ACard ANS-9010 RAM Drive

[Clanked]

I'm going to start off by saying, this is a niche product. Even among us multiboxers.
I threw a link into some random post a while about, towards this product. Now I am making an actual thread so people can read about it/search for it.

ACard put together a Frankenstein of a "hard drive." They paired eight 240-pin DDR2 DIMM slots, with a 2400 mAh (around 4 hours) battery, and a forward accessible CF slot for backup purposes. According to the manufacturer: With the push of a button it takes ~20 minutes to back up 32GB to a CF card, and ~15 minutes to restore said backup.

RAM Drives are nothing new, but the last real commercial offering was the iRAM which had a limit of 4GB. This drive has a theoretical limit of 64GB, however there are no 8GB DIMMs listed on their compatibility list.

Now you may be asking yourself: “Self, why the hell would I care about this?”
Allow me to answer: Access Time.

WoW is largely affected by access times, due to the nature of having to load lots of very small files. As you are moving about the world (of warcraft), thousands of little texture/model/sound files are being loaded from your hard drive. Each time one of these is needed, the head must find the spot on the platter where the file resides and transmit it. This is why SSDs are much quicker at finding files than mechanical hard drives, SSDs have no moving parts.

Well RAM is even faster than SSDs. In the quest to read the most small files the quickest, the ANS-9010 takes the cake. Yes, I know what you are saying: “There is no cake.” Well that’s because the ANS-9010 took it.

I don’t expect you to believe me without tests to back up what I say, so without further ado I present TechReport.com’s Benchmarks.

The first test that pertains to WoW would be the IOMeter Webserver test. This test primarily consists of small reads, and well, look at the graph. TechReport IO Meter Test: ('http://techreport.com/articles.x/16255/9')
The ANS-9010 beats the Intel X-25 M (the fastest SSD on the market) by ~30%. (I eyeballed the graph, and pulled out a calculator, so don’t quote that percentage)

The second test that is significant is the HD Tach Random Access Time test. ('http://techreport.com/articles.x/16255/10') This is what we are really looking for. When the drive was setup in Raid0 it came in at .1ms response time. (So did the other SDDs test) When it was setup as a normal drive, without raid, it clocked a time of 0.0ms.

So, the bottom line is: This drive excels at handling small files.

The downsides are: You either have to backup the drive to CF each time you turn off your computer for more than 4 hours, or keep your computer turned on.
More CPU usage than a SDD.
The drive + the RAM will be about the same cost as an X-25 E drive.
However if you absolutely must have the best, this is the way to go.

[-silencer-]

It's on my purchase list once I find some decently-priced 4GB DDR2 sticks in stock. :)

[aboron]

Finally got back from my business trip to access my testing shots on this thing.

Here's what i got with Atto using the drive as a single sata:

http://www.ggxtech.net/pics/Atto-after-single.jpg

Here's what HDTach got:

http://www.ggxtech.net/pics/HDTachnew-single.jpg

And HDTune:

http://www.ggxtech.net/pics/HDTuneAfter-single.jpg

I have to unload my camera if anyone wants shots of the actual unit, there seem to be plenty of those online though.

I also found another use for this drive - running a virtual machine disk image file off of it makes the virtual machine really fly.

If anyone has any other freely available benchmarks they'd like to see results from, I will be happy to run them against it.

[Sam DeathWalker]

Why are your tests so much worse then theirs (you have the latest firmware?):

http://www.2san.com/manual/english/9010%20performance%20test%20report.pdf

Can you try sandra, its free.

http://www.sisoftware.net/index.html?dir=dload&location=sware_dl_3264&langx=en&a=

It seems impossible that Raid0 is SLOWER then non raid, basically they are saying you are worse off running in raid, then the $245 model might be better if 12G max is ok with you.

http://techreport.com/articles.x/16255/10

I have an iram but didnt see all that much improvement but thats with 4G only. I took it out....

[aboron]

Why are your tests so much worse then theirs (you have the latest firmware?):

http://www.2san.com/manual/english/9010%20performance%20test%20report.pdf

Can you try sandra, its free.

http://www.sisoftware.net/index.html?dir=dload&location=sware_dl_3264&langx=en&a=

It seems impossible that Raid0 is SLOWER then non raid, basically they are saying you are worse off running in raid, then the $245 model might be better if 12G max is ok with you.

http://techreport.com/articles.x/16255/10

I have an iram but didnt see all that much improvement but thats with 4G only. I took it out....

You know, I didn't even think to look for a firmware update for this thing, but lo and behold they have a newer one. I will give that a shot later tonight, thanks for reminding me. (I've been thinking of it too much like a hard drive)

Also, I only have mine hooked up as a single drive, no RAID at all, since the ICH Win-Raid was acting funny and hanging now and then so it made things worse. But it seems that is one of the things this firmware fixes so i guess I'll re-raid0 it and try again. Since it restores for an image every time you restart the PC, it's very simple to change the configuration around.

[Starbuck_Jones]

What controller did you have this attached to? sata1 or a sata2?

[combhua]

It seems impossible that Raid0 is SLOWER then non raid
I recently read somewhere (maybe here) an explanation for this. Following the thought that random access times are important in our use case, it makes sense that finding a file split over 2 drives (2 seeks) takes a longer amount of time than the same file on 1 drive (1 seek). Perhaps my understanding of this isn't correct, but that's how I see it at the moment.

[Catamer]

i don't know. seems expensive for what you get.
$400 for the drive with no ram and no backup CF card.
typical CF backup cards are near $400.
the B version has 6 slots of DDR2 which if you use cheap ram 2G sticks, that's only 12G and not enough to put WoW on it.
and the 4G sticks are expensive right now ( near $200/stick)

The Intel X-25E is pretty close to this performance or better and 32G large for only $400.
and for the cost of filling this out with 24G ( or the 8 slot with 32G of memory ) and a backup CF you could easily have multiple Intel X-25E in a Raid-0 smoking it.

[Sam DeathWalker]

it makes sense that finding a file split over 2 drives (2 seeks) takes a longer amount of time than the same file on 1 drive (1 seek)

They both seek at the same time. Raid0 doubles transfer rate (with 2 drives) but does not impove access time, but should not make it worse ... unless the seek times are not sycronized for some reason.

[Sam DeathWalker]

The Intel X-25E is pretty close to this performance or better and 32G large for only $400.
and for the cost of filling this out with 24G ( or the 8 slot with 32G of memory ) and a backup CF you could easily have multiple Intel X-25E in a Raid-0 smoking it.

I dont think you need the backup card, just read the folder from a raptor before you play and return it to the raptor when done (WAY faster then the CF anyways).

if you went the X-25E route I would get two of these instead in raid0 ($320 for a pair of 16G's). "About" as good for less cost.

Ok here is a .11 ms access time ssd drive for $165 (i.e. for wow about as good as the Intel SSD): This seems a very very resonable and cost effective solution. Better to go to the Acard if you need more then 16G or you have the $400 (plus $12 per G of ram).

http://www.sandisk.com/OEM/ProductCatalo…TA_5000_25.aspx

http://cgi.ebay.com/NEW-Sandisk-SSD-SATA…%3A1%7C294%3A50

[combhua]

it makes sense that finding a file split over 2 drives (2 seeks) takes a longer amount of time than the same file on 1 drive (1 seek)

They both seek at the same time. Raid0 doubles transfer rate (with 2 drives) but does not impove access time, but should not make it worse ... unless the seek times are not sycronized for some reason.Short answer: The slower seek will make any faster seeks wait anyway. So your effective times are now always the worst of the two.


Long answer
Well, this may be more abstract than you care to know. Say average seek times are 0.100 seconds. So over the course of many seeks, the average should be 0.100. This implies that there are some seeks which took longer than 0.100 seconds and some which were faster. Still with me?

So it's just as likely that a seek taking 0.070 seconds on one drive, takes 0.130 seconds on the other. Well, the effective seek is now 0.130 seconds, even if the first drive did it in 0.001 seconds.

[not5150]

$1200 for 32GB seems absurd to me.

Cool idea, but geezsus! You guys must have a money tree in your backyards.

It's for all those people that want to say, "Nyah Nyah, I zoned in three seconds faster than you did."

For that money, I'll max out my mobo ram AND get an Intel SSD.

[Sam DeathWalker]

Long answer
Well, this may be more abstract than you care to know. Say average seek times are 0.100 seconds. So over the course of many seeks, the average should be 0.100. This implies that there are some seeks which took longer than 0.100 seconds and some which were faster. Still with me?

So it's just as likely that a seek taking 0.070 seconds on one drive, takes 0.130 seconds on the other. Well, the effective seek is now 0.130 seconds, even if the first drive did it in 0.001 seconds.

Seek times are different depending where on the disk the data is. I would assume though that raid controller designers are savy enough to know that and place data that was split at the same spots on both drives at the same time, else, as you correctly point out, seek times would suffer. Of course they cannot do anything about rotational lantecy, but I doubt that thats a lot. Ya maybe raid0 does have some access time problems, with Hard drives, but the Intel SSD has internal Raid0 and dosnt suffer from these problems. Still what you say is logical and should be given consideration when going to Raid0.

At any rate if the files are all tiny like they seem to be in wow raid is not much improvment:

Using 1 Mitron loading quake was same as NINE Raid0 Mitron SSD drives:

http://www.nextlevelhardware.com/storage/battleship/



Quoted
As you can see in all of the games, we are averaging a load speed increase of 68% over the Western Digital Raptor 150 compared to the single Mtron 16GB. The speed increase is truly incredible with this solid state drive. Now, please remember the Horsepower/Torque analogy that I discussed earlier in this article. Even though we are adding more horsepower (more drives and sustained throughput), latency and random access time (torque) remains the same. For Quake 4 we displayed an identical load time telling me that this game has a large amount of small blocks of files during load. However, for games that required a little more large file seeking on the drive we displayed minor increases in load time while scaling in raid. FEAR is the only game that actually scaled tremendously with more drives. When I loaded up FEAR on the 9 drive setup, Level 1 was pretty much loaded as soon as I clicked the mouse. Pretty incredible to say the least. Based on all of my results, not to mention having the ability to personally get a taste of all of these different test setups I am going to say the ultimate current choice in SSD technology is going to be a 2 X 16GB Mtron Pro Raid 0 setup for gaming.

[Catamer]

Sorry, I didn't find the ram drives to be so cheap when I searched the internet. you must have found a better source for them than I did.

.11 msec access time ( or latency ) ?

the intel x25-e claims to have 75uSec latency.
x25-e ('http://www.intel.com/design/flash/nand/extreme/index.htm')

and I don't have to copy the data over to the drive at startup and it's probably fast enough you don't need a raid anyway.
I know I'm biased because I'm using the x25-e and only put WoW on it and symlinked it to it's old location.
I've noticed a pretty good improvement on load times as I'm sure anyone using these ram drives would.

to each his own I guess.

[Greythan]

$1200 for 32GB seems absurd to me.

Cool idea, but geezsus! You guys must have a money tree in your backyards.

Are you serious Fur?

How much total money have you spent on your systems over the past five years? I will tell you right know the average person would say its "absurd".

I would have hoped we'd all have learned by now that spending on hardware is all relative. I'm pretty sure the average person outside this community would look at the base line fact that we're paying multiple monthly fees to multi-box as superfluous; let alone what many of us spend on hardware/software to support multiboxing.

[Sam DeathWalker]

.110 ms is 110uSec.