SPCR

And on a completely different subject, I brought up the 20,000 RPM hard drive rumor. I should say 'rumors,' since they're nothing new, but always seem more and more plausible. While Ted told me that he wouldn't talk about products in development, he was completely free to talk about products that don't exist. There's no way, with existing or planned technologies, that anyone will make a 20K hard drive. It's that at those speeds, it would be hard to control the vibration, even if the platters were only the one-inch variety. He mentioned trials (experiments, really) in hydrogen gas environments... it's just not on the horizon.

http://www.thetechlounge.com/article/52 ... of+Caviar/

Oh, and new drives from WD, including 1TB. Will be interesting to see noise/vibrations.

This was also interesting:

With all the talk about reducing vibrations, I asked Ted how hard drive suspension affected performance. "It's horrible!" Not only does it increase the chances that the heads sidetrack, in some cases, the drive hits a resonant frequency inside the case, causing massive and repeated failures. If you're Hell-bent on suspending the drive, at least benchmark it before and after: see for yourself.

With all the talk about reducing vibrations, I asked Ted how hard drive suspension affected performance. "It's horrible!" Not only does it increase the chances that the heads sidetrack, in some cases, the drive hits a resonant frequency inside the case, causing massive and repeated failures. If you're Hell-bent on suspending the drive, at least benchmark it before and after: see for yourself.

Wow, that is awesome and the first time I've heard from an expert on the topic

I like how their diagram has the WD drive bursting out of the graph with a big logo on top to make it difficult to tell where the top of the bar actually is. : P Also, the chart begins at 5000 to exaggerate the differences between the drives, and they've disabled NCQ to give their drive more of an edge. I'm sure that their three platter terabyte drive will be faster than Samsung's overall, but it should be, seeing how the competition's been out for over half a year already. : )

Well, SSD drives are finally starting to get where vendors promised they should be last autumn.

http://www.ewiz.com/query.php?categry=0 ... asterdrive
Super Talent 1.8 inch 120GB MasterDrive KX Micro-SATA Solid State Drive
# Performance:

* Sequential Read Rate: 120 MB/s (max)
* Sequential Write Rate: 40 MB/s (max)
Price: $646.75

Which looks pretty darned good if you compare with the 1.8" HD alternatives and I believe its just a matter of months before you see a 1.8" SSD that is both faster and bigger than the HD alternatives.

I believe there was promises from some vendors early last year that there would be 120GB SSDs around 600 USD in August/September last year, but we seem finally to get there. One year later than the vendors promised.

That in any case a cost reduction of some 50% in roughly a year, which is pretty good and I would not be surprised to see that accellerate as you got a whole bunch of new producers that has announced that they will enter this market this year, including Intel and some of the big memory producers.

If SSDs continue dropping in price like now, I suspect the SSD will be cheaper than the 1.8" HDD in less than 2 years as well.

If we look at the other extremes of the SSD market now, you have on the performance side, the fusion-io which plugs straight into a PCI-e slot and goes up to 360GB on a single card with 6-700MB/sec transfer rates (bloody expensive of course). This card cannot be bought yet I believe, but it does exists and indeed works. There are very real prototypes out there which I can guarantee works, although performance is not fully as promised yet, they are still faster than anything else except pure memory drives and then it must be a drive that does not use a traditional storage interface as none of those can compete on throughput and latency with PCIe

On the capacity side BitMicro say they can offer 1.6TB capacity in 3.5" with their Altima E-Disk, but I am actually not sure if they produce them or just say they will produce if someone orders. In any case, you already got a major vendor out there claiming that they can deliver SSDs with higher storage density than any available HDD today.

Look at the extremes that vendors already offer and consider that there are already far more SSD vendors (companies with products, or which have announced that they will make products) than HD vendors (which number is dropping) and it should be very likely that we will see a very competitive market in the near future.

I really think that once the big silicon producers throw themselves seriously into this fight, prices will start crashing down and value will increase as quickly (if not faster) as we have seen on the memory market the last few years.

I am not suprised if vendors can push out a 20 rpm reasonably soon if they really want. After all, 15k rpm technology is a quite old now and I would really expect that we are able to make materials that can push this up 30% by now. I would however think that it only keeps the performance HD business alive max 2-3 years more, so the question is... is it really worth developing.

I don't really think so myself, unless the technology is practically there and can be pushed out at hardly any cost.

m^2 wrote:I wanted SSDs obviously, 8GB hdds are very slow...
And are there any fast 8GB SSDs? Mtron starts at 16, Memoright at 32, OCZ at 64.

dhanson865 wrote:

m^2 wrote:I use 20GB for my system with a lot of programs and several games.
RAID5 of 4 fast 8GB drives with a good controller would be perfect for me. I don't like the fact that the latest drives are not that small anymore..

1. RAID 5 penalizes writes more than RAID 1 which penalizes writes more than a single drive.

You're right, I didn't think about it. TMS uses it (RAID3 IIRC, but it's no difference), but I wouldn't trust consumer disk wear leveling. So 1 and 10 seem to be a better option.

dhanson865 wrote:2. SSDs already have a write penalty, combining that with RAID 5 is not going to give you good results compared to your other options.

That's why I said that I want a good controller. For home use, even 32MB cache should easily solve the problem with small writes. And RAID5 has better str than 10.
Anyway, prices are still too high.

Prices are too high but when they come down why should an end user buy a dedicated RAID controller? That just offsets the price that you've been waiting on.

By the time a 128GB SSD is under $100 I expect it to already be fast enough that I won't worry about small writes or str. I expect by then my only decision will be if I want a single drive or RAID 1.

dhanson865 wrote:

m^2 wrote:I wanted SSDs obviously, 8GB hdds are very slow...
And are there any fast 8GB SSDs? Mtron starts at 16, Memoright at 32, OCZ at 64.

dhanson865 wrote:1. RAID 5 penalizes writes more than RAID 1 which penalizes writes more than a single drive.

You're right, I didn't think about it. TMS uses it (RAID3 IIRC, but it's no difference), but I wouldn't trust consumer disk wear leveling. So 1 and 10 seem to be a better option.

dhanson865 wrote:2. SSDs already have a write penalty, combining that with RAID 5 is not going to give you good results compared to your other options.

That's why I said that I want a good controller. For home use, even 32MB cache should easily solve the problem with small writes. And RAID5 has better str than 10.
Anyway, prices are still too high.

Prices are too high but when they come down why should an end user buy a dedicated RAID controller? That just offsets the price that you've been waiting on.

By the time a 128GB SSD is under $100 I expect it to already be fast enough that I won't worry about small writes or str. I expect by then my only decision will be if I want a single drive or RAID 1.

Most people don't need 128 GB for system and some (including me) are willing to pay more than 100$ if it gives substantial performance gain.

Terje wrote:Well, SSD drives are finally starting to get where vendors promised they should be last autumn.

http://www.ewiz.com/query.php?categry=0 ... asterdrive
Super Talent 1.8 inch 120GB MasterDrive KX Micro-SATA Solid State Drive
# Performance:

* Sequential Read Rate: 120 MB/s (max)
* Sequential Write Rate: 40 MB/s (max)
Price: $646.75

Which looks pretty darned good if you compare with the 1.8" HD alternatives and I believe its just a matter of months before you see a 1.8" SSD that is both faster and bigger than the HD alternatives.

OK, as long as we're talking about ultra mobile market it may be good. But almost all use 2.5" or 3.5" and there 40 MB/s just doesn't cut it. And it is MLC, so fear low lifetime.

Terje wrote:I am not suprised if vendors can push out a 20 rpm reasonably soon if they really want. After all, 15k rpm technology is a quite old now and I would really expect that we are able to make materials that can push this up 30% by now. I would however think that it only keeps the performance HD business alive max 2-3 years more, so the question is... is it really worth developing.

Well, the 20kRPM Raptor was just a hoax. It's unlikely that we'll ever see HDDs that fast. You may want to read this article from the last year.

m^2 wrote:[OK, as long as we're talking about ultra mobile market it may be good. But almost all use 2.5" or 3.5" and there 40 MB/s just doesn't cut it. And it is MLC, so fear low lifetime.

Hm, I was only trying to refer to the extremes in the market. The read performance they are talking about will most likely beat any 2.5" laptop drive anyway, so even though the writes might be a bit slower, I would think the total would be better than even a 2.5" drive.

Normal 2.5" is coming up next and might end up beating anything mechanical you can find.

Look at what for instance Samung is promising in September:
http://gizmodo.com/393198/blazing-samsu ... aiting-for
---
With a sequential read speed of 200 megabytes per second (MB/s) and sequential write speed of 160MB/s, Samsung's MLC-based 2.5-inch 256GB SSD is about 2.4 times faster than a typical HDD. Furthermore, the new 256GB SSD is only 9.5millimeters (mm) thick, and measures 100.3x69.85 mm.
---
same article also claims that this MLC will have similar reliability to SLC and you will see a 1.8" version by end of the year.

I expect delays, like normal, so I would not be surprised if we don't see any of this until next year, but its a pretty good sign that you will have SSDs that beat HDD on the performance front about a year from now and be quite competitive in terms of capacity as well (within the performance segment).

Nope, wont happen this year, but 2-3 years down the road, I would think you are more likely to get an SSD with any middle to high end laptop than HDD.

As for the article you refer to.... I doubt you will ever see hybrid drives in a server. There are already SSD caching systems out there, but they are not taking off so far and I doubt you will see them take off either.

I am not putting my money into WDC stocks. That for sure (unless I am shorting of course ).

Terje wrote:same article also claims that this MLC will have similar reliability to SLC and you will see a 1.8" version by end of the year.

I think it's because of very high capacity. MLC cells are said to have 10 times lower number of writes, but drive lifetime scales linearly with capacity. If they reduced 10 to 8...than this drive would be as durable as 32GB SLC.