What is the fastest way to load Windows XP?

  • Thread starter Thread starter Unknown User
  • Start date Start date
I trimmed down w95B to less than 9Mb, but still does not boot less
than 5 seconds; IMHO the problem is in legacy IDE drive running before
OS is full loaded to drivers kick in for DMA ...
Does anybody know a solution to enable DMA right at boot in Dos?

now I got an idea to speed it up; will check adding smartdrv to
Autoexec.bat if helps ...

Some bios do have the ability to enable DMA, I don't recall
the bios setting though. It was fairly obvious in use as I
used to Ghost several systems and some would maintain over
1.25GB/min backup while others only 300MB/min... of course
this was factoring in the CPU overhead for compressing it
too, and 2-3 years ago, today I'm sure that 1.25GB/min
figure would be quite a bit higher.
 
I like to impress my customers by my fast booting system.
First, hit the power on button, then hit the KVM switch to
the box that is already on. Quite a wow factor. Yes, I am
a bastard con artist. Actually not, because this is all a lie,
so I am just a liar. : -)


: I'm looking for a way to load Windows XP extremely fast on an x86
: computer. I guess 2-3 secs would be fine.
: I have configured 2 SCSI 10,000 rpm HDD in Rack-0, but it is not fast
: enough.
:
: Is there a way to attach a sort of Eprom, or Flash RAM from which I could
: install/boot/load Windows XP and that would use 100% Bus throughput? Do
: you have other suggestions?
:
: Thanks,
:
: --
: Using Opera's revolutionary e-mail client: http://www.opera.com/mail/
: http://www.auriance.com - http://www.auriance.net
 
thats from their website data!
normal latest big 7,2kRpm HDs have around 10MB/s (WD
Are you sure?

well maybe a little more (no Raid!); IMHO for Raid0 setups speed
almost doubles ... Sustained Data Rate Transfers (that matters!)
Here copying an entire movie (700MB) from one raid array to
the other takes less than 10 seconds.

NOT copying, Try MOVEing that file back & forth between drives, you
will get much longer times! IMHO
I found out from a review that you
can read at 55MBps and write at 33MBps. I think you can almost double this
transfer rate if you have 2 HDD in RAID-0. Here's an exerpt from the
review:

"HD Tach?s sequential read test (which calculates the average data reading
rate across a whole platter) coughed up a score of 55.1MB/s. Writing
across an entire platter was equally impressive at 33.5MB/s. That?s a
whopping 20 percent faster than yesterday?s champ, the IBM Deskstar
180GXP!".

Those are not sustained rate data transfers, but burst rate
data transfers (which occur very rarely IMHO)
I just ran Sandra to have an idea. It says "SATA150 3.5' 2x Raid0 (36GB)
Drive Index: 95MB/s" - Is this the throughput?

as I see, you did not bench that with that Sandra I gave you the
link!!! (it gives drive index in points!).

Any newer Sandras versions are too much dependable on CPU & not
suitable for comparisons! Bench again!
 
thats from their website data!

Are you sure you interpreted it correctly?
NOT copying, Try MOVEing that file back & forth between drives, you
will get much longer times! IMHO


I often move files around on my gigabit lan. From a (now
relatively old and slow) 40GB 7K2 drive like a Maxtor Plus
8, moving it to one of the 120GB or smaller drives it
averages about 26MB/s. I believe some of that is the PCI
NIC, it would be slightly faster if it stayed within same
system. Within same system moving files on larger drives
I'm sure it's averaged over 50MB/s, probably faster but I've
not benchmarked it in awhile. With smaller files I can
question it due to caching, but these were video files
ranging up to several GB in size... and the destination
system only had 256MB memory in it.
Those are not sustained rate data transfers, but burst rate
data transfers (which occur very rarely IMHO)

HDTach does what it calls a sustained figure, but like any
brief "one-test" type of benchmark it's only necessarily
valid when comparing other drives with same test.


as I see, you did not bench that with that Sandra I gave you the
link!!! (it gives drive index in points!).

Any newer Sandras versions are too much dependable on CPU & not
suitable for comparisons! Bench again!

Sandra does screwy things too, I do know that it and HDTach
don't always agree. Caching algorythms of the drives
themselves seem to effect Sandra more, make WD drives look a
little better than they might in general-purpose use, but
it's a more (still not entirely) applicable test if one were
to put the WD in a fileserver with multiple access requests.
I suppose a windows mulitasking environment could benefit
too, but the tradeoff seems to be slightly slower sustained
reads... ignoring the Raptors, I am generalizing the Caviar
series.
 
In message <[email protected]> Spajky
now I got an idea to speed it up; will check adding smartdrv to
Autoexec.bat if helps ...

smartdrv is likely a bad idea, it won't be used once Windows is off the
ground and accessing hardware directly, but smartdrv's memory will still
be reserved and unavailable for re-use.
 
Are you sure you interpreted it correctly?

Yes, that was for the 1Gb (350$) no DMA one!
(http://www.cdw.com/shop/products/default.aspx?EDC=690178)

/the other one with the Hell of a price is few times faster!/

PS.: I tried that loading of Smartdrv with Autoexec.bat for my
W98seLite & it loads tiny bit faster (a second maybe faster-not
noticeable really) but whith that my Micro w95B (w/o chipset drivers,
9Mb version), it speeds things up for about 16% faster & that is
noticeable; it loads it in 4 seconds now :-))) (only WIN without Post
time routines) on my 3y old 7,2kRpm drive ... :-)
...
 
X-No-Archive: yes

True, a loaded system with lots of configuration tweaks,
networking and software tools (in other words, a very
functional system) will take longer to boot. If one could
strip windows down to the point where it's only taking a few
seconds then it would have to be so specific in purpose that
one has to wonder if windows is even the right OS for the
job.

Win98 could be installed in a minimum configuration - the user has a
choice. This was given the go bye by Microsoft in XP and the whole
thing is installed, whether you like it or not.

98Lite made Windows installations still leaner - my 98SE, as installed
is just 110 mb.

XP has a similar program - the only difference is that you have to
install XP as well as Framework and under this use nLite to create an
XP install disk and then use that disk to install afresh an
abbreviated XP. The perhaps XP should start faster.

Look for nLite here: http://nuhi.olmik.net/

HTH
 
X-No-Archive: yes



Win98 could be installed in a minimum configuration - the user has a
choice. This was given the go bye by Microsoft in XP and the whole
thing is installed, whether you like it or not.

Whether I like "what" or not?

I have no gripes with Win98, within it's limits it's fine.
I feel most people who thought it buggy were misplacing
blame as a large percentage of the time it was drivers or
applications that were buggy. As each newer generation of
windows was released, the drivers and apps got more stable
too, and thus someone running win98 would reap many of the
stability improvements seen by winxp users by doing same
thing winxp users did- Getting modern more mature drivers
and software. There are definite limits to Win9x though,
the memory and resources, drive capacity issues and less
robust networking and security. For smallest full-featured
system win9x is still the best alternative, BUT trimming
down an OS till it loads in a few seconds is anything but a
full-featured system.

98Lite made Windows installations still leaner - my 98SE, as installed
is just 110 mb.

It'll go a lot lower too, and further, you can use
compressors like UPX et. al. to further shrink the size of
files retained. Whether or not this is better than using
drivespace I don't know, but it certainly does make things
more manageable to use compressed individual files rather
than an entire compressed filesystem volume.
XP has a similar program - the only difference is that you have to
install XP as well as Framework and under this use nLite to create an
XP install disk and then use that disk to install afresh an
abbreviated XP. The perhaps XP should start faster.

Look for nLite here: http://nuhi.olmik.net/


I"m sure it would start faster, though some other poster
already mentioned one of the most significant steps to a
fast boot- how fast the bios initializes, POSTS, enumerates
and passes control to the booting OS. I don't recall anyone
mentioning suspend-to-disk yet either, perhaps there is some
way to set the system up such that the suspend-to-disk image
is simply that of the system state after it has finished
booting, such that differences in windows load times are far
less significant than the total memory load.
 
Back
Top