• Content Count

  • Joined

  • Last visited

Community Reputation

0 Neutral

About jskubick

  • Rank
  1. Whoops. I was under the impression that the read/write heads in modern (since '96 or '97) hard drives were flexible plastic film with embedded read/write elements that not only could come into contact with the disk surface... they did. The arms that positioned them were rigid, of course, but "the last micron" (so to speak) was soft, harmless, and actually spent most of its life in contact with the disk surface anyway. So much for that idea. :oops: (*scratches head, trying to figure out where he ever got the idea that drives now used thin film heads*)
  2. jskubick

    [b]Video Capture Cards[/b]

    Well, I'll admit that once ATI finally got their act together sometime around January 2001 and released drivers for Windows 2000 that worked, they worked pretty well. But their XP drivers have sucked since day one, and unlike the Win2k drivers, haven't gotten any better since. I've been religiously letting XP report the crashes to Microsoft in the hope they'll take away ATI's WHQL certification for the November 2001 drivers, but I'm not holding my breath. What's SO frustrating is the fact that ATI's capture hardware and chipset are phenomenol. The problem is they keep releasing new chipsets and cards that are almost completely incompatible with the previous generation in every significant way, and wind up having to practically rewrite the drivers from scratch every time. Which basically means that the drivers for their "latest and greatest" chipset achieve some modicrum of stability on their one favored OS of the moment right around the time their next chipset comes out and they all but abaondon further work on it... and the drivers for any other platform they pretend to support never make it out of perpetual beta. Or, they need to spend some cash bringing in a consultant or two for a couple of weeks to teach their development team how to write proper WDM drivers. If they did it right and followed Microsoft's rules, they wouldn't need one driver for XP, another for 2k, yet another for ME, and so on down the line. Correctly-designed WDM drivers will work on everything from 98 (maybe 98SE) to XP. Just the fact that they have different drivers for 2k and XP (to give one particularly graphic example) shows they don't know what the f**k they're doing (or don't care, or aren't allowed to care by ATI's management), because properly written XP drivers WILL work with 2k. That doesn't mean that any driver that works with Windows 2000 will work with Windows XP... but it does mean that if they follow the rules, dot their "i"s and cross their "t"s exactly the way Microsoft says they should, that same driver will work with every Microsoft OS that's come out since Win95c and NT4(*). Sigh. In another 3 or 4 months I'll probably end up chucking the AIW128pro for a new card, but it really sucks because I'm totally content with its hardware capabilities... my only complaint is with the STOP errors caused by its shitty drivers 4 or 5 times a week, usually when I'm in the middle of something. :-( ------------------------------------------- (*) "So," you might be wondering, "Why does XP rarely work with Win2k drivers, especially for things like Webcams and Videocards?". Excellent question. It's because XP was the first Microsoft OS that disavowed support for anything besides WDM drivers. 98/SE/ME can use WDM drivers, but they don't require them. They can still use Windows95 drivers. Likewise, Win2k can use WDM drivers (and in fact, strongly prefers them), but it's mostly content to slough along with NT drivers instead. Prior to XP, companies had little incentive to support WDM, because it would have meant supporting a third driver model that wasn't explicitly needed by any existing OS anyway (either that, or write off support for Win95 and/or NT4 altogether). So they ignored WDM, and kept developing parallel sets of 9x and NT drivers instead. The problem is, most people will now agree that Win95 is (or at least should be) safely dead and buried, having lived a longer and more fruitful life than any legacy OS in modern history, and NT4 is equally dead as a multimedia platform. But rather than spend a week or two tweaking the "XP" (WDM) drivers so users with ANY WDM-supporting legacy OS can enjoy the benefit of new drivers as the company focuses exclusively on the WDM drivers' development going forward, they're simply ignored entirely (or worse, the company wastes time applying half-assed patches to the non-WDM drivers). Worst of all, in many cases, the inability to use "XP" (WDM) drivers is simply the fault of the installer, which determines that the OS isn't XP and simply gives up, even though the drivers might, in fact, work flawlessly with no additional work required by the development team. sigh.
  3. On the random chance that someone who works for IBM, Seagate, etc. happens to surf this site out of boredom occastionally... Consider the possibility of a 2.5" (or 3", or even 3.5" form factor as long as it's thin) drive that implements the equivalent of SpeedStep, capable of spinning at a slow power-saving speed (say, 2400RPM), a medium speed for normal battery operation(4200-5400RPM), and a high speed for A/C operation(7200, 10k, or even faster). Heat? Make the drive able to be pulled out with minimum ceremony. It doesn't have to support hot swapping, but it should at least be robust enough to survive an attempt with nothing more than Windows going down in flames and the drive/notebook not being damaged. Use some form of fast serial interface (USB2 or FireWire) as the drive's native interface. The main idea being that the drive operates at its slow and medium speed when inside the notebook, but can be pulled out and connected via a foot long cable when the notebook is connected to A/C and run at its high speed (harmlessly radiating the extra heat into the room rather than the notebook's inside). This would offer the advantage of providing convenient mobile/battery-powered operation (the drive's safely inside the notebook), but offering desktop performance when used externally. Furthermore, if FireWire or USB2 were integrated into the drive's motherboard as its native interface (probably still ATA behind the scenes), its status as a niche product for notebook OEMs would be mitigated by its appeal to third-party companies that sell external drives and would need to do little more than package the drive in a case and bundle a USB-2/FireWire cable with it to have a marketable product.
  4. jskubick

    [b]Video Capture Cards[/b]

    I'd recommend a card based on the Brooktree BT878 family from someone like Pinnacle or Hauppauge. The BT cards are popular and generic enough to have their own "home grown" drivers and utilities, so even if the manufacturer itself loses interest in supporting it a year or two from now (*cough* ATI *cough* AIW 128pro *cough* same Rage Theatre capture chips as Radeon, but still stuck with year-old driver that blue-screens almost daily *cough*) you probably won't be left out in the cold. I have a rather personal, intense disdain for anything by ATI. I made the mistake of buying an All-in-Wonder 128 Pro back when it was still their top of the line card, delighted to be getting a card with 3D almost as good as a GeForce, with MPEG-1 and MPEG2 capture and Faroudja-licensed deinterlacing to boot. Of course, it took ATI almost 9 months to finally release drivers that could capture more than 10 minutes of video under Windows 2000 on the random occasions when it didn't just crash altogether the moment capture was initiated. And of course, XP arrived about 2 weeks after they finally released THOSE drivers, and it was back to square one for another 4 months... What pisses me off more than anything is the fact that the current AIW Radeon cards have the exact f**king same capture hardware and chipset as the 128 Pro, but ATI hasn't bothered to release new 128 Pro drivers for XP in almost a year (even though the card causes STOP errors at least 3 or 4 times a week due to looping errors in the driver). Of course, it's possible that they simply haven't done stinker on the capture drivers for the Radeon either because they've been too busy trying to up the QuakeMark framerate instead, but that would just lower my opinion of them even MORE if it were true... Grrrrrr....
  5. jskubick

    "Luggable" laptop that uses 3.5" HD?

    Definitely NOT a laptop, but quite luggable and a worthy peer of ANY high-powered desktop dualie workstation: http://www.maxvision.com/maxpac/MaxPac6200.cfm * Dual Athlons * 10k or 15k SCSI hard drive * 17" TFT * Quadro 4 750, Geforce 4Ti4600, ATI 9700 Pro, and others * SECOND TFT panel can be accommodated with larger case Wow. I'm in love. I can't afford it, but I'm still in love ;-)
  6. Does anybody know of any manufacturer/models (probably one of the more adventurous Taiwanese OEMs like Kapok, Clevo, and the like) of big, fast, powerful, but nevertheless reasonably light (say, 8lbs or less)laptops that use normal (well, maybe low-profile) 3.5" desktop hard drives? It seems like something sensible to do since new laptops have room to burn thanks to the proliferation of big 15" displays and increasingly use desktop CPUs for extra speed when battery life isn't terribly important anyway. More importantly, it would open the door a tiny crack for some of the more adventurous speed freaks among us who don't mind living dangerously to perform an, *ahem*, "unsupported performance upgrade" and drop in a 7200RPM drive to augment the 2.6GHz desktop P4 with high-powered cooling system sitting nearby... ;-) Seriously, it seems like there'd be a sizeable market for laptops built using a desktop CPU (or, god forbid, two CPUs) and 3.5" hard drive, along with two bays -- normally, one with a DVD+CDRW, another with a battery intended to last 30-60 minutes (the assumption being that the user needs power and transportability, not necessarily the ability to use it w/o A/C for long periods of time), with other combinations involving two batteries (no drive), optical + floppy sans battery, two optical, etc.).
  7. jskubick

    5.25" HDD Media

    Hmmm. Well, actually, the next-best thing would be to electronically interface it to the PC as two separate hard drives (which I think is what you said you're doing). Then, the OS could do its own RAID 0 striping. The main thing would be to ensure that both "drives" could transfer data to/from the PC simultaneously without affecting one another's transfer rate (in other words, in IDE terms, make sure both virtual head-drives are the master and get their own private channel). check out www.arcoide.com -- they make a chipset/controller that transparently handles RAID 0 or RAID 1 striping behind the scenes. Basically, each drive connects to the Arco controller, then the Arco controller itself pretends to be a single EIDE drive with make-believe geometry that encapsulates both drives. Because of the way it works, it's OS-agnostic. As far as the OS knows, it's a standard single ATA/xxx drive. The last time I checked, they had two basic varieties: one that plugs into a PCI slot (using the slot only as a power supply and physical mount point), and one that sits inline between the drives and the motherboard's IDE controller (powered by one of the power supply's drive cables). I'm not sure how good/bad their performance is (the ONLY reason I even know about them is because I tripped over their booth at Comdex a while back and thought it was a cool idea), and I've NEVER seen their stuff reviewd here or anywhere else, but it might be worth checking out....
  8. jskubick

    Need analog modem; does anyone do reviews?

    The problem isn't so much a matter of percentage as one of lockstep regularity. Regardless of what the host CPU might be doing at the time, it must drop whatever else it might be doing and devote its full attention to the modem every few milliseconds. So even if Windows is 99% finished with something that needs to be done before it can update the display or give control back to a user or application, it must stop, divert its attention to the modem, and only then can it finish the task. Delaying gratification to the modem is not allowed. A good metaphor might be a late-night visit to McDonalds, with a single cashier working both the front counter and the drive-through window. Your entire order might be ready, with everything on your tray EXCEPT the fries. But then, before she can get the fries for you, someone arrives at the drive-through window, and she has to stop working on your order to take THEIR order first. Technically, the work was done... the food was cooked and ready. But until she finished the intermediate task, you were forced to stand waiting (blocked). In a more forgiving and sensible environment, she might have made the drive-through customer wait 5 more seconds so she could finish your order first and spare you the delay with minimal inconvenience to the drive-through customer. But she couldn't. She HAD to service the drive through customer immediately, just like the CPU must service the HSP winmodem with precise regularity. And thus you had to wait. In Windows terms, the blocking activity manifests itself as brief periods of unresponsiveness. A window that doesn't resize instantly. A button click that doesn't seem to do anything until a few moments later. A menu that doesn't pop out the moment it should. In the grand scheme of things, they're brief interruptions, but the cumulative effect is to make the computer SEEM a lot slower and less responsive.
  9. jskubick

    Need analog modem; does anyone do reviews?

    You make some excellent points, but the above is utter nonsense. PS Posted from an HSP modem. I won't argue TOO much over that particular point. ;-) It's just something I read online a couple of weeks ago when I was troubleshooting a friend's computer (her PCtel modem quit working). At the time, it made sense and resonated well with the reality that chipset manufacturers do take liberties with cost cutting related to legacy components, partly because compromises there usually just cause them to work slower than they should, and few people care about them or really notice anyway.
  10. jskubick

    5.25" HDD Media

    (lightbulb above head) Since you're prototyping a drive with two heads and two controllers ANYWAY, why not orchestrate the two and provide transparent RAID 0 striping? Mainstream manufacturers aren't interested because all stupid mass-market consumers (and the OEMs who make their PCs) care about is noise and cost, but if two heads are a design necessity to begin with, it seems the added expense to go all the way and just stripe the data between the two heads on the same platter would be trivial...
  11. jskubick

    5400rpm versus 7200rpm

    When I bought a 40-gig 7200rpm Maxtor last year, I was BLOWN AWAY by the performance difference between it and my old 14GXP. I expected it to be faster, but NEVER thought it would be SO MUCH faster. I couldn't believe that a drive that was a little over a year old (and the absolute best of its class at the time I bought it) could have fallen SO FAR behind in a little more than a year. But it did. Of course, it all became a moot point a few months later, about 3 weeks after the warranty expired... :-(
  12. jskubick

    Need analog modem; does anyone do reviews?

    Not all Winmodems are equally bad. There are two kinds of Winmodems: * controllerless. The host CPU handles protocol-related matters. The actual generation and translation of audio into a proper bitstream (including things like echo cancellation, noise suppression, etc.) is handled by a proper DSP. Modems of this type have the potential to be quite good, and in some ways are objectively superior to traditional modems. More on this in a moment. * host-signal processing (HSP). These are evil. Bad. There's absolutely NOTHING good you can possibly say about them. These are the $6 modems that consist of little more than a tiny PCI card with a few connectors and a single (small) chip in the center. Unlike controllerless modems, these modems don't even have a digital signal processor to handle the audio to data conversion. They LITERALLY are nothing more than a minimal soundcard with a few low-tech parts to handle the impedance matching between the phoneline and sound chip. They digitize the modem audio in realtime, and rely upon the host PC's CPU to do the millions of fast fourrier transformations necessary to turn it back into a stream of binary data. This is a lot more work than you'd think, because modem audio is NOT a single sine wave. All kinds of behind the scenes digital signal processing needs to be done to turn an audio stream resembling random white noise into something resembling 56k data. You've probably figured it out by now, but the use of a $7 HSP winmodem is the fastest way to squander half the market value of your blisteringly fast 2+GHz CPU (it'll soak up about 20% of your CPU speed when in use. Compare the current price of your CPU with the current price of one that's 20% slower). And don't even get me STARTED about how easy it is to f**k up the drivers for such a modem badly enough to require either replacing it with a new modem using a different chipset or reinstalling Windows. HSP modems to avoid like the plague are ANYTHING with a PCtel chipset, the letters "HSP" on the box, or a PCI card with just a few components and a single tiny chip in the center. As far as I know, all of Connexant's chipsets are HSP, though it's not inconceivable that they might have one or two controllerless ones that use a DSP. I believe ALL CNR and AMR modems are HSP by definition. Now, for the good news. CONTROLLERLESS modems entail very little sacrifice. The amount of CPU processing required to handle protocol-related stuff (responding to AT commands, data compression, etc.) is virtually nil. Unlike HSP, where the host CPU's undivided attention is required on a regular and frequent basis when connected, controllerless cards are entirely capable of buffering the incoming bitstream for a few milliseconds at a time when the CPU has something more important to worry about. Kind of like the difference between old CD-R drives that required a 150-600kbit/second datastream in absolute lockstep and would churn out a coaster if the CPU's attention wandered for even an instant, and new drives that can buffer the bitstream and tolerate periodic inattention from the CPU. Everyone loves RS-232 serial modems because they're painless to connect, don't require drivers, and work with everything. Unfortunately, RS-232 is probably the LEAST efficient interface you could possibly use to connect anything to a modern computer. RS-232 is an old, slow, inefficient protocol. More importantly, because it's so slow, it's one of the "sloppiest" functions of a modern mobo chipset, rivaled only by the floppy drive. Few people realize that 16550 emulation is just that -- emulation. Not as in, "a virtual 16550 is embedded in the chip and offers the same functionality," but as in, "the chip lies and claims to be a 16550, but in reality it can only buffer a few bytes at a time before overflowing because the designers (or their bosses) knew modern CPUs are fast enough to hide that little sin anyway, so nobody will really notice the difference." That's why Windows (and presumably Linux) offer the option to ignore the fake 16550 entirely. As a practical matter, Windows and Linux are usually reduced to polled RS-232 I/O, just like they were 15-20 years ago. On the other hand, a good controllerless winmodem (like the Lucent/Agere ones) can buffer a good chunk of the bitstream and periodically deliver it to the CPU with minimal overhead. In many cases, a system using a controllerless (but NOT HSP) Lucent-based winmodem performs BETTER than one with a modem connected to the db-9 serial port behind the computer, simply because the added protocol-related work is minimal relative to the hoops Windows/Linux needs to jump through to use the fake 16550-based serial port. That's not to say someone like USR can't make a PCI card with a high-quality serial port that has all the advantages of a controllerless lucent winmodem... but then you've lost the advantage of not needing drivers. And as I said, the real overhead imposed by relying upon the host CPU to do PROTOCOL-related stuff in virtually nil. A good, controllerless Lucent/Agere-based modem (WITH DSP) costs about $20 online or at a computer show. The PCI USR modem with the quality serial port and 100% hardware-based modem costs about $100. Is it worth it? If your ONLY means of getting online is via modem, "maybe, but probably not." If you've got DSL/cable and only keep the modem around to send faxes and act as a safety net to keep you sane once or twice a year when there's an extended cable/dsl outage, I'd say "no way in hell." NOTES: * Lucent/Agere explicitly supports Linux, and apparently their Linmodem drivers work quite well. They're NOT open source. Whether or not you care is a matter of personal ideology for you alone to decide. * Lucent's modem chipsets are now manufactured by Agere. I don't know whether Agere is a new division, someone who just bought it, or what, but the chips themselves are the same. * Not ALL winmodems with a Lucent/Agere chipset are controllerless and have a DSP. They DO have at least one chipset that's HSP. So make sure you do your homework before buying, especially since the retail price difference between a controllerless and HSP Lucent/Agere winmodem is a whopping $5 or so. * I'd recommend against HSP from anyone, even Lucent, but if you've got no other choice, Lucent HSP is STILL way better than PCtel, and I'd trust it before Connexant and Cirrus too. But NOTHING is worse than PCtel. * If you're feeling really evil, buy a PCtel modem and give it to someone you REALLY HATE as a gift. Then cackle with glee, knowing that they'll ultimately have their computer slowed down, and probably burn several days at some point trying to make it work when something upsets Windows and it decides to stop acknowledging its existence ;-)
  13. jskubick

    Need a SCSI Controller Recommendation

    I think the subject adequately sums it up... like lots of others, I'm thinking about getting a Cheetah to augment my existing 7200RPM ide drive as the main system drive -- booting and running Windows and apps from the Cheetah, and using the ide for bulk storage of stuff I rarely touch. I can confidently say that NO controller I might buy now will EVER be used with more than two drives. In all likelihood, it will never be used with more than one. Historically, I buy new drives every 14-20 months. At that rate, by the time controller performance became an issue, it would probably STILL be cheaper to just toss today's controller and replace it with a future one than pay top dollar for an ultra high-end card today that will suck compared to even cheap future cards anyway. SO... what's the cheapest controller capable of driving a single Cheetah at full speed, with a comfortable bit of headroom? A controller that's fast enough to where its own performance will never be the limiting factor, but not necessarily capable of much beyond what the 15.3k Cheetah itself can throw at it?