architecture, which more than doubles the available TOPS of the previous Ryzen Pro 8040 series from 16 to 50 to 55 TOPS, depending on the model. Otherwise, the Ryzen AI 300 Pro is similar to the 8040 in that it still uses the same 4nm process technology and consumes the same amount of power or between 15 to 54 watts. The new chips include the Ryzen AI 9 HX Pro 375, the Ryzen AI 9 HX Pro 370, and the Ryzen AI 7 HX Pro 360. Note the slight differences between the three Ryzen AI Pro chips, including the difference in TOPS inside the NPU. Note the slight differences between the three Ryzen AI Pro chips, including the difference in TOPS inside the NPU.AMD Note the slight differences between the three Ryzen AI Pro chips, including the difference in TOPS inside the NPU.AMD AMD AMD is banking fairly heavily on AI as a selling point, noting that the Ryzen AI Pro 300 series is currently the only X86 business processor available with the available TOPS to meet the 40-TOPS threshold to meet Microsoft’s Copilot+ PC standard. Intel has yet to announce a vPro business version of its Lunar Lake processor, which also meets the TOPS standard. Qualcomm, of course, was first to market with its AI-powered Windows on Arm Snapdragon X Elite chip, too. Like Intel, however, AMD will require an update from Microsoft to enable its AI capabilities. AMD’s new security features include Cloud Bare Metal Recovery, which helps recover the system via the cloud, AMD Device Identity, which guarantees authentic AMD chips for traceability across the supply chain, and a Watch Dog Timer function, which helps identify and recover processes that “hang” or stall out. AMD is still a minority player in the PC notebook space, but it has steadily increased share over the past year. A year ago, AMD held a 16.5 percent unit market share of all notebook processors, but it’s now just above 20 percent as of the second quarter, according to Mercury Research. AMD said it will ship more than 100 different notebook models with its partners this year that use the Ryzen AI Pro, including HP and Lenovo.