5 ms·
Just out of curiosity, what sort of serious tasks does one use 192GB of RAM for? (I'm not implying that there aren't any; just want some examples.)
by jawns 13y ago
Just out of curiosity, what sort of serious tasks does one use 192GB of RAM for? (I'm not implying that there aren't any; just want some examples.)
- jrockway 13y agoOpening more than one tab in Chrome.
- nakedrobot2 13y agoI guess the industries that benefit from "the more RAM the better, with no limit" would be 3d rendering, video editing, simulations, and large-format photography. I am doing the latter, joining together thousands or tens of thousands of images into a single seamless photo. for example: http://360gigapixels.com/tokyo-tower-panorama-photo/ http://360gigapixels.com/tokyo-tower-panorama-photo/ http://360gigapixels.com/tokyo-gigapixel-roppongi-hills-mori-tower/ http://360gigapixels.com/tokyo-gigapixel-roppongi-hills-mori... http://360gigapixels.com/london-320-gigapixel-panorama/ http://360gigapixels.com/london-320-gigapixel-panorama/ http://www.360cities.net/gigapixel/strahov-library.html http://www.360cities.net/gigapixel/strahov-library.html I am aware that in many applications, the GPU has taken over for some of the "heavy lifting", but that is not true in all cases. Given there are basically no real power or thermal considerations limiting the RAM for a computer like this, I would really like to know if this 64GB limit is real, or if it is due to what RAM modules are available on the market now. Is it a limit in the OS? Will the Mac Pro be able to have 256GB RAM in the future, or not?
- ch4s3 13y agoOff topic, but that tokyo tower panorama is super cool. You can see people in buildings and in cars really far away, and they look recognizable.
- mildtrepidation 13y agoit is 2013. 64GB is not enough for many serious tasks. You present a few valid examples. But I can't see how that represents "many" tasks; it's a very, very small subset of the fields in which people use computers for work. If there were no hardware out there that supports the RAM requirements you're talking about, that'd be a great area for development. But if you're suggesting everything -- or even most things -- should support what to most of us is an over-the-top amount of memory, perhaps your niche is not quite as major as you think.
- sn0v 13y agoI agree that it's a niche application, but imho it's one of the target demographics of a 4k+ USD workstation. I'd like to know the reason (software/hardware) behind the 64 GB limitation as well.
- scottlinux 13y agoThere are only four ram slots. The largest ECC stick available now are single 16GB sticks. Just not enough room for slots on the board, is my guess.
- melling 13y agoIt's a $4000 workstation. You buy it to run big jobs. Your department buys 10 or 20. In 3 or 4 years, the amount of "big data" grows and now you need more RAM. Do you simply spend another $100k to buy all new machines? You are just thinking about today's needs but a few years out memory needs always grow.
- bilbo0s 13y agoNot disagreeing with the memory point... just pointing out that a 12 core processor would, most likely, not be sufficient to the cause in 3 or 4 years if your memory needs grew that much. So wouldn't you buy new machines anyway every 4 years if your department were doing that sort of work???
- makomk 13y agoThe Mac Pro itself is a niche machine and priced as such, though. I mean, it uses a Xeon workstation/server-class CPU - how many computer users need Xeon and wouldn't be just as happy with a much cheaper high-end desktop CPU?
- mikeash 13y agoIt's not unheard of for Macs to be limited in how much RAM they'll recognize, but it's not common. The maximum is typically limited by the number of RAM slots and the largest modules on the market. The Mac Pro has 4 slots and 16GB modules are the biggest out there at the moment. I would put good odds on it being able to take e.g. 128GB when 32GB modules become available. But there's no guarantee.
- scott_s 13y agoIs it an absolute requirement that the machine on your desk is the powerful one? For example, lots of scientists and engineers rely on massive computing power to do their work, but that power is in a server room somewhere. They often have modest laptops on their desk.
- frozenport 13y agoWorkstation tasks involve as much ram as possible 1) Physics simulation 2) Large image manipulation ( what I do) 32GB is what my laptop has, for this reason I cannot take this as a serious machibe.
- collyw 13y agoLarge image manipulation has been done for years now, no? If we managed it ten years ago on comparably puny machines, what has changed? Images bigger I guess is one thing, but does that account for advances in RAM? (Saying that I am amazed at how my desktop slows down with 4gb of RAM these days.)
- huxley 13y agoBack in 1995, Toy Story was rendered with a server farm of 87 dual-processor 100MHz SPARCStations w/ 192MB of RAM and 30 quad-processor 100MHz SPARCStations w/ 384MB RAM. I used to do professional digital video editing on Media 100s and Avids with 80-96 MB of RAM back in the mid-90s. It's amazing how quickly top-of-the-line becomes crap.
- krakensden 13y agoBioinformatics uses lots of memory. 3-4 years ago people were buying 512GB machines for it, and I'm sure it's crazier now.
- collyw 13y agoMost bioinformaticians don't know how to code very well in my experience. (Some yes, but the majority I have worked with, no).
- scott_s 13y agoI doubt most people doing bioinformatics are doing their number crunching on the machine in their office. They're probably using a cluster that's maintained in a server room. Keep in mind that these are workstations, not servers. They are intended to be used by an individual who, for whatever reason, has interactive applications with large compute requirements.
- Fomite 13y agoMany scientific computing tasks are memory intensive - operations on large matrices, R and Python both tend to keep their data in memory rather than writing it to disk, heavy simulation, etc. I've slammed a cluster node with 1 TB of RAM pretty hard, I can easily see exceeding the 64 GB the Mac Pro comes with. Though my solution to this is admittedly "Toss it on the cluster".