I can wholeheartedly echo the sentiments in this article. My first experience was likewise a first-gen SSD - one of the old Asus Eee 901 netbooks, which had a dual SLC/MLC setup. The SLC was just barely fast enough to tolerate an OS on, and the 16GB was frightfully slow. About a year ago I replaced it with a RunCore (a major advance, and yet crushed by current SSD’s). From that moment on, my little netbook was faster in most tasks than my MacBook Pro. It boots faster, launches apps faster, and is FAR more responsive. And of course, I can slam the lid shut and toss it across the room to the couch, as there are no moving parts to worry about.
Garthy did bring up an excellent point, which is that random I/O is infinitely more important than raw throughput, and you never get those statistics outside of some of the better reveiws. Fortunately, all of the high-end SSD’s (Intel X-25, Crucial, OCZ Vertex, and likely others) perform extremely well in such scenarios.
SSDs show great promise, but I’d still say buyer beware. Only a few months ago we bought several SSDs for the office in which I work. But when we upgraded to Vista64, things went downhill fast. After having to re-install the boot volume on two production machines, we yanked all of them and went back to regular hard drives (which, in my opinion, weren’t terribly slow in the first place, given that I only run a few applications all day long).
When SSDs work, they’re amazing. But they’re still not as “slap it in and forget about it” reliable as the technology they’re replacing yet. Be sure to do your homework: check the manufacturer’s information, but also google for specific instances of people using the drive with your computer configuration, check Amazon and/or Newegg to see customer feedback on the drive, etc., etc.
If your CPU is typically maxed-out yet your other components are not, then your CPU is the weakest link and should be upgraded.
On the other hand, if you keep paging to disk because you don’t have enough RAM, while your CPU utilization doesn’t rise above 30%, then upgrading to a faster CPU is going to do stuff-all for performance.
In tough economic times, it’s hard to find room in the corporate budget for new systems. But end-users still complain that systems are slow and want a newer, faster PC. For the mere $350 cost of an SSD, the performance difference is like getting a new computer.
Instead of comparing how expensive the SSD is to the HDD, compare the cost of an SSD to replacing the whole system. In this light, the relatively low cost of SSD is a great way to extend the usefulness of old hardware. The downside is that you are sinking that cost into a computer that may now have 1-2 years of life left, where a new PC may give you 3-5 years. But the low cost and performance gains now keeps both the bean counters and the end-users happy.
The OCZ Vertex has been fantastic in my experience, but I’ve found that almost any recent SSD will give you better performance than an HDD.
Something to watch out for with SSD’s is their performance as the disk fills up. Almost all SSD’s will perform great out of the box, but mainly will significantly drop in performance with time. So bad that they fail to even compare to normal slow HD’s.
This has been fixed in the 2nd Gen X25-m’s, but I don’t know about the other drives. The following describes the issue: http://www.anandtech.com/storage/showdoc.aspx?i=3531&p=8 For sure Intel has fixed it, but I don’t know about the other manufacturers.
I bought one
13" MacBook (lowest CPU I could get: 2.26GHz - I agree with Jeff and others that you don’t need top range CPU power for dev/everyday use)
4GB RAM
128GB SSD.
Here’s the info from the “About this Mac” link on the SSD:
Capacity: 121.33 GB (121,332,826,112 bytes)
Model: APPLE SSD TS128A
Revision: AGAB0202
Serial Number: 799S1021T0RZ
Native Command Queuing: No
Removable Media: No
Detachable Drive: No
BSD Name: disk0
Partition Map Type: GPT (GUID Partition Table)
S.M.A.R.T. status: Verified
I can honestly say it’s the best ever purchase I made. The laptop boots up in 7 seconds.
There are quite a few performance reviews on the net, but the read performance is greatly improved.
The last time I checked up on SSD lifetime it wasn’t even a problem anymore. I did the math and I think the drive came out to last more than 50 years. Can anyone help me out with some documentation for this?
For constant churning systems (e.g. databases that are constantly optimized and rebuilt), MLCs still present a very real reliability issue. The whole existence of the X25-E is based upon it having a much greater estimated lifespan relative to the X25-M.
I may have 8 cores in my MacPro, but they don’t twiddle their thumbs for very long - if Boinc detects more than 3 minutes of inactivity it pegs those cores working on medical research project for the World Community Grid.
I don’t have a top of the line vid card or fast hard drive, but I do have a lot of memory (10 GB). Just right for writing code and testing/debugging it against multiple platforms at the same time.
I upgraded my primary Dev box and within a week my home PC and laptop also got SSD drives (I went with the OCZ Vertex).
I would say SSD drives are quickly replacing RAM as the “Best upgrade for your money”. 5-10 years ago if you or someone had a slow computer the best way to speed it up was to add more ram. After upgrading my drives it was very clear the memory is no longer the bottleneck and disk IO can drastically change your computing experience.
I was expecting more. But as some has point out. Seq Read/ Write dont matter. It is Random R/W that counts. And Intel beat every single current SATA products on the market. That is why Linux loves it, and state in the 2nd follow up.
Indilix is good, best value / performance. But it simply does not perform better then Intel. Because general usage pattern includes Random R/W and Seq. Intel will make up more then enough during the Random Section. So to say Indilix perform better then Intel is completely false. ( Just like Most reviews )
And to point out Fusion-IO no longer matters. I think. Because Intel manage to show off an 4GB/s Random R/W SSD prototype. Compare to current 20 - 40MB/s !!! While the thing is huge, Intel has yet to use double side of the PCI-E Card, which means we can expect half the size shown in IDF.
Intel will have True Next Gen SSD out in Q210. I think we should expect SATA 3.0 support. There are samples already shown new Controller manage to get 700MB/s in Seq. So Seq performance are nearly done on Home PC computers. Which is left is some major improvement from Random RW, then we just wait for some more price drop…
I’m looking forward to the day where it becomes feasible to install large SSD’s for storage as well. For someone like me who is running a system used for gaming and other general PC tasks, I will soon want to install much more than 128GB of apps or games that I’d prefer to run from the SSD.
I think it would kind of get annoying to eventually need to start installing apps on the slower drive and then have to pick & choose what goes where.
Thanks Dennis, I was speaking for the average consumer (desktop user), but you’re correct there are some applications where the SSD is not yet fit for the job. Also I believe I was referencing the MTBF on drives which gives figures near 140 years.
I have heard articles trying to discredit MTBF as a reliable figure for SSD’s but I’m not sure I understand their reasoning before.
@Mr Atwood
Would you happen to have figures for random reads and random write times? As stated earlier by some other commenter, for the desktop user these are practically the only figures that matter.
Another general comment, I’m glad some newer quality controllers are coming out. Competition will help get these to mass market faster.
Your comparison on throughput was incomplete. For the price of this unit, a three-drive RAID-0 of 1TB drives would be cost competitive - and provide higher throughput. And provide 3TB of storage.
SSD are still not competitive for price/performance.
Yes, exactly, you have to take workload into consideration. In the first sentence of this post Jeff complains about pairing a fast CPU with a “generic” video card. Not necessarily a bad thing depending on workload. In your earlier comment you asked the loaded question about whether it’s worthwhile to upgrade your “already-insanely-fast” CPU, which, again, depending on your workload may be completely worthwhile.
All I’m saying is that I’m tired of “Newegg-weakest-linkers” like (apparently) you and Jeff, who will look at a system, compare it with what’s on the market now, and suggest upgrades accordingly, without taking into consideration how a person uses his computer or how much actual impact the upgrade will have. “Oh, your video card is 3 generations behind, you should totally get a new one” even though the person never plays 3-D games. That sort of nonsense.
