How much faster are the SSDs?

Win 8.1, MB GA-X58A-UD7-2Rev2, GeForce GTX480Ultra, RAM
12GBOCZ3X1600R2LV6GK, IntelCore i7 970,

I installed an "Intel 530 series 120GB" SSD directly to the MB, but
the speed is no greater than any of my many HDDs.

Is there some special connection required?


Also, I have run out of Sata sockets on the MB. Can I adapt somehow
the PCIEX16 slot to give me more?

Peter

Peter Jason wrote:
Win 8.1, MB GA-X58A-UD7-2Rev2, GeForce GTX480Ultra, RAM
12GBOCZ3X1600R2LV6GK, IntelCore i7 970,

I installed an "Intel 530 series 120GB" SSD directly to the MB, but
the speed is no greater than any of my many HDDs.

Is there some special connection required?


Also, I have run out of Sata sockets on the MB. Can I adapt somehow
the PCIEX16 slot to give me more?

Peter

http://www.newegg.com/Product/Produc...82E16820167180

SSDSC2BW120A4K5
120GB
SATA III --- only needed if the device actually goes that fast

Gave up on getting a benchmark from the review section for a single drive.

http://ark.intel.com/products/75611/...-6Gbs-20nm-MLC

Launch Date Q3'13
Sequential Read 540 MB/s --- reads tend to be pretty good, always
Sequential Write 480 MB/s --- usually a function of capacity, 120GB is low end
Random Read (8GB Span) 24000 IOPS
Random Write (8GB Span) 80000 IOPS
Latency - Read 80 us (target 25us to 100us - 25us is flash delay)
Latency - Write 85 us
Power - Active 140 mW --- fanbois say it's Sandforce, cannot be this low
find a review, with power measurements
Power - Idle 55 mW --- Um, sure buddy

Best motherboard ports, in declining order:

1) Intel PCH (Southbridge) SATA III
AMD Southbridge SATA III
2) Add-on SATA III motherboard chip
3) Marvell 91xx version of (2) - limited to ~300MB/sec
4) Any SATA II port - maybe ~200MB/sec

Best rotating hard drive as of today = 220MB/sec (Seagate 6TB, non-shingled one)

A low-capacity SSD may write at 200MB/sec and read at the
SATA III bus limit. It depends on whether the controller
channel population, uses more channels for the 240GB or
480GB models, as to whether write rate saturated the
SATA III limit. (Maybe the 120GB drive has four flash
channels, and the 240G/480G SSD drives use eight channels.)
You need to find a review of the product somewhere - the
reviewers in the Newegg section, didn't give the quality
of test results I was looking for.

The IOPS (Input Output Per Second) rate should become
more apparent, if you do a Windows search for which the drive
is not indexed. That operation should go faster.

The Windows desktop file system doesn't have a lot of headroom,
which means in many situations where you'd hope to see 10,000
files a second processed, you still only see a couple hundred
files processed per second. The file system is a "bottleneck
to being impressed".

Run HDTune read benchmark, so you can feel better. That
should help tell you whether you've got a SATA II hardware
port on the motherboard...

http://www.hdtune.com/files/hdtune_255.exe

The market offerings on expansion hardware are
*really really weird*. Yes, a PCIEX16 should be
an excellent source of connection point. And, there
are some high-port-count chips for connecting
SATA. Trouble is, the manufacturer goes from
charging $10 a port on a dual port chip, to quite
a bit more on an eight port chip. The higher port
count chips are priced for usage on business RAID
cards. Which leaves a real hole when it comes
to a user solving the "me got no stinkin SATA III"
ports problem.

You can see here, $250 can give you hours of hair-pulling fun.
Check out the customer reviews. They flash the onboard firmware,
to change between RAID and JBOD (target mode). Just like Silicon Image
cards of long ago. SAS cards connect to SAS or SATA drives.

http://www.newegg.com/Product/Produc...82E16816118112

"4 Crucial M4s in raid 0 are giving me ~ 1800mb/sec"

If instead of going SAS controller based (and dealing
with drivers and firmwares), I could try SATA cards.
I see just one here, and the PCI Express interface is
x2. That means, roughly one or two SSDs might be a good
choice for this four port card. It's not making good usage
of the PCI Express slot. The market simply refuses to
put more PCI Express lanes on these chips. Just as the
two port chips, used a x1 lane interface. It cramps
your style, and makes spending a lot of money a waste
of time - lots of return postage when dissatisfied.

http://www.newegg.com/Product/Produc...82E16815158365

http://www.startech.com/Cards-Adapte...ard~PEXSAT34RH

Chipset ID Marvell - 88SE9230 --- not a 91xx chip, but research this...

Supported RAID Modes JBOD - (Just a Bunch of Disks) --- single drive, for you...
RAID 1 (Mirrored Disks)
RAID 0 (Striped Disks)
RAID 10 (1+0; Striped set of Mirrored Subset)
Type and Rate SATA III (6 Gbps)

Throughput Benchmark

1 x SSD 490.28 MB/S 120 GB --- could be slightly better...
beats the 300MB/sec 91xx generation

If you connect two SSDs and do disk to disk transfer
between them, they should do 490MB/sec, because the
PCI Express bus is full-duplex, one drive "reads",
one drive "writes", the PCI Express x2 read bus and
the PCI Express x2 write bus run at equal 490MB/sec
rates. If you run two drives in RAID0 however,
and read off the array, don't expect to get
exactly 980MB/sec from that. And if you connect
four drives in RAID0, you're likely to still be
limited to about 980MB/sec range. As the x2 bus
interface is the limitation (Rev2, 500MB/sec bus
rate per lane). That chip really should have
had an x4 or x8 interface. And bumped up the
internal processor power another notch.

Suffice to say, only the really really well prepared
home-builder, gets what they paid for. I'd probably
get screwed, if I went to my local computer store
and just did an impulse buy on the first shiny thing
I saw. I'd end up at 200MB/sec in all probability,
with just impulse buying crap. You have to be really
lucky or really good, to add crap to an old motherboard
and get 500MB/sec SSD performance from it. The market
doesn't have the hardware to offer the home users.
And hackery on a SAS controller is going to try your
patience. There isn't a large population of SAS
hackers, ready to answer your every question.

HTH,
Paul

When talking about the system's "hard drive", you should consider
overall throughput. The fact that the system "disc" is used more than
anything else throughout the day means that any significant improvement
will do wonders.



Peter Jason wrote:

Win 8.1, MB GA-X58A-UD7-2Rev2, GeForce GTX480Ultra, RAM
12GBOCZ3X1600R2LV6GK, IntelCore i7 970,

I installed an "Intel 530 series 120GB" SSD directly to the MB, but
the speed is no greater than any of my many HDDs.

Is there some special connection required?


Also, I have run out of Sata sockets on the MB. Can I adapt somehow
the PCIEX16 slot to give me more?

Peter

Peter Jason wrote:
Win 8.1, MB GA-X58A-UD7-2Rev2, GeForce GTX480Ultra, RAM
12GBOCZ3X1600R2LV6GK, IntelCore i7 970,

I installed an "Intel 530 series 120GB" SSD directly to the MB, but
the speed is no greater than any of my many HDDs.

Is there some special connection required?


Also, I have run out of Sata sockets on the MB. Can I adapt somehow
the PCIEX16 slot to give me more?

Peter


Daft question, but most people who install an SSD immediately start
praising them.
Well, the the question is this; "Did you make the SSD your boot disc?"


You should be ok with a plug-in for your PCI slot. There are lots
available; some with only internal sockets, some with only external,
some with about two of each. Here's a typical one;
http://tinyurl.com/kdbuz7b

Ed

On Mon, 11 May 2015 19:01:54 +0100, Ed Cryer wrote:

Well, the the question is this; "Did you make the SSD your boot disc?"

Say what? I never created such a thing and I can't complain about my
SSD's speed...

--
s|b

On Mon, 11 May 2015 14:40:44 +1000, Peter Jason wrote:

I installed an "Intel 530 series 120GB" SSD directly to the MB, but
the speed is no greater than any of my many HDDs.

Is there some special connection required?

SSD's are pretty fast, but be warned!

https://blog.korelogic.com/blog/2015/03/24#ssds-evidence-storage-issues

| A stored SSD, without power, can start to lose data in as little as a single week on the shelf.

--
s|b

On Mon, 11 May 2015 14:40:44 +1000, Peter Jason wrote:

I installed an "Intel 530 series 120GB" SSD directly to the MB, but
the speed is no greater than any of my many HDDs.

Is there some special connection required?

Have you set AHCI in the BIOS?

--
s|b

s|b wrote:
On Mon, 11 May 2015 14:40:44 +1000, Peter Jason wrote:

I installed an "Intel 530 series 120GB" SSD directly to the MB, but
the speed is no greater than any of my many HDDs.

Is there some special connection required?

SSD's are pretty fast, but be warned!

https://blog.korelogic.com/blog/2015/03/24#ssds-evidence-storage-issues

| A stored SSD, without power, can start to lose data in as little as a
single week on the shelf.


This is an example of the Intel input.

http://www.aecouncil.com/Workshop/4A...evan-Intel.pdf

The only anecdotal info to date, is a problem with TLC.
None seen with MLC that compares. In the case of TLC,
one particular product becomes "slow" on data which hasn't
been "refreshed" in the last three months. No data is
lost. It's almost like an ECC error correction process
is taking time. Or, the device is reading the block
a second time (less likely). It could be, that the
controller implements its powerful ECC, using software
and one of the SSD processor cores (some SSDs have
two or three cores). No data loss is involved.
I didn't see any mention of "sparing out a block"
either. No evidence it affects lifespan of the
device, over and above adding extra writes
to "freshen" blocks like that.

The cure, was two different firmware releases. The
first release didn't fix the problem (some users were
still seeing a slowdown). The second release was
relatively recent. And this was for TLC. And this
incident is why I don't recommend devices based
on TLC (3 bits per cell, eight level thresholding).
As long as there are no reports of a slowdown on
MLC drives sitting on a shelf, there is no reason
to panic.

I'm sure Intel has done their homework. But to relate
another similar prediction, years ago there was a
theory that *reads* could wear out the device. And
there is no evidence of that, from anecdotes. (Like
someone leaving a device reading repetitively for
long periods of time, and noting that damage occurred.)
So while Intels prediction is certainly conservative,
we need more stories about doom and gloom to
really believe it. If the devices are that
fragile, they have no business selling them!

Paul

On Mon, 11 May 2015 21:34:21 +0200, "s|b" wrote:


SSD's are pretty fast, but be warned!

https://blog.korelogic.com/blog/2015/03/24#ssds-evidence-storage-issues

| A stored SSD, without power, can start to lose data in as little as a single week on the shelf.


I wouldn't be very concerned about shelf life of the data, but much
more about loss of data during vacation travel when the computer's
power is off.

"s|b" wrote:

Ed Cryer wrote:

Well, the the question is this; "Did you make the SSD your boot
disc?"

Say what? I never created such a thing and I can't complain about my
SSD's speed...

But that's where SSDs shine (most brightly), as the boot disk, as the
Windows disk.

It's a good question.

Paul wrote:
s|b wrote:
On Mon, 11 May 2015 14:40:44 +1000, Peter Jason wrote:

I installed an "Intel 530 series 120GB" SSD directly to the MB, but
the speed is no greater than any of my many HDDs.

Is there some special connection required?

SSD's are pretty fast, but be warned!

https://blog.korelogic.com/blog/2015/03/24#ssds-evidence-storage-issues

| A stored SSD, without power, can start to lose data in as little as
a single week on the shelf.


This is an example of the Intel input.

http://www.aecouncil.com/Workshop/4A...evan-Intel.pdf

The only anecdotal info to date, is a problem with TLC.
None seen with MLC that compares. In the case of TLC,
one particular product becomes "slow" on data which hasn't
been "refreshed" in the last three months. No data is
lost. It's almost like an ECC error correction process
is taking time. Or, the device is reading the block
a second time (less likely). It could be, that the
controller implements its powerful ECC, using software
and one of the SSD processor cores (some SSDs have
two or three cores). No data loss is involved.
I didn't see any mention of "sparing out a block"
either. No evidence it affects lifespan of the
device, over and above adding extra writes
to "freshen" blocks like that.

The cure, was two different firmware releases. The
first release didn't fix the problem (some users were
still seeing a slowdown). The second release was
relatively recent. And this was for TLC. And this
incident is why I don't recommend devices based
on TLC (3 bits per cell, eight level thresholding).
As long as there are no reports of a slowdown on
MLC drives sitting on a shelf, there is no reason
to panic.

I'm sure Intel has done their homework. But to relate
another similar prediction, years ago there was a
theory that *reads* could wear out the device. And
there is no evidence of that, from anecdotes. (Like
someone leaving a device reading repetitively for
long periods of time, and noting that damage occurred.)
So while Intels prediction is certainly conservative,
we need more stories about doom and gloom to
really believe it. If the devices are that
fragile, they have no business selling them!

Paul


To supplement your fine response

Acronyms explanation

SLC (Single Level Cell) - highest performance, at a very high cost,
enterprise grade NAND
~ 90-100,000 program/erase cycles per cell (highest endurance)
- lowest density (1 bit per cell, lower is better for endurance)
- lower power consumption
- faster write speeds
- much higher cost (3 times higher than MLC)
- good fit for industrial grade devices, embedded systems, critical
applications.

eMLC (Enterprise Multi Level Cell) - good performance, aimed at
enterprise use
~ 20-30,000 program/erase cycles per cell
- higher density (2 bits per cell)
- lower endurance limit than SLC, higher than MLC
- lower cost
- good fit for light enterprise use and high-end consumer products with
more disk writes than MLC

MLC (Multi Level Cell) - average performance, consumer grade NAND
~ 10,000 program/erase cycles per cell
- higher density (2 or more bits per cell)
- lower endurance limit than SLC
- lower cost (3 times lower than SLC)
- good fit for consumer products. Not suggested for critical
applications which require frequent updates of data

TLC (Three Level Cell) - lower performance, lowest cost NAND
~ 3-5,000 program/erase cycles per cell
- highest density (3 bits per cell)
- lower endurance limit than MLC and SLC
- best price point (30% lower than MLC)
- somewhat slower read and write speed than MLC
- good fit for lower-end consumer products. Not recommended for critical
applications which require frequent updating of data


Also for those interested

Micron web page article has a video on SLC, MLC, TLC (What's the Difference)
http://www.micron.com/products/nand-...nd-slc-devices

Above video (direct link) on Youtube
https://www.youtube.com/watch?v=wOLkEfGT7XM


--
...winston
msft mvp consumer apps

John Doe wrote:
"s|b" wrote:

Ed Cryer wrote:

Well, the the question is this; "Did you make the SSD your boot
disc?"

Say what? I never created such a thing and I can't complain about my
SSD's speed...

But that's where SSDs shine (most brightly), as the boot disk, as the
Windows disk.

It's a good question.


I've noticed two particular improvements since I put Windows onto an SSD.
One is boot up and close down; about 4 times as fast.
The second is getting TV Guide for the Hauppauge program; about 10 times
as fast (and it sometimes used to hang under the mechanical HD).
Oh, and things like checkdisk and defrag (and similar scans of C.

Ed

On Mon, 11 May 2015 14:42:49 -0700, "Ken Blake, MVP"
wrote:

On Mon, 11 May 2015 21:34:21 +0200, "s|b" wrote:


SSD's are pretty fast, but be warned!

https://blog.korelogic.com/blog/2015/03/24#ssds-evidence-storage-issues

| A stored SSD, without power, can start to lose data in as little as a single week on the shelf.


I wouldn't be very concerned about shelf life of the data, but much
more about loss of data during vacation travel when the computer's
power is off.

Yikes, that's the first I'd heard of that issue. Do you have any idea
if hybrid drives natively provide an automatic and transparent backup
of the SSD side of the drive so that if the SSD side seems to have
gone back the "regular" side will restore it?

Ashton Crusher wrote:
On Mon, 11 May 2015 14:42:49 -0700, "Ken Blake, MVP"
wrote:

On Mon, 11 May 2015 21:34:21 +0200, "s|b" wrote:


SSD's are pretty fast, but be warned!

https://blog.korelogic.com/blog/2015/03/24#ssds-evidence-storage-issues

| A stored SSD, without power, can start to lose data in as little as a single week on the shelf.

I wouldn't be very concerned about shelf life of the data, but much
more about loss of data during vacation travel when the computer's
power is off.

Yikes, that's the first I'd heard of that issue. Do you have any idea
if hybrid drives natively provide an automatic and transparent backup
of the SSD side of the drive so that if the SSD side seems to have
gone back the "regular" side will restore it?

http://www.anandtech.com/show/5160/s...rid-hdd-review

"Seagate takes its standard 7200RPM 2.5" drive architecture and
combines it with a NAND controller, firmware and 8GB of SLC NAND.

The job of the NAND controller is to look at LBAs requested by
the host controller and pull the most frequently read data into
the NAND. A copy of the data remains on the mechanical drive,
so in the unlikely event of a failure in the NAND the data is
still safe.

The controller only caches read requests to the drive, which has
a couple of major implications. On the one hand, this greatly
simplifies the Momentus XT's architecture. It reduces traffic
to the NAND, allowing for a simpler page management algorithm and
guaranteeing longer lifespan (although using SLC NAND already helps
in that department).

On the other hand by only caching reads, it means that writes
to the hard drive are no faster than they would be on a more
conventional drive."

So if the SLC fails in the sense that it is completely
disabled by the drive controller, the platter has a
perfectly complete copy. According to that article.

If the hard drive decided to disable the SLC NAND flash,
then it isn't going to re-enable it again. Like the bad
block sparing algorithm, there would be no user control
of the behavior. Doing a Secure Erase would not turn
the SLC portion back on.

Also, as long as the Flash type is SLC, that improves
the margin to threshold wander. So this "baking in the
cupboard" problem, would not be quite as quick to occur.

*******

Research has discovered that the write life
of NAND flash would improve by a large factor, if the
chip could only be annealed. I'm still waiting for
someone to figure out a way to put an Easy Bake Oven
inside modern drives, to make this happen. The annealing
undoes all the damage. So you're no longer looking at
a 3000 write limit on the worst flash products out there.
But any heater scheme, would be many times larger than the
flash cell sitting next to it. And the density would suck.
But until memristors come along, we can always dream.

Paul

Paul wrote:

Research has discovered that the write life
of NAND flash would improve by a large factor, if the
chip could only be annealed. I'm still waiting for
someone to figure out a way to put an Easy Bake Oven
inside modern drives, to make this happen. The annealing
undoes all the damage. So you're no longer looking at
a 3000 write limit on the worst flash products out there.
But any heater scheme, would be many times larger than the
flash cell sitting next to it. And the density would suck.
But until memristors come along, we can always dream.

I'm waiting for bubble memory to come along...