Don’t these hardware manufacturers already deliberately reduce the performance of some chips just so they can sell them for cheaper without actually developing and manufacturing an additional line of chips (i.e. price discrimination)? This doesn’t seem much different.
De-rated units are usually chips that didn't perform as well as they could have. A variety of manufacturing discrepancies (defects) can cause excess heat dissipation or non-functional cores.
Selecting what a given unit should be rated to (clocked at or enabled to do) is often known as 'binning', and is common throughout the electronics industry.
> De-rated units are usually chips that didn't perform as well as they could have
"usually" is not correct here. They are sometimes chips that didn't perform as well, but usually they are actually higher-tier chips that are artificially locked down to perform at lower levels.
AMD had over 80% of their Zen1 chips and over 70% of their Zen2 chiplets coming off the line with 8 fully functional cores. Yet they probably have 60% of their demand being 6-core or 12-core parts (2x 6-core chiplets). What do they do? Lock off 2 cores and sell 8-core parts as 6-core parts and 12-core parts.
It is nice that they have something to do with the chips that actually are broken, but most of them are just gimped for price discrimination and market segmentation.
Same for the Phenom II X3 - almost all of those really had 4 functional cores and you could usually turn them back on. Why were they being sold as 3-cores? Price discrimination and market segmentation.
Or for a more recent example - the rumors of an impending 3080 Ti launch in January turned out to be true! Someone got a picture of the actual chips, with a big line through the 3080 Ti part number, re-marked with the part number for a 3090. NVIDIA decided to cancel the 3080 Ti launch, and sell the parts as 3090 - and those parts were actually 3090 tier parts that were being marked down and sold as 3080 Tis.
It is a reality in this industry that you are constantly being sold artificially gimped parts, because price discrimination is the financially optimal strategy. It's better to charge a price-insensitive customer way more, and to still have high-volume parts for price-sensitive customers, than to have insufficient inventory to supply price-sensitive customers, or to price everything "fairly" and lose the price-sensitive customers due to prices being too high. But due to the way yields and binning work (you don't want to have high-volume SKUs that are difficult to produce - you want high yields even in your top SKUs) you have the most parts available in the bins that have the lowest volume, and the lowest amount of (broken) parts available in the lower bins that you sell the most of.
That's called binning but it's not as bad as it seems and it reduces waste. The way chips are manufactured all the high end and lower end variants come from the same process, but the company can only guarantee the performance of the higher binned ones.
It’s easy to think of it that way. But here’s another way to think about it: a manufacturer has two assembly lines for two different products which are sold at two price points. As they scale up production, they realize that they could realize significant economies of scale if they combine their assembly lines and only produce the more expensive product, but then apply an “artificial” limitation to half of those products so they can still sell to both markets. Is it somehow wrong for them to offer the same two product lines to the same two markets, but just with a more efficient manufacturing process?
