1TB RAM

[rich_m]

hi heard of this:::
Amazing 1TB RAM

[distantvoices]

ONe word: Hoax.

[JoeKayzA]

ONe word: Hoax.

My opinion, too.

Had only a short look at it, but: Windows version 2002? 1TB RAM?
A company to deliver that revolutionary technology - and such a CRAPPY website?


cheers Joe

[dh]

I heard that MIT university manages such a size on a hair sized chip :O

[Candy]

From website:

Sale Price: $8500-$15000 (depends
on the completeness).

If you sell computers costing 7000-12500 euros, you could make a decent website with decent photos.

BTW, for that price, it doesn't have to be a hoax. Might just be misplaced, since I don't know of any consumer that will pay at least 7000 euros just for a computer (or it'd have to be a really nice one, but just a big hd doesn't cut it).

Create your own with an external harddisk cabinet and 10 250-GB disks. Slightly less portable, at most a third of the price.

Thinking of it, only 7000 euros for a harddisk with nanosecond access times and 2TB of storage? With 1TB of memory with 0.3 ns access time? Probably is a hoax.

[Sami_Tervo]

But oh, would that be sooo cool

[JoeKayzA]

Reading some things more slowly:

Windows XP works with 64-bits.

Ah, yes. ::) This is a statement I would expect from a Joe Average User, but not from a serious IT company. BTW, the system properties dialog says it is windows xp version 2002, again, not the so-called 'x64'.

Nowhere in the paper I can find anything about the implemented processor architecture. If it is compliant to x86(-64), they would have stated this. If not, especially then they would have stated it.

And why to hell should the storage appliances be attached to ATA interfaces? If their RAM is non-volatile (as they claim) and unused parts can be used as storage - why do we need seperate storage then?

RAM department - 200pin SODIMM. Counter statement: one word is 4x64bits. 4*64 = 256. So with address and handshake lines we would need more than 300 pins, or did I miss something in today's developments? Memory bus compression ? Wireless memory bus ;D?

Don't know - even when the subject (the device) is not a hoax - the website definitely is.

cheers Joe

[Freanan]

Haven't you seen that it claims to have a quantum processor?
It is an obvious joke!

[Eero Ränik]

I actually laughed reading through the site. It is either just a joke, a hoax meant to get someone pay those prices for nothing, or to get people's e-mail addresses for spam lists (check the "Contact us" page). Every picture on the site looks fake, or too funny to be true.

[Xardfir]

Hiya,
On a more serious note some systems vendors in Australia are now offering PC's with 1 Tbyte of Hard-disk space for under $AU 2000 (whatever that is in your local currency).
www.pc-express.com.au, www.iicomputers.com.au, www.marvelcomputers.com.au.
I'm just imagining how long all of that data will take to backup!
Although I have nothing against progress (always need more space), sooner or later the transfer speed is going to hit a wall even if transfers are done using fibreoptics or whatever.
Any thoughts?

[Eero Ränik]

I just assume that when hard drive spaces are growing, so are the speeds. I don't believe it'll hit a wall anytime soon. Besides, I doubt that you, or many other people would have a terabyte of important data they really needed to backup.

[Xardfir]

Hiya, Any kind of video editing needs big files, a terabyte is probably just enough. Ditto 3-D modelling etc.
Hard disk speeds do not keep up with PC speeds. Although SCSI can hit 320MBPS' this is still only theroretical. Once you add control data etc, the actual transfer is a lot less.

Imagine a database server or webserver with 1000 requests pending (an amount that makes RAM caching irrelevant) the hard disk has to perform each request serially (even if those requests are optimised by the OS. Transfer speed goes to hell in a handbasket, even RAID only doubles the channel width.
You can cluster servers, but this doubles your equipment cost.

Neither of these approaches addresses the real problem - accessing huge amounts of information on larger hard disks is SLOW. The bigger they are, the slower they get.
I have heard many a debate about "Oh, smart indexing will fix it." No it won't, because most of the index will be paged out to the hard disk! The hard disk is still a serial device. It's cheap sure, but unlike processors or memory that can be mulit-threaded/interleaved, has a fundamental limitation.

[Eero Ränik]

Sure, but (I think) most of the people won't backup their raw video. I didn't say it'd be hard to fill up 1 TB.
I agree that there are, indeed, limits. Though, as I said, those limits are rising constantly.
Now, the main bottleneck with hard drives is the bus they connect to. I believe you wouldn't notice a huge difference between the speeds even if you used a solid state disk instead. Which is where you could keep the paging file, that would speed up smart indexing quite a bit. Hard drives are capable for more. There is no real need to start fighting the limitation, which is fundamental.
Oh, and 1000 queries/requests is not really a good example, it could be handled without extra expenses.

[Xardfir]

Hiya,

"1000 requests could be handled without extra expenses"

OK, that was just a small figure - change that to an absurd number of requests handled by a stock exchange, geographical/scientific datacenter, telecommunications company for example where everyone wants the information 'NOW'. Each request needs to be queued, this can only be done in software not hardware unless you purchase 1000 servers, each one handling one request. Those 1000 extra servers will cost 1000 times more in order to handle the load, not to mention the cost of maintaining the environemnt of those servers (air conditioning, power etc.) Not to mention the overhead of co-ordinating all of the requests.
This is a real life scenario. Massive 'data-centers' have to be set up this way to satisfy our insatiable need for info either simply to transfer or to calculate.
The reason these data-centers are so massive in my mind is because of that fundamental limitation of hard disks.
Flash memory can make things smaller but still suffers the same fate as it too is a 'serialised' form of storage.
So that is where the extra expenses would be incurred.

For comparison, think of VRAM versus DRAM as used in video memory. VRAM allowed the video controller and video display to access memory simultaneously.
If a hard disk could do the same, you could have one controller/transferrer performing writing and reading requests whilst the second interface would be setup to satisfy only read requests which could be handled by a standard network co-processor. Potentially you could halve the amount of equipment needed thereby halving the expense needed.

[Eero Ränik]

Ah, I see what you mean (or have meant all the time) now. I agree that a hard drive could do only one thing at a time, and the hard drive cache, that is meant to access the most-used data, is not as useful as it might be. However, it'd be hard to actually implement it to something like a hard drive, and if one was to use some alternative (like solid state disks, and I mean those that are RAM-based, and not those that are Flash-based, or optical disks, where it isn't too hard to get multiple lasers to do different tasks), the price for the same functionality (speed, space) would stay about the same or even rise. Things certainly could be made better, but would it save any cash, management costs, or even space? I think that if it really did, it would've been implemented already. Feel free to comment any of my thoughts.