Lately, I’ve been shopping for a new motherboard platform to build a new daily driver desktop PC, specifically AMD X870E chipset-based. To my shock and dismay, most motherboards with that chipset only had two or three PCIe slots since many of the other PCIe lanes have been allocated for M.2 NVMe slots instead; it seems very counterintuitive and unnecessarily limiting to ATX motherboards because historically, the huge selling point for them was the freedom of card/peripheral expandability whether it be adding a sound card, extra video card, capture card, faster NIC, or in my case PCIe to U.2/U.3 adapters.
For desktop PCs, it appears that M.2 NVMe SSDs have created a bottleneck in the rate at which storage capacity can be increased and become more cost-effective. It’s highly unlikely that the average consumer needs three to five M.2 slots; if they require so many for storage expansion, U.2 or U.3 SSDs will offer better value since these formats don’t suffer as much from physical limitations in terms of circuit board size and the number of NAND flash chips soldered on. Regrettably, M.2 has become the defacto form factor for consumer desktop PC storage in recent years while U.2 and U.3 are predominantly used in data centers and enterprises.
I believe M.2 is a more sensible choice as a storage form factor for laptops and mobile devices due to their limited physical space, power conservation, and battery life importance—factors less critical for desktop PCs. It’s somewhat absurd that storage components are confined to the motherboard slot size when there’s room in the PC case to accommodate different dimensions. I would prefer U.2 or U.3. because the form factor is in a 2.5” enclosure versus M.2 which comes in form factors of 22x30mm to 22x110mm; in addition, a 2.5” enclosure because of its physical size allows significantly more overprovisioning which also significantly increases the SSD endurance as well as maintaining consistent read/write performance.
Over the past decade, SSD manufacturers have primarily focused on speed rather than storage capacity and endurance for consumer market products. However, many consumers may not realize that advertised sequential speeds are often burst-rate figures, with sustained performance varying significantly. Even M.2 NVMe SSDs can experience a dramatic slowdown when transferring large files of around 25GB or more after just 15 seconds, dropping to speeds as low as a single megabyte per second—outpaced by even older hard drive models.
For the majority of consumers, I argue there are diminishing returns on “faster” SSDs since most applications and use cases do not fully utilize these high-speed capabilities due to software constraints. From my observations, even modern Windows Update processes exceed a few hundred megabytes in both read and write operations, which is substantially below the maximum speeds advertised for recent M.2 NVMe SSDs capable of handling several gigabytes per second transfers. Few consumer applications that can benefit from these peak sequential NVMe speeds—ranging between 5-14 GB/s—include video editing tasks, moving extremely large files in tens of gigabyte sizes, and gaming through Direct Storage technology. Despite the availability of Direct Storage to gamers for over two years, its implementation remains limited, with fewer than a dozen games utilizing this feature according to Steam
Furthermore, as a somewhat enthusiast gamer, I would appreciate more widespread adoption of Direct Storage because it allows for 1-2 second load times in games similar to that of the current-generation consoles such as PS5 and Xbox Series X/S.