https://www.nextplatform.com/2022/10/20/ibms-ai-accelerator-this-had-better-not-be-just-a-science-project/ In-depth coverage of high-end computing at large enterprises, supercomputing centers, hyperscale data centers, and public clouds. Sun, 02 Jun 2024 15:58:19 +0000 hourly 1 https://wordpress.org/?v=6.5.5 https://www.nextplatform.com/2022/10/20/ibms-ai-accelerator-this-had-better-not-be-just-a-science-project/#comment-225103 Sun, 02 Jun 2024 15:58:19 +0000 https://www.nextplatform.com/?p=141387#comment-225103 Hello i noticed that there might be a miss interpretation in how IBM has communicated the data vs how they implied because IBM states Tflops at 1.6 ghz is 25.6 but in the next line IBM means to state that every watt is 1.9 Tflops not that every Tflop is 1.9 watts. The way ik this is by looking at and comparing the same measurements between Tops numbers and doing the same with A100 and what you should get is that IBM is measuring roughly 100w overall draw for the AIU between A100 at 400w (204.8 Tflops vs 312 Tflops)

]]> https://www.nextplatform.com/2022/10/20/ibms-ai-accelerator-this-had-better-not-be-just-a-science-project/#comment-202451 Wed, 21 Dec 2022 16:57:22 +0000 https://www.nextplatform.com/?p=141387#comment-202451 THE IBM AIU BOARD IS OUTDATED

]]>