DIARY 2013














Sat, Oct 26

<17:38 EEST> I have a busy weekend going on with school assignments and whatnot, but this tech update was too important to postpone. Besides, it just might kickstart a writing spree to help finish some homework. Anyway, today's theme is why Willow had its hard drive swapped.

Having played the sound guy in more than a handful of theatre pieces, my latest and fastest laptop had an odd slowness to it I had never seenheard before. It was detrimental enough to require a ramdisk workaround , which introduced its own set of inconveniences. Basically, the hard drive (a WD3200BEVT) had a lag of a couple of seconds when opening files. This was a no-no in cases where a song or an effect should be played right now upon cue. Most annoyingly, a song would often start on time, only to hang for a couple of seconds while the rest of the file would load.

My theory is that the OS generally caches some initial portion of a file into RAM when the player first loads it. After a lot of files are cached, the HD can enter some kind of sleep mode. It is the recovery from that state that takes unacceptably long. Now, this is not spindown; I rarely experience a spindown anyway, since I generally set the timeout for 10 minutes, and I have enough things going on that the HD never truly sleeps. OTOH, I sometimes notice this effect with complete spindown when playing files on a large external HD, and it is noticeably different. But "laptop" drives have a host of other features for saving power and protecting against head crashes. WD, notably, has its Idle3 feature that had given me trouble before. But even after turning it off, the same basic lag was still there.

Incidentally, I had a spare 2.5'' drive of the same exact capacity, so I decided to give it a try. This Toshiba is nominally slower, being older, and I guess it implements an older SATA standard. However, this only refers to bulk throughput and tells nothing about latency. Not surprisingly, this older drive works much better. I think it gets a little warmer and noisier when you pay careful attention, but all in all it really solved a problem.

In a grand twist of irony, this Toshiba came from hoo, for which it had been too slow, and it had been replaced by a WD. Then again, that WD is 3.5'' and being even older, it lacks the idle3 misfeature.

Now, pretty much all of my hard drives have been the cheapest and/or most easily available for their size and capacity. So far, I've had little need to discriminate against HD makers; IBM/Hitachi does come to mind, as they are very noisy in my personal experience, and they have a bad reputation in some geek circles (Deskstar aka Deathstar). OTOH, I have always found Toshiba HDs nice and quiet. Now it seems WD has made my blacklist, and I'm not the first one to give them flak about idle3.

This must be one of those cases again with a lesson about misleading marketing here. First of all, the drive with faster advertised speed turns out slower in practice. It's one thing to implement a faster SATA standard, another to deliver those speeds in bulk, and yet another to have low latencies for practical speed. Latencies are not really advertised, and even a basic seek latency figure cannot tell a whole story.

Especially if all these nice things are ruined by power saving overkill. In fact, too many "eco" features these days are doing quite the opposite. If I need to setup a ramdisk to use this HD efficiently, it's not exactly saving power. In the end, it is only saving power by sitting unused in my shelf, while I revert to older tech to get some work done.

Sun, Oct 13

<18:32 EEST> A number of small hardware updates and software developments since the last entry, with too little time to report each one in time. These days, I find myself virtually living and breathing mathematics, with somewhat more concurrent studies than springtime. Besides the couple of courses needed to finish my subject studies, I am also taking some basic courses in general linguistics, and an advanced math course. The latter is kind of borderline, though; after prioritizing time for the more essential courses, and a few nights per week for the bare minimum of theatre sound tech, I rarely manage to complete the advanced homework. I find it takes a little more for the material to sink in, as it is not immediately obvious like many of the subject courses, but in the end it is not too hard, basically continuing on the abstract albegra pathway.

On the hardware front, the major changes around Hoo involve selling my two "lesser" Radeon GPUs, the passive-cooled HD5570 and HD5770, as Litecoin mining is getting less profitable. My only genuinely profitable GPU now is a HD7970, but I am also keeping the pair of HD5870s, as there is still plenty of numerical work left in the world. Plus, with their silent Shaman coolers, they are kind of special ;)

By numerical work, I am referring to Mersenne prime search, and the software mentioned last time has taken an interesting turn since clLucas reached a working version. Thus a full primality test can now be done on a Radeon, taking about 4-5 days for current fresh exponents on a 7970. It is not (yet) quite as efficient as the Nvidia counterpart, but a drastic improvement over the usual CPU test that can take months. As done before for trial factoring, my contribution is an automatic network update script.

The other change on Hoo is going somewhat against my cool and quiet principles. Having a spare HD around, I swapped its 2.5'' drive into a bigger, bulkier one to deal with the increasing disk activity. NFS service, in particular, was getting annoyingly slow, with a multitude of cryptocoin daemons and other p2p applications thrashing the drive. The results is not even any noisier, since the Antec case, a fine example of objet trouvé, has nice removable drive bays with rubber washers.

As an advocate of using low-power "mobile" parts in "desktop" usage for lower energy waste and noise, I used to wonder why 2.5'' hard drives are so despised for being slow. We generally make computers faster by making them smaller, and a drive-head should have shorter seek latencies in proportion. In fact, some 3.5'' server drives are made with 2.5'' platters for this very reason. Alas, "mobile" components come with a bunch of features intended for power saving and/or impact protection. Besides general slowness, some of these like the Idle3 head parking in many WD drives, are outright detrimental.

Besides hardware updates, it is four weeks since my headware update, and I guess it is time to reflect on my experiences on this change. Except, there aren't really many to report. Life without hair feels quite a natural state, which is no wonder since I have donned a similar hairstyle oftentimes in the past. This might be the reason for the many little annoyances about long hair, it is not something I grew up with.

Some of the details to consider are things like tossing your hair back upon dressing up or putting a backpack on. The first few times of going out after the shave, I did give a moment on such long-hair reflexes, but I never once found myself doing any of them by accident. This, surely, is further evidence on my "natural state of hair". It also relates to the feeling of heightened awareness, which of course comes with any notable change in style, but this time I feel there is a real physical explanation, as I cannot hide behind my mane. I actually feel I am seeing more of my surroundings, which is sometimes a little overwhelming; I also feel other people can see more of me. Then again, I feel happier walking around in public when there is no web of hair trying to invade my face and orifices constantly, and I hope that feeling shows.

Risto A. Paju