Gigabyte i-ram box--fiction?

[RADCOM]

I can't seem to find these for sale anywhere? I have the RAM for it lined up. Does anyone actually have one and where did they get it?:shrug:

[TheGoat Eater]

eBay is your friend
eBay (Taiwan) (http://goods.ruten.com.tw/item/show?11080714033828)


RADCOM -!!!!!!! BOOKMARK !!!!!!! (http://shop.ebay.co.uk/items/Computing__i-ram_W0QQ_nkwZiQ20ramQQQ5fcatrefZ1QQ_flnZ1QQ_sacatZ 58058) pre-sorted for your ease in the UK eBay

will update as I go through the countries - all the international sellers on US eBay are OOS it looks like... will look at UK eBay too and scour the intewebz for ya
though I spent a lot more time on tracking it down than i should as I most likely need sleep ;) :D an had to dig out the XP install CD for Taiwanese language pack (I think that is right)

EDIT - too much looking not enough end result... SCAN (http://www.scan.co.uk/Product.aspx?WebProductId=783843) overdue since march '06 LOL:down::slapass:

Importing looks to be the ideal choice - now if you can only read Taiwanese

[IanB]

Yes, I noticed SCAN reintroduced it into their site with that Out-of-Stock label. They removed the original listing (of the PCI card version) abiout a year ago, along with just about every other etailer in the UK... a few weeks before I finally had the cash to buy. :mad:

The reintroduction, and of the boxed version, suggests they do have some confidence of getting stock eventually. I am hoping to swap my PCI card versions for the boxed versions at some point, so I was going to email them to confirm this. RADCOM, if you do, I'd be very interested to hear their response.

If anyone is interested, I did a lot of digging in the PCI specs while trying to design a power mod that would route power to PCI-card i-RAMs without them beng in a PCI slot. I was more than a little shocked by the way GigaByte implemented the slot connections, and I believe that in a system with a lot of expansion cards added, the PCI i-RAMs could be a source of instability because of the way they draw power without notifying the mainboard that they are doing so.

Specifically, there are two mechanisms for notifying the power consumption, one is a connection to two PCI pins (PRSNT1# and PRSNT2#). See the PCI base specs (http://www.i-asm.com/PCI_22.pdf), section 2.2.7:


PRSNT[1:2]# (in)
The Present signals are not signals for a device, but are provided by an add-in board. The Present signals indicate to the motherboard whether an add-in board is physically present in the slot and, if one is present, the total power requirements of the board. These signals are required for add-in boards but are optional for motherboards. Refer to Section 4.4.1. for more details.

Implementation Note: PRSNT# Pins
At a minimum, the add-in board must ground one of the two PRSNT[1:2]# pins to indicate to the motherboard that a board is physically in the connector. The signal levels of PRSNT1# and PRSNT2# inform the motherboard of the power requirements of the add-in board.
The PCI card version of the i-RAM does NOT connect these pins, even to ground, so the mainboard has no idea it is drawing power from the PCI bus.

The other mechanism is required due to the i-RAM drawing standby power from the PCI 3.3Vaux line when the PC is shut down. From the PCI power management specs (http://www.i-asm.com/PCI_pm1.2.pdf):


5.4.3. 3.3Vaux/D3cold Add-in Card Power Consumption Requirements
When the system is switched from its main supply outputs to the auxiliary power source, strict power budgeting with respect to which slots are allowed to consume full 3.3Vaux power becomes necessary. A PCI function must draw no more than 20 mA through the 3.3Vaux pin when in D3cold if its PME_En bit is cleared (i.e., 0b).

If a PCI function has been enabled for PME# generation (PME_En = 1b) prior to having entered into the D3cold state, the PCI add-in card (any single function or combination of multiple functions) may continue to draw up to 375 mA through the 3.3Vaux pin while in D3cold.

Auxiliary Power Consumption Reporting
The optional Data Register has been defined to enable the reporting of fine granular power utilization data for each of the supported power management states. In order to minimize implementation cost while also ensuring a robust architecture, a three bit field has been defined in the PMC register that is required for PCI functions that do not support the Data Register yet wish to draw from 3.3Vaux while in D3cold.

PMC(8:6), (“Aux_Current”), provides a rudimentary power reporting vehicle for PCI functions to indicate their maximum required 3.3Vaux current. Power budgeting software can then use this data to determine how many PCI functions can be configured for wakeup from D3cold.

7.4.2. Physical Connection to the 3.3Vaux Pin
Only PCI add-in cards implementing functions that support PME# from D3cold by design (i.e., functions that report PMC(15) = 1b) may physically connect to the 3.3Vaux pin (14A). All other designs must not come into electrical contact with pin 14A and must only draw current from the main PCI bus voltage rails.

The i-RAM has no on-board ROM that the mainboard can access, and is not reporting in any way to the mainboard that it is drawing power from the PCI slot either in standby (the D3cold state referred to above) or normal power. I regard this as a very significant design flaw, and for this reason I feel the boxed version of the i-RAM is "safer" than the PCI card version, as it only draws power directly through a vampire tap of the power supply cables.

[RADCOM]

Cheers Goateater, ebay is my mortal nemesis actually lol. I was thinking about your tag when I was going to post Jamaican curry goat in the food thread lol. I had looked on ebay but they only have the PCI version :(
Thanks for the indepth explanation IanB, I hadn't bothered emailing SCAN as I figured they'd update the site when they had any news. I think I'll contact Gigabyte.

[IanB]

SCAN are now showing stock of the boxed version due on 29th July. :up::woot:

http://www.scan.co.uk/Product.aspx?WebProductId=783843

[Stigma]

Hmm why, really bother with I-ram at this point in time though? For the price of the I-ram unit itself, you will be able to get an OCZ core2 SSD very soon which will be better in most aspects.

-Stigma

[IanB]

Hmm why, really bother with I-ram at this point in time though? For the price of the I-ram unit itself, you will be able to get an OCZ core2 SSD very soon which will be better in most aspects.

-Stigma

In your opinion. :rolleyes:

Why the anti-i-RAM threadcrap? If you don't want one, then go buy your SSD and have fun. But us i-RAM users will have faster streaming data, faster absolute access times and (likely) longer longevity, even if we don't have huge capacity (we just prioritize it where it counts, on the OS). That seems like better in "most aspects" that matter to me. :shrug:

[RADCOM]

I called Gigabyte UK office and they advise stock will be available in 6 weeks.

[Sunayknits]

Well I suppose now is a good time to discuss my experience with Gigabyte's iRam-Box :rolleyes:

I did a lot of research on these some time ago, then purchased 2 on EBay. They came directly from China, from seller "asiannetstore".

I was very excited about these as they offered absolute saturation of the SATA-1 bus, with near-zero latency. I was going to use 2 in RAID 0 for a total of ~7Gb of useable space. This is enough for Win XP and a pagefile but not much else. I had 2 Raptors in RAID 0 that I would use for game installs and such.

I was very curious if they worked with 2Gb DIMMS and had read a great number of conflicting reports, so I bought several sticks of 2Gb ECC DIMMS from various vendors to try this out. 2Gb DIMMS are all ECC and generally considered "server memory" so they are super expensive in comparison to
standard 1Gb units.

In any case, 2Gb DIMMS did not work. The memory controller on the board is not designed to handle more than 4Gb of address space and if you put anything beyond that on the board it will fail once you get past 4Gb.

So I filled them each up with 4Gb and did some preliminary tests. Let me tell you, just having the OS on these units makes a computer lightning fast. Even with games on the hard disk, the operation of the PC in day-to-day tasks is phenomenal. I was very pleased with performance.

However, this comes with a major cost. DRAM must be refreshed continually and uses a small amount of power to do so. If you lose power you lose it all; just like when you turn off your PC everything in RAM is trashed.

Gigabyte uses a battery backup to prevent data loss for several hours. The battery is charged by a separate connection to the +5VSB line from your power supply (they supply a pass-through dongle to facilitate this). The original iRam used power from the PCI slot for the same reason.


Here is the huge, glaring oversight they neglect to mention:

They claim you can leave your computer off, but as long as it is plugged in your data will be retained. I found that this was not the case. It's possible that I have defective units but I highly doubt it as they were brand new, and the batteries hold a charge for nearly 10 hours.

I found that once the battery is charged, it will begin to drain until it's empty, regarless of whether or not +5VSB is supplied. Once it has quit charging, if you unplug the PC for a moment or simply turn the power supply off then back on, the unit will begin charging again, until it is full then it will begin draining again.

The result is that if you don't turn on your PC or cycle power before the battery dies, you will lose your data.

I went to all the trouble to custom-mount these in my case because I had no 5.25" bays left, only to find out they are utterly useless.

You can't even use them for a page file in this fashion because you would have to format the drive each time at boot in order to use it.

This is absolutely, completely, utterly the most retarded thing I have ever seen.

I have verified with a multimeter that they were getting power. I also used a video camera to record the LEDs on the face as the unit was charging. I watched the red light simply turn off after charging was complete, then never turn back on.

I wrote their tech support, and after several days they finally replied telling me my battery was bad. Sigh.

Well live and learn I guess, I was determined to know more about these and I sincerely hope reading this will prevent others from wasting their time with these retarded monstrosities of computer malfunction.

Here are some photos of my efforts:

http://www.brimbo.com/temp/mozartbuild/iRamiSham_001.jpg
i-Ram Box splayed

http://www.brimbo.com/temp/mozartbuild/iRamiSham_002.jpg
4Gb DIMMS installed

http://www.brimbo.com/temp/mozartbuild/iRamiSham_003.jpg
Added custom mounts

http://www.brimbo.com/temp/mozartbuild/iRamiSham_004.jpg
Test fitting

http://www.brimbo.com/temp/mozartbuild/iRamiSham_005.jpg
Improvised straps made from old slot covers :yepp:

http://www.brimbo.com/temp/mozartbuild/iRamiSham_006.jpg
Ready for installation

http://www.brimbo.com/temp/mozartbuild/iRamiSham_007.jpg
My custom wiring (note the tap into the ps line for continuous +5VSB)

http://www.brimbo.com/temp/mozartbuild/iRamiSham_008.jpg
Completed installation

[Sunayknits]

But us i-RAM users will have faster streaming data, faster absolute access times and (likely) longer longevity, even if we don't have huge capacity (we just prioritize it where it counts, on the OS). That seems like better in "most aspects" that matter to me. :shrug:

It is very easy to surpass iRam performance right now by using 2 SSDs in RAID 0 on a decent controller, not to mention much more cost effective and far less of a hassle. In 6 months, a single SSD will likely provide double the bandwidth of a single iRam. Access times for the iRam are indeed faster, but this is not something you will ever notice in real life.

Furthermore, you don't have to worry about that big Li-on battery inside the iRam overheating, bulging, and eventually bursting (perhaps in flames) over time , or listen to the annoyingly loud squirrel-cage fan running at a zillion RPM inside just to keep the unit cool :shocked:

Also, the larger size allows you to install all your apps and games on ultra-fast storage too. This makes a HUGE difference in my experience.

Yes, I have reason to hate the iRam. I wasted several weeks and many dollars on this business which I can't get back so forgive me if I'm a little bitter :upset: Frankly I'm surprised that it's taken me this long to post about this absurdity as I've been rather upset about it for some time ...

[m^2]

It is very easy to surpass iRam performance right now by using 2 SSDs in RAID 0 on a decent controller, not to mention much more cost effective and far less of a hassle. In 6 months, a single SSD will likely provide double the bandwidth of a single iRam. Access times for the iRam are indeed faster, but this is not something you will ever notice in real life.

Furthermore, you don't have to worry about that big Li-on battery inside the iRam overheating, bulging, and eventually bursting (perhaps in flames) over time , or listen to the annoyingly loud squirrel-cage fan running at a zillion RPM inside just to keep the unit cool :shocked:

Also, the larger size allows you to install all your apps and games on ultra-fast storage too. This makes a HUGE difference in my experience.

Yes, I have reason to hate the iRam. I wasted several weeks and many dollars on this business which I can't get back so forgive me if I'm a little bitter :upset: Frankly I'm surprised that it's taken me this long to post about this absurdity as I've been rather upset about it for some time ...

Flash is poor when it comes to small, random writes, things like temp would work best placed on RAM and I think that it should be to whole OS, as even relatively small amount of writes has huge impact on Flash performance.

RAM has lower access time too.

Side note: You seem to get one thing wrong - iRAM is a SSD (http://en.wikipedia.org/wiki/Solid_state_disk).