วีดีโอ

Any specific topic of interest ?

Glad you enjoyed it!

No clue why

You are welcome!!

I recorded this live a while ago.

Nice 👍🏾

Thank you for your interest! I'm using an Intel(R) Xeon(R) Gold 6244 CPU @ 3.60GHz, which supports AVX-512 instructions, for this project. Here’s a breakdown addressing your queries: Type of Computer: Processor: Intel(R) Xeon(R) Gold 6244 CPU @ 3.60GHz Reason for Choice: The cloud platform I am working on uses this chipset, so I don't have the option to use something else. Nevertheless, this is a very capable server CPU with the AVX-512 instruction set, providing powerful vector processing capabilities. Compatibility with M Series Environment (Apple Silicon): The Apple M series processors use a different architecture (ARM-based) compared to Intel’s x86 architecture. While you can run similar computational tasks on Apple M series chips, you'd need to use ARM-optimized libraries and possibly rewrite some parts of the code. Apple’s M1 and M2 chips are very capable, but they don't support AVX-512, so performance characteristics will differ. Compatibility with AMD 7800X3d: The AMD 7800X3d is a powerful CPU with strong multi-threading performance. It does support AVX-512, so you should be able to leverage these instructions for HPC workloads. Using AMD GPU for Matrix Calculations: AMD GPUs can be excellent for matrix calculations, especially with libraries like ROCm (Radeon Open Compute) and frameworks like TensorFlow and PyTorch, which support GPU acceleration. Leveraging the parallel processing power of a GPU can significantly speed up matrix operations and other compute-intensive tasks. Future Plans: I am planning on working with the Intel GPU Max Series on the Intel cloud platform to see how it performs. I will also try to leverage their Gaudi 3 accelerators for some AI work I am planning. This will be featured in one of my next videos, so stay tuned! Summary: While Intel’s Xeon CPUs with AVX-512 provide high performance for these tasks, both the Apple M series and AMD 7800X3d can handle HPC workloads well. Code running on Intel Xeons would run on the AMD 7800X3d with little to no modification due to the compatibility of similar instruction sets. Using an AMD GPU for matrix calculations is also a great idea for leveraging additional computational power. Hope this help!

@@antshivrobotics Thanks this is very helpful.

@@antshivrobotics That Xeon Gold CPU sounds high end. If only Intel released this instead of their of that I9 for gaming and its only 150watts that is so much better.... Does cache matter much for this HPC workloads? Is it better to use the GPU or the Processor for matrix calculations or there any algorithms where it makes sense to use a multicore CPU over a GPU? For instance the Threadripper had been featured on several channels like Gamers Nexus and Linus but not understanding why its so great if you can just buy a GPU like a 4090? Do you think the M series is better for the High Performance Computing since the memory is closer to the CPU? Or is it mainly just going to lower power consumption? When doing these high work loads would a higher end silicon chip like XEON and Threadripper work better than even a M series Pro since they are designed for higher work loads to begin with? A little confused why XEON is so much better than a I9 to begin with... Cheers.

@@RickBeacham Cache absolutely matters for HPC workloads. By memory aligning and using prefetch mechanisms, you can significantly increase speed and reduce the fetch cycles of the CPU from DRAM to cache. Generally, GPUs and accelerators are better for matrix calculations as they support many matrix multiplication (matmul) operations and have multiple cores (analogous to CPU cores) to perform these operations efficiently. Newer versions of Xeons, like the 4th gen and beyond, not only have AVX-512 instructions but also AMX instructions specialized for advanced matrix operations. These instructions are particularly beneficial for AI workloads or intense mathematical computations. For gaming, you likely won't need these instruction sets. Regarding the M series, having memory closer to the CPU is a significant advantage. In multi-CPU configurations (having two CPUs on one motherboard), there is a concept called NUMA (Non-Uniform Memory Access). It involves placing your data in memory closest to the CPU processing it, so yes, this is an advantage for consumer-grade computers. When comparing one CPU to another, the M series may seem better. However, the power of Xeons becomes apparent when scaling them across multiple clusters, running Kubernetes or other orchestration tools to manage these clusters with minimal performance degradation. Running Linux on these systems, which has very little overhead compared to most consumer-based OS, is another advantage. Most Xeons don't have a graphic unit and need to boot using the network (like PXE) or via SSH from another computer, making them less user-friendly for the consumer market. They are designed to be server systems that rarely need a graphical user interface and just work efficiently. Some Xeons do have integrated graphics or you can attach a GPU to get graphic output and modify it to be a gaming module. However, they are primarily designed to operate in server configurations, removing many bells and whistles that consumer hardware needs for convenience. So, while the i9 is excellent for consumer tasks, including gaming, the Xeon is designed for high-performance computing, server tasks, and workloads that require robust, scalable, and reliable performance.

@@antshivrobotics Oh I see so the XEON has been optimized for Server work loads. This is why they are using less power.

I haven't tried to program this board and I also don't think you can. Are you talking about interfacing with this board using python? If you are using a linux system, you can follow the official documentation - You can follow these instructions: ww1.microchip.com/downloads/aemDocuments/documents/OTH/ProductDocuments/SoftwareLibraries/Firmware/MCP2200_MCP2221_CDC_Linux_Readme.txt and then perhaps use pyserial?

sure. I will add that to my list of videos to do.

Thank you so much 👍

Sure. Let me upload the code Shortly.

I have the repo in github and posted the link in the description too. github.com/antshiv/antennaDesign

sorry, I haven't tried interfacing any lcd yet.

I am not sure to be honest. I believe you can simulate different settings and find the one appropriate to your use-case.

I don't think so. The Nrf53 should automatically program the end device if the debug out pins are connected.

I haven't programmed much in C++ tbh.

Yeah, Zephyr is definitely complicated. FreeRTOS is a great alternative and a lot easier to get started with. Use the best tools for the job is my recommendation.

Are you referring specifically to how matter works or in general how those protocols work. For matter, Bluetooth is required to make the initial connection between the matter controller and the matter devices. It programs the necessary nodes, protocol etc. Once that is established the edges are controlled using the protocol established (thread, wifi, etc )

I would also watch some deep dives into particular subsystems

Thanks, will do!

Few videos that might interest you: Mathematical model of a Quadcopoter - Part 1 - th-cam.com/video/9eyQyCqHOyg/w-d-xo.html&ab_channel=ANTSHIVROBOTICS All about hover - th-cam.com/video/rAVJOXzKGKw/w-d-xo.html&ab_channel=ANTSHIVROBOTICS Profile Power - th-cam.com/video/CPQsVjCgq4Q/w-d-xo.html&ab_channel=ANTSHIVROBOTICS