The new RTX 3000 series GPUs from Lambda offer significant improvements in deep learning performance and efficiency. However, they also come with challenges such as limited power supply, cooling requirements, and potential throttling issues when not properly managed. The RTX 3090 is the most powerful of the three, but its unique dimensions and high power consumption require careful consideration when building a workstation around it. The RTX 3080 and RTX 3070 also offer impressive performance gains, but their power consumption and cooling requirements are more manageable than those of the RTX 3090. Liquid cooling may be necessary for some configurations to mitigate heat buildup, while blower edition cards can provide an alternative solution. Ultimately, building a deep learning workstation with these GPUs requires careful planning and consideration of available space, power, cooling, and relative performance.