LSI announces SandForce SF3700 SSD controller – faster and more reliable

vbimport

#1

We’ve just posted the following news: LSI announces SandForce SF3700 SSD controller – faster and more reliable[newsimage]http://static.myce.com//images_posts/2013/11/myce-sandforce-sf-3700-95x75.gif[/newsimage]

LSI today unveiled the third generation of its SandForce flash controller line. LSI claims that its SandForce SF3700 controller provides increased levels of flash performance, better reliability and improved endurance.

            Read the full article here: [http://www.myce.com/news/lsi-announces-sandforce-sf3700-ssd-controller-faster-and-more-reliable-69567/](http://www.myce.com/news/lsi-announces-sandforce-sf3700-ssd-controller-faster-and-more-reliable-69567/)

            Please note that the reactions from the complete site will be synched below.

#2

Hope it will be soon on the market, the 22xx are to old to buy :wink:

But I wonder that Micron, Toshiba and Sandisk will use it, these brands have some nice SSDs with Marvell-Controllers


#3

Next year doesn’t sound soon enough since 1.3GB/s will have become norm by then.


#4

Don´t see that. At least not with interesting prices


#5

[QUOTE=Wischmop;2709812]Don´t see that. At least not with interesting prices[/QUOTE]

It was far more difficult to move from 1MB/s to 100MB/s and then 100MB/s to 550MB/s.

The chips are there already. The bottleneck’s the SATA interface, not NAND.

4GB next year would still be expensive. That’ll be the domain of enterprises and high-end power-hungry users. 550MB/s to 1.3GB/s is far less demanding, far less costly, far less exciting.

Apple’s Mac Pro already has that power. It is expensive not because the chips and controllers are expensive. It is expensive only because Apple brand name costs something like half a trillion dollars.


#6

Yeah, I know Apple politics :wink:

But if the notebook/Mainboard-manufacturers don´t use these faster interfaces on their components the SSD-manufacturers don´t push these standards

And don´t forget, there maybe only less ppl at least need these speeds. A guy that only start Windows and some programs don´t need such a speed. I dunno whether the 4K-speeds are getting much higher with newer standards? Only with more powerful controllers. I think the NAND write speed isn´t limited by the SATA-interface if you use drives with max 512GB

The sequential speed is only important for ppl who have more than one fast drive in the PC because where will you get data from outside the PC so fast?

This means:

  1. You need faster NAND to have an advantage when writing
  2. You need an more powerful controller
  3. No consumer-chipset or onboard-chip supports it, an add-on-card added also costs
  4. How many PCIe-lanes are needed if you want to use many drives? Do you need then a expensive board with biggest chipset?

So I think you underestimate the costs for the complete package of all the things you need :cop:

I have some PCs and laoptops, own over 10 SSDs, I´m not the guy who is spare for these things, but still I think all the Pro-models are to expensive. :wink:

So I maybe won´t spend much more money for things that aren´t give me a big boost in real world scenarios

And I don´t believe the ppl who only have one SSD and one HDD would run and buy those things.

Less ppl buy it = costs won´t fall fast

We will see :flower:


#7

Did any of you really, really need 500MB/s in a desktop at home? For gaming? For encoding movies for personal use?

Storage Review community forums are filled with posts screaming “WHY WOULD ANYONE WANT SOMETHING FASTER THAN HDD?”

That was the justification to block the entry of SSD into popular markets in the first place for as long as possible.

People DO NOT need 500MB/s speed. They do not need 10,000,000 IOPS. They do not need 10,000 IOPS. 100 IOPS was more of the norm for most VERY SATISFIED users.

But the same people still bought SSD. Intel’s, Samsung’s, Toshiba’s, and even Apple’s. People will again buy newer generations of SSD products if they provide 1.5GB/s and 6GB/s for less than what they paid for 100MB/s and 500MB/s.

See what I posted 8 years ago on the coming years of SSD at CDFreaks. Most people were against SSD both here and at Storage Review (and just about everywhere else where English was spoken) and most people claimed SSD was too slow to replace HDD. They said so simply because the sequential speed of writing movie files from SSD to SSD seemed not to reach 100MB/s. Only very few took notice of drastic improvements in IOPS, or the times to read and write thousands to millions of smallest files. Forcing two chips, four chips, even sixteen chips to write at once is something HDD could not match as disks and heads are far less maneuverable, so it took not long to increase sequential write speed from like 1MB/s or 10MB/s or 30MB/s to 100MB/s and then again to nearly 600MB/s. Try to find if there was one single post on the millions of computer hardware websites and technologist forums filled with people proud to be early adopters and self-appointed gurus and power users that predicted this in 2005.

Anyway, to answer more specifically,

  1. You need faster NAND to have an advantage when writing
  2. You need an more powerful controller
  3. No consumer-chipset or onboard-chip supports it, an add-on-card added also costs
  4. How many PCIe-lanes are needed if you want to use many drives? Do you need then a expensive board with biggest chipset?
  1. NAND chips are faster, and are fast enough already and are only going to get even faster. That’s simply the result of investment and more investment and continued investment for 10 years to 30 years. NAND chips are capable of writing at 1GB/s and 4GB/s.

  2. Controllers what you now term as “more powerful” are already in samples and being mass-produced.

  3. Chipsets change and people buy new motherboards with changed chipsets. Virtually all new and future generations of motherboard chipsets emphasize support for at least 2GB/s interface for SSD.

  4. PCI lanes are not exactly physical. Unlike physical HDD drives and ODD drives, SSD drives are easy to design in more than one form factor and with more varied number of chips. Worrying about maximum number of physical drives in a PC is something that was started in the days of 8-inch drives. LP was based on rotating disks of 30cm diameter. NAND can be made into products based on something of 1cm diagonal length. Manufacturers make big boards with expensive controllers out of vanity, or expensive boards with big controllers out of greed. If Apple, a company full of vanity and greed and no technology, can make something like the 2013 MacBook Air with a preinstalled PCIe SSD that can write at 0.8GB/s and with a built-in Wi-Fi chip that can transmit wirelessly at 0.1GB/s in 2013, why can’t anyone create in far better ways in the next years? And even MacBook Air is priced at just a little over US$1,000.

Conclusion: I expect fast and faster PCIe SSDs to be ubiquitously priced at more like US$50, not US$500 or US$5,000. That can’t happen overnight, but you never said it should have been done 8 years ago. Why complain it’s coming so late and so slowly if you have been resisting faster speeds for at least several years?


#8

NAND itself isn’t that fast, the fastest stuff only being able to do 400Mbps. That’s only about 40MB/s. To get the speed several NAND chips are accessed simultaneously.

You can think of the NAND array in an SSD as an 8 - 16 device RAID 0 array.
Depending on your usage pattern, despite SSDs being much faster than an HDD, the SSD is still the bottleneck.


#9

I always look for faster and better stuff, that´s natural.

I wasn´t one of the first SSD-owners, I still decide is it worth buying -for me- or not. But if I thought the price is ok for me I buy it.

I have some 2TB-HDDs in use (not every day but often enough), if I copy some hundred GBs it felt very slowly with 80-150MB/s, so I say Yeah, GIVE ME MORE SPEED :wink:

But for other ppl who don´t do that: The average read speed for SSD is about 500MB/s, even with budget-models. I´m sure in the case you could get 300MB/s-SSDs for much cheaper, most ppl would buy it :wink:

If I look in the recommendation-threads, most ppl buy the budget-ones, not the Pros.

I believe faster controllers are in mass production, but that means it´s cheap. The standard-SATA-controller is on cheapest mainboards.

I don´t see these faster controllers on “normal” Mainboards, but this is important for new technologies.

I know ppl who would spend much money for CPU, much for videocard, much for monitor and drives, but they won´t buy an add-on-controller card

I bought 256GB-SSDs for about 140€ one year ago, and it wasn´t the cheapest. If I look 2day at prices here, the cheapest one is about 130€

That means the prices for “slow” SSDs going down very slowly and for many ppl it´s still to expensive.

I would hope you´re right with your price-prediction but I still think it will not go the fast way.


#10

[QUOTE=Dee;2710263]NAND itself isn’t that fast, the fastest stuff only being able to do 400Mbps. That’s only about 40MB/s. To get the speed several NAND chips are accessed simultaneously.
[/QUOTE]

Kenshin sure knows that, me also :iagree:

That´s why I mention: I think the NAND write speed isn´t limited by the SATA-interface if you use drives with max 512GB

If you take more channels it will get faster, but you need to find enough ppl who would pay for it :stuck_out_tongue:

But the chips will be bigger with every new step and won´t get fast enough to compensate this in near future.

That´s the reason why my M4 128GB do 190MB/s writing and the newer 120GB M500 only 120MB/s :wink:


#11

[QUOTE=Dee;2710263]NAND itself isn’t that fast, the fastest stuff only being able to do 400Mbps. That’s only about 40MB/s. To get the speed several NAND chips are accessed simultaneously.

You can think of the NAND array in an SSD as an 8 - 16 device RAID 0 array.
Depending on your usage pattern, despite SSDs being much faster than an HDD, the SSD is still the bottleneck.[/QUOTE]

I said NAND is fast and faster because people look at it the way they look at HDD when NAND has very little to do with HDD. NAND chips read and written at once is very much different from RAIDed HDD. They are on the same physical board just like the 8 DRAM chips on a single DRAM module.

After all, most people, and I mean 99.99% by that, cannot understand what NAND is. What you can find on the web on the differences among SLC, MLC, and TLC, can be more confusing than telling. The same data can be interpreted in several ways. Some people interpreted data available on the web on NAND to mean NAND is slow and meant early deaths. Well, 99.99% of people were wrong. The data meant little. You just chose to believe something and then chose the part in the published data as evidence.


#12

Yeah, I´m one of the 0,01 who understand that :slight_smile:


#13

[QUOTE=Wischmop;2710932]Yeah, I´m one of the 0,01 who understand that :)[/QUOTE]

What I actually meant is lab engineerss (working to invent and innovate on NAND) rarely understand and analysts of the largest firms usually scratch the surface only, just repeating words from others forever.

When you see results like 554.04MB/s for certain files on certain configurations, it means a certain performance for a certain combination only. It needs not be interpreted to prove anything else.

Most of the good’s and bad’s are mostly subjective, and can often be proved the opposite by people with opposing agenda.


#14

LSI just got bought out by Avago,

Semiconductor company Avago Technologies has agreed to buy storage and networking company LSI for US$6.6 billion in an effort to expand its data center offerings.
The deal involves Singapore-based Avago paying $1 billion in cash, an equity investment of $1 billion by Silver Lake Partners and $4.6 billion in term loans from banks.