Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
Intel about to launch NVMe SSD.
CalmOceans
Member UncommonPosts: 2,437
Intel countdown clock: http://www.intelgamingpromo.com/intel15b/ssd/notice
They're not SATA express SSD, so you should be able to use them if you have a PCI Express slot.
You're going to need either Windows 8 or 10 I believe though, Windows 7 nor consumer Linux supports NVMe boot. You won't be able to boot from a NVMe if you use Windows 7. Windows 7 might get support in an update, but I am unsure when, I don't believe it is out yet.
The current SATA 3.0 bus is capped at 6Gb/s, in reality current SSD top out at around 540MB/s read, there abouts. It's not down to the SSD though, SSD are faster than that, we just need a new standard to unlock the potential reads. It was initially going to be SATA Express, a true SATA controller, but Intel and Samsung seem to be going for NVMe I believe. There's just many more people with PCIe and the speed is much higher.
It will require an extra controller, but as you can see from this Samsung slide, NVMe will allow a massive increase in bandwidth over SATA 3.0.
There have been PCIe SSD before of course, but none were standardized, and many suffered from lag, because of the extra controller that wasn't optimized. While reads were faster, you lost performance in random reads and writes, NVMe should fix that.
Comments
Btw, their pro grade server NVMe they launched a couple of days ago looks like this.
They also showed a consumer prototype at the PAX gaming event thing a few weeks ago, but nothing official.
No one knows what the consumer SSD will look like, but since it's a NVMe SSD, you should expect to see something similar, it will have a PCIe connection instead of the planned SATA Express connection.
It's possible that SATA Express won't be used at all in the future, and all SSD will look like this for at least a few years until maybe another standard comes along.
It could simply look normal too though, like a SATA SSD, but with a slightly weirder cable that hooks onto your PCIe slot. Anyway, as you can see from this site: http://www.nvmexpress.org/ most of the industry is now behind NVMe, so we should start to see much faster SSD soon.
This was the consumer prototype they showed at PAX, as you can see, it looks like the pro server verion, but it's black and says "prototype".
You can also see it is fitted in what must be one of the PCIe slots below the GPU.
http://be.hardware.info/reviews/6008/5/intel-ssd-750-12tb-review-met-ruime-afstand-de-snelste-consumenten-ssd-ooit!-iometer-4k-random-readwrite
http://www.anandtech.com/show/9090/intel-ssd-750-pcie-ssd-review-nvme-for-the-client/4
2 versions, PCIe and SFF-8639 connector
That's ridiculous to say. NVMe SSD are the future, all newest SSD will use PCIe.
People playing MMO with load zones much faster, it's sequential read is over 2000MB/s, SATA SSD top out at 540MB/s.
The OS will load much faster, files will transfer much faster, games will load faster, doing any type of video editing will be much faster.
This SSD is amazing, no matter how you measure it. Like HW info said in their review. "This is the fastest consumer SSD on the market by a very large margin".
No, all of those things will be limited by something else even on a SATA 2 SSD. Loading a zone isn't purely storage; the CPU has to do quite a bit of work, too. Same with loading an OS. Even a SATA 2 connection is massively faster than gigabit ethernet or 802.11ac, so you only speed up file transfers if you copy from an SSD to an SSD in the same computer without having to go over any network--or if you're on a peculiar network that is massively faster than anything consumer-grade.
If something would have taken a CPU 10 seconds of work and a hard drive 30 seconds of work, then with a hard drive, you sit and wait for 30 seconds. If the SSD can do the storage portion in 3 seconds, then you're waiting on the CPU and so you wait 10 seconds. If you get a faster SSD that can do the storage portion in 1 second, you're still waiting on the CPU, and so you still wait 10 seconds. Faster storage only helps if storage was the bottleneck. With hard drives, that's often the case, but with SSDs, it's not.
There's a good chance that PCI Express SSDs will be more popular in the future. If you can get a PCI Express SSD or a SATA SSD of the same capacity for the same price, then sure, you probably go with PCI Express. But for double the price, when the performance difference amounts to a rounding error? No.
The cost is in the NAND flash. Having a PCI Express SSD that uses the same NAND flash as a SATA SSD doesn't make it cheaper to buy the NAND for the SATA SSD. So no, this isn't going to cause price drops. If it's just as cheap to make a PCI Express SSD as a SATA one, you might see SATA SSDs mostly disappear, though.
It's not as cheap to make as SATA SSD, because NVMe SSD require an extra controller to talk to the PCIe bus, you can see it on the board, it's a bit more complex than what you would see on a SATA SSD, it's a CH29AE41AB0 controller.
That's one of the reasons this is $380 for 400GB, which, considering it's performance, it's a fairly ok price in my opinion.
A lot depends on how much it costs to make that chip. There are plenty of chips on motherboards that cost under $1. And chips that are expensive as low volume enterprise parts can become cheap as high volume consumer parts. The cost of making the masks for a chip is the same regardless of the volume, but spreading it across far more chips allows it to be much cheaper per chip.
Regarding the game loading performance, I haven't seen any benchmarks off that, my thought is that it will knock quite a few seconds off of loading times. I have been waiting for a few weeks for this announcement, reviewers have had these boards for a while, but were under NDA. We should see some loading times benches probably.
The performance you see from this board, is the future of all upcoming SSD. SATA express has pretty much been dropped in favor of NVMe at this point.
So even if you can't afford this $380 board, 2,000MB/s reads and insane random 4k performance will be the norm soon. Because SSD have been capable of extremely high performance for a while, the only thing holding them back was SATA, now that there's an agreed upon standard, SSD will become much faster.
Anyone else reminded of these?
I still have no idea what you are talking about.
These aren't "rounding errors"
4k random from 20 IOps for SATA to 93 IOps for NVMe
sequential read from 500MB/s for SATA to 2000MB/s for NVMe
Intel said about 4 times the performance of SATA at PAX, and they're right. It's a 400% increase in performance.
The only way these are "rounding errors" is if you're drunk.
You know why they focus so heavily on synthetic benchmarks? Because they're the only consumer workloads where you can tell the difference. When I upgraded my own SSD from an OCZ Agility 1 to a Seagate 600, it was twice as fast in synthetics, but no discernable difference in real life usage. (The upgrade was for reliability reasons; at a little over four years, I didn't trust the old SSD anymore. Better to replace storage a year too early than a day too late.)
anyway..
selected motherboard compatibility tests from Intel:
http://www.intel.com/support/ssdc/hpssd/ssd-750/sb/CS-035484.htm?wapkw=750+ssd
more rounding error reviews:
http://hexus.net/tech/reviews/storage/82147-intel-ssd-750-12tb/?page=4
over 5 times sequential read performance over SATA 3.0 SSD
2,700MB/s
If by "rounding error" you mean "synthetic benchmarks unrepresentative of real-world use", then yes. Because the set of programs that a consumer might plausibly run where you'll be able to tell a performance difference of that over a good SATA SSD consists of synthetic benchmarks.
Yes, it is extremely fast in synthetic benchmarks. And also in some non-consumer uses such as hammering on a database all day. But that's not the stuff that consumers care about for reasons of real-world performance.
Really, you called them rounding errors 5 minutes ago.
The performance difference in any real-world circumstances that a consumer is going to care about for its own sake is a rounding error. It's only the synthetic benchmarks where the difference emerges.
Really? Here is simply copying a file in Windows. Real world enough for you?
3 times the speed of SATA drives is a rounding error again?
The only drives that can keep up are non NVMe PCI SSD, but they still get left in the dust by the NVMe solution.
SATA SSD are much slower.
Assuming you're talking about the AnandTech review, it's three synthetic benchmarks that the site created itself at various times, a bunch of Iometer, some AS-SSD, and that's it. Nothing remotely similar to "let's time how long it takes to zone in a game". Review sites gave that up long ago because the results were always "all of the SSDs are essentially tied".
It's possible to make a synthetic benchmark out of consumer programs, and that's what AnandTech does. For example, you could have a bunch of web pages that are heavy on lots of files and big sizes but light on computations, then have a program load them one after the other in a browser, moving on to the next as soon as one loads. And then make your benchmark do that for six separate programs simultaneously and see which SSD goes fastest. But that's not remotely representative of how people will actually use a web browser or any other consumer software.
Because what people tend to do all day is to copy enormous files from somewhere on an SSD to somewhere else on the same SSD. I can't remember the last time I did that; the most recent time may or may not be "never". Remember that if you're copying over a network, you're limited by the speed of the network, which for anything consumer-grade, is going to be massively slower than any decent SSD. Roughly the same applies for copying to or from a hard drive or (non-SSD) USB flash drive, which are the sort of file copies that a consumer might well do a lot of and have to wait on. For small files, an intra-SSD copy is going to be done faster than you're ready for it.
Well , this exactly thing, you could have said it when SSD's have first appeared. "Why someone needs a hard with 540MB/s?"
Let me tell you something, I have seen a BIG increase in daily use ( Gaming and Programing ) when I first bought a SSD , and I am pretty sure that I will see a BIG difference too when the PCIe's will be released.
Just imagine , what this could mean for games ( let alone other departments ). Devs will not be so limited with details anymore. We could see hundreds of player's on screen with little difference in lag , etc.
We already saw "Megaservers" which I heard is capped at 10k players per server, up from 5k ( or so ).
Let's see if we will see a fantasy game in the near future, with EVE like core server ( 50k concurent users in 1 server ) .
That's BIG!
Reporter: What's behind Blizzard success, and how do you make your gamers happy?
Blizzard Boss: Making gamers happy is not my concern, making money.. yes!
The target audience is consumers, especially gamers, high end PC and small workstations. (People at home doing Photoshop or 3D).