Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Review of Intel's New X25M SSD (same one Torvalds has) in a new MacBook Pro (paulstamatiou.com)
63 points by PStamatiou on Oct 31, 2008 | hide | past | favorite | 28 comments


The new controller on this drive from Intel is fantastic. I look forward to seeing this technology become ubiquitous. With filesystems designed for random-access persistent drives and good controllers, well... rotational disks are going to look extremely antiquated extremely quickly.


Yeah, from my reading on this drive the biggest gains were made in the controller, which is orders of magnitude more efficient at wear leveling and small-file performance.

For a really good read on why the Intel drive is way better than the competition: http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=34...


NCQ was first introduced in SATA disk drives to prioritize disk read/write operations based on the current location of the head, and therefore eliminate superfluous revolutions and head movement.

Actually, SCSI (and SAS) had it first, and still beats SATA's implementation in terms of queue length. (256 entries vs 32 I think)

EDIT: It's called Tagged Command Queueing (TCQ) on SCSI, and the 256 entries are a limitation by most SCSI disks/controllers, the system can handle way more.


An exciting effect is the possibility of new algorithms to take advantage of this different set of performance characteristics (and vindication of neglected, maligned crackpot schemes). There are avenues that intelligent people simply don't explore, because they wouldn't help solve the problems they face, given the fundamental nature of hardware.

For example "So disks are not random access any more?" asked of Jim Gray in the fascinating "A Conversation with Jim Gray" ACM, June 2003 http://www.acmqueue.org/modules.php?name=Content&pa=show...

But when fundamental nature of hardware changes...


SSDs are the future. This is where the next big computer upgrade is going to be. The result, I think is that when rotational parts get completely removed from computers, we will see desktop software programmed for instant reaction. Using SSDs, most programs could be rearranged in a way that almost every standard usage happens in less than 1 second.


I would imagine the blazing random reads and writes of this drive would drastically help database performance (and possibly allow operations too expensive in time or number of IOs to become more feasible). Remember you can always push the limit of what's possible technically because hardware becomes cheaper, faster and more reliable over time.


Database systems are already optimized for disk i/o that has blazing fast sequential reads and a high penalty for seeking. SSD's put the seeking penalty to zero, but with worse sequential read speed compared to your typical raided disks.

Which is the long way of saying, it's an expensive option that is unlikely (at least right now) to improve database performance that much.


Plenty of database workloads involve a lot of random I/O (e.g. anything that does an index lookup is typically doing random I/O). In theory, you should be able to tell the query optimizer about the relative costs of random vs. sequential I/O on your hardware, and have it adjust the costing of query plans accordingly (e.g. PG allows this, at any rate -- it might even be possible to infer that automatically, although not easy).

Aside from the query workload, WAL is designed under the assumption that sequential writes are much cheaper than random writes, so you could imagine changing that when using SSDs...


Good point, I didn't really think about that. Perhaps somebody could build a new kind of datastore that takes advantage of that?


They should sell a small, maybe 8GB version to use for swap or for installing a DB. What do you think?


They should sell a small, maybe 8GB version to use for swap

It's called RAM. If you want 8G of RAM, get 8G of RAM. It's cheap.


RAM is volatile, SSDs are not. Obviously not relevant in the case of swap, though.



Watch out; that's a different controller than the X25.


Before and after benchmarks of "unibody" MacBook 2GHz/4GB upgraded to an Intel X25-M 80GB SSD: http://tinyurl.com/5g2afd


While not technically hacker related, I figured it would be relevant since most hackers have MBPs and would like to hear about the benefits of the ~fastest SSD on the market.


This drive is faster and lower power than a typical hard drive. Assuming that purchase price is not a factor, what does this do for building out supercomputer clusters on commodity hardware? Could this change the way companies like GOOG and AMZN build their computing utilities?

p.s. And what part of an 80GB drive that consumes way less power than the drive it replaces and is way faster than the drive it replaces is not hacker-related? Techno-fetishism is always hacker-related.


SSDs in the data center are definitely coming up. CCP, the creators of Eve Online, use SSD drives for their unsharded persistent game world.

http://www.superssd.com/success/ccpgames.htm has a testimonial about their SSD experiences.


Didn't know that. Interesting!!


Now that's a HUGE assumption : "since most hackers have MBPs".


That's awesome, but:

How fast does it start Word 2008?


Okay I just tested it - to open Word and no document it takes 4 seconds. To open a 5 page document from being closed it takes also about 4 seconds.


I'm encrypting the SSD at the moment so I wont get accurate results if I do it now, but I'll do it later today and report back.


A little off topic but how do you like the glossy screen on your MBP?


It is a pretty big pain in the ass most of the time, but I don't have as much a problem with it as most people do. The glare is only troublesome if you focus on it, I just "filter it out" and I don't really have an issue with it. Granted, a lot of my MBP coding is done on a couch in my batcave-like living room so I don't have much glare to begin with.

While on campus and in class though, the fluorescent lights do make the glare an issue and I was able to see the kid behind me staring at me through my screen the whole class. But on the upside, now I will never get stabbed from behind.


Why didn't you get a refurb 2.6Ghz MBP with a matte screen instead?


Because, I assume, he, like me, wanted the new hotness. I cheaped out and bought the macbook, but I could've gotten a faster matte MBP.

I haven't minded the glossy yet. Its waaay brighter than my old matte screen and I usually forget its there. I still feel a bit torn about it though...lotsa pluses and minuses. Matte screens don't have such noticeable minuses unless you switch back.


agreed. why would i get an _old_ mbp? Most of the time I'm hooked up to my 24-inch dell anyways




Consider applying for YC's Summer 2026 batch! Applications are open till May 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: