Review Request: More extensive RAM-driven solid-state drive review

[davecason]

I would like to see a head-to-head comparison of the following:
* Hyperos hyperdrive IV
* Gigabyte iRam (gc-ramdisk) v 1.3 or box (box is available on eBay from Hong Kong)
* Acard ANS-9010 (if you can find one)
* Any high-end software RAM drive on a system with tons of ram. Something like 4-8 gig for the OS and 4-8 gig for a ram drive.

I would also like to see them compared using multiple controllers using multiple disks.

To date, I have not seen these compared side-by-side anywhere (toms hardware had an opportunity but compared flash-based ssd's instead).

I have not seen these compared using raid 3/5/6 or in any instance with more than two of the devices (stand-alone and raid 0 has been done).

I would like to see the performance of these things in higher capacity situations that include redundancy. Like to explore using at work with database, business intelligence, and high-end data mining applications (splunk, vertica, kx, etc.). You could even get a trial version of splunk or vertica to try.

You could use the microsoft sql io benchmark for one or more tests.... maybe even create a sql server 2005/2008 database or an oracle 10i to do some crazy queries that exceed the regular OS's ram... You could compare 32-bit vs. 64-bit os to see how they fare with relational databases when I/O bottlenecks are mostly removed.

 

[davecason]

Finally! Pictures of the ACARD ANS-9010:
http://babelfish.yahoo.com/tra..._en&btnTrUrl=Translate
AND
http://translate.google.com/tr...%3Dcom.microsoft:en-us


The device is like an iRam (or iRam-Box) with DDR2 in a 5.25" drive bay container. It has both a battery for maintaining the data AND a compact flash slot for backing up the data to a CF card. Up to 32GB of storage. Fantastic!

This is the second computex in-a-row where it is has been discussed. They mention a possible release date of Q4 2008 (but I am not holding my breath).

There is supposed to be a 3.5" version, the ACARD ANS-9011, but I could not find a picture.

 

[JJChicken]

your best bet is PM DerekWilson or e-mail the main site.

 

[davecason]

Acard just sent me an email. They expect to have samples on the ANS-9010 and ANS-9011 in August 2008 and a official release date of September 2008. We will have to wait a few months.

Last week I called Gigabyte in the United States to see if they could help me with a new battery for one of the two iRam's I have (the battery didn't explode, but it doubled in size by about 50% so I took it out and recycled it since it looked so scary). They did not have any more batteries and they did not expect any new ones in stock any time soon (this was not a plesant phone call, btw). While on the phone, I decided to make the most of my call and I asked about the iRam-Box. They said they have no plans to sell it in the United States so we could go with the old iRam 1.3 (which might actually be the same) in the review.

I would really like to see the Hyperdrive IV compared to the ACARD and Gigabyte products, too... and an old Cenatek Rocket Card. Besides an actual RAM Drive made of real ram (maybe a big one using a 64-bit version of UNIX or Windows) and a good SSD, I don't have any other ideas for comparison.

 

[davecason]

Acard sent me another email and said the ANS-9011 will not be available until Q1.

There will be two versions of the ANS-9010:

1. One with two SATA ports (acting like two drives) with 8 RAM slots.

1. One with only one SATA port

The pictures I found show the version with two SATA ports.


I thought of a couple of other drives that might be fun to also test:

• (PCI) Cenatek Rocket Drive (not the micro)

• (PCI-e) ioDrive from FusionIO

• (PCI) Platypus/Virtual Logic QikDrive (any variant)

• (PCI-e) ddrdrive x1 or x2 if Christopher George really has a product (I heard he sold the x1 to Intel)

 

[supremelaw]

Hello Dave,

Are you having any luck finding products like these
that are actually being distributed in the USA?

We ordered 2 x Gigabyte i-RAM Box units from PC Club, but
after waiting 2 months, Gigabyte decided NOT to distribute
the i-RAM Box in the USA.

RSVP here, because we're presently gathering as much technical
information as possible about RAMDISKS that use DDR2 or DDR3.

FYI: I have a patent pending on a much denser, and faster, enhancement
to Gigabyte's i-RAM Box.


Thanks!
Link removed, check your PM box.
-Schadenfroh (AT Mod)

 

[semo]

finally! something gigabyte should have done 1.5 years ago.

i wonder if these would be as speedy as as the iram.

do you know what is the difference between the ANS-9011 and 9010 davecason? is it just max ram size and physical form factor? also, how are these devices powered?

i still don't get why have gigabyte sat on their hands all this time. didn't they notice that dd2 memory has been steadily going down in price for a long time now? and no doubt consumer interest has been increasing in direct proportion.

 

[supremelaw]

Word on the street is that Gigabyte's Project Manager quit
to take a job with another company.

This rumor has not been confirmed, to my knowledge, however.


p.s. ANS-9010 uses standard DDR2 desktop DIMMs;
ANS-9011 uses laptop SO-DIMMs (Small Outline DIMMs).

DC power comes in via a standard 4-pin Molex connector,
as far as I can tell from the AND-9010 photos I've seen.

I've only seen 1 photo of the ANS-9011, and I can't quite
make out the connectors.

The ANS-9012 uses a standard SATA power connector,
NOT a 4-pin Molex.


Sincerely yours,
/s/ Paul Andrew Mitchell, Inventor and
Webmaster, Supreme Law Library

All Rights Reserved without Prejudice

 

[semo]

is power still supplied to the molex connectors when the computer is shut down?

i thought only the pci slots did that.

 

[davecason]

9010 is like the iRam box... but it has two SATA interfaces and two sets of memory banks for a maximum of 8 sticks of RAM. Acard may release two varieties of this device but they were not clear about what the differences will be.

9011 is a smaller device, the size of a hard drive. I can't remember if is a 2.5" or a 3.5" hard drive emulation.

 

[davecason]

Acard told me that the 9010 would be out in the 4th quarter of this year.

The best thing about these devices, as compared to the iRam (I have two) is that they use DDR2 which is much less expensive than DDR.

 

[davecason]

The 9010 has a battery and a compact flash backup.

 

[semo]

so would that mean that the acard would backup the ram drive to a cf (on battery power) when you switch off the computer?

you won't have the quick boot times like the iram if that's the case.

and yeah, ddr2 has been cheaper than ddr for a long time (not that long after iram came out). it's also easier to find higher capacities and i think ddr2 consumes less power too.

 

[supremelaw]

> is power still supplied to the molex connectors when the computer is shut down?
> i thought only the pci slots did that.


The answer to your good question above depends upon the power source:

Yes, if the 4-pin Molex connector and cable originate at an ATX-style power supply,
the DC power to those cables is turned OFF whenever the system is shutdown,
and of course whenever the ON-OFF switch on the PSU itself is turned OFF.

On the other hand, one solution which we have been incorporating into designs
is a separate AT-style power supply, ideally in a chassis that accommodates
2 power supplies -or- in an external enclosure with its own power supply
such as this inexpensive external enclosure we found at Newegg:

http://www.newegg.com/Product/...x?Item=N82E16811164106

We wouldn't recommend the PSU that comes with this product, however;
you would do better to purchase a 300W AT-style PSU from a quality
company like Antec:

http://www.antec.com/us/productDetails.php?ProdID=23004

The advantage of an AT-style PSU is that they come with their
own ON-OFF switch which does not need a motherboard's ATX
logic at all. You can see the ON-OFF switch in the Newegg photos
of these products:

http://www.newegg.com/Product/...x?Item=N82E16817707002
http://www.newegg.com/Product/...x?Item=N82E16817338024
http://www.newegg.com/Product/...x?Item=N82E16817338010
http://www.newegg.com/Product/...x?Item=N82E16817338025

We keep one AT-style PSU in our parts inventory, to do routine testing
of fans and such.

Thus, a second PSU with its own ON-OFF switch will permit
volatile solid-state RAM to remain powered ON, even if
the rest of the system is powered OFF completely.

Here's one of our first component drawings, which incidentally
was REFUSED by Microsoft -- because it did not meet
the pixel dimension rules required of such submissions:

http://www.*****/patents/prototype.1.GIF

The latter drawing assumed either 4 x Gigabyte i-RAM Box units
or 4 x HyperOS HyperDrive4 units, in addition to a modern RAID
controller with ports wired to Multi-Lane "Infiniband" brackets --
for connecting to mass storage devices in external enclosure(s).

Here's one of THE best PCI brackets we have found,
because the SATA cables connect IN-LINE rather than at 90 degrees:

http://*****/systems/...cket%20Backplane.1.jpg
http://*****/systems/...cket%20Backplane.2.jpg
http://*****/systems/...cket%20Backplane.3.jpg
http://*****/systems/...cket%20Backplane.4.jpg

Latter PCI bracket is available from www.cooldrives.com .


The 90-degree connectors crowd adjacent expansion slots,
like this one from Addonics:

http://*****/systems/....multilane.ad4saml.gif



Sincerely yours,
/s/ Paul Andrew Mitchell, Inventor and
Webmaster, Supreme Law Library

All Rights Reserved without Prejudice

 

[supremelaw]

Furthermore, using a second PSU, preferably using AC input from a
quality UPS, effectively renders the on-board SSD and on-board battery
quite superfluous, imho.

Workstations should be protected by a quality UPS in any event;
so, adding more battery(s) necessarily increases the costs of the
ANS-9010 by forcing unnecessary redundancy on the user.

Moreover, software is available, like Symantec's GHOST, Acronis etc.
to create a "drive image" of an entire Windows drive letter.

So, adding a port for inserting a flash SSD also increases costs
of the ANS-9010.

Workstations can be easily augmented with 2.5" flash SSDs
that are now available from numerous memory suppliers,
if one MUST HAVE a solid-state disk to store such drive image files.

A 2.5"-to-3.5" adapter will allow any 2.5" HDD to fit into a
standard 3.5" drive bay: this item is available from Newegg
for about $19 + S&H:

http://www.newegg.com/Product/...x?Item=N82E16811993005


Nevertheless, writing drive image files to rotating disk drives
is already available using software like GHOST, Acronis, etc.,
and the restore tasks don't absolutely require a flash SSD
when reading the drive image file.

For all the above reasons, we have concluded that both
on-board battery(s) and on-board flash SSD ports
are superfluous, and will only increase the manufacturing
costs, and hence the final retail prices; plus, these add
more discrete components, each of which can fail,
reducing MTBF (statistically speaking).

For example, we have recently found complaints on the Internet
about failures of the battery used in Gigabyte's i-RAM units
(NOT GOOD, if true):

http://www.*****/syst...dead.i-RAM.battery.jpg


Sincerely yours,
/s/ Paul Andrew Mitchell, Inventor and
Webmaster, Supreme Law Library

All Rights Reserved without Prejudice

 

[davecason]

I have two iRams and one of the two had a battery die... it took about 14 months. It did something really scary... after I noticed it was not working, I pulled the card out and the battery had expanded to nearly 150% of its original size. I was afraid it might explode so I took it to a Batteries Plus store to have it recycled. I have not yet been able to get a new one from Gigabyte ( I have been trying for a couple of months). They never have any in stock.

To speak to the CF card backup: I have no idea how this works. It may run at shutdown, it may be manual (you push a button), or you may be able to use software to initiate it. I find this option only mildly useful because, if you want to back the whole drive up, your CF card has to be bigger than the amount of RAM you have installed... and that will not be cheap.

 

[supremelaw]

Good points about the CF card backup:

As for your objection to the CF capacity required,
Symantec's GHOST software does compress the
drive image file output during a backup task,
and it permits different levels of compression too.
Compressing the backup file(s) may reduce the
cost of the CF disc required.


To extrapolate upon your other excellent observations,
I see no reason to bypass or duplicate the logic already existing
in the BIOS software installed in modern motherboards:

If one MUST HAVE a drive image file stored
on a non-volatile medium, the restore tasks
that come with Symantec's GHOST for example
permit reading drive image files from SATA
and IDE devices already connected to the motherboard,
and/or from CD and DVD media via optical drives.

One of the ONLY WAYS I could honestly justify such a
CF card backup, is by using a power loss to the
Molex DC input cable as the "event" which automatically
triggers a complete backup of the RAM to the CF card,
with that on-board battery powering both the RAM
and the circuitry that writes RAM contents to the CF card.

Because I generally distrust these "automatic" solutions,
I much prefer a more deliberate and scheduled downtime,
like this:

Let's be systematic and create a "decision tree"
that handles all possible scenarios.

If the ANS-9010 is NOT being used to store the Windows C: system partition,
the more robust battery in a modern UPS will provide plenty of power
to maintain the volatile contents of the RAM -- and the main system too --
long enough to create a routine drive image file of that ramdisk
using software like Symantec's GHOST, Acronis etc.

And, if the ANS-9010 is being used to store the Windows C: system partition,
the worst possible scenario is that the RAM must be powered down too.

Under this latter scenario, after the RAM is powered up again,
a task like GHOST Restore is run to restore the C: system partition --
by reading from a permanent storage medium like an optical disc
or a conventional rotating hard disk drive. This task is already
routine and fully debugged: I've done it several times already myself.

The latter can be accomplished with only the logic in the BIOS;
XP need not be loaded, because the GHOST Restore task hosts
a mini-DOS environment with support for I/O to and from devices
like HDDs and optical discs.

If the RAM is NOT being powered down too, the data remain available
for the next Restart, and under this latter scenario, it's irrelevant
whether the drive letter is C: hosting the system partition,
or some other drive letter hosting non-system files.

Our "mandatory" UPS, with a large and robust battery to maintain
and condition power to the entire system, is far superior to these funky
and failure-prone on-board batteries that we are seeing included
with the i-RAM and the ANS-9010.

Finally, it is possible to connect 2 independent power supplies,
and keep the AT-style PSU powered ON while we are doing maintenance
to the motherboard etc. that requires the ATX PSU to be powered OFF
completely i.e. by flipping its power switch OFF.

One user liked the idea of installing the AT-style PSU in a chamber
near the top panel, like the one in the discontinued Cooler Master
CM Stacker chassis. Then, the internal wiring can be physically
separated from the cabling that connects the motherboard & other
peripherals to the ATX-style PSU, installed near the bottom panel.

The Lian-Li PC-A70 also accommodates 2 x PSUs, and one of
those 2 bays is right below the top panel.

But, EVEN IF both PSUs are installed in bays near the bottom panel,
careful cable routing should minimize the possibility of accidental
shocks and/or short-circuits of the DC power being fed to the
ramdisks by our AT-style PSU. I mean, if I were to build such
a system, I would REMEMBER that the ramdisk power circuitry
is still ON because I designed it that way. Cheesh!!

From a marketing point of view, I think ACARD is assuming
that its users possess less than average intelligence,
so their CF backup procedure will protect these users
from their own ignorance. I think the contrary is the case:
the users who would want to buy these ANS-9010 units
will have ABOVE AVERAGE intelligence, and those are
the kinds of users we want to impress -- with engineering
quality, simplicity and reliability.


I hope this helps. And, I appreciate our correspondence here, very much.


Sincerely yours,
/s/ Paul Andrew Mitchell, Inventor and
Webmaster, Supreme Law Library

All Rights Reserved without Prejudice

 

[supremelaw]

> I was afraid it might explode so I took it to a Batteries Plus store to have it recycled.
> I have not yet been able to get a new one from Gigabyte ( I have been trying for a couple of months).
> They never have any in stock.


And, Gigabyte are using the very same battery in their i-RAM Box product!

http://www.*****/systems/gigabyte/

http://www.*****/syst...igabyte_ramdisk.22.jpg

http://www.*****/syst...e/GO-RAMDISK-BOX.9.jpg

http://www.*****/syst.../GO-RAMDISK-BOX.11.jpg

http://www.*****/syst.../GO-RAMDISK-BOX.15.jpg

http://www.*****/syst.../GO-RAMDISK-BOX.17.jpg

http://www.*****/syst.../GO-RAMDISK-BOX.25.jpg

http://www.*****/syst.../GO-RAMDISK-BOX.26.jpg



What you are reporting is NOT GOOD!!

And, am I correct that i-RAM does NOT work at all without that custom battery
installed, charged and fully operational?


Sincerely yours,
/s/ Paul Andrew Mitchell, Inventor and
Webmaster, Supreme Law Library

All Rights Reserved without Prejudice

 

[davecason]

For the question:
And, am I correct that i-RAM does NOT work at all without that custom battery
installed, charged and fully operational?


No. The iRAM works just fine without a battery... but if I lose PCI power (i.e. my server is unplugged), I lose the data.

With a pair of the iRAMs in raid 0, restoring from a ghost image takes only seconds, but my box is not backed up in real-time so this is sometimes annoying. The last time I lost any data was while I was out and I lost power for long enough to kill my UPS.

I am more annoyed than anything. I wish there was some way to get battery status from an API so I could write something to watch for battery problems. If the battery had exploded, it would have burned up a lot of very expensive equipment. Between this battery problem and the lack of DDR2 progression in the product line, I am pretty down on Gigabyte. I also cannot believe they did not release the BOX in North America. Maybe we complain too much about battery problems.

 

[wmaciv]

Dave, I currently own two I-RAM "box" style ram drives, along with a Platypus QikDrive8. The qikdrive maxes out around 96 MB/sec; the I-RAM's about 120 MB/sec (on my platform, at least).

 

[davecason]

I picked up an Acard ANS-9010 and a 9012. They are both pretty neat.

Here is a quick ANS-9010 vs iRam comparison:
http://www.wideopenwest.com/~dcason6634/Acard.html

If anyone is interested in a contact at Acard/2SAN so they can buy their own, please post a reply. The ANS-9010 is essentially a newer version of an iRAM box that uses DDR2 and uses a different backup strategy.

If anyone at Anandtech is interested in borrowing my drives for a review, please send me a personal message or an eMail.

-Dave

 

[davecason]

wmaciv,

My iRam's top-out at around 120 MB/s, too, but it does vary with the SATA controller I attach them to.

The Acard is a bit faster but, again, it depends on what is attached to. In addition, the Acard ANS-9010 seems to perform very poorly in forced SATA I mode (but I am still testing this).

From my screen shots, I can't figure out why the burst speed is so high on the RAID0 benchmarks. I'm assuming that some sort of controller caching is affecting the results. Theoretically, using two drives together should no more than double the speed of an individual drive.

-Dave

 

[davecason]

Another article:
http://www.techreport.com/articles.x/16255/1

I read the above article but they didn't test with very good hardware (less than decent SATA controller). I bought a pair of Dell Perc 5i cards and upgraded them with the latest LSI MegaRaid 8840E firmware which really made my ANS-9010 cook. See chart number 4:

http://www.wideopenwest.com/~dcason6634/

You definitely need a decent SATA controller to get the most out of one of these.