Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
..and then there were three.
What do you guys think? Acceleration towards full SSD or something actually good for mass storage? ^_^ Discuss.
Comments
I think Seagate just wanted to get a jump start into SSDs.
Not a bad move on either company's part, Samsung wasn't a huge hard drive player, and it gives Seagate an immediate boost in the consumer SSD market. I think before they just had the hybrid drive, and an enterprise line.
You say that as though there are still people who buy hard drives.
More seriously, the hard drive market is dwindling, and all of the major players know it. The days of every computer needing a hard drive are over. If you don't count tablets or phones, then most computers still need a hard drive, but even the days of that are coming to an end. That will take some years to play out, but it's coming, and the major hard drive manufacturers all know it. That surely played a big role in the decisions of Samsung and Hitachi to get out. It wasn't just that things were bad today; it was also that there was no real prospect of things getting better in the future.
But not all portions of the market will disappear at the same speed. I don't know what goes on with enterprise hard drives, but I'd imagine that if the days of buying large RAID arrays of 15K hard drives aren't over, they will be soon. When a single SSD completely smokes a large RAID array of 15K hard drives in the one thing that the RAID array is heavily optimized for, namely, IOPS, what's the point of the large RAID array anymore? That hurts Seagate a lot more than Western Digital, as Western Digital isn't a major player in that market. This is kind of like how, if professional graphics cards suddenly became unnecessary, Nvidia would get hurt badly by it while AMD would shrug it off as not that big of a deal. Enterprise will surely still need large RAID 6 or RAID 10 arrays for reliability and/or sequential throughput, but without the high performance requirements of a single drive, that might not drive the same profit margins as the 15K hard drives.
On the consumer desktop side, I count two major players now: Western Digital and Seagate. Not three, as Toshiba isn't active in that space. There, Western Digital has the better products (Caviar Black) and can charge a price premium. That leaves Seagate in a bad spot, as they have to sell their products at a discount to have much of a point, but their products probably aren't meaningfully cheaper to produce. Compare that to the bind AMD is in with desktop processors now: it costs as much to make a Phenom II X4 as a Core i5 2500, but the Phenom II has to sell for a lot less to get any customers.
Now, on the consumer desktop side, I see hard drives being around for a long time. Too many people accumulate too much junk that they don't have any real use for but don't want to get rid of, and so they need a hard drive to make space. The power, noise, and fragility issues aren't a big deal in desktops, either. So Western Digital could survive and even thrive on hard drives for quite some time to come. Not forever, but that market has a lot of life left in it.
The laptop side is different, however. Here, power consumption is a meaningful issue, and the fragility of hard drives is a big problem. Furthermore, far less capacity is needed in laptops. I'd argue that if you need more than 120 GB of capacity in a laptop, you're probably using it for something that you shouldn't use a laptop for. What happens when that 120 GB SSD costs $100? What happens when it costs $50? Will people still buy hard drives for laptops? Actually, some people probably will out of inertia for quite some time, not realizing that using 50 GB out of a 120 GB SSD is far preferable to using 50 GB out of a 1 TB hard drive.
But still, I see the market for laptop hard drives drying up much more quickly than that of desktop hard drives. Western Digital has the performance advantage here, too, with their Scorpio Black hard drives. Seagate would counter that their Momentus XT is better, but they're going to have to come up with more than 4 GB of NAND flash and use it a lot better to have much of a case there. And Toshiba? Their hard drive division is surely in trouble, too, and I wouldn't be surprised if they want to sell.
-----
So where does this leave everyone? Western Digital is still sitting pretty in the hard drive market for a while. They'll need to adapt to hard drives becoming less important, but they've got some years. The problem is that their entries into the SSD market have been a complete fiasco. They bought up SiliconSystems for their SiliconDrive SSDs, which cost a fortune and don't seem to have any real point. Next, they tried to sell a rebranded JMicron SSD as the SiliconEdge Blue. The problem is that these drives were thorough obsolete the day they launched, not only trailing far behind older Intel and Indilinx drives, but prone to completely choke on some write workloads and lose badly to a simple hard drive.
Seagate, on the other hand, is in a worse spot on hard drives, and not a better spot on SSDs. Seagate has launched their Pulsar SSDs based on an unknown controller, and the drives promptly got ignored. Pulsar is targeted at enterprise, and Seagate hasn't yet tried to sell any SSDs to consumers.
Samsung, on the other hand, is in a better position. They're out of hard drives, so what happens to hard drives isn't their problem. Samsung is one of the major NAND flash vendors, so greater adoption of SSDs helps them. Samsung also makes their own SSD controller. Samsung is thus able to build their own Samsung 470 SSDs completely in house, and they're pretty good SSDs. They trail behind SandForce and Marvell drives in performance, but they're still nice SSDs.
Hitachi is out of the hard drive market, so what goes on in hard drives isn't their problem. Hitachi may well be out of storage and stay out of storage now.
And then there is poor Toshiba. They're the third biggest hard drive vendor in a market with two important ones. They've got the fifth best SSD controller in a market with four good ones. They got Super Talent to use their SSD controller, and then Super Talent apparently disappeared from the SSD market. The got Apple to use their SSD controller, and then Apple switched to Samsung. They got Kingston to use their SSD controller, but Kingston's SSD naming scheme is such an incomprehensible mess that it's inadviseable to buy any SSDs from them, for fear of ending up with a JMIcron piece of junk. I'm not sure what Toshiba's plan is, but I wouldn't be surprised if they want to sell off their hard drive unit for the same reasons as Samsung and Hitachi.
Going into NAND flash production probably isn't a serious option for Western Digital or Seagate. Creating their own SSD controllers might be. Surely they have extensive experience in creating hard drive controllers and firmware, and it's plausible that that could help with creating SSD controllers and firmware. Or it might not.
Nice post Quizzical. Here is my reply. (thanks for your honesty)
Do you know the cost of each 1TB HDD when bought in bulk of 1000 ordered directly from the manufacturer as part of an affilitate program is around $20 per drive? Its around $30 for 2GB HDDs. The big guns of the industry rely on these deals. WD has special deals for their 10K RPM HDDs and Velociraptors too that are quite cool when bought in bulk...and as long as they hold the industry wont side with SSDs.
There is something special in my book that is happening too....
The size of video games are really being normalized. There is a limit to really how big a game can be based on the time and energy required to program them. I use SSDs for OSes, Enterprise Class HDD for Internal Storage and everything else for external storage. So a HDD has its place....
Remember, I know the two of you hate it when I mention ramdisks...an old screenshot comparing a 75% Capacity vertex II OCZ Drive vs a 4GB at 0% capacity Ramdisk all under stock speeds. (From my SLI system)
The sad thing about SSD performance is that not even today's SSDs beat out DDR-1 Memory in performance.
For the last 8 months I've ran my laptop under an OS which boots completely from RAM and partitions a 1GB Ramdisk. If I want to save anything permamently I use a flash drive. Its preconfigured, There is no HDD or SSD inside and I get more performance there than what SSDs give me.
By the time a 1TB SSD exists that has an excellent controller, is mainstream and can be bought for $100 - $200, I will probably have 64 - 256GB RAM in my computer, enough to simply RAMDISK every major program I throw into it. The size of a game hasn't been increasing as much as System Memory and Drive Space has.
HEY!!! What do you guys think of Hybrid Drives?
Think they can work?
Sure, system memory is vastly faster than solid state drives. The problem is that system memory is far more expensive in $/GB and cannot act as permanent system storage. As you know, when you lose power, any data on a hard drive or SSD is fine, while any data in SDRAM is gone in a second or so, whether it's system memory, video memory, or whatever.
The real question is how much speed you need for a given task. The various processor caches, system memory, and video memory all need tens of GB/s in bandwidth to perform their function properly. They also all need latencies in the tens of nanoseconds at worst, with L1 cache needing latencies around 1 ns in order to serve its function.
Long-term storage doesn't need that kind of bandwidth. I'm of the view that 100 MB/s is fast enough for main system storage. Sure, 1 GB/s is faster, and likely noticeably faster, but not dramatically so. 10 GB/s for system storage would likewise be faster than 1 GB/s in benchmarks, but for ordinary desktop use, the difference would be imperceptible. Ramdrives have their uses, but trying to use them in place of main system storage isn't one of them.
Hard drives can often go over 100 MB/s in sequential transfers, and that's fast enough. The problem is that for random transfers, they can't even do 1 MB/s. And that's way too slow. Many realistic workloads involve loading enough small files that the hard drive will only be able to deliver a few MB/s, and that's not enough. Good SSDs can offer tens of MB/s under artificially odd workloads designed to reduce performance, and over 100 MB/s in nearly any realistic workload. That's fast enough.
Hard Drives and SSDs will load into memory, unfortunately those files that are unusued will be erased from memory , forcing you to reload them. In RAMDISKS, the files are part of physical memory and mapped there. In short they NEVER are deleted so slowdown is kept at a minimum.
HDDs today still are very useful...Data Storage is still an important thing and most people with their wallets will disagree with you. Why buy a $100 - $200 SSD that sports 60 - 65GB when you can buy TBs for $100 - $200 is what most would think...Specially with for $200 I can buy also 16GB of RAM, Ramdisk 8GB and put most major games through it and have greater performance than even the strongest SSD in the world today.
What I can truly be sure of is that while controllers will get better, there never will be a generation where an SSD is stronger than the current memory in a computer. Computer Memory Sizes will increase you know and eventually a point will come when SSDs while they will overtake the HDD market will still not outperform pure memory itself. To do so means asking the SSD to double as System Memory if possible...
In doing such things....all SSDs are doing is making System Memory Stronger by pushing data into System Memory a lot faster than HDDs. Which is not a bad thing...In the very end of it....It will save more time to one who is smart. Currently it takes me 30 seconds to load all of Crysis to an Ramdisk and around 10 seconds to launch a game, choose a file and play on a map.
Something cute I actually do on Civ V do is map the heavy part of the game to a ramdisk and also MAP saved games to the ramdisk, I copy the save game I want to run and then I load the game much faster
In fact it takes me 20 seconds to copy that data to the Ramdisk where normally it would take me 30s - 90s to load a game, skip the intro video and FINALLY....LOAD THE GAME ITSELF through a file...and it normally takes me around 10 - 15 seconds to actually get into a map and start playing. Turn Ends are much smoother too 
note: the large time is for HUGE MAPS over 300 turns in late game.