homer

Member
  • Content count

    8
  • Joined

  • Last visited

Community Reputation

0 Neutral

About homer

  • Rank
    Member
  1. In rare cases where you miss (or disable) OS-level disk cache, but get a cache hit in the drive, the read will complete faster. It's hard to see how that would have a noticeable impact for the end user, except in some rare corner cases and specific circumstances. In common usage, the only difference would be a slight reduction in read/write latency. However, write latencies are usually hidden by the OS and read latencies are usually dominated by rotation and/or seeking. Perhaps it would be possible to measure this on a fast 10k drive, but it should be fairly negligible. Now for some hypothetical stuff with lots of cavaeats. I'm not entirely sure what's the maximum SATA packet payload, but some ACHI docs I found seem to imply it's 4 MB. Now, if a drive reassembled this chunk before starting to send it to the controller (again, I don't know if SATA works like this), you could be talking about 13.3 ms to send it via SATA 2 vs. 6.7 ms to send it via SATA 3. But if drives really worked like that, then controllers or the OS would probably break such large reads into smaller ones, so that the first parts of the read could be transmitted while the drive accessed the later parts.
  2. [Apologies if this seems misplaced. I don't see a forum on RAID configuration.] I'm configuring a RAID on SSDs. It happens to be 3 drives in a RAID 5, but this is a fairly generic question. I had the idea that I could reduce stripe read-modify-write operations and write amplification by using a segment size of 4k (which equates to a stripe size of 8k, in my case). Then, I build the filesystem with a block size that matches the stripe size. The only downside I can see is the overhead of using such a small stripe size, if the controller is too dumb to combine a sequence of 4k reads into fewer, larger reads. The reason I care about performance of small writes is that this filesystem will be used for software builds, among other things. This involves frequently creating large numbers of small/medium-sized files. From what I can tell, this isn't a very common practice, but I suspect the tendency towards large stripe sizes is a legacy of mechanical disk drives and simple controlers. My RAID "controller" is Linux software RAID (mdadm). Any thoughts?
  3. homer

    A survey: What is correct:

    I think you mean "precisely define". As far as I know, "Scientifically" implies use of the scientific method, where a hypothesis is formed and experiment is designed to support it (by trying to disprove it). Yes, many people in all countries probably don't realize that they may be seeing slightly distorted graphics by using 1280x1024 on a 4:3 CRT. Some people don't even realize they should run LCDs at native resolution. Anyhow, it's not absolutely necessary to have square pixels. If software has access to the screen aspect ratio and resolution, it can compute a correction factor and render graphics correctly. Since this is extra work (and can add a great deal of complexity to some graphics code), programmers often take the lazy way out and assume square pixels. Probably the most notable case of non-square pixels is NTSC video, where a digital representation of a frame typically consists of 704x480 active samples, yielding a 1.1 pixel aspect ratio. These numbers were carefully chosen (see ITU-R BT.601), based on aspects of the NTSC and PAL analog signal formats and hardware design practices. Fortunately, HDTV resolutions do have square pixels (when used on HDTV monitors). Now, if they could have just gotten rid of interlacing...
  4. What are the various platter counts? That could account for the differences. Heh, read/write throughput is an area where loosing the platter density race actually works in your favor (assuming equal drive capacity).
  5. homer

    A survey: What is correct:

    ** "homer" creeps around with ruler, measuring every LCD monitor he can find ** Hey, you're right!! Well, I guess I learned something. I'm still not convinced that's the case for all 1280x1024 LCD monitors, throughout time. I'm pretty sure I once actually measured a 17" at 4:3, but I can't check that, right now. But, I'll certainly accept that's true of all current models. I'm a little happier that these LCD monitors have square pixels... but the problem with this approach is that lots of games and video playback software have fullscreen modes that assume a 4:3 monitor aspect ratio. I guess the best thing to do (if you're a monitor mfg and even use that resolution, in the first place) is to prioritize square pixels. Lots of developers are too lazy to support non-square pixels, since that can be a lot of work. It's the windowing system's job to provide an interface for finding out the screen aspect ratio - and with widescreen monitors gaining popularity (with aspect ratios like 16:9, 16:10, and 5:3), developers pretty much have to query that. Well, the situation should improve as windowing systems move to embrace 3D & vector display technologies, leaving pure raster systems in the dust. Relax. If HamaZ feels bad about this, it should be embarrassment from making the case so authoritatively, in spite of being quite wrong - not from harsh criticism (which may leave HamaZ feeling defensive, instead of reflective). If we don't at least help HamaZ understand the errors, this thread will have been a complete (if insignificant) waste of time and energy. The point about the tone of this whole thread is a good one. To start out as a seemingly innocent question, such as seeking advice or understanding, and then turn around and start bragging that many people are wrong is just poor manners. I would have never replied, had I known this thread wasn't started out of genuine curiosity.
  6. homer

    Please recommend hard disk cooler

    Well, you can add a case fan, blowing out the back. Then, you could slap a heatsink or two right on the drive (use thermal tape, instead of thermal paste). Many PC modding sites sell heatsinks for cooling ram or motherboard components. PC Power & Cooling makes (or used to make) a product called DriveCool, which is a bit like a CPU cooler, except that it's smaller and shorter. I know neither of these is exactly what you asked for, but perhaps they'll give you some ideas.
  7. Does anyone know of external USB2 enclosures that support SMART (assuming the drive inside does)? Recently, the only drive I've not been monitoring is in such an enclosure. Since that drive just died, the importance of this matter has been escalated (I still need an external drive). Thanks.
  8. homer

    A survey: What is correct:

    Both are common resolutions, though I don't know if 1280x960 is "standard". 1280x1024 originated from the workstation world, in the 80's (though I don't know exactly how or why those specific numbers were used, since the monitors were presumably 4:3). 1280x960 is something I've seen more recently (since the late 90's), and has the benefit of providing square pixels on 4:3 monitors. Since graphics systems mostly use pixel-oriented dimensions, you may notice slightly distorted (short & fat) graphics or video, when viewing in 1280x1024. For this reason, if I have to use a 1280 resolution, I always go for 960 vertical. The one (big) exception to this would be on LCD monitors or projectors that natively support 1280x1024 (unfortunately, this is quite common). There's no good technical reason for LCDs to have 1024 lines, instead of 960. However, if you want to support a resolution between 1024x768 and 1600x1200, nearly universal support for 1280x1024 in graphics hardware and windowing systems is sure a strong argument. And why ask this on a Storage site?