Intel new SSD 34nm?

Status
Not open for further replies.
serbiaNem said:
Similar, but have a dedicated channel for erasing data which cannot be used for writing. So the drive is SIMULTANEOUSLY erasing a section in anticipation, or during idle time well before the writing can cause a bottleneck by waiting on erases (this should be able to improve copying to and from the drive also).
Click to expand...

Largely, this is what the drives are doing. An erase can happen in one chip while another chip is being written to or read from. Two operations can't happen in the same chip, AFAICT.

The problem is that erasures happen so much more slowly than writes. That means that write operations from the host will slow down if they need to wait for any erasures. It'll always be possible to overwhelm the drive, since there's a finite resource and a time to wait.
 
You know thats an interesting concept for the controllers.
have two chips, one for writing, one for erasing. There are engineering obstacles, but pretty much, they would use the same MFT/FAT for knowing which bits they can write/erase. and would report once an operation is complete and only once it is complete. So the write can only write when there is a write command. and the delete will delete (TRIM) once a delete command has been issued. They would never overlap, since the write does not know that it can write to a location until the delete cleared it as empty, if its working on it, then its the same as occupied in fact if it is working on a sector, then that entire secor is reported as occupied until the operation is completed

but again, this assumes that you can send two signals to the storage bits at the same time.
 
Reisen said:
Interesting how this thread has turned into the general SSD discussion thread. I don't mind, but find it funny.
Click to expand...

Yeah, therefore I thought to post it here that OCZ is extending the warranty on its Vertex drives. This is good news, 3 years is quite nice compared to what it has previously been. Competition is hardening up. ;) Link to Dailytech.
 
Intel's warranty has always been three years.
Megalomaniac said:
but again, this assumes that you can send two signals to the storage bits at the same time.
Click to expand...
Which you can't do. I linked to the timing diagrams for a typical flash chip earlier in the thread. If you know the chips a particular drive uses, you shouldn't have a hard time finding the data sheet and confirming the matter for yourself.
 
extide said:
SATA 2 isnt quite the limiting factor YET, but we are getting close...
Click to expand...

How do you then explain the outstanding performance of SSD with PCI-E interface?
Here is the specs of the fusion-io below.

This is way way faster than any SATA SSD, and I don't think the performance bump can be attributed to just the controller or the performance of the NAND flash used.

Large 64-K packets

* 730 MB/s random read
* 630 MB/s random write

4-K packets, random

* IOPS in excess of 102,000 reads per second
* IOPS in excess of 101,000 writes per second

1-kB packet, random

* IOPS in excess of 140,000 reads per second
* IOPS in excess of 115,000 writes per second

Medusa Labs Test Tool Suite proprietary testing software:

4-K or smaller packets, sequential

* 141,000 IOPS for read
* 110,000 IOPS for write

50/50 mix reads and writes on 4-K or smaller packets, sequential

* More than 101,000 IOPS
Click to expand...
 
deanx0r said:
How do you then explain the outstanding performance of SSD with PCI-E interface?
Here is the specs of the fusion-io below.

This is way way faster than any SATA SSD, and I don't think the performance bump can be attributed to just the controller or the performance of the NAND flash used.
Click to expand...

if the fusion IO was this good, why dont they send one to kyle so he can test it out.???
 
deanx0r said:
How do you then explain the outstanding performance of SSD with PCI-E interface?
Here is the specs of the fusion-io below.

This is way way faster than any SATA SSD, and I don't think the performance bump can be attributed to just the controller or the performance of the NAND flash used.
Click to expand...

Actually it can largely be attributed to the controller and the flash used. The fusion-io is way more than just a larger X25-E with a PCI-E interface.
As we've seen with the Jmicron mess, as well as Intel's outstanding performance relative to the others, the controller makes a lot of difference, as does the flash configuration (a 128GB SSD with 2 64GB channels is a lot different than one with 8 16GB channels).
 
deanx0r said:
How do you then explain the outstanding performance of SSD with PCI-E interface?
Here is the specs of the fusion-io below.

This is way way faster than any SATA SSD, and I don't think the performance bump can be attributed to just the controller or the performance of the NAND flash used.
Click to expand...

All of those basically use several SATA drives raided up and then shown to the system as a single logical drive. You can do that with regular SATA2 right now. There still are no (individual) drives on the market that can max out SATA 2.
 
deanx0r said:
How do you then explain the outstanding performance of SSD with PCI-E interface?
Here is the specs of the fusion-io below.

This is way way faster than any SATA SSD, and I don't think the performance bump can be attributed to just the controller or the performance of the NAND flash used.
Click to expand...

As said, such SSDs use RAID to boost speeds. Each Flash chip by itself is much slower than a single HDD, but if you RAID multiple of those chips, say about 10-20 or so, you get much faster speeds. As a necessity it would have to be like RAID 0, so there better not be any chips suddenly failing. Better keep backing things up :)
 
mjz_5 said:
if the fusion IO was this good, why dont they send one to kyle so he can test it out.???
Click to expand...

I had no problem getting my hands on a couple, Kyle just needs to ask. They are amazingly fast/and just as expensive.
 
It's a BIOS issue - if you don't have a "Boot from PCI-E device" option, then how can you boot from it ? Their only option would be to set up a fake SATA device to support it...
 
mjz_5 said:
an interesting article comparing 4 intel SSDs to a fusion IO

http://hothardware.com/Articles/Fusionio-vs-Intel-X25M-SSD-RAID-Grudge-Match/?page=1
Click to expand...

Looks like they used the old drivers for the ioDrive. You can be much more granular with the capacity vs performance adjustments now.

1: Max Capacity, 2: Improved Write Performance (at the cost of approximately 50% capacity) and 3: Maximum Write Performance (at the cost of approximately 70% capacity). We low-level formatted the ioDrive before each new benchmark test with option 1, since we felt this would likely be the most common usage model
Click to expand...
 
extide said:
All of those basically use several SATA drives raided up and then shown to the system as a single logical drive. You can do that with regular SATA2 right now. There still are no (individual) drives on the market that can max out SATA 2.
Click to expand...

Well, if they used several SATA drives raided up, doesn't that prove my point that SATA (2.0) is currently the bottleneck of SSDs? I have seen several comparisons where the X25-M is put to the test against RAM disks which are supposedly much faster than NAND flash, and yet, all the drives seem to have reach a threshold in performance.

My guess is that they reach the SATA 2 ceiling which is around 300MB/s.
 
Again there is no single drive that can max out SATA2. Its no different than hard disks, just because people are RAIDing them up to make them faster doesnt show that the INTERFACE is the bottleneck,
 
extide said:
Again there is no single drive that can max out SATA2. Its no different than hard disks, just because people are RAIDing them up to make them faster doesnt show that the INTERFACE is the bottleneck,
Click to expand...

Sure there is, the SSDs are pretty close as 300mb/s is theoretical speeds, not actual speeds.
 
I honestly think spending any money on current or future Intel SATA SSD drives is a waste. Don't get me wrong, I'm aware of all the Intel reviews and recognize it's the best SATA SSD drive - it's just that I think the enthusiast market is set to shift to PCI-e SSD in the next 6 months, with a step change in performance.

Here's a link to Supertalent's upcoming PCIe SSD featured at Computex a few weeks back http://www.gadgetfolder.com/super-talent-raiddrive-super-fast-pcie-ssd-up-to-152gbs.html. Uses 4 Indulux controllers, PCIe 2.0 8x, and 1GB of cache memory to achieve sequential read speeds of 1.52GBs and sequential write speeds of 1.33GBs. That's over 5x faster than the Intel X25-E (not X-25M). Read that again...5x faster. Best part is they come in capacities up to 2TB, with the 1TB model under $1000. Say that again, 1TB for under a $1000... which is less than $1 per GB. The problem with SATA SSD is they cannot be internally raided for performance above the SATA bandwidth cap (for the current SATA 2 spec that is 300MBs, for the upcoming SATA 3 spec that is 600MBs). That's why your'e not seeing SSD with better performance, even though it's possible just by adding more controllers and RAIDing the hell out of internal memory modules. And even if you RAID a bunch of SATA SSD together (so you're bypassing the individual SATA bandwidth cap) the Intel ICH10R chipset cap is about 700MBs (which has been well documented, including by Kyle...just search the hard reviews) so you're still stuck.

So even if Intel releases an improved drive that offers 300MBs, 400MBs, or even 500MBs (i.e. double) performance, what's the point? SATA SSD performance can't touch what's possible with PCIe SSD. Now before you get all over me about how it's not about sequential read/write speeds, and more about random read/write and IOPS, You are right!! But mark my words, the new crop of PCIe drives released in the next 6 months are going to walk all over SATA SSDs. Just look at the Fusionio PCIe SSD specs someone posted above...the random read/write and IOP specs are an order of magnitude higher than any SSD available on the market today (including the Intel X25-E). Now that Supertalent and others are pushing PCIe cards, we're going to see this peformance pushed down to an affordable level.


Iodrive random read/write and IOP specs:

4-K packets, random

* IOPS in excess of 102,000 reads per second (Intel X-25E only get 35,000....that's 3x more for the PCIe SSD)
* IOPS in excess of 101,000 writes per second (Intel X-25E only get 3,300...that's 30x more for the PCIe SSD)

1-kB packet, random

* IOPS in excess of 140,000 reads per second
* IOPS in excess of 115,000 writes per second

50/50 mix reads and writes on 4-K or smaller packets, sequential

* More than 101,000 IOPS
 
Last edited:
jdyoung75 said:
I honestly think spending any money on current or future Intel SATA SSD drives is a waste. Don't get me wrong, I'm aware of all the Intel reviews and recognize it's the best SATA SSD drive - it's just that I think the enthusiast market is set to shift to PCI-e SSD in the next 6 months, with a step change in performance.

Here's a link to Supertalent's upcoming PCIe SSD featured at Computex a few weeks back http://www.gadgetfolder.com/super-talent-raiddrive-super-fast-pcie-ssd-up-to-152gbs.html. Uses 4 Indulux controllers, PCIe 2.0 8x, and 1GB of cache memory to achieve sequential read speeds of 1.52GBs and sequential write speeds of 1.33GBs. That's over 5x faster than the Intel X25-E (not X-25M). Read that again...5x faster. Best part is they come in capacities up to 2TB, with the 1TB model under $1000. Say that again, 1TB for under a $1000... which is less than $1 per GB. The problem with SATA SSD is they cannot be internally raided for performance above the SATA bandwidth cap (for the current SATA 2 spec that is 300MBs, for the upcoming SATA 3 spec that is 600MBs). That's why your'e not seeing SSD with better performance, even though it's possible just by adding more controllers and RAIDing the hell out of internal memory modules. And even if you RAID a bunch of SATA SSD together (so you're bypassing the individual SATA bandwidth cap) the Intel ICH10R chipset cap is about 700MBs (which has been well documented, including by Kyle...just search the hard reviews) so you're still stuck.

So even if Intel releases an improved drive that offers 300MBs, 400MBs, or even 500MBs (i.e. double) performance, what's the point? SATA SSD performance can't touch what's possible with PCIe SSD. Now before you get all over me about how it's not about sequential read/write speeds, and more about random read/write and IOPS, You are right!! But mark my words, the new crop of PCIe drives released in the next 6 months are going to walk all over SATA SSDs. Just look at the Fusionio PCIe SSD specs someone posted above...the random read/write and IOP specs are an order of magnitude higher than any SSD available on the market today (including the Intel X25-E). Now that Supertalent and others are pushing PCIe cards, we're going to see this peformance pushed down to an affordable level.


Iodrive random read/write and IOP specs:

4-K packets, random

* IOPS in excess of 102,000 reads per second (Intel X-25E only get 35,000....that's 3x more for the PCIe SSD)
* IOPS in excess of 101,000 writes per second (Intel X-25E only get 3,300...that's 30x more for the PCIe SSD)

1-kB packet, random

* IOPS in excess of 140,000 reads per second
* IOPS in excess of 115,000 writes per second

50/50 mix reads and writes on 4-K or smaller packets, sequential

* More than 101,000 IOPS
Click to expand...

that looks really good.. but i doubt it that a 1TB version will only be 1000 dollars.. Maybe a 512GB version.. which will still be amazing
 
jdyoung75 said:
I honestly think spending any money on current or future Intel SATA SSD drives is a waste. Don't get me wrong, I'm aware of all the Intel reviews and recognize it's the best SATA SSD drive - it's just that I think the enthusiast market is set to shift to PCI-e SSD in the next 6 months, with a step change in performance.
Click to expand...

I don't see this happening, for three reasons. The first is that as discussed a page or two back, PCI-E by itself doesn't automatically a drive faster. It offers a bigger pipe, but the drive still needs the ability to pump enough data to fill that pipe. The PCI-E drives now can do this because of (more expensive) higher performance controllers and NAND chips. SATA drives will be pushing the 3Gbps barrier soon, but we'll have 6Gbps soon enough that I don't see it being a major problem for a while.

The second is that the ATA control set is currently a well developed standard that all BIOSes and OSes can work with. PCI-E drives need to implement special drivers and control chips so that the OS can actually make use of them (note that many of the current ones aren't bootable). Time would be needed to develop and settle on a new standard, all of which costs money.

The third is that a big market (and an even bigger slice of the potential growth segment) for SSDs is laptops. While desktops may still be the focus around here, the overall market has overwhelmingly shifted to laptops, and it's in laptops where SSDs give the biggest advantages over existing tech. The performance delta is that much more when compared with current laptop drives, and lower power and no moving parts are big advantages in laptops, whereas they aren't as much for desktops. We may see an eventual move to using PCI-E as the interface for laptop drives, but I doubt that will happen for quite a while. For one thing, there's a huge established base of mechanical HDDs that use the standard SATA interface, and manufacturers will be loathe to switch away from that as long as HDDs hold a price and capacity advantage. Two interfaces adds additional cost, and design complexity. Sticking with SATA means that a manufacturer can offer any drive they want in any standard system. That will be the driving force.
 
jdyoung75 said:
I honestly think spending any money on current or future Intel SATA SSD drives is a waste. Don't get me wrong, I'm aware of all the Intel reviews and recognize it's the best SATA SSD drive - it's just that I think the enthusiast market is set to shift to PCI-e SSD in the next 6 months, with a step change in performance.

Here's a link to Supertalent's upcoming PCIe SSD featured at Computex a few weeks back http://www.gadgetfolder.com/super-talent-raiddrive-super-fast-pcie-ssd-up-to-152gbs.html. Uses 4 Indulux controllers, PCIe 2.0 8x, and 1GB of cache memory to achieve sequential read speeds of 1.52GBs and sequential write speeds of 1.33GBs. That's over 5x faster than the Intel X25-E (not X-25M). Read that again...5x faster. Best part is they come in capacities up to 2TB, with the 1TB model under $1000. Say that again, 1TB for under a $1000... which is less than $1 per GB. The problem with SATA SSD is they cannot be internally raided for performance above the SATA bandwidth cap (for the current SATA 2 spec that is 300MBs, for the upcoming SATA 3 spec that is 600MBs). That's why your'e not seeing SSD with better performance, even though it's possible just by adding more controllers and RAIDing the hell out of internal memory modules. And even if you RAID a bunch of SATA SSD together (so you're bypassing the individual SATA bandwidth cap) the Intel ICH10R chipset cap is about 700MBs (which has been well documented, including by Kyle...just search the hard reviews) so you're still stuck.

So even if Intel releases an improved drive that offers 300MBs, 400MBs, or even 500MBs (i.e. double) performance, what's the point? SATA SSD performance can't touch what's possible with PCIe SSD. Now before you get all over me about how it's not about sequential read/write speeds, and more about random read/write and IOPS, You are right!! But mark my words, the new crop of PCIe drives released in the next 6 months are going to walk all over SATA SSDs. Just look at the Fusionio PCIe SSD specs someone posted above...the random read/write and IOP specs are an order of magnitude higher than any SSD available on the market today (including the Intel X25-E). Now that Supertalent and others are pushing PCIe cards, we're going to see this peformance pushed down to an affordable level.


Iodrive random read/write and IOP specs:

4-K packets, random

* IOPS in excess of 102,000 reads per second (Intel X-25E only get 35,000....that's 3x more for the PCIe SSD)
* IOPS in excess of 101,000 writes per second (Intel X-25E only get 3,300...that's 30x more for the PCIe SSD)

1-kB packet, random

* IOPS in excess of 140,000 reads per second
* IOPS in excess of 115,000 writes per second

50/50 mix reads and writes on 4-K or smaller packets, sequential

* More than 101,000 IOPS
Click to expand...

Any tech purchase is a waste if we all used your logic.
 
PCI-E SSDs will be a niche and always a niche.

I don't see this as a problem as the SATA 3.0 spec just came out, and future SSDs will no doubt be using it in their next reiterations.

Also, I would be buying Intel because:
1) better controllers
2) promising longer life than most competitors

I rather risk my money on having a better controller, better QC, and better life. Not to mention with frequent firmware updates with the other companies.

Still waiting for prices to go down and having most drives ~128GB being the smallest.
 
at newegg, they have 512GB SSD out for a while, and intel is still sitting at 160Gb, how much gain in GB will this new tech. move them up to?
 
512GB, but what type ?
1) JMicron type, which is not worth even to look at (A-DATA XPG).
2) PCI-E type, which costs a fortune and can't even boot.
 
Violin Memory designed a flash SSD with balanced random R/W IOPS in Q4 2008 using a combination of factors.
.

* over-provisioning and fast garbage collection ensured a constant flow of pre-erased flash.
.
* their non blocking RAID RAM cache architecture enables a read to be done from a virtual flash block while an erase-write is still in operation on the same flash block.
 
mjz_5 said:
supposedly intel will release their new drives this tuesday

http://www.theinquirer.net/inquirer/news/1433725/intel-ssds-launch-tuesday
Click to expand...

Previously, high costs, limited capacities and sometimes questionable reliability have made SSDs something of a niche offering, but our sources tell us Intel's upcoming babies will not only be super fast and ultra dependable, but will be able to undercut everyone else in the market price wise.
Click to expand...

That's what I've been waiting for.
 
Previously, high costs, limited capacities and sometimes questionable reliability have made SSDs something of a niche offering, but our sources tell us Intel's upcoming babies will not only be super fast and ultra dependable, but will be able to undercut everyone else in the market price wise.
Click to expand...

I'll sell my 80GB X25-M if a new 60-80GB unit shows a major improvement and it's under the ridiculous price of the X25-E.

Has the rumor mill produced any ideas about stats for a 60-80GB unit using the new tech?
 
Old Hippie said:
I'll sell my 80GB X25-M if a new 60-80GB unit shows a major improvement and it's under the ridiculous price of the X25-E.

Has the rumor mill produced any ideas about stats for a 60-80GB unit using the new tech?
Click to expand...


From what I have read (and conversions of foreign currency to US Dollars), the new "Postville"
named SSD's will be in the $250 range (for the 80GB model) and about 10-15% faster than
the current X25-M.

Not really a breathtaking speed increase
or a stunning price reduction if this holds to be true.
 
aamsel said:
From what I have read (and conversions of foreign currency to US Dollars), the new "Postville"
named SSD's will be in the $250 range (for the 80GB model) and about 10-15% faster than
the current X25-M.

Not really a breathtaking speed increase
or a stunning price reduction if this holds to be true.
Click to expand...

An increase in speed and reduction in price.....sounds like a pretty realistic assessment to me.

Even if they could sell them cheaper, they'll price what the segment will bear.

At that price point, I'll probably focus on my i7 build and wait for the next SSD tier......unless I just can't stand it! :D

Thanks for the info!
 
Sounds like a step in the right direction to me!!! MORE COMPETITION!

Hopefully it's not just an increase in performance, but some more robust controller logic towards self-maintenance.
 
Status
Not open for further replies.
Back
Top