A little over a year after the first 512GB microSDXC card was unveiled, SanDisk has introduced the first 1TB card. Micron has one too.
The SanDisk Extreme microSDXC UHS-I card should be available in April for $450 and it supports read speeds up to 160MB/s and write speeds up to 90MB/s.
The 1TB Micron c200 microSD card, meanwhile, should be available in the second quarter of 2019 and supports read speeds up to 100 MB/s and write speeds up to 95 MB/s.
These tiny storage cards have four times as much storage as my laptop… but it’s a lot slower than the PCIe NVMe SSD in my laptop. But the SD Association’s got us covered there as well. The group has just introduced a new MicroSD Express standard that will enable microSD cards to hit speeds as high as 985 MB per second.
MicroSD Express uses the PCIe 3.1 interface and NVMe 1.3 protocol to deliver higher data transfer speeds. Cards using the new technology are also expected to consume less power than existing microSD cards.
Of course you’re not just going to be able to pop one of those new cards into your old phone and triple or quadruple your data read/write speeds. You’ll need new hardware that’s compatible with the new standard to take full advantage of the new, faster cards.
But they are backward compatible. So if you have, for example, a camera and/or USB card reader that supports the MicroSD Express protocol and a phone that doesn’t, you can still use the same card with all of those devices. It’ll just be slower in some devices than others.
There’s no word on when the first microSD Express cards or devices that can read them will arrive — so if you’re looking for news you can use out of MWC, I guess you’ll have to settle for the impending arrival of 1TB microSD cards that cost more than a mid-range smartphone.
“… read speeds up to 100MB/s and read speeds up to 95 MB/s”
Interesting typos aside, at a TB of space, that’s like 3 hours to completely fill or read the entire thing. We already have similar bottlenecks with multi-TB HDDs. New gigabyte per second standards can’t come soon enough!
We can pack more than 500x microSD cards in 3.5″ HDD space, so theoretically, we could have 500 TB disc.
Except it would melt in 10 minutes under heavy write load. BTW, I first read your comment as 3.5″ FDD. It took me a second attempt to grasp that it’s hdd, not fdd
I’d like to see a motherboard, where these are soldered on just around the perimeters of the CPU for some quick/direct access stored data.
The CPUs could be soldered on the board, next to RAM that is soldered on for quick access.
The RAM would be encircling the largest entity on the board, which is the GPU that is soldered on.
Data can flow freely from Storage-CPU-RAM-GPU backwards and forwards. A much better APU solution for phones/tablets/laptops/consoles.
What would be the point of direct access? The point of RAM is to have something really fast to read and write to. DDR3 was around 10+ GB/s, and DDR4 can go over 20. So what’s the purpose of having 90-985 *MB*/s storage directly accessible? You’re talking about SATA-like speeds for NVMe level costs. Why not just use soldered-in SSDs in that case?
I think you misunderstood.
I was talking about a PC with the parts unified within the board for a logical layout, instead of the current modular layout. There are a lot of efficiencies at lost using a motherboard to try to connect different parts of the PC together. The best solution would be to have tiny NVMe-life flash chips surrounding the CPUs, not necessarily relying on microSD which is designed to be a user removable medium.
As an off-hand example, the new Chinese Console (Subor Z) actually has a very fast CPU and GPU operation, faster and more efficient than expected/traditional, because they are unified on the board by the 8GB GDDR5 memory. However, it also happens to have less Memory, GPU, and CPU (bottlenecking) resources than your typical Modern PC, which brings its performance down to unimpressive metrics.
Even if I misunderstood, the point still stands. In context, I thought “…these are soldered…” referred to the earlier commenter talking about microSD cards, which is also the context of the article. How was I to infer otherwise?
Regardless, isn’t NVMe still an order of magnitude slower than RAM? You’re making my point by saying that the Subor Z has fast GDDR5 memory, NOT some sort of flash technology. I’m saying that volatile memory (ie RAM) is currently going to be faster than nonvolatile (ie flash) storage.
I don’t think the issue with speeds is about whether they’re soldered down; the signaling will be fine as long as it meets the spec (unless you’re doing crazy overclocking). The bigger issue is what technology you’re using, and for the moment volatile beats nonvolatile.
I’m not a fan of having a single memory solution that does both volatile and non-volatile tasks, because ultimately they have different characteristics (cold storage, write limits, etc etc).
Believe it or not, having a mesh system (storage-processor-memory-graphics) is a faster AND more efficient solution. Basically a complete system, connected together like AMD’s InfityFabric right there on the board/die. The downsides are repairability (difficult) and modularity (extremely limited).
I mean, think back how many times have we increased the spec of connectors on a motherboard just as we’ve hit the bandwidth limits (Sata, DIMM, PCIe, etc etc). That’s just the legacy way of doing things, but its putting us as a distinct disadvantage.
Comments are closed.