When we talk about computing and performance in data centers we must put it in relation to another very important topic, that of efficiency. Have the hardware we need in terms of power to meet the needs of our company is important, but efficiency is key to maximizing its value reducing costs and improving sustainability of our activity, a very important topic that we have delved into in an interview with Alexander Troshin from AMD.
The division of questions that we have asked in this interview is not a coincidence, they have a very important reason for being that will help you understand better. why efficiency is so important When we talk about data centers, to discover what are the most important challenges that a company faces in this regard, what advantages it brings to a company, what advances AMD has made in terms of efficiency and how it faces present and future challenges.
In the answers you can also find multitude of examples and success stories that perfectly illustrate the importance of efficiency in data centers, whose reading I highly recommend. Before getting into the matter, I want to thank Alexander Troshin for sharing his time with us to carry out this interview. Without further ado, he left you with it, happy reading.
(MC Pro) 1.- What are the most important challenges that a company faces in terms of energy efficiency?
(Alexander) Beyond general sustainability goals and plans to reduce CO2 emissions or achieve Net Zero, the key issue is How to keep pace with increasing amounts of data and computational demands while controlling power consumption and data center size.
If to this we add power limitations and rising energy costs, the challenge is considerable. However, it is also a great opportunity, above all, to adopt new technologies and consolidate IT, and we have many products and use cases that demonstrate precisely this.
(MC Pro) 2.- Why is efficiency important in data centers?
(Alexander) Efficiency is the most important thing. Whoever has a computer asset, the ideal would be to use it and will add value in the most constant way possible. In addition, many of the companies’ final products or the processes involved in their production or sale are digitalized. So having state-of-the-art architecture and computing capabilities can directly translate into revenue or time to revenue when taking into account product development cycles, etc.
Now, having a lot of hardware used inefficiently reduces profits. Therefore, having an efficient, high-performance data center can directly translate into faster time to market, more revenue, or higher margins. Of course, this is all at a high level, but we have examples of clients who have already addressed this issue in a similar way. As an example, DBS, a leading digital bank, switched to AMD EPYC CPU-based servers, reducing consumption by 50% with a 10x improvement in growth capacity.
In Europe, STMicroelectronics recently shared the story of how AMD EPYC could help drive a 33% reduction in electricity consumption, that too with better performance and better value per core at the same time. Or look at Emirates NBD Bank who can now run the same number of virtual machines on fewer servers thanks to AMD EPYC technology. There may be different examples, but what unites them all is the ability to leverage the leadership of AMD technology and use the consolidation possibilities that come with it.
(MC Pro) 3.- What advances has AMD achieved in this regard and what technologies have been key to achieving the most important leaps?
(Alexander) Our goal is to drive the sustainability of data centers in general. Our 30 X 25 goal is to multiply the energy efficiency of AMD processors and accelerators by 30, which power servers for HPC and Artificial Intelligence (AI) training between 2020 and 2025. By the end of 2023, we have achieved a 13.5-fold increase in energy efficiency of AMD processors and accelerators from the 2020 baseline, using a configuration of four AMD Instinct MI300A APUs (GPUs with integrated 4th Gen EPYC Genoa CPUs).
There are multiple initiatives and products that are having an impact on this goal and we would like to highlight our “Zen” architecture. With “Zen” we had three main goals: maximum performance, maximum scalability and maximum efficiency. Leveraging leading manufacturing processes with our technology partners and innovative design, we were able to bring the radical idea of x86 processor chiplets to market, and for the past 6 years we have been bringing this innovation and world-leading performance to our partners and clients.
We now have over 300 world records across many tests and workloads, including, of course, energy efficiency, and many customer success stories. In addition, AMD powers eight of the ten most energy-efficient supercomputers in the world according to the November 2023 Green500 list, highlighting how important it is for us to improve performance without losing sight of energy efficiency.
(MC Pro) 4.- What advantages and benefits can a company obtain by improving the efficiency of its data centers?
(Alexander) Considering the required performance, imagine that you can perform the same tasks with the same speed and efficiency, but consuming 30% less energy. Great, now you can also have fewer racks and systems to monitor and manage, while saving on previous results.
You may wonder: how can all this be achieved? Well, at least using the appropriate computer architecture for your use case or requirements. The example above is what we can now offer with just AMD EPYC CPUs versus alternatives on the market. AMD EPYC 4th Generation processors power the most energy-efficient x86 servers, delivering exceptional performance and helping to reduce energy costs.
Of course, the final results and specific figures They depend largely on the plethora of factors, However, we have examples with a 15:3 and even 17:1 ratio. Additionally, we have a wide range of data center products, from CPUs to GPUs, FPGAs, DPUs, specialty accelerators, etc., and a broad ecosystem of solutions from our OEM and ODM partners, each with their unique technology and portfolio. .
(MC Pro) 5.- How does AMD address the medium and long-term challenges that are key to continuing to improve PUE values in data centers?
(Alexander) Regarding computer equipment, when developing new products and technologies we take into account at least two aspects: performance and energy efficiency. We strive to offer with our products the highest computing capacity per energy used, from the traditional perspective of performance per watt.
In terms of non-computing energy use, supporting the ecosystem and the different solutions, and developing partnerships. We already have many technology partners that offer different types of cooling solutions, including liquid cooling and immersion coolingand we have success stories.
I would like to cite two examples: firstly, our great partner LUMI, the fastest supercomputer in Europe according to the Top500 list November last year, powered by AMD CPU and GPU technologies. LUMI uses 100% carbon-free hydropower for its operations, and its waste heat is reused to heat surrounding houses in the town of Kajaani, Finland, the home of LUMI 2ª.
Another example is Shell, which is using our 4th generation AMD EPYC processors with immersion cooling from our technology partner Penguin, increasing performance and, at the same time, support your business and sustainability goals. Its Houston data center already uses 100% renewable energy supplied by Shell Energy North America.