The SSD Relapse: Understanding and Choosing the Best SSD | Hacker News Reader

The SSD Relapse: Understanding and Choosing the Best SSD

anandtech.com

51 pointsblasdel20 comments

I've been trying to get an 80GB Intel X25-M G2 for a few days. They are going for well over double MSRP ($600 versus $230) at several retailers (e.g., http://www.newegg.com/Product/Product.aspx?Item=N82E16820167...)

Newegg has an autopricing algorithm that ramps the price up into the stratosphere if a lot of people are buying something. It takes a while for the price to settle back down to 'reasonable'.

Either find it elsewhere, or wait it out... Amazon has had good prices on hard drives recently. The magnetic kind; I haven't been paying attention to SSDs.

They are pretty surprisingly hard to get. At the beginning of the summer I tried to buy, I think, 20 drives, to upgrade all the developers to SSDs and I remember being surprised that I cleared out Amazon AND NewEgg completely to get the drives I needed.

Yeah, and those were the G1s.

After the whole bad firmware thing, the G2s are even worse. Amazon doesn't have them in stock at all, they are $300 over MSRP at Newegg, and I haven't seen them for less than about 150% MSRP anywhere.

According to his benchmarks, compilation of C code (Pidgin at least) doesn't seem to benefit from SSD, even the 5400RPM drive performs the same.

SSD for developers is usually waste of money from technical POV, as the working set usually fits in the fs page cache.

30" IPS panels, OTOH, would be money well spent.

Developers often do more than simply compile code. For example if I could cut the start up time of the app I'm developing in half that would be really quite nice indeed, even if compile times stay the same.

I don't know about that, it depends on the build tree I guess. Our IDE build tree (which includes compiler, RTL and lots of other stuff) writes about 600M of intermediate files, 1G of executables / debugger symbol files / map files etc., and reads probably several G worth of source. A rebuild (clean and build) currently runs in about 13 minutes on my 10K drive, but I'm fairly sure it could be cut down with faster I/O.

Actually Spolsky had already found that to be true for their compile times back in March. But decided the benefit was still worth it for his developers. See http://www.joelonsoftware.com/items/2009/03/27.html

As I recall, this was because Intel had a manufacturing problem and stopped making the X25-Ms until only about a month ago.

My understanding is that it was simply a firmware problem. Reflashing the buggy drives fixed the issue.

I was looking for this same drive today and the suppliers in Australia are under the impression that Intel will ship them end of next week (for a pretty reasonable 400 AUD - approx $300)

http://www.solidstatecentral.com.au/store/index.php?main_pag...

I put an order in with Toshiba Direct for one since they had non-insane pricing.

http://www.toshibadirect.com/td/b2c/adet.to?poid=450025

We'll see how long it takes to ship.

One minor quibble: as far as I can tell, Apple is actually mostly shipping drives manufactured by Toshiba, which don't seem to be rebrands of Samsung's tech. The way to tell is to look in System Profiler; Toshiba drives will be labeled "Apple SSD TS128" or "Apple SSD TS256", while Samsung will be "Apple SSD SM256" (they only seem to show up in the 256GB option, and even most of those are Toshiba).

Anyway. I recently picked up a new MacBook Pro and went for the (256GB, ended up with Toshiba) SSD option, and I've got to say I don't ever plan to go back to a mechanical drive if I have the choice. It's not just the flashy-but-pointless things like boot time or first-time application launching. It's the fact that most of my life consists of reading, paging and grepping through lots of files, and that is stupid fast with the SSD. I don't worry too much about write performance, because typically I'm in an incredibly read-heavy workflow.

After several hours of researching SSD options, I recently bought a Corsair P128 from Amazon. Firmware version was VBM18C1Q (which supports background garbage collection as the article points out).

Price is currently identical on both Amazon and NewEgg ($330.99), but NewEgg has tons of user reviews. Affiliate-free links:

http://www.newegg.com/Product/ProductReview.aspx?Item=N82E16...

http://www.amazon.com/exec/obidos/ASIN/B002CI41US/

I can't wait for the new X25-M. Recently threw a 160GB X25-M (first gen) in a datamining database server to replace a 6x15k SCSI RAID 10 array, and it was 2.5x faster. One disk faster than six.

Interesting! Can you elaborate a little bit on the usage pattern and how you measured the speedup, please?

It was a very simple, unscientific test: ran a MySQL query with the old drives, then ran it on the new drive. Old drives took 5 minutes, new drive took less than 2 minutes.

I'd add at least another disk and build a raid1-array. With a single-disk setup you miss the greatest feature of RAID, redundancy.

"There’s a Mac Edition of the Vertex, unfortunately it’s no different than the regular drive - it just has a different sticker on it and a higher pricetag."

As a longtime mac user, this kind of thing drives me crazy.

Also the article talks about Windows XP and Windows 7 but doesn't mention anything about TRIM support in Mac OS X. Is this something that needs to be enabled at the kernel level? Does anyone know the status of this command with the Snow Leopard kernel?

IIRC older kernels blocked any unknown ATA commands, including TRIM. Newer kernels no longer block TRIM, so someone is working on a third-party utility.

http://www.ocztechnologyforum.com/forum/showthread.php?p=409...

Reply on news.ycombinator.com