Coding Horror: The State of Solid State Hard Drives
codinghorror.com
codinghorror.com
"The new Indilinx controller models, such as this Crucial 128 GB SSD, are just as fast as the X25-M. And, best of all, they're cheaper, while also offering a not-insubstantial bump to 128 GB of storage!"
This kind of hand waving irritates me to no end. There are a ton of different ways of measuring i/o performance, and I'm sure the author knows this. To say that the Indilinx based SSD drives are "just as fast" as the Intel SSDs is simply not true. Let's take my favorite way of measuring drive performance, 4KB random write speed.
Intel: 5923 IOPS (23.1 MB/s)
Indilinx: 1275 IOPS (6.47 MB/s)
Source: http://www.anandtech.com/storage/showdoc.aspx?i=3535&p=3
Granted this article is a few months old, perhaps the firmware for the Indilinx controller has gotten faster, or perhaps the Crucial 128GB drive uses a different Indilinx controller than the 'barefoot' model profiled in the article, but I doubt it.
If so, the Crucial 128GB SSD doesn't look half bad versus the Intel SSD. If this isn't the tests, the rest of the tests in that particular review have the Crucial doing really good.
It's an amazing and surprising difference, one that reminds me of my first dual-processor machine. It's not that your system feels faster when doing one thing (although it does), but that no matter how much you throw at it, it just doesn't slow down - a tremendous difference from the chug I experienced on the 250GB drive that was previously installed.
Easily the best $450 I've spent on an upgrade in a long, long time - far more effective than replacing my MBP would'be been.
I can't remember the last time I had to wait on a file to do the whole moving/copying/deleting dialog.
When I bought the G2, it was reasonably competitive with the other drives - I think the pricing on competitive drives has come down quite a bit since then, but the G2 is fast and stable - I have no complaints.
On that note, the OCZ forums are awesome, that company deserves heaps of praise for it's communication with the community. Developers are constantly releasing beta firmware and pre-release (fresh builds) there.
I've put the same disk in my MacBook Pro. Strongly recommended.
Some of my computers are cpu-bound on disk operations now.
If you run an engineering team, make the investment.
I'm moving to a workflow where I import to the SSD, do my work, then immediately move the raws to an external drive to save space.
How do you ask Disk Utility to call TRIM?
You can't.
Does writing 0's over every free block actually have the same effect?
No. (We might like it to, but apparently SSDs don't actually look at the data they're storing and thus it would cost more for them to detect zeros.)
Executive summary: it doesn't have to be an Intel drive, just stay away from drives based on those old JMicron chipsets.
drool
A couple months ago I bought a new MacBook Pro and picked up the 256GB SSD option on it. The drive's Toshiba (Apple ships Toshiba and Samsung drives -- I don't know model numbers).
So far it's been wonderful. Everything in this machine (processor, RAM, etc.) is a significant bump up from what I'd had before, but the SSD performance boost is still easy to notice; even though it's not the fastest drive you can get, the impression I have from watching the various videos of people with Intel drives is that the difference -- for the sorts of things I do -- is basically negligible (in order to notice it I'd have to be regularly writing huge files and timing the process).
Typical time from turning the machine on to having a desktop in front of me is around 15 seconds; that goes up if I have lots of peripherals plugged in (as I often do at work, where I end up using every available port), but still makes my older laptop painful even to contemplate.
Of course, both boot time and cold launch of applications are really just for show (though it's all very fast). The real point, for me, is the sorts of mundane but read-intensive things I do all the time: grepping over big log files or large codebases, getting status and diffs out of local checkouts, etc. There's just no contest here; that stuff happens stupid fast, and by itself would make the SSD worth having.
It should be noted that I haven't really paid attention to or particularly cared about write performance; the only time it would be noticeable is when pulling lots of stuff from a remote repo (lots of small writes to a bunch of files), but any write-performance problems with that get drowned out by things like network latency.
I'm also not particularly worried about wearing the drive out or anything like that; Toshiba's estimates for the lifetime of their drives are almost certainly a bit optimistic, but a more conservative and realistic number (confirmed by various places where I've seen it discussed) is that I'd need to sustain a pattern of writing around 20GB/day to have a significant chance of wearing out the drive in five years.
Go to newegg and sort the reviews by date on some of the most popular ones.
However, the next drive I bought was an 80gb Intel X25m. The speed difference of these drives are truly amazing
Buy yeah, I now back up my drive to an external disk.
Oh, yeah, BTW, you do realize that it's not a question of whether your magnetic drive will die but when, right? No fair comparing SSDs to a hypothetical magnetic drive that will "never" die. If you're actually worried about data integrity, I'd trust a well-manufactured SSD over a magnetic drive any day, though whether or not the current crop qualify as "well-manufactured" is something that is hard to determine. (I have no data either way.)
Why would you buy an SSD only to do large sequential transfers with it? Just use a 1.5TB HDD to store your media files.
However, if you need a large hard drive for a database and not as a media storage, obviously random access is just as important as sequential access.
Personally, I can only watch one TV show at a time...
Online editing is still quite CPU intensive so I don't think that SSDs alone will change that workflow but it might still help somewhat at offline editing.
Visual FX, video transcoding etc. Rendering times can often be disk-bound, and eat up a lot of time. It becomes aggravating once you start editing HD video, even on a quad-core. For that, the Crucial drive mentioned by Jeff is a win because of the price - the lower random-access performance compared to the Intel drive is not that big of an issue for this application, and even with that limitation it'll smoke my existing system drive too.
Sure wish Apple would ship the Intel drives. Perhaps they can't get a low enough price on them.
Or, alternately, that Samsung would get with the program and do the kind of controller work required to make these things go faster. (Whatever Intel's doing--see the anandtech.com massive article.)
Or, perhaps both Apple and Samsung could get together and implement the critical TRIM command (see http://www.anandtech.com/storage/showdoc.aspx?i=3631&p=8).
They need to let the market decide.
Why not make one of everything like Dell and let the market decide?
Offering the _best_ of everything is. So shipping a crappy SSD is dumb. (Personally I strongly suspect it's because they can get closer to that 100% hardware markup they love on a lower-end model, than because it's greener. But I'm sure "green" sounds better.)
But seriously, when was the last time an Apple computer has the fastest CPU? Or the largest laptop screen. Or anything with a card reader. Or blue ray... Just pointing out that they are entirely willing to avoid the absolute top of the line and have done so consistently.
Maybe it's a margin thing as you say, I don't know...
In what sense is the SSD you haven't paid for that they do not offer, 'yours'?
You make a false assumption that halogen-free directives are about being green. It's as much about safety as anything else:
For some applications there is a concern that halogenated plastic materials will release corrosive and toxic gases if ignited in a fire. The corrosive element of these gases has the potential to damage electronics wherever the smoke travels, and the toxic element can be potentially hazardous to persons if they cannot easily evacuate from the area.
Applications where corrosion potential is of particular concern include communications data centers and phone switching stations with large amounts of expensive electronics. As a case in point the fire called “the worst disaster in telecommunications history” occurred on May 8th 1988 in a Hinsdale, IL Central Office facility for Illinois Bell. The fire cut local service to 35,000 suburban Chicago phone customers and was estimated to have cost many millions of dollars in recovery cost. In the after effects of the fire the central processor of the phone switch, which was not directly involved in the fire, had to be replaced reportedly due to the effects of acid corrosion.
The toxicity of smoke is of highest concern within enclosed spaces where means of escape by persons are limited; examples include mass transit rail cars, ships, and Offshore Oil and Gas platforms. There have been several tragic fires that have occurred involving transit vehicles over recent years that have brought heightened concern to this issue. In a February 2003 incident, news reports of a fire involving two trains in a Daegu, South Korea subway station told of how toxic fumes and heavy black smoke prevented firefighters from making a quick rescue of those trapped. 120 were killed and many others injured in that incident.
http://www.gocsc.com/uploads/white_papers/E47B142616A84C3F90...
For accurate price info you should know that the second generation 160 GB X-25Ms originally listed for $450 but shot up in price due to excessive demand. It's only in stock at stores that price gouge and MIA everywhere else. Amazon, not price-gouging, has the 80 GB for $260 and the kit version of the 160 GB for $500.
Is this the market-economics version of political correctness?
I did say that. The issue is that I need 2 names: a name for the stores charging $450 (and are out of stock) and another name for the stores charging $600+.
I chose "gouging" and "non-gouging". Concise, and everyone who read it understood what I meant to say. I'm not seeing any real alternatives being offered, except to replace clear, concise, universally understood words with tedious explanations.
Why the nice name for a supplier who doesn't, you know, supply?
Discount camera stores are infamous for advertising things at great prices to get you in the store but being out of stock. It's a crappy way to do biz and should have a crappy name; "non-gouging" doesn't qualify.
Again with the criticize-without-offering-alternative.
Discount camera stores are infamous for advertising things at great prices to get you in the store but being out of stock. It's a crappy way to do biz and should have a crappy name; "non-gouging" doesn't qualify.
Right, that's called bait and switch. No relation to price gouging.
Actually, I didn't describe bait and switch. And, even if I did, the question isn't whether that bad behavior has any relationship to price-gouging, the question is whether suppliers who engage in crappy behavior should be given good labels.
Suppliers who advertise at prices and don't come through are crappy, regardless of what they do after not coming through. Calling them "non-gouging" is inappropriate.
Others are replying, but I don't know that there's any idiom needed. Raising the price to whatever people willingly pay is just being a merchant.
Is this the market-economics version of political correctness?
I don't think it is. Political correctness applies when people feel they can't plainly say something that's true. In this case, using "price gouging" adds nothing to the statement that prices are high except a sense of outrage, and my response was merely that I think the outrage is misplaced.
Of course it does. It says:
1. the subject is a retail product with a standard price, as opposed to a market product where the price is determined by bids and asks.
2. the price is temporarily higher than the standard price due to some sort of unforeseen disruption.
3. the price will return to the standard price when circumstances return to normal.
In the present example the product that Jeff says costs $600 will cost $450 within 3 months. Just as it cost $450 for a brief moment 1.5 months ago.
Edit: Also, you are speaking for yourself about the outrage. I certainly don't have any outrage. Not does "price gouging" give me any sense of outrage.
The people who are outraged by the practice will be outraged by it no matter what you call it. And people, like myself, who have no problem with it will have no problem with it no matter what you call it.
> Price gouging is a term for a seller pricing much higher than is considered reasonable or fair.
or
> pricing above the market price when no alternative retailer is available
That's my point, actually.
Price gouging is a term for a seller pricing much higher than is considered reasonable or fair.
I'm only interested in the non-value-judgement related meaning. "Reasonable" and "fair" are meaningless.
What other people are saying is "you can't use that word because it's loaded with value judgement." I'm saying "OK, give me an alternative that captures the meaning without the value judgement. And is concise, clear, expressive, and widely understood."
pricing above the market price when no alternative retailer is available
This, also, is meaningless.
Scarcity may be the term you're looking for.
So by using the term, you've added confusion, including the need for your extra verbiage, to the effect of "...not that there's anything wrong with it!"
(I also think you're giving way too much weight to MSRP. Actual prices to the consumer always vary, even without a bid/ask 'market'. The MSRP -- as a coordination mechanism among producers and retailers used to loosely enforce pricing discipline -- is arguably a more-unfair practice than this market-clearing 'gouging'. So you're using a mildly-disapproving term, 'gouging', for the praiseworthy practice of maintaining a sellable inventory, and an mildly-approving term, 'standard', for a potentially-unfair collusive practice. The connotations of your chosen language are backwards!)
See also Obama's elementary explanation of why lowering gas taxes doesn't lower gas prices when gas is in limited supply: the market clearing price is still the price at which the amount of gas that people want at that price equals the supply of gas. Setting the price lower than this will result in shortages. It doesn't matter whether gas taxes go down.
"Price-gouging" really is a terrible phrase for what you do in a supply-bound market. The people who raised prices have drives if you want one, and if nobody wants one, they lose their bet - have unsold inventory. The people who set lower prices sold out their inventory, but now they don't have drives if you want one, which isn't exactly a favor to you either. There are four methods of allocating resources and the method that our hunter-gatherer brains don't understand is market prices, so people think it is "unfair".
Those are definitely not the phrases I'm looking for. They don't contain enough information. I hope I can unwrap this without being too confusing.
First, you're preaching to the choir here. I'm pro-market-prices. I don't think there's anything unfair about the practice of price gouging. I support it 100%.
The issue here is you, and the others, aren't playing Gordon Gekko: "Gouging is good". You're playing Margaret "society doesn't exist" Thatcher: "When supply is constrained then prices go up. That's just the market at work. There isn't even such a thing a price gouging. Stop hallucinating."
Do I really need to spell out why and how it's different than a market-fluctuation? I got into it a little bit in a neighboring message, but I didn't go into the manufacturer/retailer dynamic. I think just the fact that it makes so many people irrationally angry is evidence enough that there's something unique about it.
Thought experiment: if gas prices doubled overnight would anyone get upset at gas station owners? Exactly. That's why it's not price gouging.
All I'm doing is defending the concept's existence and saying that because it exists I need a way to refer to it. If "gouging" is too pejorative then give me another word that actually captures all the relevant non-value-judgement-related meaning and is not just a sleight of hand attempt to pretend there's no there there. Because there is.
Something's not price gouging when people get upset? Or maybe you weren't paying any attention to sentiment when gas prices were soaring a few years back? People hated gas station owners, and there certainly were accusations of "gouging".
That was a bad example. But those accusations were obviously made by idiots.
My point was that if the manufacturer doubles the price no one gets mad at the retailer. Except idiots.
For those of us not as current on SSDs as you are, actually correcting his mistakes might be useful.
The only problem is that it started reporting S.M.A.R.T. errors.
See (http://blog.gtuhl.com/2009/03/26/ocz-vertex-ssd-in-a-17-sant...) for install and before/after xbench runs.
http://www.ocztechnologyforum.com/forum/showthread.php?t=634...
For example, intense BitTorrent downloading (and uploading) on a RAM disk frees my hard disk from the constant reads and writes so I can have high speed downloads (and verification of d/led data) and still do disk intensive activities in other applications.
I can't wait for the day where my mass storage can have that kind of speed! (and it looks like it's closer than ever now)
Can you go into a bit of detail? How do you do that?
As long as you have atimes disabled, the only thing RAM disks will speed up is writes -- recent FS operations are cached in memory automatically (and never paged to disk), so reading a 'warm' file doesn't hit the disk at all.
If you really need such a thing, tmpfs is vastly superior -- it operates directly at the VM page level, instead of idiotically mallocing a ton of address space up front (though I don't think anything like it is available for OS X).
The speed benefits look enormous but I don't want to have to go through the trouble of replacing this in a year or two.
I have those dreaded pauses when my VMs(typically 2 at a time) are busy doing something and I decide to untar a file for example. Does anyone have a similar problem?
I still run on a 32bit environment so use only 3GB of ram.
edit: DirectCanada has disks cheaper than NCIX by almost 2¢/gb as well.
In comparison the average price for an SSD is CAD$375 for ~120gb, so let's call it $400 to make things easy, it's going to cost more with tax anyway. So it's around $3/gb or $3.50/gb which is 35x the cost.
However if you are building a new box and spending $1500+ anyway then an SSD is probably a very wise choice.
Call me a young whippersnapper if you like, but I am not paying anywhere near $1000 for one now. Same idea with SSDs.
When the G2s were released, I put a G2 in my desktop and the G1 in my laptop. Having a SSD in my laptop is great, but the improvement from the SSD in my desktop is actually much more noticeable.
It'll be the next upgrade I do to my laptop when it starts feeling a bit sluggish.