• Welcome to OS2World OLD-STATIC-BACKUP Forum.
 

News:

This is an old OS2World backup forum for reference only. IT IS READ ONLY!!!

If you need help with OS/2 - eComStation visit http://www.os2world.com/forum

Main Menu

Compact Flash Cards (and any OS)

Started by RobertM, 2008.07.13, 02:08:46

Previous topic - Next topic

RobertM

Hey all, reading an interesting article via Slashdot.org:

http://hardware.slashdot.org/article.pl?sid=08/07/12/1851251

Apparently most CF cards have problems with DMA transfers due to a poor chipset design. Guess that explains the weird issues I had with WSeB and a CF card when I tried DMA transfer mode...



|
|
Kirk's 5 Year Mission Continues at:
Star Trek New Voyages
|
|


ddan

Some may have missed this: a PCI RAID controller using Compact Flash.

http://news.cnet.com/8301-13580_3-9803084-39.html?hhTest=1

It mentions 300X UDMA Compact Flash.

"We have tested a Transcend 250X industrial CompactFlash card and were getting close to 40 MB/sec sustained data transfer. When we striped two of these (in a RAID configuration), we achieved read/write speed close to 80 MB/sec," Kwong said.

The striped situation is not too shabby, though my 3 year old garden variety Maxtor 60G here benchmarks at an average 40MB/sec. In any case, OS/2 will fare better than, uh, some others because doesn't do anywhere near as much drive thrashing for swap file and    other, mostly needless, access.

kim

Well, better support for the "bulk" raid such as Promise would be great and I wonder if this shouldn't be quite easy to obtain - the Dani drivers already now picks up the Promise raid controllers but the drivers doesn't support the software based raid feature of the cards. I have Win XP system that I use edit recorded video on and I've using a Promise TX card with 4x200gb Maxtor drives in striped mode. This setup gives a quite good spin when working with larges files and as well the system doesn't feel chunky or slow at all when doing several things on the same time.

So I would guess the the Dani drivers would pick up the CompactFlash RAID card, but won't support the raid options of the cards since for $50 I'm quite sure that it's a software based raid such as the cheaper Promise controllers. But, really interesting since I've myself been using CF-cards on couple of smaller systems.

RobertM

Quote from: ddan on 2008.07.13, 04:28:03
The striped situation is not too shabby, though my 3 year old garden variety Maxtor 60G here benchmarks at an average 40MB/sec.

In any case, OS/2 will fare better than, uh, some others because doesn't do anywhere near as much drive thrashing for swap file and    other, mostly needless, access.


Ummm... yes and no. Windows (contrary to any technical doc's claims) seems to serialize disk writes. OS/2 on the other hand, seems to have no compunctions about asking a disk drive (or even other devices) to try to do multiple things at once.

If this hardware can handle such, then yeah, OS/2 will fly on it and take advantage of it... if it is marginal in it's support, then OS/2 will be slower or not even work in such a setup.

When I tried having my CF card work under OS/2, it definitely did not like being asked to do more than one thing at a time.

Sometimes, OS/2's advantages or design superiority, can be a disadvantage on hardware no longer designed with such in mind. Network cards (that are primarily designed for Windows) are another area that comes to mind... might work great in a Windows server... not so great in an OS/2 server of the same power that can handle a lot more traffic from a lot more ports all at the same time.

Robert


|
|
Kirk's 5 Year Mission Continues at:
Star Trek New Voyages
|
|


ddan

To RobertM: Uh, YES. On this very platform, even with 512M of memory, XP used
a swap file (300M or more) ALL THE TIME. But OS/2 NEVER uses its even with the
same mix of applications running: several Firefox windows, a command line or
two, MP3 player, ctorrent (5 streams at the moment). It's not just that OS/2
apps are smaller because not so GUI, it's that Windows is FOCUSED on the swap
file even when it has unused RAM. The reason I deduce is that they're
anticipating use and reserving RAM, rather than USING it right now for the
convenience of the user. Linux / Unix prefers this strategy too (though can be
somewhat tweaked), not coincidentally. It's a SERVER strategy, and I guess
that's why the comparision seems a bit ambiguous to you.

I want a PERSONAL COMPUTER, responding NOW to MY input, not ready to respond
to net traffic.

Concrete example: When I used the same antique DOS word processor with Windows
on this 2800+, very lightly loaded while I'm focused on typing (no streaming,
just browser windows open), even my slow typing frequently got ahead of
Windows updating the screen, then it'd put characters up in a burst, HIGHLY
annoying because you think keys have been missed. Meanwhile, XP would be doing
SOMETHING nearly all the time to the drive (and NO, NOT a virus -- besides
Windows). NO, problem is not the antique GUI, it's Crimosoft's antique "cooperative" method of multi-tasking.


ddan

Whoops. Alt-s is used in my w/p to save, so I hit it frequently by habit, and posted, er, prematurely. "GUI" in the last line should be "program". Rant continues:

You can easily observe Windows FAKING multi-tasking with its "cooperative"
scheme: tasks visibly freeze while others are updated. Windows is busy nearly
all the time because that's the only way it can try to ensure that each task
gets serviced. (It can be better for gaming because one task CAN hog the whole
system, but I'm sure even THAT virtue was entirely accidental.)

I pinned that down when trying to clone the w/p using Turbo Pascal 5 (so, a
DOS task, but YES it's still valid for comparison), same code, giving up time
slices properly implemented for both OS/2 and Windows. Under OS/2, the dynamic
adjustment that limits keyboard polling would cut in quickly, and my program
loop through its keyboard / internal timing checks three or four times a
second (yet be instantly responsive to a key). Under Windows, the loop ran
HUNDREDS of times a second, which is NOT FASTER, it's only WASTING time with
needless task switching (and sometimes doesn't keep up with typing).

WASTE is what Windows does best. With NTFS, it's ALWAYS doing something with
the file system, mostly unnecessary, having a bunch of "features" that track
usage, apparently. Frequently, even when just idle, XP has heavy disk
activity, and when its own screen saver starts, EVERYTHING ELSE STOPS while it
swaps out to disk. XP is even rabid about its "swap file integrity" between
sessions, will COMPLAIN if it's been changed or deleted (as I did during some
off-line clean up), which MIGHT be useful for security if hibernating, but I
never did that, so the swap file shouldn't have ANY relevance then.

So that's the kind of disk activity that I refer to.

(From what I read online, Vista is KIND OF happy with 3 gigabytes, but still
uses its swap file frequently and wants yet more on a SSD for "Readyboost". In
characteristically insane manner, they attempt to overcome bad design by
loading everything even vaguely likely into RAM, then having run out of RAM
due to staggering bloat, page it all back out to the HD.)

Saijin_Naib

My pagefile is off. I wonder what you are doing wrong. I also have the UDMA drivers for my Intel IDE controller, and disk access seems to be quite fine in UDMA mode, certainly not serial like it would be in PIO mode.

RobertM


Quote from: Saijin_Naib on 2008.07.14, 18:34:15
My pagefile is off. I wonder what you are doing wrong. I also have the UDMA drivers for my Intel IDE controller, and disk access seems to be quite fine in UDMA mode, certainly not serial like it would be in PIO mode.

Saijin:
Perhaps not serial... but a lot closer to serial than OS/2? Or perhaps I should say than the OS/2 machine I tried it on that was also running a web, ftp and SQL server on the card...

I have noticed that if I am copying more than one thing from drive to drive (ie: not from one place to another on the same drive), Windows seems to stop or near stop all but one of the copies... and both take horrendously longer. I have a feeling there is something NTFS isnt doing - especially with the added "overhead" HPFS/HPFS386 have with trying to ensure that there is no fragmentation.

Regardless, OS/2 seems happy to spawn new threads to do lotsa things at once... including the plethora that just the OS uses - which doesnt help.

Dunno... even doing little with the CF installed resulted in corrupt or lost data when DMA was enabled. But to be fair, I havent tried *running* Windows from it... just accessing it (and doing multiple reads/writes at a time). So, there may be the difference...


...thus, I'd probably disregard that part of my previous post.


ddan:

Add to that, OS/2 seems to know how to empty the swap file when it isnt being used anymore, and seems a LOT quicker to move DLLs back to RAM from the swap file when it does so.

I have a laptop (ancient, 288MB RAM) where Firefox (ancient 1.5 version) uses LOTS of RAM (and only has 180MB available when I start it - the rest is being used by HPFS386 disk caches and other things I am running). After a few hours, it hits the swap file... after that reaches 100MB or so, the system still seems to respond decently. Closing FF results in near instantaneous freeing of the swap file, and the delayed but quick loading of any DLLs or whatever that were swapped into it to make space for what Firefox needed.

I think part of the problem is Windows still doesnt handle many threads and processes well (which I think you hit on - it still seems very much "cooperative" multitasking once a certain threshold of running threads and processes are started - which also seems to be Vista's problem with handling "background" thumbnail creation - which slowed down file copies horrendously).

That was something else I didnt understand... OS/2 (for years) has simply been moving the file with it's attached EAs (which included the thumbnail)... And MS still couldnt follow that lead? It's not even like NTFS doesnt support it - it does. So then their kludge for something that was already there but they chose not to use doesnt run properly because the system cant seem to manage a copy thread and a thumbnail thread.

Funnily, WPSWizard, even launching the DLL calls that do the work via REXX (I think), manages it with so much less resource usage.


Sorry gang... jsut got back from updating user rights on a Windows Server 2003 box... not fun. I miss drag-n-drop-n-done WSeB management... but that's what our client has. Whenever I return, I am tempted to change user rights on my server just for the heck of it because it is so simple and so intuitive...

-Robert


|
|
Kirk's 5 Year Mission Continues at:
Star Trek New Voyages
|
|


Radek

2ddan, the "bursts". This seems to be a "feature" of eXtreme Problems. DOS sessions (or screen refresh in DOS sessions) seem to have so low priority that even a simple typing a text does not pass. You get the "bursts" you have written about. They prevented me from using a DOS plain text editor (kedit) under eXtreme Problems. Typing anything was impossible. Note that DOS 98 was not featured by this property.
I don't know whether the problem can be fixed (boosting the DOS session priority or refresh rate). I have made a mild attempt do fix it and I failed. Because I don't care about winblows I haven't tried further.

ddan

To Saijin_Naib: you've proved my point exactly without knowing it. You have Windows pagefile turned off. -- WHY? Because otherwise it uses it! I bet so even with your 2GB. How about giving that a test and report back?

To RobertM: as I mention elsewhere, frequently, Crimosoft appears to have deliberately crippled their own products: thumbnails in EAs are a good illustration (ha, I crack me up) of this. I doubt there's any reason Crimosoft couldn't have used HPFS in total -- they only designed the system and WROTE the HPFS code, after all -- but instead they went with FAT32 and eventually NTFS, both requiring a  crude separate file, the hidden: "thumbs.db" which now and then shows up on my OS/2 system from ZIP packages.  I can't come up with any compelling technical reason for FAT32 (though drive makers like it because it's destroyed millions of drives by wearing out the un-relocatable FAT tracks), so can only conclude that their reasons were Bill's ego and bank account.

Saijin_Naib

I have it turned off to save space (which is at a precious premium) on my 40gb drive. I have 13gb of music, and about 7gb of pictures, plus an assortment of games, so you can see why I felt I had no need for the pagefile's services any more. It does however, cause me issues when using the Adobe Creativity Suite because they DEMAND a paging file on a volume, no matter the ammount of physical RAM. For this reason and others, I've switched to Paint.net for my image manip needs.

Also, having no pagefile allows me to do an online defrag without any "system/unmovables" issues, as well as afford myself quicker response time when using large applications such as games or video editing/rendering software, which as is their wont, will page as they please if it's available.

I have it disabled in eCS as well, but only because I have yet to see a need for the 2gb of RAM on that system as all I can run are rudimentary web apps. That and eCS only occupies an 8gb HDD, so again, space is at a premium. If ever there comes a day when I can game and render in eCS, I may very well find myself in need of its "superior" thread and page management, but I sincerly doubt that day will ever come.

ddan

To Saijin_Naib: well to save space, you should also have Windows "System
Restore" turned off, it's just another benefit for drive manufacturers because
typically wastes 10% of an HD, and is useless so far as I've seen (own
experience and repair shop), particularly after malware. By the way, as I
recall the pagefile isn't deleted merely by turning it off. For purposes of
this thread, both those items argue against Windows and its heavy disk use.

Defragmenting is another hassle caused by Crimosoft deliberately avoiding all
the good ideas that THEY had implemented in OS/2, such as HPFS pre-allocating
file space. This seems to me particularly important for downloading. NTFS is
supposedly derived from HPFS, but has enough "improvements" to mangle it. I've
seen downloads to an NTFS partition result in THOUSANDS of fragments on some
files. I know that only because Crimosoft finds it necessary to include a
defragger that shows it to me. OS/2, except for JFS, doesn't even mention
defragmentation (but it's easy to do for HPFS with xcopy if you have a spare
partition). NTFS creates its own busy work because blocks must be allocated on
the fly while downloading, each allocation taking processor time and several
disk writes to link the new blocks into the lists, instead of just one write
of data to a pre-allocated location. The problem multiplies with several
simultaneous streams, hence the thousands of fragments. All that busy work
seems likely to use up the limited writes of a CF, though this doesn't seem to
be happening yet in practice, at least I've not seen any reports of failures.

Now back to philosophical differences: some of us LIKE text and "rudimentary
web apps", the more rudimentary, the better. I do NOT like web "features" and
turn off as many as possible, particularly anything automatically installed.

As for games, well, the graphics have gotten better since I played "Temple of
Apshai" on a TRS-80 (128x48 pixels, as I recall, monochrome), but basic game
play is still just random monsters to hack-and-slay, so I lost interest WAY
back. I've attempted to write my own games, but the lack of any real point to
them helped me to divert my programming time to something more useful.

You may not remember a time when it was promised that computers would expand
our minds, not narrow them to trivial thrills, spectacle, and the weird.

Even when you've "won" a game, what have you accomplished except wasted time?
What little I do is put toward some lasting value, however slight. Instead of
games, I spent time learning a smattering of French and German, and with that
better understand English, and can even sometimes get the gist of written
Italian and Spanish. I've dipped into programming and electronics a bit too,
always sticking to "rudimentary" almost as a design goal.

The goal of life is to improve YOU, not to divert and dissipate into empty
entertainments whatever talents whoever has endowed you with. I advise you to
cease being a slave to the now evil, but thoroughly inept, empire of
Crimosoft. Even Bill had his merits before pursuit of money corrupted him.

RobertM

Quote from: ddan on 2008.07.21, 16:41:55
As for games, well, the graphics have gotten better since I played "Temple of
Apshai" on a TRS-80 (128x48 pixels, as I recall, monochrome), but basic game
play is still just random monsters to hack-and-slay, so I lost interest WAY
back. I've attempted to write my own games, but the lack of any real point to
them helped me to divert my programming time to something more useful.

You may not remember a time when it was promised that computers would expand
our minds, not narrow them to trivial thrills, spectacle, and the weird.

You've obviously never played Final Fantasy VII!!!  ;D

On a serious note, you are correct. The sad thing is the game designers have lost touch with how to write a truly mentally immersive game - now it's all about better graphics. And history hasnt been a good teacher. FF7 was (and often still is) rated as one of the best games of all time... today it isnt rated that because of it's "neat graphics" - the graphics are abysmal by today's standards... it's all about the amazing story... one so good that FF7: Advent Children was made to critical acclaim. One so deep that anyone who has finished the game had a true understanding of the story and characters - and saw how that flowed near seamlessly into "Advent Children"

Yet still today's gaming companies cant figure out why current games are quick hits that die just as quickly... while people are begging Square to "recompile" FF7 for the PS3 (heck, there are petitions out there even!).

Eye candy rules! Sadly that is the case in most markets.


|
|
Kirk's 5 Year Mission Continues at:
Star Trek New Voyages
|
|


RobertM

Quote from: ddan on 2008.07.21, 16:41:55
Even when you've "won" a game, what have you accomplished except wasted time?
What little I do is put toward some lasting value, however slight. Instead of
games, I spent time learning a smattering of French and German, and with that
better understand English, and can even sometimes get the gist of written
Italian and Spanish. I've dipped into programming and electronics a bit too,
always sticking to "rudimentary" almost as a design goal.

The goal of life is to improve YOU, not to divert and dissipate into empty
entertainments whatever talents whoever has endowed you with. I advise you to
cease being a slave to the now evil, but thoroughly inept, empire of
Crimosoft. Even Bill had his merits before pursuit of money corrupted him.

Entertainment always has value... it's what keeps us sane, helps us wind down after a stressful day, and spawns our creativity. I dont think games are the only answer, but they can be a piece of that puzzle - for those it works for. For me, games are a small part - but so is reading (I read on average 50 books a year), and occassionally watching TV. Well written games can also stimulate the mind (good RPGs and story based ones) - so there are even other benefits than just relaxation. This is an area that I think is covered by "to each, their own"


|
|
Kirk's 5 Year Mission Continues at:
Star Trek New Voyages
|
|