Information

Dear forum members,
We are delighted to inform you that our forums are back online. Every single topic and post are now on their places and everything works just as before, only better. Welcome back!



Discussion

Discussion on Article:
Kingston Speeds Up Cost-Efficient SSD Line with New SandForce Controller and NAND Flash.

Started by: sanity | Date 11/15/12 01:54:57 AM
Comments: 6 | Last Comment:  11/24/12 02:31:46 PM

Expand all threads | Collapse all threads

[1-4]

1. 
"The new drives provide out-of-the-box sequential read and write speeds of up to 450MB/s as well as maximum random read input/output operations per second (IOPS) of 85000."

so whats the big deal here,so what we can finally buy a few MB/s more throughput over a slow PCi-E bus interconnect this year, we already know that antiquated limited write NAND Flash SSDs are reaching their limits already and IMO should have never been used as NV ram or at least retired and transitioned already given the http://www.jedec.org/news...ndard-wide-io-mobile-dram already exists today from several vendors Samsung etc and was known about years before.

the real question in 2012/13 should be so when will we get the real deal extreme power efficiency of "Wide I/O" Mobile unlimited read/write MRAM/FRAM/other real Dram type speeds and super low latency's of NV ram in Wide I/O form to beat even that current up to 17GB/s per single channel spec and beyond never mind the even faster proposed "Wide I/O 2" for general purpose mass markets at a reasonable consumer price ?

hell even ARM inc have a real "AMBA4" Open Specifications http://www.arm.com/produc...ache-coherent-network.php 1 Terabit/s bus and matching CoreLink DMC-520 Dynamic Memory Controller block's already being tested in the worlds foundries http://www.arm.com/images..._CCN-504_system_large.png ,
did the desktop dinosaurs BOD's sleep through all this while milking the PCi-E bus specs for all its worth as their general purpose slow bus interconnect

LOL at the down vote with no reply to counter the speeds and feeds i point out, so i take it you are happy with limited write slow speeds in 2012/13 Kingston want you to buy here then, if so i have a nice XT IDE mode 20gig hard drive for sale , that should suit you down to the ground, you are actually old enough to know what an "XT IDE mode HD" speed is i assume, the futures already here you just don't know it yet.
http://semiaccurate.com/2...-makes-st-mram-a-reality/
0 2 [Posted by: sanity  | Date: 11/15/12 01:54:57 AM]
Reply

2. 
If I'm reading this right, it would seem that these drives are only rated for about 500 P/E cycles. ...yeah, no thank you.
1 0 [Posted by: hsew  | Date: 11/15/12 05:26:27 AM]
Reply

3. 
All good points, Sanity!

Yes, the RACE IS ON for non-volatile RAM.

However, don't forget SATA and SAS cables: the Storage Developer Conference accepted our paper proposing overclocked data storage channels e.g. variable 8G, 12G and 16G "SATA-IV" transmission clocks and replacing the 8b/10b "legacy frame" with the 128b/130b "jumbo frame" now in the PCI-E 3.0 standard:

8G/8b = ~1.0 GB/second per SATA cable.

If you think WAAAAY back, the legacy PCI bus "synced" with ATA-133: 32 bits @ 33 MHz = 1,056 Mbps/8 = ~133 MB/second. Therefore, at a minimum, all future editions of the PCI-E bus should also "sync" with all future SATA and SAS data cables and integrated device controllers.
1 0 [Posted by: MRFS  | Date: 11/18/12 08:19:59 PM]
Reply
- collapse thread

 
im not sure what [our]paper your referring too there, got a link and a timeline handy ?

and what time scales do the "Storage Developer" ? Conference [Council] see for this to be in general availability ?

(im seriously not trying to be negative here, honest) while thats true and its a nice small bump to get ~1.0 GB/second per SATA cable presumably soon...

i thought it time to go and look up the latest RAM speeds and feeds specs for this rough (
PCI-E 3.0 to PCI-E 4.0 standard timeline on the assumption that your ~1.0 GB/second per SATA cable and the x2 ~2.0 GB/s at the 16G "SATA-IV" would generally not be available until close to the PCI-E 4.0 general availability) timeline.

as it happens there is now a reasonable write up posted November 15th, 2012 here
http://semimd.com/blog/20...15/mobile-memory-madness/

regarding next [LP]DDR4 and "wide I/O 2" that is rather jaw dropping, not in the fact the JEDEC membership screwed us yet again, but the fact although "wide I/O" is supposedly available 2013 as originally stated they label it niche-oriented parts now, just so they can insert an extended and slower LPDDR3E part and delay the real mass uptake by 2 years.

they are also artificially restricting "wide I/O 2" speed to the lower spec throughput 25.6-GB/s of LPDDR4-based mobile DRAMs rather than its real 34.4 GB/s at it's base followed by parts at 51.2-GB/s rather than the real dual channel 68.8GB/s.

as you can see, these speeds delayed for 2 or 3 years in the 2015 timeline make that ~1.0 or max ~2.0 GB/second per SATA cable seem lets say rather limited for taking advantage of MRAM type drives... especially by 2015 when you would expect at least 32nm MRAM and other NV ram then.

as i said im really trying to see the bright side here on getting far faster generic data throughput , wheres the cheap self contained home install kit PAN/LAN "fiber link" in a match box for the SOHO/Home desktop, that might just cover or at least improve these 2013-2015 expected speeds throughput to get a lot closer to the top speed perhaps...

as your not likely to get 20/10 raid0 SATA3/4 ports on an average SOHO/Home motherboard....
0 1 [Posted by: sanity  | Date: 11/24/12 01:53:28 PM]
Reply

4. 
p.s. It's also my opinion that storage vendors are trying to ride out the Recession in the USA -- by squeezing as much profit margin as possible out of Nand Flash storage. But, I also fear that they are setting themselves up for a price war, particularly if Intel gets uppity. At last count, there were about 100 Nand Flash vendors competing in that same niche, and almost all are hitting the same "6G Ceiling" with their SATA-III devices. Add to that the reality, well documented by this website, of serious performance degradation on systems that do not support TRIM, and on most RAID arrays even if one's OS does support TRIM. Want TRIM with RAID 0 arrays? Then, your choices are limited to the most recent Intel chipsets, but NOT (yet) supported by the leading add-on RAID controllers.
1 0 [Posted by: MRFS  | Date: 11/18/12 08:32:02 PM]
Reply
- collapse thread

 
yeah id also be of the same opinion on storage vendors but for the fact if you think about it the generic Ethernet vendors have done exactly the same thing for a lot longer, milking the high end, while providing nothing better than crappy 1GBit/s for years now to the SOHO/Home users, that needs to change ASAP.

its also clear that the Nand Flash vendors have put way to much time, effort and, cash into trying to make antiquated limited write NAND Flash SSDs and hybrid HD's work,and cant seem to stop themselves now..and call it a day, where a simple hybrid generic DRAM+generic HD would provide far better throughput since NAND Flash appeared never mind today.

its not like you even need a battery on any hybrid DRAM+generic HD setup as the PC spins up the drive before it can read/write to/from it anyway and writes the ram to the drive before powering down to save power if set that way... and thats been obvious for a long time, IMO , and now we have DDR3 NV MRAM to use everywhere... and since 2004 when Freescale first commercially made it for hybrid HD's , its been out there in real deep space sat's storage for a very long time now for instance.
0 0 [Posted by: sanity  | Date: 11/24/12 02:31:46 PM]
Reply

[1-4]

Back to the Article

Add your Comment