Stories
Slash Boxes
Comments

SoylentNews is people

posted by chromas on Thursday September 30, @03:54AM   Printer-friendly [Skip to comment(s)]
from the but-still-no-gpus dept.

AMD wants to make its chips 30 times more energy-efficient by 2025

Today, [AMD] announced its most ambitious goal yet—to increase the energy efficiency of its Epyc CPUs and Instinct AI accelerators 30 times by 2025. This would help data centers and supercomputers achieve high performance with significant power savings over current solutions.

If it achieves this goal, the savings would add up to billions of kilowatt-hours of electricity saved in 2025 alone, meaning the power required to perform a single calculation in high-performance computing tasks will have decreased by 97 percent.

Increasing energy efficiency this much will involve a lot of engineering wizardry, including AMD's stacked 3D V-Cache chiplet technology. The company acknowledges the difficult task ahead of it, now that "energy-efficiency gains from process node advances are smaller and less frequent."

What does it mean?

In addition to compute node performance/Watt measurements, to make the goal particularly relevant to worldwide energy use, AMD uses segment-specific datacenter power utilization effectiveness (PUE) with equipment utilization taken into account. The energy consumption baseline uses the same industry energy per operation improvement rates as from 2015-2020, extrapolated to 2025. The measure of energy per operation improvement in each segment from 2020-2025 is weighted by the projected worldwide volumes multiplied by the Typical Energy Consumption (TEC) of each computing segment to arrive at a meaningful metric of actual energy usage improvement worldwide.

See the 25x20 Initiative from a few years ago.

See also: NVIDIA CEO Jensen Huang to unveil new AI technologies and products at GTC Keynote in November


Original Submission

Related Stories

AMD Succeeds in its 25x20 Goal: 2020 "Renoir" Over 31 Times More Efficient than 2014 "Kaveri" Chips 12 comments

AMD claims to have improved performance by about 5x while cutting power use to about 1/6th, when comparing 2014 "Kaveri" mobile APUs to 2020 "Renoir" mobile APUs. This exceeds a goal of improving efficiency by 25x by 2020:

The base value for AMD's goal is on its Kaveri mobile processors, which by the standards of today set a very low bar. As AMD moved to Carrizo, it implemented new power monitoring features on chip that allowed the system to offer a better distribution of power and ran closer to the true voltage needed, not wasting power. After Carrizo came Bristol Ridge, still based on the older cores, but used a new DDR4 controller as well as lower powered processors that were better optimized for efficiency.

A big leap came with Raven Ridge, with AMD combining its new highly efficient Zen x86 cores and Vega integrated graphics. This heralded a vast improvement in performance due to doubling the cores and improving the graphics, all within a similar power window as Bristol Ridge. This boosted up the important 25x20 metric and keeping it well above the 'linear' gain.

[...] The jump from Picasso to Renoir has been well documented. Our first use of the CPUs, reviewed in the ASUS Zephyrus G14, left us with our mouths open, almost literally. We called it a 'Mobile Revival', showcasing AMD's transition over from Zen+ to Zen2, from GF 12nm to TSMC 7nm, along with a lot of strong design and optimization on the graphics side. The changes from the 2019 to the 2020 chip include doubling the core count, from four to eight, improving the clock-for-clock performance by 15-20%, but also improving the graphics performance and frequencies despite moving down from an silicon design that had 11 compute units down to 8. This comes in line with a number of power updates, adhering to AHCI specifications, and as we discussed with Sam Naffziger, AMD Fellow, supporting the new S0ix low power states has helped tremendously. The reduction in the fabric power, along with additional memory bandwidth, offered large gains.

AMD accomplished this while using refined "7nm" Vega GPU cores in its APUs, instead of moving to a newer architecture such as RDNA2.


Original Submission

Display Options Threshold/Breakthrough Reply to Article Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
(1)
  • (Score: -1, Offtopic) by Anonymous Coward on Thursday September 30, @06:46AM (1 child)

    by Anonymous Coward on Thursday September 30, @06:46AM (#1183055)

    #Freearistarchus!!!! Is this actually spam, if it is true?

    • (Score: -1, Offtopic) by Anonymous Coward on Thursday September 30, @06:54AM

      by Anonymous Coward on Thursday September 30, @06:54AM (#1183058)

      OMG, Really? No body here? Hello?

  • (Score: 2) by Rich on Thursday September 30, @07:14AM (2 children)

    by Rich (945) on Thursday September 30, @07:14AM (#1183059) Journal

    The Three-letter Abbreviation Inventiveness (TAI) is strong in this one. Actual Watt numbers for full load situations would make it a lot less bullshitty.

    • (Score: 0) by Anonymous Coward on Thursday September 30, @10:42AM

      by Anonymous Coward on Thursday September 30, @10:42AM (#1183076)

      True but reducing energy consumption at under-utilization is more efficient in terms of scale than targeting reductions in consumption at heavy loads. So even if this assumption were correct, is an aggregate reduction any less of a reduction?

    • (Score: 3, Funny) by Anonymous Coward on Thursday September 30, @11:55AM

      by Anonymous Coward on Thursday September 30, @11:55AM (#1183085)

      AMD - Advanced Micro Devices

      CPU - Central Processing Unit

      CEO - Chief Executive Officer

      GTC - GPU Technology Conference

      GPU - Graphics Processing Unit

      The other TLA's are already explained in TFS.

      TLA - Three-Letter Acronym

      TFS - The Fiendish* Summary

      *STT

      STT - Season To Taste

  • (Score: 0) by Anonymous Coward on Thursday September 30, @10:38AM (6 children)

    by Anonymous Coward on Thursday September 30, @10:38AM (#1183074)

    This would help data centers and supercomputers achieve high performance with significant power savings over current solutions.

    Do you really think those data centers and supercomputers will actually spend 1 watt less than they did before? I bet they just push more calculations on those systems, to process more than reduce power usage.

    • (Score: 1, Insightful) by Anonymous Coward on Thursday September 30, @11:11AM

      by Anonymous Coward on Thursday September 30, @11:11AM (#1183078)

      Do you really think those data centers and supercomputers will actually spend 1 watt less than they did before?

      Actually yes, they do. Like an IBM mainframe that used 10kW supply and now uses 4kW in same frame is probably going to reduce the energy consumption.

    • (Score: 0) by Anonymous Coward on Thursday September 30, @11:51AM (4 children)

      by Anonymous Coward on Thursday September 30, @11:51AM (#1183084)

      Car analogy:
      AMD is getting ahead of (or giving lip service to) the coming fuel economy regulations (CAFE in USA) for computers.

      More generally, how long before the regulators of car efficiency start to aim at the computing industry? Don't say it isn't going to happen--that's what the car industry said at first, about 50 years ago, but society (and thus government) said different.

      • (Score: 2) by takyon on Thursday September 30, @12:07PM

        by takyon (881) <reversethis-{gro ... s} {ta} {noykat}> on Thursday September 30, @12:07PM (#1183086) Journal

        So is Intel, with Alder/Raptor/Meteor/Etcetera Lake *****mont efficiency cores.

        Also, the car is a computer:

        Intel pushes the European car business [eenewseurope.com]

        Intel chief executive Pat Gelsinger is pushing car makers to more advanced process nodes with the promise of more chip making capacity in Europe.

        In a keynote speech to the IAA Munich car show this week, he sees semiconductors accounting for over 20 percent of the total bill of materials of a premium vehicle by 2030, up 500 percent on 2019 driven by the need for more data processing from cameras and Lidar sensors. Gelsinger predicted the total addressable market for automotive semiconductors will nearly double by the end of the decade to $115 billion, accounting for more than 11 percent of the entire silicon market and he wants Intel to be a significant provider of that silicon in Europe.

        [...] The company announced plans to establish committed foundry capacity at its fab in Ireland and launch the Intel Foundry Services Accelerator to help automotive chip designers transition to advanced nodes. For this, the company is launching a new design team and offering both custom and industry-standard intellectual property (IP) to support automotive customers.

        --
        [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
      • (Score: 1, Insightful) by Anonymous Coward on Thursday September 30, @04:38PM (1 child)

        by Anonymous Coward on Thursday September 30, @04:38PM (#1183127)

        Parent said:
        "Don't say it isn't going to happen--that's what the car industry said at first, about 50 years ago, but society (and thus government) said different."

        "Society" didn't do any such thing. Ask the public if they are demanding another 1 mpg, regardless of addition to vehicle cost to achieve this, or reduction in vehicle interior space, or increase in cost to repair vehicle. Like just about everything, the govt mandates are pushed by small pressure groups.

        • (Score: 1, Touché) by Anonymous Coward on Thursday September 30, @04:42PM

          by Anonymous Coward on Thursday September 30, @04:42PM (#1183128)

          Let's not forget sluggish acceleration to improve MPG.

      • (Score: 2) by shortscreen on Thursday September 30, @08:50PM

        by shortscreen (2252) Subscriber Badge on Thursday September 30, @08:50PM (#1183192) Journal

        After CAFE the public went out and started buying massive SUVs, which seem to outnumber cars these days in many places. Not exactly a resounding success.

        In the computing space, hardware has become orders of magnitude more efficient over time without government intervention. And the public has already responded by running software that is orders of magnitude less efficient to cancel that out. It's too late for the government to step in and turn it into a clown show because it already is one. If they were honest (but not necessarily sane) they would try to regulate the software. That would at least be entertaining to watch.

  • (Score: 0) by Anonymous Coward on Thursday September 30, @07:22PM (1 child)

    by Anonymous Coward on Thursday September 30, @07:22PM (#1183161)

    It would be nice if these improvements trickled down to consumer desktop and laptop processors so nano scale solar would be more practical for off grid living, etc.

    • (Score: 2) by takyon on Thursday September 30, @11:46PM

      by takyon (881) <reversethis-{gro ... s} {ta} {noykat}> on Thursday September 30, @11:46PM (#1183233) Journal

      IDK what nanoscale solar means for your power budget but laptops and desktops are obviously much more performant and efficient these days. You can even get a half decent desktop experience from smartphones, e.g. Samsung DeX, which is being copied by Xiaomi [youtube.com] and others.

      x86 CPUs/APUs in the 5-65 Watt range have advanced tremendously and will continue to do so for at least a few years. AMD and Intel are hanging around the "7nm" FinFET nodes and haven't even gotten to "2nm" gate-all-around field-effect transistors (GAAFET) yet.

      Both AMD and Intel are going to put machine learning accelerators [digitaltrends.com] on die because you don't really need too many CPU cores anymore. It remains to be seen how useful that will be but features [9to5google.com] using it are available. Integrated graphics performance will also go way up, e.g. up to 192 EUs on Meteor Lake desktop chips, and possibly 320 EUs on mobile (Arrow Lake).

      You can pick up a 5700G, 5600H, 5500U, i5-11400, etc. today and have a decent experience. Alder Lake (mobile) and Rembrandt in 2022 will be great upgrades. Whatever is available in 2025 will be incredible, and I think that would still be pre-GAAFET.

      TL;DR get trickled on.

      --
      [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
  • (Score: 2) by Tork on Thursday September 30, @10:14PM (2 children)

    by Tork (3914) on Thursday September 30, @10:14PM (#1183212)
    Question: What does an AI Accelerator do? What sort of problem is being sped up? Is it that different it can't be done on a modern 3d card?
    --
    Slashdolt Logic: "24 year old jokes about sharks and lasers are +5, Funny." 💩
    • (Score: 2) by takyon on Friday October 01, @12:35AM (1 child)

      "AI accelerators" are typically more efficient at accelerating machine learning training and algorithms than general-purpose GPUs. To the point where smartphone SoCs have separate AI/ML/NPU/TPU/whatever cores (taking up their own die space) in addition to the GPU cores.

      For example, look at Apple [wikipedia.org]:

      A11 Bionic (2017): 2-core, 0.6 trillion operations per second (TOPS)
      A12 Bionic (2018): 8-core, 5.0 TOPS
      A13 Bionic (2019): 8-core, 5.5 TOPS
      A14 Bionic (2020): 16-core, 11.0 TOPS
      A15 Bionic (2021): 16-core, 15.8 TOPS

      On-device machine learning capabilities are arguably more relevant to smartphones and augmented reality, but will still be coming to new consumer desktop and laptop CPUs over the next couple of years.

      AMD's Instinct AI accelerators are the competition to products like Nvidia's A100 [nvidia.com].

      For AI training, recommender system models like DLRM have massive tables representing billions of users and billions of products. A100 80GB delivers up to a 3x speedup, so businesses can quickly retrain these models to deliver highly accurate recommendations.

      The A100 80GB also enables training of the largest models with more parameters fitting within a single HGX-powered server such as GPT-2, a natural language processing model with superhuman generative text capability. This eliminates the need for data or model parallel architectures that can be time consuming to implement and slow to run across multiple nodes.

      [...] On a big data analytics benchmark for retail in the terabyte-size range, the A100 80GB boosts performance up to 2x, making it an ideal platform for delivering rapid insights on the largest of datasets. Businesses can make key decisions in real time as data is updated dynamically.

      For scientific applications, such as weather forecasting and quantum chemistry, the A100 80GB can deliver massive acceleration. Quantum Espresso, a materials simulation, achieved throughput gains of nearly 2x with a single node of A100 80GB.

      --
      [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
      • (Score: 2) by Tork on Friday October 01, @03:23PM

        by Tork (3914) on Friday October 01, @03:23PM (#1183392)
        Thank you!
        --
        Slashdolt Logic: "24 year old jokes about sharks and lasers are +5, Funny." 💩
(1)