[PCWorks] Fix? What's the fix for getting 1394b (firewire 800) to work on XP SP3??
- From: "Clint Hamilton-PCWorks Admin" <PCWorks@xxxxxxxxxxxxxxxxxxxxxxxx>
- To: "PCWorks@xxxxxxxxxxxxx" <pcworks@xxxxxxxxxxxxx>
- Date: Tue, 2 Feb 2010 07:03:43 -0600
I guess I'm the only person at this list (and a forum) that
wants to or trying to use 1394b/FireWire800 on XP. <confused>
Below are some things I did after this post:
//www.freelists.org/post/pcworks/PCWorks-Whats-the-fix-for-getting-1394b-firewire-800-to-work-on-XP-SP3,1
that either fixed the problem, or at least helped. Note that
in that post, the M$ "fix" for this on SP3 not only made things
EVEN SLOWER, but screwed up other things as well (and I
shouldn't be surprised).
I pasted all of this below, so that it may help another. At
the bottom is my original post.
OrangeWare Corp
(http://orangeware.com/developers/dev_prod.html#5) claims their
drivers will give you true S800 1394b speeds, but that's what's
bundled with the controller card! Unibrain
http://unibrain.com/Products/DriverAPI/ubcore.htm also claims
this, and I tried their drivers...same crap! What they claim
is also BS.
I got to thinking about trying the 1394a port on the PCI card,
to see if I could find out if it were maybe the enclosure, or
the PCI card. So I hooked up the enclosure's 1394a port to my
mobo, and the tests were about the same (actually a bit better
in some, slightly worse in others, essentially the same). So I
then hooked up the enclosure's 1394a port to the 1394a port on
the controller card, and, the tests were bad, significantly
slower!
So this points to the controller card "having something wrong"
with it. But the question there is; is it its drivers causing
it, or the hardware? Or, is it the PCI slot/bus? I'm more
inclined to believe it's not the PCI bus because it's more than
fast enough for FW800. OR, is it the 1394b drivers and issues
with XP with it??? I have no way of knowing if this problem
would still exist with a different PCI card, or even with a
mobo with integrated 1394b.
I emailed Koutech to flat-out ask them does their FireWire
cards REALLY WORK at 1394b speeds on XP SP3, since
two of their cards were two I was considering. Get this,
he told me all of their 1394b controller cards **require a
64-bit PCI slot in order to achieve full 1394b FW800
speeds!** He also went on to say ALL 1394b cards are
that way! Huh??
First I have to say, IF that is the case, then why are there so
many of the cards that are only for 32-bit PCI slots? Why do
these cards mention nothing about this? Why do no 1394b
enclosures mention anything about this?
Furthermore, I've been reading all of the user reviews I could
find for all of the PCI cards I could find, and not a single
person mentioned anything about needing a 64-bit slot to get
FW800 speeds. In fact, I was familiar with some of the mobo's
they mentioned and they did not have 64-bit PCI slots! (I also
read where many of them said they had to use the Unibrain
drivers to get FW800).
Having a rather "obsessive" personality about certain things
like this, and never wanting to give up, I decided to start all
over again. Reading all the user reviews for 1394b cards, and
reading all the manuals for the 1394b cards, I learned a few
things. It would seem most of these controller cards either
come with the Unibrain drivers, or the manufacturer actually
"suggests" downloading the Unibrain drivers (if they don't come
with the card). In the case of them not coming with the card
and they are linked at the manufacturers' websites, on some
occasions they were even linked as "FireWire 800 drivers" or
something like that.
And again, in reading the reviews I saw that many mentioned
Unibrain drivers having to be used in order to get FW800 or
"faster" speeds(1). On one of the Syba cards, someone from
Syba TS actually responded to one of the reviews saying, (to
paraphrase), "3rd party drivers are available for FW800 from
Unibrain".
I also saw in some of the manuals that the drivers should be
installed FIRST before the card is installed.
Like I said, I tried the Unibrain drivers and they didn't work
any better, and the card's original drivers are by OrangeWare
(which also claims FW800 speeds). BTW, my card is only a
32-bit card. Many of these cards say they work in both a 32 or
64-bit slot. So since some of the cards are indeed 32-bit, and
claim FW800, it would seem a 64-bit slot isn't really
required.(1)
So this all points to actual FW800 speeds being possible on
XP (after SP1).
I'm going into details so that this may help others trying to
do the same thing.
I uninstalled the card/drivers from the Device Manager (both
the 1394 and LAN portions of it, LAN was "Disabled"), I
uninstalled the Unibrain drivers, then did the ERUNT restore.
Then I **first installed the Unibrain drivers.** They
"complain" about not being able to find something via 2 yellow
marks in the Device Manager for "Unibrain PC". I put the card
in a different PCI slot (slot #5, is only shared with slot #1
which has never been used), the started the PC. The yellow
marks were resolved.
Drag 'n copy timed tests were previously inconsistent, probably
because the HD was about 60% full, and we all know that even
after defrag'ing files can still be all over the place and
actually vary from time-to-time. So I reformatted the HD and
left it empty.
After installing the drivers, I checked all "1394" areas in the
Device Manager and saw no mention whatsoever of "1394b"
for any of the entries' Properties dialogs.
After installing the drivers, in the Program Files folder for
Unibrain, there's a "Tools" folder and a "ubSwitch" app.
Clicking that puts an icon in the System Tray which gives one
the ability to actually immediately switch between M$ and
Unibrain drivers on-the-fly via left or right click. It would
appear that by default, M$ drivers are used for one of the
entries. I had two for some reason; "OHCI compatible board"
and "IEEE 1394b (FireWire 800) adapter". The latter defaults
to the M$ drivers. I made sure both were Unibrain, and then
the DM showed "1394b" and "FireWire 800" as a description for
one of the areas, "1394 bus host controller".
I then proceeded to do the (synthetic) benchmarks again with
the 1394b tests of course being on the controller card, but
with the 1394a tests being on my mobo's integrated 1394a
controller. (I wanted to give 1394a an even better shot at
being fast, see 5th paragraph above for details). As before,
1394b was significantly faster than 1394a. So I then did the
"real world" right click drag 'n "copy" and drag and "move"
timed tests. I also did the right click and drag 'n copy tests
**on the drive itself,** so the target HD would be doing both
the reading and writing at the same time (a good torture test
for a HD since those times are usually slower).
This time, unlike before, the 1394b tests WERE a good bit
faster. I haven't done all the math yet, but it would appear
the timed tests are anywhere from 30 to 50% faster for 1394b,
depending on the type of test. For one example (same for 3
tries), dragging a 2.91gb folder of .rar files from my HD103SJ
to the FW HD (write) took 110 secs for 1394a and 71 secs for
1394b. That's about 26.45MB/sec Vs. 41MB/sec, and those are
pretty impressive for write times. Right click, drag, and
"Move"(ing) the folder back to the HD103SJ was 97 secs for
1394a and 88 secs for 1394b.
Where 1394b really shined, was copying the folder onto itself;
right click, drag, and "copy here" in the same root directory.
1394a took 5:12 & 5:17, 1394b took 3:34 and 3:25. That's a
pretty big difference.
I then did the time tests with a more difficult 1.35gb folder
consisting of all kinds of file types (116 folders, 564 files,
of .txt, .html, .mht, .jpg, .gif, .bmp, .doc, .rtf, .reg, .zip,
.exe, etc., etc.). Dragging and copying that folder from the
HD103SJ to FW on 1394a took 52 secs (26MB/sec). 1394b
took 36 secs (37.5MB/sec). (Both avg over 3 tries, all within
1 sec of each other). Then doing the right click and "copy
here" for that folder (using the 1st one dragged over), 1394a
was 2:24-2:30, 1394b was 1:46-1:43.
I also ran the FC-Test tests
http://www.xbitlabs.com/articles/storage/display/fc-test.html
(As you can see on the page, it's better than "benchmarks"
because it actually simply uses Windows to test HD's, by simply
creating many thousands of files in the write test, then
reading them back for the read test). This was confusing,
1394b was still faster, but not by as large of an amount like
the other tests. But its time-to-completion margins increased
as the test sizes increased.
Previously I had not run PCMark for the FW tests, but this time
I did:
PCmark04
Results shown as 1394b / 1394a (In MB/sec)
Total score: 4435 / 4053
XP startup: 8.031 / 7.635
App loading: 6.906 / 6.422
File copying: 25.315 / 19.716
General HDD usage: 5.733 / 5.392
PCmark05
Results shown as 1394b / 1394a (In MB/sec)
Total score: 4447 / 3602
XP startup: 8.068 / 7.67
App loading: 6.869 / 6.472
General HDD usage: 5.748 / 5.387
Virus scan: 47.652 / 31.870
HDD file write: 47.141 / 29.289
For those that aren't familiar with PCMark tests, tenths
actually make a much larger difference on their tests than is
normally the case with other benchmarks. A drive that is
actually literally twice as fast as another, will NOT get twice
as high scores.
I can post the other benchmarks comparing 1394a Vs. 1394b if
anyone is interested.
(1)-So the question remains, are the "faster speeds" actually
really 1394b FW800 speeds, or something just in between?? In
keeping with the THEORETICALLY "twice as fast speeds" going
from one protocol to a newer one and the newer one ***not ever
being actually twice as fast,*** FW800 should not actually be
twice as fast as FW400. Because SATA300 isn't anywhere near
twice as fast as SATA150 and SATA6 isn't anywhere near twice as
fast as SATA300. UDMA66 was not twice as fast as UDMA33, and
UMDA133 is not twice as fast as UDMA66, etc. So I might
"guess" that FW800 may actually in reality maybe be only 40-50%
faster than FW400, and maybe I am getting true FW800 speeds??
I'm still wondering if a different controller card (or 64-bit
card) would be faster (again, see 5th paragraph as to why), or
if doing some XP driver file hacks would make it even faster.
I've seen some that replaced some XP 1394 driver files with
some from Vista, and got better results. I've also replaced
some XP SP3 files with SP2 files to get back certain features
that the M$ morons removed from SP3. So that's why I had asked
about possible 1394 file changes in Win7, wondering if I could
find out that list and possibly get those files and try them on
XP. More so because from what I've heard the 1394b issue is
still a bit of an issue on Vista.
But at least now at this point, I am getting significantly
better performance with 1394b over that of 1394a on ALL tests.
I don't know if it was due to not installing the card-supplied
drivers and using the Unibrain drivers, or installing the
Unibrain drivers first before installing the card, or using a
different PCI slot, or a combo of all of the above. I'm more
inclined to believe it was due to only using the Unibrain
drivers and installing them first. (And again, be sure you put
the "ubSwitch" icon in the System Tray and the Unibrain drivers
are loaded).
The Unibrain page doesn't specifically state the types of cards
supported, but they appear to work with most 1394b PCI
controller cards that use Texas Instruments chips.
-Clint
----- Original Message -----
I just installed a 1394b PCI card
http://www.siig.com/ViewProduct.aspx?pn=NN-8G0012-S1
and hooked up an external 1394b HD case
(http://eshop.macsales.com/item/Other%20World%20Computing/MEFW912AL2/).
I ran benchmarks on it, and it's only "a little" faster than
the exact tests I ran on the same HD in a 1394a enclosure.
Not anywhere near twice as fast and I didn't think it would
be, but I still thought it would be faster than this, but it IS
significantly faster.
So I then started with the timed file transfer tests which is
simply copying of large folders with all kinds of files in
them. 1394b is SEVERAL TIMES SLOWER! How can this be
possible? How can the benchmarks be FASTER, yet it takes over
***TWENTY MINUTES*** to copy a file that took 90 SECONDS with
1394a??????????????????? (It said it would take 24 minutes and
I just disgustingly canceled it).
I started checking into this to see what could be the problem,
and to my utter horror, I came across this:
http://www.tomshardware.com/forum/148542-45-firewire-1394b-mbps-driver
Look at the bottom of the page. WTF?????
So I then found this "fix" at the M$ site:
http://support.microsoft.com/kb/885222 But the freakin' patch
is ONLY for SP2! It won't install on SP3! I do have the path
it mentions:
"HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Enum\PCI\"
....but I do not have that info "1394_hc_hw_id" after that path
anywhere. I guess because I can't install the patch. ???
I also don't have that file "cstupd1394sidspeed.dll" anywhere
on my PC, but I DO have the other file Ohci1394.sys. These
files are in the unpacked patch but it doesn't say where
cstupd1394sidspeed.dll should go. I don't even know if
manually adding that file to the "correct" location would work.
The first thing that's beyond my comprehension, is HOW ALL of
the many benchmarks I ran can be FASTER for 1394b, but when
dragging and copying files it's a 15-20 times slower than
1394a?!?!?!?!?!?!?
The next thing that's beyond my comprehension, is how these
retards at M$ can not only pull a stunt like that, but, also
how can anyone SELL ANY 1394b hardware for XP (after SP1)
when it DOESN'T WORK?!??!?!?!?!?!?!?!?
Then finally, how can I incorporate that fix above to work on
SP3? Just add those registry keys? Or does SP3 not even need
the fix? If SP3 does not need the fix, then why is it so slow?
I can't be the only person that's trying to use 1394b on XP SP2
or SP3.
-Clint
God Bless,
Clint Hamilton, Owner
http://www.OrpheusComputing.com
http://www.ComputersCustomBuilt.com
http://www.OrpheusComputing.com/cheap_reliable_web_hosting.html
=========================
The list's FAQ's can be seen by sending an email to
PCWorks-request@xxxxxxxxxxxxx with FAQ in the subject line.
To unsubscribe, subscribe, set Digest or Vacation to on or off, go to
//www.freelists.org/list/pcworks . You can also send an email to
PCWorks-request@xxxxxxxxxxxxx with Unsubscribe in the subject line. Your
member list settings can be found at
//www.freelists.org/cgi-bin/lsg2.cgi/l=pcworks . Once logged in, you have
access to numerous other email options.
The list archives are located at //www.freelists.org/archives/pcworks/ .
All email posted to the list will be placed there in the event anyone needs to
look for previous posts.
-zxdjhu-
Other related posts:
- » [PCWorks] Fix? What's the fix for getting 1394b (firewire 800) to work on XP SP3?? - Clint Hamilton-PCWorks Admin
