will hard drives ever NOT be the bottleneck?

M

MetalDwarf

[H]ard|Gawd
Joined
Jan 20, 2002
Messages
1,072
even with SATA2, hybrid drives and the advances in Flash hard drives and storage are still the biggest bottleneck in any given system. when if ever do you think they will stop being the slowest part of the system. i hear Intel is working on some fancy new flash-esque technology but thats still years away.
 
Maybe when we have super cheap nonvolatile RAM. I'm talking about developments way down the line. When I was still in college, one of my chemistry professors had a colleague researching molecular NVRAM, but all this cool stuff is years away, at best.
 
While I'm eagerly awaiting advances to affordably better my Raptors, I'm really awaiting the advances in the Enterprise sector. There will hopefully be a day, soon, where large SQL / DB clusters and large Exchange environments won't need rows and rows of complex and expensive SAN hardware.
 
They'll always be a bottleneck because of what they're designed to do. Long term storage. Even when they come out with super duper lickety split drives, it'll still be slower than the rest of the machine. The cost of making a drive subsystem fast enough to not be a bottleneck makes it not really worth the benefit, considering that you'll only benefit during loading times (OS, games, programs), which already makes up the smallest portion of computer-use time.

There are tools today that enable you to build a system where the drive isn't the bottleneck, but you're talking about increasing the percentage of the build price that goes to the drive system.

I would like to see faster drives, and I think that companies should focus more on it, as the need for the largest drive available seems to be reducing dramatically, A 5TB drive costing $1000, I'm sure, would sell far fewer than a 500GB drive with 5x the performance of a Raptor for $1000. But even then, a drive that's 5x the speed of a Raptor (think of having a slew of Raptors in RAID 0 or 10) would still be the bottleneck of a fully modern DDR3 Quad core system.
 
Unfortunately, due to the nature of what the hard drive does, it will always be the slowest part of the system. You have to store, hold and retrieve massive amounts of data. It is just not cost effective to make the bulk storage as fast as the main memory, video card, and processor.

They make it bigger and faster, but the other parts of the system will continue to outpace the bulk storage as far as speed is concerned.

Don
 
They will NOT always be the bottle neck. Eventually, not in my lifetime, cpus and such will reach their limits and only be able to be pushed so far and the technology of Harddrives will continue to advance and catch up. I mean look at SSD drives at 120mb/s that is nearly 50% faster then a raptor and 100% faster then a standard drive. Advancements will continue, just slowly.

Truthfully eventually we won't have any HD's at all. Everything will either be streamed or my theory is your brain will actually store the information. Turning US into the harddrive.
 
You do realize they have already started implanting hearing devices into the surface of the brain and the brain begins to interupt the signals and the person learns to hear again? My theory is not too far off..
 
Eventually, not in my lifetime, cpus and such will reach their limits and only be able to be pushed so far
Click to expand...

No, they will always get faster and faster and faster. There will never be a time when processors reach their "limits" and can't go any faster.
 
visionviper said:
No, they will always get faster and faster and faster. There will never be a time when processors reach their "limits" and can't go any faster.
Click to expand...

+1

Look at what's happening now. When we started having issues with increasing clockspeed, they started increasing the number of processors. There's processor specialization, like GPUs and APUs. There's also faster bus designs, like HT. Nanotechnology is the hot new thing, but we can't predict what will come after that.
 
Plus carbon nanotubes can do pretty much anything... at least if you believe all the stuff the scientists say. Then there are optical and quantum and bio computers down the line.
 
We use DDR SDRAM because SRAM is too expensive(only used for CPU cache.) We use Hard drives partly because RAM is too expensive(as well obviously, as volatile.) My point is, should some insanely cheap, dense, and fast non-volatile RAM be developed, we'd have no need for hard drives. The entire concept of long term storage would be pointless, as all data would be equally accessible.

More than the technology itself, Windows will also need to be coded to recognized this(or rather stop from differentiating the data from Hard drives,) as RAM drives are limited to something like 32MB. And you couldn't connect it with SATA, or even PCI Exp x1-4. You need the full CPU/FSB bandwidth to fully utilize it. Utilizing a single storage medium, RAM, you would not have confusion between RAM and Storage, it would all be the same.

I don't know about specific research, but I see no purpose for hard drives in the "ideal" future PC. They will always be the bottleneck, because they are compensating for another system limitation already. In order not to be the bottleneck, they'd need to:
1. be Random access - precluding "drives" in the traditional sense
2. have transfer rates matching or exceeding that of system RAM - what are we approaching now? 12GB/sec?
3. have a method of transferring that data to the CPU and RAM that also has that bandwidth( ie SATA8)

So I was thinking one time before, why can't AMD extend their HyperTransport bus to more devices? Add a short external cable even. Fill a chassis with RAM, connect with HyperTransport, viola! Instant access to TB's of RAM/Storage at HT speeds. Perfect for the Library of Congress. :D

Until then, no.
 
verteron said:
We use DDR SDRAM because SRAM is too expensive(only used for CPU cache.) We use Hard drives partly because RAM is too expensive(as well obviously, as volatile.) My point is, should some insanely cheap, dense, and fast non-volatile RAM be developed, we'd have no need for hard drives. The entire concept of long term storage would be pointless, as all data would be equally accessible.

More than the technology itself, Windows will also need to be coded to recognized this(or rather stop from differentiating the data from Hard drives,) as RAM drives are limited to something like 32MB. And you couldn't connect it with SATA, or even PCI Exp x1-4. You need the full CPU/FSB bandwidth to fully utilize it. Utilizing a single storage medium, RAM, you would not have confusion between RAM and Storage, it would all be the same.

I don't know about specific research, but I see no purpose for hard drives in the "ideal" future PC. They will always be the bottleneck, because they are compensating for another system limitation already. In order not to be the bottleneck, they'd need to:
1. be Random access - precluding "drives" in the traditional sense
2. have transfer rates matching or exceeding that of system RAM - what are we approaching now? 12GB/sec?
3. have a method of transferring that data to the CPU and RAM that also has that bandwidth( ie SATA8)

So I was thinking one time before, why can't AMD extend their HyperTransport bus to more devices? Add a short external cable even. Fill a chassis with RAM, connect with HyperTransport, viola! Instant access to TB's of RAM/Storage at HT speeds. Perfect for the Library of Congress. :D

Until then, no.
Click to expand...

HT is designed to work with add-in cards too. Called the HTX slot http://www.hypertransport.org/
 
yea, but have you actually seen one? Now we just need APU's(audio) and SPU's(storage) to go with our CPU's and GPU's. I would gather that Intel is resisting, and pressuring others not to participate. Honestly, I think that PCI Express is a waste. Every slot needs to be 16x in physical size at least, let the software assign the channels dynamically. If I buy an x4 add-in card, I can't use my x1 slot, and my x16 is taken already. Seems like the HTX solution would be better.
 
You must log in or register to reply here.
Back
Top