Before we know it SSDs will exceed mechanical drives in terms of price, capacity and density, then continue to grow in a way HDDs never could to the point where we have more fast small storage than we know what to do with
I'm still waiting for 1TB SSDs to go down to around £100. Speed wise, I don't think we need 2Gb/s speeds etc. At least not for home use. These drives with 550MB/s sound good enough, don't think I would notice a difference between 550MB/s and 2000MB/s SSD loading OS.
To replace system RAM it would need not only to be faster but, also have considerably improved write durability.
Agree. 1TB should hit 150quid this year. At times there are already 199 specials. Two 2TB drives will see me drop my spinning disks completely and I'll be all flash for primary NAS. I already have a 1TB 840 Pro and 1.6TB Intel drive due to swapsies for other hardware. It's random IO, 4K speeds and latency you need more than raw bandwidth, but I couldn't tell you how much is 'enough'. NVMe certainly helps cut latency vs ACHI and the 4K/IO are increasing in line with raw MB/s thanks to better controllers and 3D NAND.
Well if I had enough dosh I would get in line for these beauties. Seriously tho IMHO nobody really need 2500 R/W for games or for storage. Standard SATA3 will do fine (SATA 4 anyone?). Like I mentioned in some older thread, running mix of drives in my PC. NVMe is all fine when you work on it, but it's border line lunacy to get it to play games. Unless I dump really big data load between NVMe drive (750s) and RAIDed SSDs (for the moment 4 Mushkin Strikers 960GB in R10) there is no difference in performance when doing that between 2 750s. There is fine smooth progress which is something special. Often it's contest how quickly can I blink. Problems start when I need to quickly copy something from drivepool (spinning HDDs) to NVMe or SSD RAID. That's very painful to watch. I sincerely hope that SSD pricing will plummet further. My wallet is ready, uhm well it's not really, but I will gladly double or triple my SSD count if I can. Plenty of space left on sas expander for more. :evilgrin:
I'm not convinced that there are many home users who need the levels of performance offered by NVMe. It's nice to have, sure, but I'd rather see current levels of performance and lower prices than insane performance and prices to match
So.... why isn't this a SAS drive? At least then the target market would make more sense. What home/office user really needs a 13TB SSD? I'm not saying there aren't people who don't need that much storage, but as a single SSD? Most individuals who need that much storage need it for media collections or a NAS, which hard drives are perfectly fine for. Anyway, seeing as most motherboards these days have a stupidly excessive amount of SATA ports, I think it makes more sense to just get a bunch of smaller SSDs in RAID, JBOD, or whatever else.
It is probably just a sample of what they can do, how mch can they squeeze in. Of course nowadays 2-4TB drives are enough for average home user to let's say put 2 or 4 of those in NAS drive and be happy. I'm sure we will see a price drop on SSD drives every year. Like I said, IMO SATA3 speeds are enough for us home users. Hopefully 1TB drives will cost £100-150 by the end of 2016. Imagine 2TB drives for £200 - I would drop all HDDs and switch to SSD entirely.
Considering Fixstars isn't exactly a well-known company, it doesn't seem like a good business practice to develop something something like this "just because". Since this isn't an enterprise-level product, it will not sell well and anyone could have told them that before it left the drawing board. But what do I know - maybe just selling 1 or 2 of these covers the costs of this. 1TB is way more than enough for the average home user or workstation. 2TB is plenty sufficient for the average media collector (maybe excluding Blu Ray...) or amateur video editor. If you don't play many games or store media on your disks, you can easily get away with less than 120GB. I do agree that the prices will continue dropping and that SATA3 is plenty fast enough for home users. In most everyday tasks, nobody fully saturates the bandwidth of either their SSD or SATA controller. Doing RAID0 with SSDs rarely has any measurable performance increase, except in sequential reads/writes for massive files. Makes me wonder where the bottleneck really lies.
"Cool". I think that's exactly the sentiment this deserves. It's an interesting example of a possible future use case for SSD's, but obviously there isn't really a market for a device like this. Yet. A few years from now, when SSD's are a bit more durable and affordable, these sort of drives could have a few important advantages over HDD's. The idea of a high-capacity storage array without any moving parts, low power consumption, low heat output, and lower space requirements is very appealing. Currently the price difference between HDD's and SSD's far outweighs these advantages, but eventually this will change. I think the first major use case for devices like this will probably be in scientific computation, not necessarily in supercomputing environments, but in high-power workstations. Using monster GPU's like the GTX TITAN series it's possible to put an unbelievable amount of processing power into a desktop, enough to burn through multi-terabyte data sets in a reasonable time frame. I personally worked on a project which needed a 92TB storage array to store the data set we were processing with a desktop unit armed with a high-power GPU. Moving that data from the external storage device into the GPU's memory at the same rate the GPU could process the data was a serious challenge (I actually left the project before we solved that, so I don't know if they ever figured it out). Replacing that external array with 8 drives like this would have been great.
No it won't. RAM chips need to be close to the CPU/GPU that's accessing them. Having them on the mobo but not on the CPU die is already a performance limitation. Putting them on a data access bus would cripple performance. Maybe one day we'll have an architecture where the relationships between CPU + Cache + RAM + SSD/HDD Storage aren't as they are today, but that day is a LONG way off as you're no longer talking about an 'upgrade' to an SSD but the obsolesce of computer architecture as we know and understand it.
Board mounted non-volatile storage that is physically similar to ram is already a thing. If the performance of that storage is sufficient I don't think the the leap to CPU + Cache + Storage is as huge as you make out. If you look at it from a different perspective as a large non-volatile RAM disk then it doesn't really require a fundamental change in architecture. I'm not saying soon, but I do think that if solid state storage increases in performance at a wholly greater rate than mechanical storage ever did the fabled "universal memory" could potentially be on the horizon.
