video_OCP2026_緯穎_運算電源連接散熱創新_20261015

Hello. >> Okay. Good afternoon, everyone. I'm Lentis from Wiwynn. I'm responsible for the thermal solution development and the new technology. And today, I would like to give you a quick update about Wiwynn's technologies and the what we are bringing to OCP this year. And we also have some technology sessions. So, let's get started with the Wiwynn's technology directions. We have four technology pillars: cooling, power, and interconnect, and the manufacturing. For power, we are developing the solution for the HVDC and the vertical power delivery for the future high-power AI chip and rack with better power efficiency. For cooling, we are working on the 6 kW cycle pay, and we are also working on the on-chip microchannel. And we also have the lightweight liquid cooling solution. For the interconnect, we are working with the industry partners on the optical integration and also the next-generation connectivity. The last one is the manufacturing. We invest in more in the new manufacturing technologies, and we are also expanding our production capacity by more than two times to support the future AI market. These technologies are not just concepts. You can see them at the Wiwynn's booth. At our booth, we have three areas. First is the next-generation AI solutions. We have the NVIDIA Barrel be in Nvidia 72. And we also have AMD Helios rack. The second is the CPO solution. We are working with the industry partners. And we are showcasing a regular CPO solution. This is a lab demo. And the for the cooling solution, we have the double-sided cold plate. And we also have the diamond microchannel. And then we will also show the lightweight liquid cooling solution. So, in the next few slides, I would like to introduce each demo in more detail. On this page, I would like to introduce our demo for the CPO solution. We are collaborating with the industry partners, including IIR labs, GUC, the IC designer, and the same company, Corning. We integrate the CPO accelerator in a full liquid cool rack. This rack supports 64 uh CPO XPU with 800 TB optical bandwidth. This is not just a concept. We have a lab demo at our booth. So, please visit our booth. You will see the real optical engine in action. On this page, I would like to introduce our demo for Nvidia's AI products. Wiwynn and the Wiwynn strong are working together on Nvidia's AI portfolio. We are also the pilot view and the power and partners. This means we can start optimizing the AI platform from an early design stage to mass production. And then we also focus on optimizing the validation and the manufacturing process. This gives us strong readiness for NVIDIA's next generation AI platforms. At our booth, we are showcasing NVIDIA MVL 72. And then we also have various CPU rate. And then we are also showcasing the Spectrum 6. So, please visit our booth. You will see a wide range uh of NVIDIA's AI product from Wewin and the Wishop. So, for the AMD AI solution, we are showcasing a Helios rate and the compute tray. So, there are 72 uh AMD MI250 GPUs on a double-wide the rate. Each tray supports four GPUs. And then we are working very closely with AMD and hyperscale customers. Our work covers from PCBA to compute tray and then the full rate integration. We also build the validation and the manufacturing process for the Helios rate. So, most importantly, Wewin is one of the leading ODM to deliver the Helios rate to the hyperscale customer. So, please visit our booth. You will see our double-wide AMD Helios rate and the compute tray. About the cooling solution, we are showcasing from chip level to module level and the rate level. At chip level, we have on chip microchannel. And then we are also showcasing the advanced material, the diamond microchannel. And then for the module level, we have 6 kW double-sided cold plate for vertical power delivery on a cavity PCB. This is a live demo. We will show you the 6 kW cold plate running at our booth. And then we also have the advanced the cold plate solution. We have the topology optimized cold plate and then we also have two-phase cold plate solution. And then for the system and the rack level, uh we have the lightweight liquid cooling solution. So, at our booth, we will compare the traditional copper cold plate with the lightweight cold plate. So, beyond our booth, we are also sharing our technologies in several technology sessions, including the future technology symposium and also the cooling environment. And then we still have several and the two sessions about the CPU interconnect. And also have the cold plates. So, in the next few slides, I would like to introduce each of the session in more detail. For the future technology symposium, our topic is diamond microchannel. Compared with copper, diamond has two to five times thermal conductivity. So, this is shows the potential high cooling performance for the future high power AI chip. So, we bond the diamond microchannel to the silicon without the traditional thermal interface material. And then compare with the uh compare with the traditional copper cold plate, the thermal resistance can reduce by 38%. This shows a potential high performance for the future high power chip. You can learn more from our presentation and the poster. We also share this technology in the technical white paper. So, you also can scan this QR code to download the white paper about the diamond microchannel. On this page, I would like to introduce two breakout sessions. The first is about the lightweight liquid cooling solution. Because there are more GPUs on a rack, the total rack weight can exceed 4 tons. And it's around 50% weight can come from liquid cooling parts. So, by our lightweight design, we can reduce the weight of the liquid cooling part by more than 30% without compromising the performance and the reliability. This is very important for the future AI rack deployment in the data centers. The second is about the electro-thermal co-simulation. We are working with Microsoft on this simulation. We combine power and the thermal simulation, and then we can get a more accurate result from an early design stage. So, if you are interested in these two sessions, please join these two sessions and have discussion with us. On this page, I would like to introduce two sessions about the CPU solution. We are working with IAR Labs. The first is about scaling the CPU from SK level to class level. We will share design about 1024 CPO cluster. The second is about high power ELSP. We will discuss the key design and then we will also share our co-pay design for the ELSP. So if you are interested in the CPO solution, please join the two session. The last two session is about uh the two-phase liquid cooling. The AI chip power is increasing. The two-phase cooling is one of the potential solution for the future. So the first session is a panel discussion with our industry partners. We will discuss the key challenges and then we will also discuss how we can work together on the industry standards. The second session is about the considerations in transitioning from single-phase to two-phase. So if you would like to learn more about the two-phase solution or you would like to learn about the high cooling performance solution, please join these two session and have discussion with us. So Wiwynn booth is at C74 and then near the central park. Finally, you can scan these two QR codes to learn more about Wiwynn's product and also scan the QR code to download Wiwynn's white paper. Thank you for your time. Hope to see you all at the Wiwynn's booth. Thank you.