It seems to me that the only thing nowadays that is slowing down the speed of the computers is the hard drive. Even though most now are at RPMs of 7200 and even SCSI spins at 10,000, they still cannot keep up with the demanding speed that current CPUs are thinking. What can we look forward to in the future as far as what will hard drives eventually contain to make them faster? Magnetic media is too slow to keep up with the rest of the computer system (GHz cpu chips, DDR RAM, etc).
p.s. - How soon can we expect a change in the current hard drive designs?
I kinda forgot to mention that even though they can spin faster, the transfer rate is still being limited by the ATA modes. 133 is good, but it still slows down the whole computer when loading the operating system, larger programs, movies, etc.
Well, there’s serial ATA now, which can achieve 150MB/s, but if what I’m reading is correct, the interface isn’t the bottleneck; it is the accessing of the data on the platters.
Solid state is the future, I’m sure of it - the price just needs to keep coming down.
Solid state, like huge flash cards? That would be sweet.
A related Question;
Why is there such a concern with making everything smaller, such as chips and HD`s. When I open my computer up, there is all manner of room for a big arse chip. Why do they need to shrink them down as they make them faster? Faster and a little bigger would work for me.
Flash memory can’t be used for main storage due to its poor reliability and EXTREMELY high cost per MB. Magnetic RAM (MRAM) has a shot, but it’s still a long way off. There’s really nothing on the horizon for replacing HDDs, just better HDD technology. By switching to the Serial ATA interface, increasing HDDs to 10,000rpms in mainstream products, and increasing data densities, we can look forward to larger, faster HDDs that are much easier to work with. Mangetout is also correct, in that current HDDs could all use the ATA/66 interface without suffering performance loss. However, SATA allows other SCSI-like features that will help things.
If you want more performance today, the cheapest way would be to buy two very fast HDDs, such as Western Digital Special Edition 80GBs or WD Raptor 10,000rpm 37GBs, and use a product called a RAID card to combine them using RAID-0 (called striping) into one very large, very fast drive, effectively doubling the performance. The card will run you about $100, but you can easily get some INSANE performance using this method. The only problem is that if one drive fails, ALL of your data is gone, and non-recoverable.
scr4: Nope, flash memory is quite fast. You can buy flash drives that connect to the IDE bus, they’ll easily max out any interface you give them.
whuckfistle: The obsession is with reducing the size of transistors and wires on a chip, which allows you to cram many more of them in the same space. We can’t easily make the chip bigger because of production costs and yield problems (for example, 10 chips on a wafer will be defective whether you cram 100 or 1000 onto that wafer. If you make 100 larger chips, you’re eating a 10% loss instead of 1%). There’s also the issue of power usage; current high-end CPUs consume over 100W, which is exceedingly hard to deal with. If we doubled that power usage by making the chip twice as large, the problem would become intractable.
The secret to ultra high data transfer rates for hard drive data is drive arrays. One I saw some years ago used one drive per bit for a 32-bit wide array. So instead of the 133 Mb/s data transfer rate you get with current technology, you would get 4256 Mb/s, which is a transfer rate that will overwhelm almost any bus in the industry. Throw in a few more drives for SECDED bits (single error correction/double error detection), and a few more for hot spares, and you have one kick-ass drive array with no risk of data loss (the failure of one drive merely exercises the SEC circuitry and forces a rebuild onto a hot spare).
RAID 0 (striping) is the same idea, just not done on as grand a scale and without the error recovery.
Another reason for going smaller is the speed of light. At 3 GHz, your signal only gets to travel about ten centimeters per clock pulse (if I’ve done my math correctly). The time it takes for a signal to traverse the chip is now a significant factor in chip design; the smaller the chip the smaller this time is and the easier it is to manage.
Another thing coming up is Blu ray & it’s gonna be great but I think its for video? (I Think you can buy it now):
“Blu-ray Disc enables the recording, rewriting and play back of up to 27 gigabytes (GB) of data on a single sided single layer 12cm CD/DVD size disc using a 405nm blue-violet laser.”
Derleth: Maxtor’s newest models hold 80GB per platter (40GB per side), making them the drives with the highest data density. A little back-of-the-envelope math indicates that this is roughly 2.6GB per square inch. In reality, I’d guess it’s about 3GB per square inch, maybe even a little more, to account for the hub at the center of the platter and other such things.
Really? Who makes those? The ones could find with a brief Google search (like this and this) are pretty slow, but maybe I haven’t seen the high-end stuff.
I am not sure what you are going at. HDDs have never been able to keep up with CPUs, or main memory (RAM) for that matter. Besides, it is not how fast the patters spin matters, it’s the transfer rate that makes the difference.
SCSI can now reach 320MB/s, and there is Serial SCSI in the works that is going to be even faster.
Well, I have just checked Maxtor’s Dimond Plus 9 datasheet and it claims a 150MB/s transfer rate for SATA interface. Do you have a cite for your assertion?
Not sure if it does. The PCI bus has a top speed of 133MB/s.
Urban Ranger, that’s not out of one IDE hard drive. It’s out of 32 IDE hard drives running in parallel with a custom controller. You’d obviously want a higher speed bus from the controller to the computer than anything in the Wintel platform line. IBM has had gigabyte-speed buses on their top-end computers for over a decade. Just because it’s never been implemented in the PC market doesn’t mean it doesn’t exist. The PC market is up to a decade behind best technology in some areas.
While Alereon makes some good points, the real issue is that smaller and faster are inherently tied together in chip design. If you want to make it faster, you have to make it smaller. The conductor lengths are smaller, so less travel (and charging) time. The transistors are smaller, so they switch faster, etc.
This is one of the odd miracles of solid state devices. They get smaller, faster, more powerful, more energy efficient, and cheaper to make-all at the same time. The only negative issue is pushing the limits of technology. But the tech seems to keep getting better.
Urban Ranger: That’s the interface speed, not transfer rate. The fast IDE HDD made, the 10,000rpm Western Digital Raptor HDD (also a SATA drive, though, like all currently made drives, not NATIVE SATA), tops out at about 60MB/sec. See this Anandtech Review. And, also, I’d consider above 100MB/sec in a workstation “insane performance.”