backtop

Hardware Gigabyte Displays i-Ram Successor
Anh Tuan Huynh (Blog) - June 7, 2006 11:19 AM
Print E-mail del.icio.us 58 comment(s) - last by electriple9.. on Dec 6 at 1:17 PM
SATA 300MB/s and up to 8GB of DDR2

At Computex 2006 this year, Gigabyte is displaying the successor to its i-RAM storage device, the GC-RAMDISK. Gone is the PCI slot power interface and instead Gigabyte has made the GC-RAMDISK a 5.25” drive bay that relies on power from a Molex connector.

The Serial ATA interface remains with added compatibility for 300MB/s transfer rates, though it is unknown if the GC-RAMDISK will support SATA 3.0Gbps features such as native command queuing -- though with access times so low the only real advantage of the new feature set is the increased data transfer. DDR2 memory is supported this time around instead of DDR of the previous i-Ram. Supported memory capacity has been increased to up to 8GB from the previous 4GB which is just enough to make the i-Ram useful as an OS drive. Since DDR2 memory is not non-volatile (NAND), a battery feeds power to the memory when the system is off to prevent data loss.

As the GC-RAMDISK is still in development, availability is still a few months away. Pricing information is unavailable at this time but expect similar pricing to the previous i-Ram.

Comments     Threshold


This article is over a month old, voting and posting comments is disabled
Money saving?
By Quasmo on 6/7/2006 12:36:38 PM , Rating: 2
If I bought 10 of these and put them in a raid to run AVID off of them, edit at 1:1 compression, OMG sweet. 80GB with almost zero latency would be crap my pants good.




RE: Money saving?
By Quasmo on 6/7/2006 12:37:04 PM , Rating: 2
Ignore title


RE: Money saving?
By epsilonparadox on 6/7/2006 1:09:41 PM , Rating: 2
Won't the performance be limited by the SATA transfer speeds?


RE: Money saving?
By shecknoscopy on 6/7/2006 1:23:20 PM , Rating: 3
quote:
Won't the performance be limited by the SATA transfer speeds?


Theoretically, yes - but do any current HDD's actually come close to saturating the 3Gb/s rate? If this baby did hit the speed limit, that'd still be a step up, right?

Heh - I wonder if you can up the voltage and overclock it. Probably not, and probably not much of a performance gain if you did, but just imagine the bragging rights. "Oh, so you did a voltage mod on your video card and got an extra 10MHz from it? That's nice and all - but the real enthusiasts overclock their hard drives."


RE: Money saving?
By Scabies on 6/7/2006 2:20:12 PM , Rating: 2
correction, ten. the original poster was saying 10 of these ramdiscs. Who has 10 SATAII connectors?


RE: Money saving?
By peternelson on 6/9/2006 6:42:54 PM , Rating: 2
There's a new board coming that anandtech says has 14 SATA connectors ON BOARD.

AS for me, I would prefer an ARECA PCIE x8 raid controller with 8 SATA ports.

When added to the typical 2-4 SATA off the main chipset, you have achieved your 10 drives.

This HAS been done using an ARECA controller with 8 I-RAM cards already.

A similar setup using the new iram v2 should have no probs either.


RE: Money saving?
By Scabies on 6/7/2006 2:19:25 PM , Rating: 2
Eight 3gbps theoretical pipes in a RAID? Are you complaining?


RE: Money saving?
By Trisped on 6/7/2006 2:30:01 PM , Rating: 2
I don't think SATA controllers can actually go over the 3.0GB/s max transfere speed, even with a RAID confige. I am no expert, but I think that the RAID total speed bonus is 3.0GB/s so the only benifit of RAID with these is increased speed. There would be a risk of increased latincy with the RAID controller though.


Too bad it's DDR2...
By glynor on 6/7/2006 11:51:38 AM , Rating: 2
It's too bad they made this DDR2. With AMD's recent switchover to DDR2, those of us who upgrade to AM2 will probably have a small pile of perfectly good DDR sticks lying around. I'd love to take the 1GB I have already (and the 2GB more I'd have if I switch to AM2) and throw it in one of these to use as a swap partition in Windows.

Hopefully Gigabyte will see the usefulness of having a DDR1 version of this. Otherwise, this looks to be much more workable than the older PCI-slot-hogging version.




RE: Too bad it's DDR2...
By bunnyfubbles on 6/7/2006 12:42:19 PM , Rating: 1
I think the idea is that DDR2 is going to offer higher capacity for cheaper prices. With 2GB sticks being reasonable on DDR2, you can finally get a decent capacity out of such a drive to use it as something like an OS drive (perhaps even with a game, should you be really into competition for a particular game such as BF2 or CS:S)

Although I agree, a cheap model just like this that accepted DDR1 (perhaps only 2 slots) would be perfect for use as a swap partition. Although the price would have to be pretty cheap, the current i-RAM goes for ~$125, steep and pretty unacceptable for just a swap partition. $50 would probably make things interesting...


RE: Too bad it's DDR2...
By highlandsun on 6/7/2006 7:12:33 PM , Rating: 2
Well, that hasn't happened yet. The only way to get the full 8GB capacity is using 2GB DIMMs, which are still incredibly expensive. And, they seem to only be available in registered/ECC versions; there's no mention here of whether registered DIMMs are supported or not.

The largest non-registered DIMMs I can find online are 1GB, in either DDR or DDR2. The cheapest 2GB DIMMs I can find are $400 for both PC2100 and PC2-3200, ECC/registered.


RE: Too bad it's DDR2...
By Johnmcl7 on 6/8/2006 10:22:09 AM , Rating: 2
Here is some 2Gb memory which is unbuffered, non registered and non-ECC:

http://www.crucial.com/uk/store/partspecs.Asp?IMOD...

Even laptops have 2GB DDR2 sodimms now although still pricey

John


RE: Too bad it's DDR2...
By raskren on 6/7/2006 1:05:54 PM , Rating: 1
Why would you "upgrade" a competant AMD 939 system to AM2? That's throwing money at a problem that doesn't exist. Just for DDR2 and that 1-3% performance gain?


RE: Too bad it's DDR2...
By glynor on 6/7/2006 1:59:46 PM , Rating: 2
I wouldn't, but I have 2 other Athlon XP systems (on nForce3 Ultra boards) that could certainly be upgraded. They use plain old DDR400 RAM too!

Who said I'd be upgrading a socket 939 system? My A8R32-MVP is staying just as it is thanks.


RE: Too bad it's DDR2...
By Crassus on 6/7/2006 2:05:10 PM , Rating: 2
nForce3 was actually also S754/939, so I guess you mean nForce2?


RE: Too bad it's DDR2...
By glynor on 6/7/2006 2:41:59 PM , Rating: 2
Yep. Oops.


RE: Too bad it's DDR2...
By Trisped on 6/7/2006 2:19:44 PM , Rating: 2
You can stick all that in the older iRam, the only loss would be SATA bandwidth. Plus the largest sizes of DDR RAM are 1G chips, so the max you could fit would be 4GB.

Then there is also the fact that 1-2GB DDR2 is less then the same DDR.

I am sure there are a lot of Intel users who will love being able to use their old 667MHz RAM in this thing.

Whith 8 GB you could put it in a RAID 0 and get some actual ize out of it. That would be fun.


This would have been awesome
By creathir on 6/7/2006 12:31:56 PM , Rating: 2
Had they used a PCIe slot instead of SATA. With this thing using SATA, it makes the performance of the RAM almost moot due to the limitation on data transfer. As with the first version, this one will be nothing more than an expensive SATA hard drive. I really wish they would come out with a PCIe version of this baby. I could really handle Windows running on a RAMdisk at several GB/s access to the CPU. Could you imaging video editing at these speeds?
- Creathir




By bldckstark on 6/7/2006 12:49:41 PM , Rating: 2
I'm betting PCIe was not in widespread use when they started working on this thing. Maybe the next version.


RE: This would have been awesome
By gmyx on 6/7/2006 1:12:25 PM , Rating: 2
If it were PCIe, you would need some drivers for this. Just like the I-RAM, you don't need drivers - it looks to the OS like a regular hard drive.


RE: This would have been awesome
By creathir on 6/7/2006 1:43:58 PM , Rating: 2
I would rather have drivers... and the speed...
- Creathir


RE: This would have been awesome
By Fallen Kell on 6/7/2006 2:06:15 PM , Rating: 2
You do know that you will still get 5-6x more speed from this then any regular hard drive don't you? This will be the first device to actually truely test the speed rating of SATA (I OR II for that matter). Your best hard drives (10k-15k disks) have a transfer speed of 70-80MBps. This will theoritically have a full 300MBps if they built the pipe correctly.


RE: This would have been awesome
By Trisped on 6/7/2006 2:27:11 PM , Rating: 2
Don't forget access and seek rates of practicly 0.

Performance increases won't be high unless you have a program that uses the hard drive a lot. In most cases it is better to add to system RAM. Still, if you have 8GB of RAM laying around why not set your self up?


By Anh Huynh on 6/7/2006 2:33:06 PM , Rating: 2
You guys also forgot for it to be a storage device and PCI-E it would need to have an onboard SATA II controller which negates any benefits of PCI-E anyways. The controller used has a SATA interface, I don't think anyone makes a DDR2 to PCI-E Storage controller.


By drank12quartsstrohsbeer on 6/7/2006 2:45:16 PM , Rating: 2
This isn't the only solid state disk drive on the market, you know.
Ones like you want already exist, just be prepared to pay a lot, and i mean a lot, more.


Stability?
By pnyffeler on 6/7/2006 3:46:08 PM , Rating: 3
Anybody have any idea how stable information on one of these will be for the long-term? I mean, some errors have to creep into the data after some time. There is no error correction system that I'm aware of.

It would be great to install an OS on it and reap the benefits of blazing fast boot times and program load times, but if I have to reinstall WinXP and Office every week, you can forget it.

Also, would normal error correction even work? The way I understand it, information that is read from memory is checked for corruption when error correction is used. However, what happens if there is an error found? My guess is that the information is just reloaded from the hard drive or whereever it came from, but what happens when the Ram drive is the only data source to go to?




RE: Stability?
By cochy on 6/7/2006 5:13:55 PM , Rating: 3
They reviewed the i-Ram at Toms Hardware a little while back. Stability was fine, but performace benefits from an OS installation on the drive wasn't very large. WinXP booted in about 3 seconds faster on the i-Ram than a Raptor. Now even 8 gigs isn't enough space for a proper OS installation. I think Vista will need at least 15 gigs.


RE: Stability?
By Larso on 6/7/2006 6:10:09 PM , Rating: 2
Stability of data in ram modules should not be a problem. If it was, you would have stability problems with the ram modules in normal usage as well. Btw. it is usually the data transfer to the ram modules that cause errors, and in this case there is no need at all to push the ram timings, so stability ought to be fine :)
quote:
Also, would normal error correction even work? The way I understand it, information that is read from memory is checked for corruption when error correction is used. However, what happens if there is an error found?

Well "normal error correction" ram modules (ECC) enables catching errors in the data transfer and is able to retransfer data if something went wrong.

And "normal error correction" in its usual sense would be to use algorithms able to detect and correct errors to some degree, with the price of data redundancy.


RE: Stability?
By Larso on 6/7/2006 6:19:53 PM , Rating: 2
not sure if I'm completely correct on the ECC ram though :)


RE: Stability?
By highlandsun on 6/7/2006 6:58:42 PM , Rating: 2
Standard ECC in DRAM will detect any two wrong bits in a word (with no possibility of correction) or detect and correct any single-bit error in a word. I think a "word" is 32 bits in that sense; since DIMMs are now 64 bits wide that ought to be doubled.

If their memory controller supports ECC it should all be fine; if not, they should go back and add ECC to their controller. For a device like this that is going to store up to 8GB of data for long periods of time, it's pretty important.


Molex Power?
By Mclendo06 on 6/7/2006 11:52:30 AM , Rating: 2
One of the nice things about the PCI i-ram is that when you power off the pc, the 5v aux is still feeding the i-ram so you aren't using the battery. It only kicks in if you switch off your power supply/unplug/have a power outage. I am guessing that this will either run on a Molex or a SATA power connector. Does anyone know if these connectors will provide power even when the system is switched off?




RE: Molex Power?
By johnsonx on 6/7/2006 12:01:02 PM , Rating: 3
No, there's no power at all on the standard molex connectors when power is off. All standby power goes to the mainboard through the 20/24-pin connector.

If they want standby power to support the RAM when the system is off, they'll have to find a way to tap a PCI slot, or the keyboard port, or in some cases USB. The problems there are that not all motherboards provide standby power to the keyboard port and/or USB ports; in some instances it's set by a jumper that defaults to off.


RE: Molex Power?
By johnsonx on 6/7/2006 12:03:07 PM , Rating: 2
To add a little more, I suppose they could come up with a little plug adapter that sits between the main power cable and the power plug on the mainboard, and drop the needed power lines off that... such a thing is bound to be bulky and expensive though.


RE: Molex Power?
By yuethomas on 6/7/2006 12:04:36 PM , Rating: 2
Perfect for a DIY project.


RE: Molex Power?
By Trisped on 6/7/2006 2:21:22 PM , Rating: 2
Or they could just release an AC/DC adaptor that you run through the back of your case to power the drive with.


Only 4 slots...
By EODetroit on 6/8/2006 10:09:34 AM , Rating: 2
This only has 4 memory slots... that kind of stinks. I was hoping for 8. It will be prohibitively expensive to put 8GB on it using 2GB modules instead of 1GB modules... well not for everyone, but for a lot of people.

By the way, the batteries may not last as long as advertised. I have two I-Ram 1s in my computer right now, in a raid-0 config. Some work was being done on my apt. by my landlord while I was going to work, so I shutdown my computer and turned off my UPS because I didn't want him touching it. I was gone 9 hours, when I came back and powered up my computer, the data on at least 1 of the I-Rams was lost (as you know, in raid-0, losing 1 drive means you lost them both). The batteries are supposed to last 14-18 hours as I understand it on I-Ram 1. (Fortunately I have backup images saved to my hard drive whenever I make significant changes to the data on my I-Ram drive).

I-Ram 2 is totally dependant on the battery, no stand-by power like the I-Ram 1 can use. I hope they have an alternative way to preserve the data on there, 'cause I don't trust the batteries for any length of time.




RE: Only 4 slots...
By daschneider on 6/8/2006 11:53:42 AM , Rating: 2
It's kind of hard to tell from the picture but I count 2 rows of 4 slots each, a total of 8.


RE: Only 4 slots...
By peternelson on 6/9/2006 6:45:13 PM , Rating: 2
Yeah the STUPID photographer should have placed the camera where we can actually SEE the dimm slots properly.

As it is I'd be reluctant to say for sure.

8 slots would be infinitely preferable.


RE: Only 4 slots...
By peternelson on 6/9/2006 6:48:20 PM , Rating: 2
On close examination, I'd say that board has only FOUR slots.


score
By Scabies on 6/7/2006 11:58:40 AM , Rating: 2
I've had my eye on this thing since they came out with the first iteration of the i-ram (http://www.gigabyte.com.tw/Products/Storage/Produc...)

originally, the battery would charge for 6 hours and could keep the dimms happy for about 16. In other words, dont take your computer to college unless you have a UPS to feed it en route.
The first i-Ram's memory controller supported up to 8gb, but giga-byte said not to go over four.
We can still put old-skool DDRs in this right?

As long as you dont hit the switch on the PSU itself, the i-Ram would draw power from the PCI socket it lives in. Since this lives in a 5.25" bay and uses a molex... I dont know whether Molex connectors are "hot" in soft-off

I have two 5.25" bays empty... two of these stocked 8gb each in raid-0 (they do support RAIDing) would kill. KILL.




RE: score
By yuethomas on 6/7/2006 12:03:12 PM , Rating: 2
Molex connectors are not hot on soft-off. The best bet, as far as I can see, would be to take +5VSB off a spare motherboard USB connection (easy project), since it doesn't take much current to keep battery cells trickle-charged.

I guess I'll wait until DDR2 prices are a bit more reasonable in Canada to seriously consider this. At $75/GB (for 4x1GB value sticks) it's too rich for my blood.


RE: score
By GrammatonJP on 6/7/2006 2:19:36 PM , Rating: 2
Thats why before you leave you ghost the little 4gb and you can power off... 4gb takes like 2 minute to make and restore


Flash Backup
By daschneider on 6/7/2006 3:00:24 PM , Rating: 2
I'd like to see a couple of SD slots on this thing so that I could have the option of using a FLASH card for back-up when the power is off. If I could do that I'd buy one of these in a heartbeat.




RE: Flash Backup
By Trisped on 6/7/2006 3:14:03 PM , Rating: 2
They have hardware that does this using your hard drive.


RE: Flash Backup
By daschneider on 6/8/2006 11:58:01 AM , Rating: 2
Where did you read that? I don't see anything in the article about it. And I'm not sure it could work in all cases. Say the power in my house suddenly goes out. My hard drive is going to power down right away, before data from the RAM disk could be copied to it. An onboard FLASH device, powered by the battery, could be used to save the data before to the battery is completely drained.


native command queuing?? for RAM???
By fanbanlo on 6/7/2006 11:44:11 AM , Rating: 2
When seek time is not an issue, why's NCQ even needed?




RE: native command queuing?? for RAM???
By Scrogneugneu on 6/7/06, Rating: 0
By slunkius on 6/8/2006 5:29:42 AM , Rating: 2
Did you read the post you are replying to? sheesh...


NON?
By Netopia on 6/7/2006 11:45:34 AM , Rating: 2
quote:
Since DDR2 memory is non-volatile, a battery feeds power to the memory when the system is off to prevent data loss.


If it were "non" volatile, it wouldn't need any power at all, would it?

Joe




RE: NON?
By Anh Huynh on 6/7/2006 11:47:54 AM , Rating: 2
Oops, I meant to type volatile. Its still a bit early in the morning :)


I, too, amd puzzeled by the molex connector
By Dubb on 6/7/2006 1:52:25 PM , Rating: 2
Isn't that just limiting the device to people who never turn their computers off?

I suppose you could buy a low power, fanless PS just to feed the drives...but that's just going to add to the complexity and cost.




By Trisped on 6/7/2006 3:12:51 PM , Rating: 2
Why buy a PSU when you can by an AC/DC adaptor for $6. If necessary you could pick up a Molex connector for $1.00 and a basic switch for $.20, then make an adaptor that would run power from the PSU but when switched it would run from the power adaptor. That is unless they include a DC port on the drive somewhere. If they have a port, you can just plug in with out any hassle.
http://www.radioshack.com/product/index.jsp?produc...


How about an actual software solution?
By Fox5 on 6/7/2006 6:53:29 PM , Rating: 2
What SHOULD be done is have native support from the OS and motherboard for stuff like this.
Enable a portion of your already installed ram to be used as a harddrive. Have that portion of the ram be continuously supplied power, even when the PC is off.
Then, whatever data is on that drive can have a backup copy on the main harddrive. Upon bootup, the computer can look to the ram drive for data, and if nothing is found, load from the harddrive.
Maybe someone could make a bootloader that does this...




By highlandsun on 6/7/2006 7:01:26 PM , Rating: 2
Almost. System RAM is already limited enough as it is. This should be a separate SATA controller card with 4 DIMM slots and standby power. The SATA controller should automatically cache accesses to any drives you plug into it. This new GC-Ramdisk is slightly better than the previous iRAM, but it still misses the mark for actual usefulness.


Non-volatile?
By TheBaker on 6/7/2006 11:46:40 AM , Rating: 2
I think you mean volatile. Kinda makes a big difference.




By JPH1121 on 6/7/2006 2:24:04 PM , Rating: 2
DDR2 has lower power requirements than DDR1 which means the life of storage will be extended drastically when the power is off.

Memory access times will be pointless with this device as well as the throughput with crappy RAM will still fully saturate the 300mbps limit.

This thing is going to put the previous i-RAM to shame.

Also, don't neglect, that putting this into a PCI-e slot would cut out a majority of the potential buyers. 500mbps would be nice, but as long as the people buying this thing are enthusiasts with mostly dual-slot video cards (7900GTX, 7900gt w/upgraded cooler, 1800xt, etc) who don't have a PCI-e slot to use. A dual-slot video card will cover the PCI-e x1 slot.

This was the best possible solution imo.




Timescale
By peternelson on 6/9/2006 6:56:39 PM , Rating: 2
"As the GC-RAMDISK is still in development, availability is still a few months away. Pricing information is unavailable at this time but expect similar pricing to the previous i-Ram."

"Few months away" - seriously? What REALISTIC timescale?

Remember iram 1 was about 18 months from announcement to widespread availability.

Even in the UK, the prices were until recently £277 and only just dropped to £120. Meanwhile the americans enjoy the $120 RRP which is still cheaper.

At £120 the IRAM1 is worth it.

However, I want some REALISTIC dates.

Remember initial iram1 reports were saying $50 by summer 2005.

Note the label "designed for massive storage access application".

Is this a realisation that the REAL target market for iram is not who they thought it was. ie people want these to accelerate big databases more than help enthusiast gaming?

If they have realised the database managers are their customers might they charge more money for iram2 because they think database managers have fatter wallets?

I hope they can keep prices about the same, but the main question is WHEN will this come?

If they say a year away I will get some IRAM1s now. If they say 2-3 months away I may choose to wait.




Release date
By electriple9 on 12/6/2006 1:17:13 PM , Rating: 2
Any news on the release date.
Thanks




"There is a single light of science, and to brighten it anywhere is to brighten it everywhere." -- Isaac Asimov









botimage
Copyright 2010 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki