Invalid input. Special characters are not supported.
Scott Lang
"My days of breaking into places and stealing [stuff] are over! What do you need me to do?"
Hank Pym
"..I want you to break into a place and steal some [stuff]."
Scott Lang
"...makes sense."
Reed, P (Director). (2015). Ant Man [Film]. Marvel Studios
Table 1: Options for increasing SSD capacity
Life imitates art?
I have spent many years involved in form factor development. With EDSFF, I have been involved since its inception and have watched it grow to become a large part of the storage market. Recently, a lot of focus on EDSFF has been more on new features or higher speeds with no form factor changes. So, when I was asked how to get a lower cost per capacity by doubling capacity outside of the NAND cadence, the quote above popped into my head. I needed to start working on a new form factor.
You can’t handle the truth.
When looking at doubling capacity, there are a few options available. Each option has tradeoffs, as shown in Table 1. While this table is not exhaustive, it captures the high-level conclusion that a new form factor is the best solution. assuming certain vectors are met. It will be the cheapest to implement if enough space is allowed for 64 NAND packages and support components. It is thermally the simplest and most efficient if there is a single PCB and enough volume. The downside is that a new form factor with a new chassis is needed.
Are you talking to me?
The argument against EDSFF is that there are too many form factors. This is the downside of a specification where the form factor itself depends on the application. We have three form factors today: E1.S (compute/AI focused), E1.L (1U storage focused), and E3.S (enterprise server focused). There is also E3.L for higher capacity enterprise, although not widely deployed. Adding another form factor is going to incite some controversy, so there needs to be industry support.
Figure 1: E2 high level dimensions
Look, if we built this large, wooden badger…
A new form factor is a compromise between the host system and the device. The dimensions of the device need to satisfy the “ifs” mentioned previously. However, the dimensions need to also satisfy what the host can use based on their box architecture and dimensional constraints. This also requires industry alignment as one company’s design may not meet the needs of others. Over the past several months with other industry partners, conversations boiled down to the following requirements and led to the dimensions shown in Figure 1.
Requirements:
- 64 NAND packages
- NVMe, PCIe, EDSFF (don’t invent something drastically new)
- Leverage what has already been done when possible
Figure 2: Prototype of an E2 SSD
It’s Alive! It’s Alive!
I recently presented this form factor as part of an E2 panel at OCP Storage Tech Talk on May 14, 2025 (1:52:00 mark, LINK). The panel also included Lee Prewitt from Microsoft, Ross Stenfort from Meta and Peter Choi from Pure Storage. I highly recommend watching this panel, but to summarize:
Both Microsoft and Meta discussed the need and requirements for this device. In addition, Meta shared a 3D CAD image of an E2 system.
There is specification work happening in SNIA called SFF-TA-1042 enterprise and data center 2U form factor (E2) and is at revision 0.0.2 (LINK). The goal is to standardize this summer.
Both Micron and Pure Storage showed prototype devices. The latest Micron prototype is shown in Figure 2.
If you build it…
E2 is coming. Over the last few months, the need for a new form factor to solve the problem of the highest capacity at the lowest cost has become very clear. With the industry’s attention to making this a reality, you can expect to see more news and announcements about this form factor soon.
