It’s only fraud if they sold you or marketed to you on those specs. But at least for things like reflashing your router, short of a few explicit opener vendors (like glinet) and Linksys AFTER releasing the WRTGL version, router manufacturers aren’t usually advertising on how much ram or flash memory space they have, any more than car manufacturers are advertising how much flash memory is in their ECUs. It’s not an intended or marketed purpose, so they’re not going to be changing model numbers just because they made an internal update.
Changing the flash in a router is pretty understandable. Changing a router's CPU is going to affect core performance, and so does changing parts in an SSD, and core performance should totally count as being used to sell the product.
“Core performance” only matters relative to what the company is selling you though. For example let’s say a company sells 2 tiers of switch. One does 10G and the other 1G. For whatever reason when they start selling these, it’s cheapest for them to sell the same internal hardware, but with the internals underclocked in software. Some hardware hackers discover this and start unlocking the 10G capabilities of the 1G units. Later down the road, the company finds a cheaper implementation of hardware for the 1G that still can do 1G but even if up-clocked can now only do 2.5 at best. That’s a change to “core performance” but it’s also not fraud. They didn’t advertise or sell you a “switch that starts at 1G but can be unlocked for 10G”, they sold you a “switch that can do 1G”. As long as that’s what they’re still selling you, everything else is ancillary.
I agree with the upclocking example, but "what the company is selling" goes beyond what's on the box. If the old 1G model can do 500k packets per second, but the new one can only do 200k, that should not qualify for the same model number. There are a lot of situations where that's going to cause real problems on stock settings, after people tested the capabilities and made purchases based on those tests.
I want the most important performance characteristics that would be on a good datasheet to be maintained, even if there is no datasheet.
If you can build a plausible case that you did this (eg. simply making your fw image smaller justifies using a smaller eMMC chip), and provide a few benchmarks that demonstrate equivalent performance in those scenarios, you'd be of the hook in any legal mandate to keep the performance the same even if your new hw revision ships with weaker hardware.
This is even a common product development strategy: ship to market asap, optimize the margins later.
