Overclocking.guide's der8auer has recently posted a story regarding the recently discovered ability to "liberate" AMD's RX 460's shaders - from the Polaris 11 architecture's stock 896 shaders / 56 TMUs to a grand total of 1024 stream processors and 64 TMUs. We did some quick testing and found the mod to be working as promised.

The process is straightforward enough. First, make sure to grab TechPowerUp's own GPU-Z, and save a copy of your original BIOS by clicking the arrow next to the "BIOS Version" field, so you have a fallback in case things go wrong. Then, follow the source link towards overclocking.guide's RX 460 tested BIOSes (currently only for the ASUS STRIX O4G and the Sapphire Nitro 4G), and download the appropriate one. Then run "flash unlocked bios.bat" to flash the BIOS, and in about 15 seconds, the process should be complete, granting you about 10% of extra performance. In our own testing, using the power testing setup we use in graphics cards reviews, we saw a 4 W increase in peak power consumption.The unlock is currently only tested on 4 GB versions of the RX 460, but considering the only difference stands in the memory amount, no reason should exist for those not to be (eventually) unlocked as well. Due to AMD's newfound interest in verifying modded BIOSes (on Polaris architecture cards) with their latest ReLive driver release, it is hard to say whether or not this might pose a problem for your driver installations in the future. For the time being, and in our own testing, attempting to run AMD's Crimson ReLive netted us a negative response, with the failed hash check leaving us only with the default display driver being loaded. This went away with AMD's previous 16.11.5 driver suite, so take care if you have the Crimson ReLive driver software suite installed on your system.The fact that this is even possible also begs the question as to why is it so - are the yields not good enough on the RX 460 chips that AMD is being forced to "lock" portions of a higher-end chip? I would wager this isn't likely, since Polaris 11 "Baffin" is such a small chip already that yield issues shouldn't have such effects. Most likely, we'd guess that AMD is re-purposing the mobile versions of the RX 480 (RX 480M, with some now suspicious-looking 1024 stream processors) that haven't passed power requirements for the 35 W those are rated at - but which surely correctly function at the RX 460's "up to 75 W" TDP. That might mean that AMD designed the original "Baffin" with 1024 stream processors for mobile devices, and carved the RX 460 as a way to recycle eventual defective dies that would surface for that particular SKU, not passing certification. But yields being better than expected, and demand for RX 460 cards outweighing the defective chips' output, AMD was forced to soft lock the extra stream processors on the "Baffin" SKU to satisfy demands.