ready boost

[Paul]

PLEASE note: Flash Drive is NOT the same as an SSD.

Secondly, there are 2-types of flash memory: slow and fast...

Flash Drives are typically slow, and SSD's are typically fast.

A plug-in flash drive is typically slow. I researched and couldn't
find any external SSD's as yet?

The average read on my Flash drive was 23MB/sec (20-33 range), whereas
my 7200 rpm high-performance hdd was 200 (ranged from 150-300 in
burst).

IMHO,

MJR

I think you missed his point. The USB flash stick supports a higher
IOP than the disk drive. It has a 1 millisecond access time, which
is advantageous in situations where you're grabbing small bits
of data from all over the place (like when there is file fragmentation).

If the required accesses could be perfectly sequential, the USB flash
would lose. Once the accesses become random, and small quantities
of data are being read in each case, the USB flash begins to win.
The hard drive, eventually, can't support even 30MB/sec read.
For example, if you read 4KB starting at sector zero on the
hard drive, then next read 4KB from location 0x1FFFFFF, it takes
8 to 15 milliseconds for the heads to move out there on the
rotating hard drive. On the USB flash, each read of 4KB, takes
1 millisecond for the USB packets to be sent, and an answer to
come back. That's the basic idea, and why they would bother
with ReadyBoost. The sustained transfer rate is unimpressive,
but under a random I/O scenario with small sized reads, the
USB flash wins.

A SATA SSD is even better on random reads like that, because
the access time is 0.1 milliseconds or less. SATA can be
better, because the packet communications scheme is much
better than USB. USB is a polled protocol, and that is
part of the problem with it. (It's possible the
protocol on USB3, will have some differences, but
I've seen zero information on how USB3 actually works
with regards to protocols. There is actually a new
protocol and driver being introduced, that promises to
get a bit more speed from USB external devices.)

As an example of state of the art for SSD, this slide
shows the Sandforce controller coming out this year,
will support 60K IOPS Random Read at 4KB size. With a
hard drive, you'll get only a fraction of that.

http://images.anandtech.com/reviews/storage/SandForce/SF-2000/specs.jpg

Another, indirect proof of these kinds of limits, is start a
Performance plugin, monitor read bytes and write bytes per
second, then start a Defragmentation of your hard drive.
The disk transfer rate is pure crap, and that's because
the disk head is flying all over the place. Defragmentation,
uses relatively small transfers, in order to ensure that no
disk corruption will occur, if the power goes off in the
middle of a defrag. If it didn't have to be "crash-proof",
they could be more aggressive at moving stuff around on
the hard drive, and get better rates.

If you want to test and compare your hard drive, to your
USB flash, get a copy of CrystalDiskMark.

This USB3 flash stick, does 4.353 MB/sec reads of 4K each in
CrystalDiskMark.

http://www.cravingtech.com/kingston-usb-3-flash-drive-datatraveler-ultimate-3-0-review.html

This hard drive, gets about 1/10th the speed on random I/O reads
of 4KB each (0.319 MB/sec).

http://cdn.ilaptopreview.com/wp-content/uploads/2010/05/50965.jpg

It's actually hard for me to find the right two benchmarks
for comparison, as there are so many of them posted out
there. And everyone is testing USB3 now. In any case, the
message is, you can get better *random* I/O, with flash
based devices, either your USB stick, or with an SSD. Hard
drives don't like random I/O, as they're slow to move the
head around.

Paul

 

[Tim Slattery]

I'm not sure if RAM + virtual RAM can exceed 3GB either.

By virtual RAM do you mean the swap file? These are completely
different things, one does not affect the other.

A 32-bit computer has a 4GB address space (2**32). That 4GB is the
amount of physical RAM that it can address. That has to encompass both
the system RAM that you install, as well as RAM on your video card,
your BIOS, and a few other things.

Each process running in your machine gets its own virtual memory
space. The 4GB address space limit also restricts how large these
spaces can be. That means that each program thinks that it has 4GB all
to itself. The OS has to manage things so that this works out for all
the processes. To do that, the OS divides both physical RAM and the
virtual spaces into 4KB pages. It also keeps a swap file on disk which
is divided into 4KB pages. The OS has to keep track of where each page
of each VM space is. If the program requests a page that's currently
in the swap file, the OS has to decide which page it can write to the
swap file to make room for it. Yes, this is a *very* complex system.

The swap file is just a disk area that's used to store these pages. It
can be as big as you or the OS want to make it, subject to filesize
limitations the file system being used (and we know that NTFS allows
for HUGE files).

 

[Brian Gregory [UK]]

There are many times C: gets writes--registry changes, new restore
points, pagefile updates, hiberfile updates, program installations,
program updates, etc.

Yes agreed.

 

[Brian Gregory [UK]]

The write limit varies. The 1000 I'm pretty sure refers to the
standard data usb flash drive. High speed SSD's used for the OS, have
different write specs, but still finite.

I'm sure the figure varies massively between different flash drives too.

Typically, the internal SSD's are used in conjunction with a second
high capacity HDD.

OS is on the SSD, and all other programs/data are installed on the
other HDD.

You can install programs on the SSD too if there is enough space.

 

[Gene E. Bloch]

A plug-in flash drive is typically slow. I researched and couldn't
find any external SSD's as yet?

There are USB, USB3, FireWire, and eSATA enclosures and docking stations
available, so you could create your own by adding an SSD. Performance
would be an issue for the slower interfaces.

The drives are all pretty large (like a 2.5" disk drive), so one that
looks like a thumb drive isn't going to happen (this year!).

But I suspect I'm a bit off the track of what you meant.

 

[Gene E. Bloch]

Typically, the internal SSD's are used in conjunction with a second
high capacity HDD.

I'd replace "typical" by "sometimes".

For example, check out:

MacBook Air
http://news.cnet.com/8301-17938_105-10241140-1.html (Toshiba)
http://www.dell.com/us/p/laptops#facets=133614~0~4774326&p=1 (Dell)

There are sixteen computers with four different model numbers on the
Dell page, none of which list a standard hard drive.

That's after just a quick search.

 

[Ken Blake]

I'd replace "typical" by "sometimes".

I'm not sure whether "typically" or "sometimes" is a better word, but
it's what I do. My main machine has a 120GB SSD (C, and D: and E:
are 650GB hard drives.

 

[Gene E. Bloch]

I'm not sure whether "typically" or "sometimes" is a better word, but
it's what I do. My main machine has a 120GB SSD (C, and D: and E:
are 650GB hard drives.

Did you buy it configured like that or did you set it up yourself?

BTW, to be honest I can't really disagree with your first statement
about the better word

I am trusting my memory about some of the things I've said in this
thread, but so far I haven't been shamed into verifying much of it

 

[Ken Blake]

Did you buy it configured like that or did you set it up yourself?

Neither, but closer to the latter. I had it built for me like that.

 

[TheGunslinger]

Neither, but closer to the latter. I had it built for me like that.

One of the more interesting HDD combo's I've read about sports an
SSD+HDD in one package. I think it had a 64GB SSD in tandem w/ a 1-TB
HDD.

Used the SSD as temp storage for active files and/or swap.

IMHO,

MJR

 

[Ken Blake]

One of the more interesting HDD combo's I've read about sports an
SSD+HDD in one package. I think it had a 64GB SSD in tandem w/ a 1-TB
HDD.

Do you know the brand name/model number? Price?

Used the SSD as temp storage for active files and/or swap.

Using it for the page file is probably a very poor use of an expensive
device like that. More RAM would be much cheaper and, since it would
cut down or eliminate use of the page file, much more effective.

 

[Paul]

Do you know the brand name/model number? Price?




Using it for the page file is probably a very poor use of an expensive
device like that. More RAM would be much cheaper and, since it would
cut down or eliminate use of the page file, much more effective.

This is what you want for the page file. If your system is x32
and limited to 4GB installed, then you could extend with a nice page
file like this.

http://dl.acard.com/manual/english/ANS-9010Q(CEJ).pdf

The other item on today's novelty list, is this thing. It's
a way of combining an SSD and a hard drive. It's too bad
the chip ID is unknown, because the article really doesn't
do a good job of explaining how it works.

http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/29176-silverstone-hddboost-review.html

Paul

 

[Ken Blake]

This is what you want for the page file. If your system is x32
and limited to 4GB installed, then you could extend with a nice page
file like this.

I don't agree. It depends on what apps you run, but very few people
need more RAM than 32-bit Windows allows them.

 

[Paul]

I don't agree. It depends on what apps you run, but very few people
need more RAM than 32-bit Windows allows them.

I see behaviors here, which shows the OS is abusing the page file.
And then, if the page file worked faster, these defects in design,
would be less egregious. That's my point. An OS doesn't have
to be under stress, to make a bad job of using the paging file.
If the operations on the paging file are 4KB in size, and
distributed all over the paging file at random, these
are things my C: drive isn't going to handle well. A storage
device which is RAM based, with a high IOP, would handle
that kind of thing a lot better.

Paul

 

[Ken Blake]

I see behaviors here, which shows the OS is abusing the page file.
And then, if the page file worked faster, these defects in design,
would be less egregious. That's my point.

OK. My point is that few people running Windows 7 with 4GB of RAM use
the page file at all. The remaining people, those who use the page
file, mostly have 2GB of RAM (or even less), and their money would be
better spent on additional RAM than on an SSD used for the page file.

 

[Ken Blake]

I have a laptop running windows 7 32 bit and a desktop likewise.
I tried a 4 gb flash memory in their usb ports to see if they would speed up
using ready boost.
I set it at the rate of 3170 recommended by Microsoft but it did not seem to
make a lot of difference.
Does anyone know if ready boost actually increases the performance of a
computer?

In my experience, ReadyBoost does little to nothing except on
computers with insufficient memory.


And for those computers, adding memory is a much better choice than
using a thumb drive for ReadyBoost.

 

[TheGunslinger]

Do you know the brand name/model number? Price?




Using it for the page file is probably a very poor use of an expensive
device like that. More RAM would be much cheaper and, since it would
cut down or eliminate use of the page file, much more effective.

Sorry, I don't recall the specifics, but I would start w/ MaximumPC
mag since I have a subscription, and read cover2cover each month.

I, also, have online tech subscriptions. So, if you don't find the
article archived at MaxPC, try googling best guess.

IMHO,

MJR