6 ms·
I was not expecting the CT scan to be able to identify the cores within the die. But now I think about it, the metal layers (especially traces carrying power)
by phire 11d ago
I was not expecting the CT scan to be able to identify the cores within the die.
But now I think about it, the metal layers (especially traces carrying power) are probably think enough, and different enough to show up. And even just power distribution is enough for the shape of the cores to be visible.
- xpct 11d agoYou can typically see the L1/L2 SRAM cache regions just by looking for continuous patterns, then try to divide it into the number of cores or core groups you're looking for.
- jaen 10d agoThis is only applicable to die photographs of the silicon layer. For CT scans, which use X-rays, the visible structure is very likely due to the metallic (copper) interconnect, clock- and power-delivery layers, which are generally quite different from the underlying transistors. These layers are larger-scale, and copper also absorbs X-rays much more strongly than silicon.
- xpct 10d agoAh, I see. Thank you for the correction.
- userbinator 11d agoYou don't even need a CT to identify the cores; they are already quite obvious in regular die photos as repeated areas. https://commons.wikimedia.org/wiki/File:Intel_Core_i9-13900K_Labelled_Die_Shot.jpg https://commons.wikimedia.org/wiki/File:Intel_Core_i9-13900K...