Thomas Ryan must be smiling. We have arrived at the “Adolescence of P-1”
I really, really hope everybody who is excited (and Jeff, although he’s commited to the cheap solution) reads the Anandtech articles being linked to here BEFORE they buy. Get educated from people testing this stuff in real-world desktop/laptop usage before getting excited by the folks who highlight only the raw read speed (not Jeff.)
So does anyone have any experience on hybrid systems with SSD and HDD? It would be great to have an SSD for everyday things and OS, and another 1TB HDD for media and sort, but does performance gets hit by this?
I’m not sure where anyone’s advocated blindly upgrading without considering what effect it will have.
For the typical desktop user, the hard drive is likely to be the bottleneck while starting applications and switching between them. Thus it makes sense to recommend a faster/lower-latency storage device as something to consider upgrading to.
I can agree that telling someone to upgrade their graphics card without considering what they use the machine for is a bit silly, though.
I also bought the 128GB Crucial M225 about 2 weeks ago (UK£250 - but price went up the next day). Its installed in an Intel P8400 powered 3GB laptop with Windows 7, and its performance with a HD WEI of 7.3 changes everything I do. This machine now replaces my noisy overclocked desktop - the only thing I hear now (in my living room) is my monitor buzzing.
As instructed by various threads, I’ve disabled all the pre-fetch/windows-search stuff - most done automatically by Win7. Side-benefit: switching this stuff off also releases lots of RAM that was previously used by Win7.
Now all we need is the (promised) firmware upgrade to take advantage of Win7 TRIM.
@anon A three-drive RAID-0 has about 1/3 the latency of a single drive (three parallel I/O), providing a dramatic improvement over a single drive for random I/O. It’s not as great as an SSD, but it will make you very happy in terms of application load times and file read/write - typical desktop real world use.
The only place where an SSD really provides benefit is in a mobile platform - laptop, netbook - where the price premium is buying a small footprint and lower power consumption. If you have a desktop, SSDs are a spectacular waste of money.
I’ve been looking sideways at these SSDs for a while, too… The big pros are, of course, the speed and the no moving parts… so they’re great for using on the bus. But the biggest con that I’m aware of is the limited number of read/writes.
The nature of the beast is that because it’s solid state, you can only alter that bit in the same place so many times. Until they figure out a way around that, I’ll still be a bit leery. At the most, I’ll have a combination of SSD and mechanical HDs. Boot from the SSD, and store data on the tried and true HD.
Ok - the amount of posts about how the limited number reads and writes of these drives being the limiting factor for them really bothers me. A hard disk drive (what is currently mainstream) has a finite life as well. This average life could be expressed in reads and writes as well. The fact is that most people say well a drive usually lasts me about “x” years. In my experience x, is typically 5-10 regarding HDDs, 10 being very good.
As I stated in an earlier post, the lifetime of an SSD will greatly exceed the life of a HDD. The MTBF of an SSD is 140 years. Even if MTBF isn’t a perfect unit of measure, I think you can quit worrying about the “finite” life of SSD’s - you’ll move on to more advanced technology long before the drive stops working.
Are you aware that I could measure your life having a finite number of writes? You know, programmers typically die after 10 gazabillion keystrokes. I bet you express a person’s average life in years - not keystrokes, yet the MTBF of an SSD is greater than the average persons life. So stop saying “I don’t believe in the programmers race as a reliable one because it has a finite number of keystrokes.”
Nice article Jeff, and very interesting comments from Dennis Forbes too.
Very tempted to try one of these in my Vostro 1700 (C#/Windows forms developer machine).
As I stated in an earlier post, the lifetime of an SSD will greatly exceed the life of a HDD.
For an average user, absolutely. Solid state wins virtually every time.
As mentioned earlier, though, there is a reason the X25-E line exists based upon the lower-density, more expensive (due to the prior point), but more reliable SLC technology.
An MLC cell – the very best in the world – can handle about 10,000 rewrites before it is “used up”. Magnetic hard drives can rewrite the same area unfathomable numbers of times, which is good because many systems actually do rewrite and churn over the same data endlessly.
Of course SSDs are wise to this, and they constantly move data around to perform a sort of wear leveling, so you could write the same logical place billions of times and in not have a problem as the SSD shuttles it around. But MLCs still aren’t good for enterprise uses where GBs are written and rewritten and rechurcned and reordered and repruned and parsed, etc, daily.
So I agree with your core point for home users. And arguments that data is non-recoverable from the SSD are bunk, because for all intents everyone MUST rely upon the assumption that their hard drive is non-recoverable too, because not only is that often the case, but even where it is the costs are prohibitive.
I would sell that Crucial ASAP and buy an OCZ Agility or Vertex. OCZ just released new FW that supports Trim or GC (you choose via FW)
OCZ has the best support for Indilinx drives and a small community has formed around these drives on their forums.
I think the aversion to SSD degredation is a largely psychological thing.
For a platter hard drive, we think in our heads, “Well, it might accidentally pooch itself sometime between now and 10 years from now.” But we don’t think of it as an inevitability, even though it is. We think of it as an accident that just as easily might not happen.
For an SSD it has an expiry date. We think of it as “Slowly killing itself”. Even if that expiry date is decades years in the future, we can’t escape the horror that we are using this thing up!
Being the irrational human beings we are, this makes a long-lasting, predictably degrading SSD seem better than a short-lasting, unpredictably self-pooching platter drive.
According to this article, a modern ssd can survive 3.4 terabytes a day for over 12 years:
http://www.bitmicro.com/press_resources_debunking.php
As for me I could still run three terabyte drives in RAID 0 and beat that performance for the same price, at least until I run out of SATA slots.
Basically, disabling the swap file achieves more or less the same goal on windows xp. Basically, windows xp swaps a lot. Needlessly most of the time. You can literally have 3GB ram free, alt tab away from Firefox and it will happily ‘conserve’ memory for you and waste your time twice: first spending time swapping to disk and second swapping back into main memory when you alt tab back. Disable the page file and this won’t happen again. Net result: after you’ve loaded your applications once, they will stay in memory until you close them. Alt tabs will be fast. Having more ram, will ensure that the remaining ram reduces disk access even further because it is used as a cache.
You might think, hmm I might run out of memory. And yes, this is true except that windows has a maximum per process memory size of 2GB and a system with 4GB should basically be able to satisfy all your needs unless they involve running two such processes (and good luck trying that with swapping turned on) or running a lot of processes using more than 4GB. In other words, on most modern 32 bit PCs with the memory maxed out to 4GB having a swap file is a waste of time and you are better off closing applications than having to wait for windows swapping them in and out of memory constantly.
Highly recommended, especially on slow laptop drives. The reason SSDs are so popular is because they make swapping less painful. Not swapping at all is even less painful.
Now macs and linux machines are a lot smarter about swapping and you might want to leave swapping on there. But even there, you might stop and wonder who’se needs are actually being served by the OS moving bits of memory to and from disk all the time when memory is essentially cheap and plentiful.
I would say SSD drives are quickly replacing RAM as the “Best upgrade for your money”.
Not quite yet, unless you’ve already got plenty memory. Sadly a lot of people are still buying machines with 1 or 2GB of memory which is almost unforgivable (unforgivable for the vendors, not the consumer).
But yeah, once you get to the point where you aren’t likely to fill your memory with stuff you are using ‘right now’, the SSD is going to be the biggest bang for the buck.
According to this article, a modern ssd can survive 3.4 terabytes a day for over 12 years:
A modern SLC, enterprise-quality drive, sure. A consumer MLC drive is a different beast.
Their “3.4 terabytes a day” would be rewriting the whole 160GB drive 21 times per day, or 7665 times a year. The 12 years mark is based upon about a 100,000 cycle life. With MLC you’re looking at a 1,000 - 10,000 cycle life (that’s why the drive costs 1/4 the price of a X25-E that is 1/2 the size).
So now your hypothetical 3.4TB a day has the drives life down to just over a year before it’s complete toast, presuming it’s a really good quality of MLC that reaches the max reliability, though of course some cells will fade faster than others taking out whole blocks.
Oh, but you only have a 64GB drive? Well then you’re down to 1/3rd of a year because the cell rewrite is higher. With a 32GB drive, your daily 3.4TB (Facebook logs?) has it exhausted in two months. If it’s a real-world MLC, halve each of these times.
I don’t think wear is a concern at all for normal consumers, and no one is writing 3.4TB daily to a consumer drive. Yet bogus stats like the bitmicro ones don’t do anyone any favors because it isn’t based in reality. It further waves its hands towards absurdity when it says that they didn’t even consider wear leveling, alluding that it would prove even better reliability, when wear leveling would do squat for that scenario.
The “weakest link” theory of computer performace is wrong and it pains me to hear it from people who are supposedly knowledgeable about computers.
Look at the benchmarks–faster memory makes at most a couple percent difference. A better motherboard makes a percent or two difference, if that. A better graphics card makes literally no difference unless you’re playing video games, which many of us don’t do.
People who realize this can build a computer with cheap-o parts that is, practically speaking, as fast as a computer with high end parts costing $300+ more.
I suppose the psychological motivation is “hey, I bought one high-end thing, I shouldn’t mess it up by connecting it to low-end things.” This doesn’t take empirical evidence or computer architecture into account.
That being said, I do love fast hard drives and SSDs. In fact I have a 1st-gen SSD in my Mac and it’s crazy fast. Most of the sputtering in these early drives is due to Windows’s lame disk cache flushing policy and the fact that the OS frequently blocks on I/O operations–two things OS X doesn’t suffer from.
You didn’t metion the Windows 7 Trim command and the other optimizations for SSD’s.
I’m waiting until the tech stabilizes a bit more. The SSD technology leans too far to the “write-once” model right now for my taste.