LinuxQuestions.org (/questions/)
-   General (https://www.linuxquestions.org/questions/general-10/)
-   -   How big is 4 exabytes of ram? (https://www.linuxquestions.org/questions/general-10/how-big-is-4-exabytes-of-ram-877321/)

 darkstarbyte 04-27-2011 12:27 AM

How big is 4 exabytes of ram?

When I say big I mean in physical size how big would it be. Would it take up an entire room for that much ram or what?

A better question might be what is it's volume?

I say 4 exabytes because I hear that is the limit on 64 bit computers, as around 3.5 gigabyte to 32 bit computers.

I only ask because it is a seemingly uncompromisable amount of ram.

 smeezekitty 04-27-2011 01:10 AM

4 billion gigabytes.

 SigTerm 04-27-2011 01:11 AM

Quote:
 Originally Posted by darkstarbyte (Post 4337617) When I say big I mean in physical size how big would it be. Would it take up an entire room for that much ram or what?
It isn't hard to calculate it, you know. There's gnu bc, wolfram alpha, and RAM manufacturer's websites.
A volume of random kingston DDR2 2GB ram module is 30 x 133.3 x 2.7 mm3. That's 0.0000107973 m3.
4 exabytes == 2^62 bytes. 1 2GB plank == 2^21 bytes. 2^62/2^21 == 2^41 planks.
0.000107973*(2^41) roughly equals 23743513.8 m3.
According to wolfram alpha that volume roughly equals to 5% of all water in sydney harbor, or to a sphere with the radius of 173 meters or a cube with size of 287 meters (3 american football fields, according to wolfram alpha). I.e. a cube with the size larger than 3 football fields, filled with 2GB DDR2 memory modules.

I do not think people will need THAT amount of memory any time soon, though.

 Coresay 04-27-2011 01:25 AM

I calculated it for a HDD... 205 ft diameter HDD.

 dugan 04-27-2011 01:34 AM

Depends on the shape, size and density of the RAM chips. RAM doesn't have to come in DIMMs and it doesn't have to fit any established form factor. You can take advantage of that if you're, say, designing a supercomputer.

@ dugan- true, you dont need to use DIMMs. Though if I was designing a suercomputer I'd consider using standard DIMM sticks to save money.

Quote:
 Originally Posted by SigTerm (Post 4337651) It isn't hard to calculate it, you know. There's gnu bc, wolfram alpha, and RAM manufacturer's websites. A volume of random kingston DDR2 2GB ram module is 30 x 133.3 x 2.7 mm3. That's 0.0000107973 m3. 4 exabytes == 2^62 bytes. 1 2GB plank == 2^21 bytes. 2^62/2^21 == 2^41 planks. 0.000107973*(2^41) roughly equals 23743513.8 m3. According to wolfram alpha that volume roughly equals to 5% of all water in sydney harbor, or to a sphere with the radius of 173 meters or a cube with size of 287 meters (3 american football fields, according to wolfram alpha). I.e. a cube with the size larger than 3 football fields, filled with 2GB DDR2 memory modules.

You've used the stick only size, to be more accurate you probably should consider the extra size from mounting slots, and some extra room for airflow. You dont want your 4EB getting heat issues.

2GB DDR2 sticks, fair enough, but 8GB DDR3 sticks are easy to get. Being DDR3 the airflow requirements would be lower as well. If the 16 or 32GB sticks that samsung announced in 2009 are actually avaible, that would drop the required volume for 4EB a huge amount.

 silvyus_06 04-27-2011 02:31 AM

haha lol i wish we at least had HDDs as big as that :\

 Coresay 04-27-2011 02:50 AM

I wonder if 4 exabytes is enough for a Star Trek transporter buffer to hold the pattern of 1 human.... checking wolfram... it is! Unfortunately, the space required to store a human pattern by today's standards is several orders of magnitude larger than the body itself... bummer.

One way to assess that is to answer a "simple" question... how much information does one Human being contains... ?

I do not know any information-theoretic procedure to assess that, but considering a frozen state of all cells, and their metabolic state ( is is still a "macroscopic snapshot" ) i donīt think an exabyte is enough...

Quote:
 Originally Posted by cascade9 (Post 4337710) @ dugan- true, you dont need to use DIMMs. Though if I was designing a suercomputer I'd consider using standard DIMM sticks to save money. 2 minor additions to your maths- You've used the stick only size, to be more accurate you probably should consider the extra size from mounting slots, and some extra room for airflow. You dont want your 4EB getting heat issues. 2GB DDR2 sticks, fair enough, but 8GB DDR3 sticks are easy to get. Being DDR3 the airflow requirements would be lower as well. If the 16 or 32GB sticks that samsung announced in 2009 are actually avaible, that would drop the required volume for 4EB a huge amount.
Quite correct... you want to maximize surface area ( where heat is lost by convection ) to volume ( where heat is generated by joule effect )

A "solid" shape of zero volume to infinite surface is the Menger Sponge or a Sierpinsky Cube, but this is highly impratical to fabricate... a good enough shape is the cylindrical annulus with diameter equal to height...

This is how Israelis in Dimona complex store weapons grade plutonium-Indium alloys, maximizing the surface ( neutron and heat loss ), minimizing volume ( heat generation and probablility of neutron interaction ) for a given mass...

 Coresay 04-27-2011 03:32 PM

Quote:
 Originally Posted by Alexvader (Post 4338169) One way to assess that is to answer a "simple" question... how much information does one Human being contains... ? I do not know any information-theoretic procedure to assess that, but considering a frozen state of all cells, and their metabolic state ( is is still a "macroscopic snapshot" ) i donīt think an exabyte is enough...
WolframAlpha says that there are 1x10^14 cells in the human body. An exabyte provides 1x10^18 bytes... 4 orders of magnitude more. At least a macroscopic snapshot would work if you only wanted to map the cell-type, and it's 3D coordinates. 1 enum and 3 longs or BigLongs. Plus, whatever few extra bytes needed for whatever collection you choose to store the data (a multi-linked list, an array?). Ya, if you want to record more info from there, such as the state of the mitochondria, you may end up with a data explosion. But, you can prolly get away with generalizing on the cell-types. And, you only need to store one copy of the DNA sequence.

 SigTerm 04-27-2011 03:49 PM

Quote:
 Originally Posted by cascade9 (Post 4337710) You've used the stick only size, to be more accurate you probably should consider the extra size from mounting slots, and some extra room for airflow. You dont want your 4EB getting heat issues.
I was talking about tightly packed memory sticks that aren't plugged in.
Aside from packing issues, heating issues, this amount of RAM is currently useless for any sane practical purpose. It will take many years to simply walk through every byte. For example, if BIOS POST will take 1 millisecond to check every 2GB of RAM, then it'll take 70 years to check all available memory.

 Coresay 04-27-2011 05:10 PM

Quote:
 Originally Posted by SigTerm (Post 4338471) It will take many years to simply walk through every byte. For example, if BIOS POST will take 1 millisecond to check every 2GB of RAM, then it'll take 70 years to check all available memory.
I think your calculations are a little off. It's more like 555.55 hours, or 23 days... using your figures.

 darkstarbyte 04-27-2011 09:01 PM

I think one of these days we could take ram to the atomic level, but even then 4 exabytes would be very massive. I wonder if ram could be more densely packed than 2GB per stick and still maintain volume?

Edit: Going with that star trek thing. tele-portation might just be impossible with only a 64 bit cpu you may need a 256 bit cpu so it could coronate everything on the x, y, and z axis. 256 bit should pull it off though I think you would need a lot more ram.

With 4 exabytes of ram you should be able to store a backup of every website currently in existence.