Maximum memory for W98
ngreplies
installed
The motherboard is an ABIT BD7 motherboard.
Please could anyone tell me what the maximum RAM is which I W98 can use if I
insert the maximum this mobo can take.
Thanks in advance
Terry
philo
"ngreplies" <ngre...@voidtdrd.freeserve.co.uk> wrote in message
news:dmqhh3$rv0$2...@newsg1.svr.pol.co.uk...
adjust your vcache setting before adding the extra ram
http://www.frugalsworld.com/tweaking/vcache.shtml
a machine running win98 should be able to accept 1.5 gigs of ram...
or possibly 2 gigs
Rick Chauvin
> a machine running win98 should be able to accept 1.5 gigs of ram...
> or possibly 2 gigs
That may sound good on paper but is an inaccurate assumption in todays world,
and the current day truth is that W9x is just not designed to handle more than 1
GB of ram.
That 1.5 & 2 GB assumption was often repeated from the old school of thought
where at the time they did not even have the hardware to even test that theory,
but now that we do and it's been shown over and over that old school assumption
is essentially just not true. I wish all the websites that repeat that phrase
would update their sites because it's just not a reality! The reality is if you
yourself or anyone took a thousand different 9x hardware setups, you won't see
many if any actually being able with any trusting stability to run 1.5 GB, and I
doubt you will ever see one run with 2 GB of RAM at all on a W98 machine. In
reality 99.999% of the time 1 GB is the best you can do with any piece of mind
and stability with W98, and so my advice is not to go exceed that - unless just
for testing purposes - and if you do test it's imperative to consider that you
also have complete OS partition backups for a way out of what can result out of
that instability.
1 GB of ram in 9x needs a system.ini edit to match this:
[vcache]
MaxFileCache=512000
..also for any tech wanting to test more than 1 GB of ram will need to also edit
the system.ini file to prevent windows from attempting to use more than 1 GB of
the total ram with this:
[386enh]
MaxPhysPage=40000
..so why even bother if you can't use it! ..unless of course you are
multi-booting on the same MB and want to use more than 1GB in W2K/WXP
I personally don't recommend more than 1GB for 9x no matter what though ..go
ahead though and try it <g>
Lots of this is covered at these other websites, but again if it's listed on any
websites pay no serious mind who advises to use 1.5 & 2 GB on 9x.
Here is some good talk about the subject though at these websites:
http://onlinehelp.bc.ca/tips.htm#64mb
&
http://aumha.org/win4/a/memmgmt.htm
Rick
Noel Paton
news:OYdUaEC%23FHA...@TK2MSFTNGP12.phx.gbl...
....and the EXACT truth is that it depends more on the motherboard than upon
anything else......
SOME motherboards will allow Win98/SE to run quite happily with 2GB of RAM -
some will not allow Win98/SE to run with anything more than 512MB (despite
an apparent limit of 2GB)
Whatever the motherboard, there is one first step - to limit the [vcache]
MaxFileCache to less than 512MB. The actual number used for optimum effect
can again vary with the motherboard, but usually a setting of
MaxFileCache=512000
will work
- although some people need to set it as low as 384000.
Settings above around 524000 (512MB in decimal) will ALWAYS fail, as the RAM
runs out of address-space at that point, and as soon as the 'need' for more
vcache exists, the OS will crash before freeing up lower memory, as it is
supposed to.
'Old school of thought' - Win 98/SE is an OLD operating system, and what
worked once on it still does, as nothing significant has changed in its
memory-management since the day it was released, and it is still based on a
16-bit DOS
I know at least two people who regularly run Win98 using 2GB - and one who
has 3GB installed, but throttles it back to 2GB (in [386enh]) so that
Win98/SE will still run as dual boot, when he needs it to
There should be no need to set a MaxPhysPage value, unless the motherboard
has problems with RAM above the 1GB threshold - and it can indeed be
counter-productive in Win98, as it may take the limit literally, rather than
running with (say) 768MB instead of the 1GB that your value of 40000
indicates.
--
Noel Paton (MS-MVP 2002-2006, Windows)
Nil Carborundum Illegitemi
http://www.crashfixpc.com/millsrpch.htm
Please read http://dts-l.org/goodpost.htm on how to post messages to NG's
Rick Chauvin
[....]
> ....and the EXACT truth is that it depends more on the motherboard than
> upon anything else......
>
> SOME motherboards will allow Win98/SE to run quite happily with 2GB of
> RAM - some will not allow Win98/SE to run with anything more than 512MB
> (despite an apparent limit of 2GB)
Hi Noel, could you list what those MB's are that will work please?
It would be great to have a listing of them to reference.
> Whatever the motherboard, there is one first step - to limit the [vcache]
> MaxFileCache to less than 512MB. The actual number used for optimum
> effect can again vary with the motherboard, but usually a setting of
> MaxFileCache=512000
> will work - although some people need to set it as low as 384000.
> Settings above around 524000 (512MB in decimal) will ALWAYS fail, as the
> RAM runs out of address-space at that point, and as soon as the 'need'
> for more vcache exists, the OS will crash before freeing up lower
> memory, as it is supposed to.
>
> 'Old school of thought' - Win 98/SE is an OLD operating system, and what
> worked once on it still does, as nothing significant has changed in its
> memory-management since the day it was released, and it is still based
> on a 16-bit DOS
I understand your point with the way you worded that about old school. My saying
old/new school was meaning that I don't know of any new MB's that will let W98x
work with more than 1GB of ram. Even Jim Eshelman mentions that now MS flatly
stated that W9x is not designed to handle more than 1 GB of ram, and has been my
experience too, but I'm sincerely always open to a new understanding and truth.
> I know at least two people who regularly run Win98 using 2GB - and one
I'm encouraged that you personally know 2 others who run it.
It would be great if you could list their MB's & Hardware setups (video card,
etc) ..of those 2 people, I would appreciate it. I would like to test duplicate
that setup unless their setups are too old now for me to obtain.
> who has 3GB installed, but throttles it back to 2GB (in [386enh]) so that
> Win98/SE will still run as dual boot, when he needs it to.
Very nice to know that too, and look forward to hearing of what their hardware
setups are.
Appreciate your post,
Rick
Richard G. Harper
run Windows 98 with 1.5gb of RAM on a P3 motherboard. Sorry, don't remember
the brand or model, that was some time ago and I only did it for a few
days - long enough to prove you could. :-)
There's almost as much sense running Windows 98/Me with 1gb or more of RAM
as there is putting a Ferrari engine in a Ford pick-up truck. It can be
done, but there's no real sense in it. There are very few situations where
the RAM will be anything more than ornamentation.
--
Richard G. Harper [MVP Shell/User] rgha...@gmail.com
* PLEASE post all messages and replies in the newsgroups
* for the benefit of all. Private mail is usually not replied to.
* My website, such as it is ... http://rgharper.mvps.org/
* HELP us help YOU ... http://www.dts-l.org/goodpost.htm
"Rick Chauvin" <jus...@nospamz.com> wrote in message
news:OCLQEzD%23FHA...@TK2MSFTNGP11.phx.gbl...
Galen
Richard G. Harper <rgha...@email.com> had this to say:
My reply is at the bottom of your sent message:
> I doubt Noel knows - I know darn well I don't. For a very short time
> I did run Windows 98 with 1.5gb of RAM on a P3 motherboard. Sorry,
> don't remember the brand or model, that was some time ago and I only
> did it for a few days - long enough to prove you could. :-)
>
> There's almost as much sense running Windows 98/Me with 1gb or more
> of RAM as there is putting a Ferrari engine in a Ford pick-up truck. It
> can be done, but there's no real sense in it. There are very few
> situations where the RAM will be anything more than ornamentation.
>
>
> "Rick Chauvin" <jus...@nospamz.com> wrote in message
> news:OCLQEzD%23FHA...@TK2MSFTNGP11.phx.gbl...
>> Noel Paton wrote:
>>
>> [....]
>>
>>> ....and the EXACT truth is that it depends more on the motherboard
>>> than upon anything else......
>>>
>>> SOME motherboards will allow Win98/SE to run quite happily with 2GB
>>> of RAM - some will not allow Win98/SE to run with anything more
>>> than 512MB (despite an apparent limit of 2GB)
>>
>> Hi Noel, could you list what those MB's are that will work please?
>> It would be great to have a listing of them to reference.
In addition I once loaded ME (not really 98 but still basically 9x) on a 64
bit AMD (obviously in 32 bit mode) and started off with the second RAM chip
removed. After tweaking I tossed in the second one and, I gotta tell you, it
ran VERY sweet. One thing I could not dig out of the system was that on
reboot I was constantly getting errors. Problem was, it seems, that it was
actually shutting down faster than it could handle??? I never found the
hotfix for it nor was I able to make that problem go away. I understand that
it was a common problem though as ME was out when the faster CPUs started
becoming more readily available. I can say that it was quite stable and
speedy on a GigaByte board with the nVidia chipset with the exception of the
errors (this was about a year ago so the exact error's long since escaped my
memory - blame beer) on boot it ran like a champ. Oh my did it run... It
jumped up and danced on command and suffered nary a problem the entire two
weeks or so of testing. Those error messages blew by at the speed of light
(AMD64 3400 if I recall, that box is now the backup file server) so it was
almost worth it. If ME had NTFS support and thus the additional security of
user accounts I'd have said "bugger it" and just dealt with the errors. It
was like a tweaked out chamber maid in a mid-range inner city hotel... So,
it CAN be beefed up and I strongly suspect that the same would be true for
98 and 98se... I have considered it as an idea for Windows 95 just 'cause
it'd make an interesting article or something but I don't think I'll ever be
that bored or ever not too busy to stoop that low.
--
Galen - MS MVP - Windows (Shell/User & IE)
http://dts-l.org/
"My life is spent in one long effort to escape from the commonplaces of
existence." - Sherlock Holmes
ngreplies
some point in the near future I want to upgrade to either XpPro or the new
Windows OS (Longhorn?? Whistler???).
At the moment I am on an Abit BD7 2Ghz:
Processor is Intel Pentuim(R) 4 2,00GHz(100MHzx20.0)
I am not certain if it is really worthwhile updating the RAM on this model
as it must be virtually obsolete now and I have no idea if it can be
upgraded with any other processor at all.
Terry
Rick Chauvin
> I doubt Noel knows - I know darn well I don't. For a very short time I
> did run Windows 98 with 1.5gb of RAM on a P3 motherboard. Sorry, don't
> remember the brand or model, that was some time ago and I only did it
> for a few days - long enough to prove you could. :-)
I'd like to know if it can be done on a modern spec'd MB with P4 setup is all.
I've gots some older hardware I could scrape together but there's no 'going
forward' point in testing that really. I have 3 machines here in front of me
and each triple boots 98SE/2K/XP and all of them are very modern performance
spec'd hardware, and I have 2 GB ddr ram for each but 'unfortunately' only leave
1 gb ddr in them since I often find myself needing to test software's on 98SE
and have to boot to that, and there is no way in heaven will any of these setups
let me have more than 1gb of ram installed and run with any 'stability' for 9x.
You may recall just last year when we spoke via email and yourself and Ron
Martell gave me all of each of your 'tricks of the trade' and advice to try and
get my setups to run more than 1GB in them - but the long story short was if you
remembers that it was not possible with the modern rockin' hardware I was using
to do so. I see no beneficial reasons of building old p3 machines with older
MB's and hardware anymore since the advantages of the modern ones dwarf the old.
> There's almost as much sense running Windows 98/Me with 1gb or more of
> RAM as there is putting a Ferrari engine in a Ford pick-up truck. It
> can be done, but there's no real sense in it. There are very few
> situations where the RAM will be anything more than ornamentation.
I understand your meaning Richard I do.
Your word ornamentation though can be 'in some instances' less than true. For
me I have noticed in specific 'important to me instances' where using 1gb
performance wise it's quite noticeable as compared to using less. It comes into
play for one instance when I use my software reg/file trackers taking full
system file/registry comparison snapshots where on much less ram the procedure
takes much much longer....Once a first compare snapshot is done and retained in
memory then with installation/changes made - and for a second compare takes
just 10 seconds since the large RAM can hold and compare all the info 'in hand'
and RaPiDlY process the resulting differences within itself, whereas using much
less ram does not have enough room for that full process and has to rely on
reprocessing all the information again instead; and so I say 1gb of ram can make
a big difference 'if' you have a process that can utilize it.
Granted not many people use W98x anymore and much of my time is on W2Kpro or
WXPro, but the same applies to those OS's too where the difference between 1GB
of RAM can be substantial 'if' you have applications that can benefit from it;
otherwise, yes as you say it's only ornamental yes. Thinking of my relative
just now though reminds me of how I use to tease him about him using 4GB of RAM
in his work machines and I always chided him about it at first that it was
ridiculous and overkill and all in his head, but he then one day showed me in
his specific usage of processing/editing verrry large video files that it does
make a difference for the same reasons I cited with its ability of instantly
being able to accommodate and retain and process verry large information's
within itself - and so video intensive applications is another scenario and good
example where more than 1 gb of ram can be very beneficial ...heh, and you can
bet the pro's don't use W98 to do it with either. <g>
Rick
Noel Paton
WinXP, and Longhorn was the original codename for Vista) It's not due for
release
until Q3 next year, so you have a rather long wait!
According to
http://www.abit-usa.com/products/mb/techspec.php?categories=1&model=36
<quote>
- Intel Socket 478 Pentium 4, up to 2.6GHz (400MHz FSB)
- Intel Socket 478 Pentium 4 Celeron, up to 2.0GHz
</quote>
Ron Martell
>Richard G. Harper wrote:
>> I doubt Noel knows - I know darn well I don't. For a very short time I
>> did run Windows 98 with 1.5gb of RAM on a P3 motherboard. Sorry, don't
>> remember the brand or model, that was some time ago and I only did it
>> for a few days - long enough to prove you could. :-)
>
>I'd like to know if it can be done on a modern spec'd MB with P4 setup is all.
>I've gots some older hardware I could scrape together but there's no 'going
>forward' point in testing that really. I have 3 machines here in front of me
>and each triple boots 98SE/2K/XP and all of them are very modern performance
>spec'd hardware, and I have 2 GB ddr ram for each but 'unfortunately' only leave
>1 gb ddr in them since I often find myself needing to test software's on 98SE
>and have to boot to that, and there is no way in heaven will any of these setups
>let me have more than 1gb of ram installed and run with any 'stability' for 9x.
>You may recall just last year when we spoke via email and yourself and Ron
>Martell gave me all of each of your 'tricks of the trade' and advice to try and
>get my setups to run more than 1GB in them - but the long story short was if you
>remembers that it was not possible with the modern rockin' hardware I was using
>to do so. I see no beneficial reasons of building old p3 machines with older
>MB's and hardware anymore since the advantages of the modern ones dwarf the old.
>
That is the exact circumstance where you will find the MaxPhysPage
entry to be most useful.
Just add the MaxPhysPage=40000 entry to the [386] enh section of the
System.ini file for Windows 98 and then you can increase the RAM to
whatever you want for Windows XP. Windows 98 will only see 1 gb of
RAM and all should be well.
Good luck
Ron Martell Duncan B.C. Canada
--
Microsoft MVP (1997 - 2006)
On-Line Help Computer Service
http://onlinehelp.bc.ca
Rick Chauvin
Hi Ron,
Yes I know it's useful in many situations especially with the older hardware,
just not my newer stuff. You and I went over this last year as we emailed if
you remember, but on my setups no matter what value of MaxPhysPage I entered
while the 2 GB of sticks (even 1.5) were plugged in still would not stop the
memory and numerous fatal, etc, errors. A few times after crashing and upon
reboot only to find it had neutered the OS even - of course I had a partition
image backups so it was no big deal to re-image. I tried every trick in the
book to get it to run more than 1GB of RAM but there was no way. I GB though
works flawless and is 100% stable and so that's good enough for 9x. If by
chance I need more while booted into 2K or XP I just plug in the extra 1 GB
stick ...it's really no big deal for now and in no time I won't even bother
with booting to 9x anyway and only do it to help on the forums here and there,
and so doing that it's nice to have it booted to 9x being able to accurately
check out the questions/responses in real time.
These days I would say that with some of these newer Intel MB's and these newer
6600GT VideoCards, etc installed, there's just no way to go over 1GB of System
RAM while booting W98; I guess there's just not enough address space remaining,
or whatever, with all the modern spec'd hardware installed to still have it work
right, and maybe that's why the older MB's and hardware was more lenient in
letting some of you get more that 1GB of RAM to work at that time. For GP, I'd
like to know if anyone with brand new spec'd hardware is able to do it though,
but this is really not important though as time goes on since W98 is long going
by the wayside, but I only bring it up now and then to stay current with it and
for general conversation about it.
Rick
SteveSch
<<snip>>
> otherwise, yes as you say it's only ornamental yes. Thinking of
my relative
> just now though reminds me of how I use to tease him about him using 4GB
> of RAM in his work machines and I always chided him about it at first
> that it was ridiculous and overkill and all in his head, but he then one
> day showed me in his specific usage of processing/editing verrry large
> video files that it does make a difference for the same reasons I cited
> with its ability of instantly being able to accommodate and retain and
> process verry large information's within itself - and so video intensive
> applications is another scenario and good example where more than 1 gb
> of ram can be very beneficial ...heh, and you can bet the pro's don't
> use W98 to do it with either. <g>
I ran a simple test on my machines with extra RAM. I was very surprised by
the difference.
http://www.linuxquestions.org/questions/showthread.php?s=&threadid=377255
Steve
Rick Chauvin
[.......]
> I ran a simple test on my machines with extra RAM. I was very surprised
> by the difference.
>
> http://www.linuxquestions.org/questions/showthread.php?s=&threadid=377255
>
> Steve
I noticed the same differences with extra RAM too and especially when large size
processes are needed like you do, and the information can be drawn direct from RAM
instead. We spoke of that earlier in the thread you are replying to; however,
what would increase that even 'more dramatically' on your 'desktop' computer is if
you used a PCI controller card. You will not believe the difference until you try it
and see for yourself. Besides the PCI controller card commonly used to bypass
large hard drive support issues, a controller card offers many other Excellent
advantages little realized. Less important is the extra ports allowing various
capabilities not available otherwise, but more importantly using the card especially
with the faster processors these days (but even older ones) ..the card
gives the remarkable ability (similar to an application accelerator but much better
without propriety limitations) ..to Dramatically improve overall data transfer
speed system wide; actually W2K & WXP show quite an improvement as well.
I can't say enough good things about it.
I recommend the latest version of the Promise ULTRA133 TX2, and here's just
one of many places to buy one:
http://www.newegg.com/Product/Product.asp?Item=16-102-007
...of course needs an available pci slot...
And so I see you like testing Steve, then give this one a try and.. well, you'll
see.. ..and then come back and tell us about it :)
Rick
John John
> I recommend the latest version of the Promise ULTRA133 TX2,...
I agree, that is an excellent choice for a PCI card controller. A good
way to give a significant boost to a tired pc.
John
Noel Paton
see you attempting to correlate two totally dissimilar systems???
Why are you surprised that they performed differently?
--
Noel Paton (MS-MVP 2002-2006, Windows)
Nil Carborundum Illegitemi
http://www.crashfixpc.com/millsrpch.htm
Please read http://dts-l.org/goodpost.htm on how to post messages to NG's
"SteveSch" <THIS_...@IS.FAKE> wrote in message
news:pan.2005.12.24....@IS.FAKE...
cquirke (MVP Windows shell/user)
>SteveSch wrote:
PMFJI...
>I noticed the same differences with extra RAM too and especially when large size
>processes are needed like you do, and the information can be drawn direct from RAM
>instead. We spoke of that earlier in the thread you are replying to; however,
>what would increase that even 'more dramatically' on your 'desktop' computer is if
>you used a PCI controller card.
Whoa! Do you mean this PC is running HD off an ISA card???
Let's get back to basics and build forward from there (sorry if I
re-hash old ground, I am jumping in late I know...)
Are you kept waiting by a peripheral or other component, e.g...
- dial-up networking
- slow storage device e.g. CD, 1.44M
- other device communications
...? If so, address that problem.
Is the HD LED off while you are waiting? If so, then you aren't
hitting disk, and adding more RAM is unlikely to help what may be a
raw computational processing power problem. Forklift upgrade?
OK, let's assume you've filtered out those things and are left with
things that spill over to HD if they can't fit in physical RAM.
Unless you have an atypical situation where the task is hard-coded to
be fitted into RAM so that paging isn't relevant, you have two
approaches to spend less time waiting for HD:
1) Hit the HD less often
This is the premise on which "add more RAM for speed" is based. A
task that fits purely in RAM and does not require any swapping to disk
will be far faster, and having more RAM to act as disk cache will help
too, though eventually the law of diminishing returns will set in.
How to tell when further gains will be poor? A good way to perdict
would be to use System Monitor to watch "Swap file in use". If that
is zero, no further gains from more RAM can be expected unless these
stem from more effective disk caching.
2) Make the HD cheaper to hit
You do this by:
- using a faster HD
- using a faster HD interface
- using partitioning to reduce head travel
- ensuring the HD has no failing sectors (retries)
- unlinking system vs. data access by using two HDs
- storing more data per cylinder to further reduce head travel
- avoiding long, slow-to-traverse subdirectory chains
- defragging the file systems
You want to stop the heads traversing the entire file set from FAT and
OS code at the front to the far end where new files are being created.
The way to do that is to debulk the file set, by moving everything
that is large and seldom-used to a different volume on the same HD;
this is the key principle behind intelligent partitioning. I'd use a
small C: for OS, small code, swap file and temp, and ensure this has
4k clusters for more efficient paging; this keeps most head travel
within the first few cylinders of the HD irrespective of how much junk
you keep lying around. It's the best way to maintain speed.
If your application involves fat linear data access while the system
code hops around, then it may pay to unlink these two sets of disk
activity by locating the data on a physically separate HD. Then the
system HD can skip around to do its lightweight gulps, while the data
HD's heads can stay in a narrower band of travel and pass through a
fatter throughput of data.
Once you've curbed useless head travel by keeping most activity within
a small C: of fixed size, the next step is to have that fixed-size C:
volume on as few cylinders as possible. HDs generally have a similar
number of cylinders, using extra platters and boosted linear density
to get more data on each cylinder. So the larger the HD, the more it
gets on each cylinder, and the better this approach will work - hence
the paradox; big HD, small C:, high speed.
A sick HD will massively clobber preformance, beacuse each access may
involve multiple retries - so aside from the obvious data risk, you
should dump and replace any HDs with "just one bad cluster", or that
prompt Scandisk surface scans on startup, or that have dodgy looking
details in SMART even if the SMART summary is "OK". Do a DOS mode
surface scan, and chuck the HD if the progress counter visibly pauses
while scanning and this cannot be explained by CPU thermal retreat.
The FATxx file system uses linear lookup when locating files in a
directory, so the more entries in a directory there are, the worse the
performance. This is a big problem with temp files and web cache,
where the entire directory has to be checked to ensure a new file will
not clash with a name that exists. A large slow-growing directory
will also be fragmented, increasing the HD travel to traverse it, and
widing the critical window during which file system corruption may
occur. So you want to shrink web caches, purge these and the Temp
location, and then defrag so that the directory chains shed the old
entries and are small and contiguous again.
Only after doing the "free" stuff, such as paurging temp, defragging,
ensuring HD isn't failing, and perhaps redoing your partitioning,
would I look further to replacing HDs and/or controllers.
If your HDs are puny and slow, then head travel is going to clobber
you; fix that first. If you have already done that, so that head
travel is already minimized, only then would I look to improving raw
data throughput by ensuring I have at least a UDMA vs. PIO mode to
free up the CPU. From there, I might consider RAID 0 if I wanted to
fatten up that data per cylinder beyond what is possible in one HD.
But I'd be hesitant to over-spend on a Win9x PC, especially when it
comes to adding RAID 0 controllers. In particular, there is no point
in using a hero UDMA-133 add-on controller card if you already have
PCI HD controller and the HDs are so old they are only cabable of PIO
or UDMA-33 anyway. Note: Any mobo with PCI slots and integrated HD
controllers is already going to have PCI interface to IDE.
>--------------- ----- ---- --- -- - - -
First, the good news: Customer feedback has
been clear and unambiguous.
>--------------- ----- ---- --- -- - - -
Rick Chauvin
..or even a brand new one since It's not a question of being tired or not that
directs its advantages.
Rick
>
> John
Rick Chauvin
> On Sat, 24 Dec 2005 15:23:07 -0500, "Rick Chauvin"
>>SteveSch wrote:
>
> PMFJI...
>
>>I noticed the same differences with extra RAM too and especially when
>>large size processes are needed like you do, and the information can be
>>drawn direct from RAM instead. We spoke of that earlier in the thread
>>you are replying to; however, what would increase that even 'more
>>dramatically' on your 'desktop' computer is if you used a PCI controller
>>card.
Hi Chris,
I think you are speaking to Steve I'm sure with your reply underneath although for
whatever reason you have my quoted text on top there and replied to it. I myself
didn't concentrate on all those specifics Steve was outlining since he was talking
about different systems and I didn't concentrate or get their connections since it
was a first read confusing, and so my response was in generally saying I surly notice
a difference with 1 GB of RAM verses less - and that's a given. About using the ATA
PCI Controller card, well for most systems in the last 5 years anyway, using one
makes a dramatic difference in overall across the board system transfer speed, and
that's a given as well and not even debatable really - it's a no contest.
I'll leave all those other questions you are directing towards asking Steve's for
him.
Rick
SteveSch
> It may be alcohol interfering with my cognitive senses here - but did I not
> see you attempting to correlate two totally dissimilar systems???
>
> Why are you surprised that they performed differently?
I was surprised at how much faster the system ran with an extra 1 Gig RAM.
Started slower than a dual 750MHz in some areas then was faster after the
RAM. I was surprised at how fast a dual 750 MHz is compared to a Sempron
2800 for some tasks.
I was not surprised they performed differently.
Steve
SteveSch
wrote:
> Whoa! Do you mean this PC is running HD off an ISA card???
ECS KT600-A onboard IDE.
Running Linux at this time.
Steve
SteveSch
> I recommend the latest version of the Promise ULTRA133 TX2, and here's
> just one of many places to buy one:
> http://www.newegg.com/Product/Product.asp?Item=16-102-007
> ...of course needs an available pci slot...
I have an older Promise Ultra100 TX2 that I have played with. Needed it on
my old machine to use a larger hard drive. Haven't played with the 133s.
Steve
SteveSch
> ..or even a brand new one since It's not a question of being tired or not that
> directs its advantages.
My system has Ultra DMA 133 and SATA. Would the card be faster than the
onboard 133 too? How about the SATA?
MB=ECS KT600-A
Steve
cquirke (MVP Windows shell/user)
>> On Sat, 24 Dec 2005 15:23:07 -0500, "Rick Chauvin"
>>>SteveSch wrote:
>>>...what would increase that even 'more dramatically' on your
>>>'desktop' computer is if you used a PCI controller card.
>Hi Chris
Hi!
>I think you are speaking to Steve
In newsgroups, I speak to issues first, with scant regard to who is
writing about them - it's a "kick the ball, not the player" thing :-)
>I didn't concentrate on all those specifics Steve was outlining since
>he was talking about different systems ... so my response was in
>generally saying I surly notice a difference with 1 GB of RAM verses
>less - and that's a given.
I would not call that a given, if I didn't know anything about the OS
and application load - especially with Win9x. It's quite possible to
go past the point of diminishing returns, where RAM goes. It's even
possible to lose performance if the hardware fails to cache the extra
RAM and the load is light enough to fit in smaller RAM without paging.
>About using the ATA PCI Controller card, well for most systems
>in the last 5 years anyway, using one makes a dramatic difference
>in overall across the board system transfer speed, and that's a
>given as well and not even debatable really - it's a no contest.
Once again, I'd disagree; I can't see how that can work as a blanket
recommendadtion if you have no idea what the existing controller is,
and what "PCI" card you are supplanting it with.
If you do add on a separate card, you may need special drivers to
support the device, or (in the case of NT-based OSs that do not start
from a DOS-on-BIOS boot phase) even boot the OS. I would have to have
very, very compelling reasons to create this extra fragility,
especially in the XP era, where doing so may break compatibility with
Bart CDR-booted maintenance OS.
There are 6 possible advantages to adding a controller card:
1) Resolution of compatibility issues
2) Additional device connections
3) Faster bus interface
4) Faster transfer modes
5) On-board cache RAM
6) RAID
1) is the biggie, e.g. where system BIOS suffers an addressability
limitation that locks the system when "too large" HD is detected.
This doesn't apply at the 512M barrier, so you can use an MBR-based
DDO instead, but at the 32G barrier, the system locks up before MBR is
processed, so an add-on card is the only way to get around the problem
if there's no suitable BIOS upgrade available.
2) used to apply in the 386 and early 486 era, when IDE CD-ROM drives
started coming out, but standard IDE controllers were still
single-channel. Usually, the sound card provided the CD-ROM
interface, from when this was proprietary, and later sound cards had
an extra IDE connector as well. More often than not, the reverse
problem would arise; the mobo already has primary and secondary IDE,
and the extra one on the sound card either clashes, or is set up as
"tertiary" or "quaternary", usually to be ignored by the OS.
3) applied in the VL-bus era, midway through the 486 generation,
before motherboards started including IDE controllers etc. For a
while, VL-bus systems were often built with ISA IDE controllers for
reasons of cost containment, so upgrading to a VL-bus controller made
sense. The built-in controllers were always VL-bus, and by the time
PCI arrived, the inclusion of PCI IDE controllers in the motherboard
was standard, removing needs (1) to (3).
4) started in the Socket 7 days, when either the i430VX or i430TX
debuted the first of several DMA modes. These allowed data transfers
between hard disk and RAM to be mediated by the IDE controller with no
ongoing processor involvement, thus freeing up the processor, in
contrast to the earlier PIO modes where the CPU transferred the data.
After the original DMA mode, came UDMA-66, -100 and -133, each of
which raised the limit on how fast data could be transferred between
hard drive unit and RAM. These limits were generally way ahead of
what could be transferred from the HD's disk platters, so the
real-world improvement was often negligeable; it did help smooth out
peak needs, reduce the demand on on-HD cache RAM, help when multiple
AT(API) devices had to share the same channel, and may have been
needed to get the most out of high-speed RAID.
If the HD is of the same vintage as the motherboard, then chances are
it will have a similar maximum UDMA support. It's only when you have
an old motherboard (and thus an old built-in IDE controller) and want
to use a newer HD that supports newer UDMA modes, that an add-on
controller begins to make sense.
Note that UDMA modes -66 and higher require an 80-pin IDE data cable
and are generally incompatible with removeable HD brackets.
5) is more useful with OSs that have poor inbuilt disk caching
ability. These controllers have dedicated RAM on board, that cache
the HD data at a hardware level. For recent versions of Windows,
adding system RAM is generally more cost-effective, especially as this
allows flexible load balancing between cache and program allocation,
as well as facilitating execution from within cache.
Today's HDs not only transfer the actual disk interface from system to
HD (that's what IDE means), they also contain dedicated cache RAM on
board. Old disk interface standards would have sent raw analog data
signals to the drive, so the equivalent use of cache RAM would have
had to be built into the controller instead.
6) is an obvious reason to add a HD controller card, though RAID 0
poses data risks and makes more sense when doubling the capacity
within the same HD is no longer cost-effective.
So I would certainly contend your assertion that blindly adding an
add-on HD controller will always produce a dramatic increase in
performance. If the built-in controller is limited to UDMA-33, and
the hard drive is limited to UDMA-33, I can't see how adding a
UDMA-133 card (with no hardware RAM caching, and not using RAID) is
going to make any difference at all.
In fact, once you get to today's systems, PCI is itself the limiting
factor. Historically, parallel busses were preferred to serial
arrangements because (at the modest data speeds of the time) they were
faster. Today we can pump serial data faster than parallel
arrangements can handle without unacceptable cross-talk or expense, so
the move has been to Serial-ATA and thus UDMA-150 and onwards. by
this stage, the transfer rate of the IDE interface is bottlenecked by
PCI, so modern motherboards interface the S-ATA ahead of the PCI.
Adding an S-ATA as a PCI card would walk back into that limitation.
So far, I have yet to need an add-on IDE controller for any post-PCI
system. Where RAID has been needed, it's been more cost-effective to
choose a motherboard that includes RAID, and typically I used Jetway's
i875P that has both native Intel S-ATA plus an additional Promise
S-ATA with RAID 0/1 integrated. I haven't needed RAID since moving
from 8xx to 9xx chipset generations.
>---------- ----- ---- --- -- - - - -
Don't pay malware vendors - boycott Sony
>---------- ----- ---- --- -- - - - -
John John
> There are 6 possible advantages to adding a controller card:
>
> 1) Resolution of compatibility issues
> 2) Additional device connections
> 3) Faster bus interface
> 4) Faster transfer modes
> 5) On-board cache RAM
> 6) RAID- - - -
> 7) ... Data transfer is passed on from the cpu to the card
controller, allowing the cpu to concentrate on other duties...
John
Rick Chauvin
In older setups the 133 may not give you that much over the 100 really if your
hardware is not geared with modern specs, but the cost is about the same and so if
buying new then obviously the best bet is using the 133. There would be no sense of
upgrading the current 100 on that setup unless you just desired to do so. The
updated 100's are a fine card and I even have a few, but have more 133's.
Rick
>
> Steve
Rick Chauvin
> My system has Ultra DMA 133 and SATA. Would the card be faster than the
> onboard 133 too? How about the SATA?
I have not tested SATA drives even though my boards support it.
My 133 card now runs Ultra DMA 6
I've stayed with PATA for now, but SATA is positioned and capable with future
developed to give better performance than the max'd out PATA development can give
now, but as of 6 months ago were not realized yet and were still performance wise the
same and the PATA's, and at that time were much cheaper per GB ...but the future will
be SATA yes, unless something else better takes the lead.
Rick
> MB=ECS KT600-A
>
> Steve
Rick Chauvin
> On Sun, 25 Dec 2005 14:47:45 -0500, "Rick Chauvin"
>>cquirke (MVP Windows shell/user) wrote:
>>> On Sat, 24 Dec 2005 15:23:07 -0500, "Rick Chauvin"
>>>>SteveSch wrote:
>
>>>>...what would increase that even 'more dramatically' on your
>>>>'desktop' computer is if you used a PCI controller card.
>
>>Hi Chris
>
> Hi!
>
>>I think you are speaking to Steve
>
> In newsgroups, I speak to issues first, with scant regard to who is
> writing about them - it's a "kick the ball, not the player" thing :-)
Yes that's right I remember you always told me that.
I've always told you though when you tried, that I'm not going to get in the ring
with you though to play kick the ball (tech talk) :) ...I'll concede now <g>
>>I didn't concentrate on all those specifics Steve was outlining since
>>he was talking about different systems ... so my response was in
>>generally saying I surly notice a difference with 1 GB of RAM verses
>>less - and that's a given.
> I would not call that a given, if I didn't know anything about the OS
> and application load - especially with Win9x. It's quite possible to
> go past the point of diminishing returns, where RAM goes. It's even
> possible to lose performance if the hardware fails to cache the extra
> RAM and the load is light enough to fit in smaller RAM without paging.
I personally would still call it a given for any computer from a P2 and up.
I'm not going to list details, only to say that using them on all sorts of
setups have proven easy setup and only positive results every single time.
>>About using the ATA PCI Controller card, well for most systems
>>in the last 5 years anyway, using one makes a dramatic difference
>>in overall across the board system transfer speed, and that's a
>>given as well and not even debatable really - it's a no contest.
>
> Once again, I'd disagree; I can't see how that can work as a blanket
> recommendadtion if you have no idea what the existing controller is,
> and what "PCI" card you are supplanting it with.
That's okay I respect your opinions always.
I stand by what I said though and don't think it's a complicated issue and the card
will pass all the criteria. The card takes over and does a so much better job that
even the latest motherboards do in their area of function. Drivers and 80 wire cable
for the card of course come with it! ..for instance look a the Image Viewer of it on
this website: http://www.newegg.com/Product/Product.asp?Item=16-102-007
..and download the pdf talking about the card from here: http://tinyurl.com/737gb
..it's compatible with almost everything mentioned. I'm not personally talking RAID
or SATA here either just a standard PATA setup; although I've heard only positive
things about the SATA version of the same controller card.
Rick
PCR
I believe he covered that in item 4...
cquirke (MVP Windows shell/user)
That's what UDMA modes are about, and there's no difference to thuis
being done by built-in controller or add-on controller - and this will
not happen if the HD does not support it, regardless of controller.
Within that, afe further nuances; faster modes that require 80-pin
cables, and the elimination of PCI itself as the bottleneck.
cquirke (MVP Windows shell/user)
>cquirke (MVP Windows shell/user) wrote:
>>>I didn't concentrate on all those specifics Steve was outlining since
>>>he was talking about different systems ... so my response was in
>>>generally saying I surly notice a difference with 1G of RAM verses
>>>less - and that's a given.
>> I would not call that a given, if I didn't know anything about the OS
>> and application load - especially with Win9x. It's quite possible to
>> go past the point of diminishing returns, where RAM goes.
>I personally would still call it a given for any computer from a P2 and up.
>I'm not going to list details, only to say that using them on all sorts of
>setups have proven easy setup and only positive results every single time.
I'd see 1G on a PII-era PC as absurd overkill, both in terms of rarely
being useful and in difficulty in finding that much ancient-spec RAM.
It will get you into the territory where Win9x's scaling assumptions
fall apart and manual settings start to be required.
On my own hard-working Win98SE PC - abandoned a few years ago only
because it died - I'd change from 256M down to 128M and back,
depending on whether I needed the 128M DIMM somewhere else. Frankly,
I could hardly tell the difference. This is on a PC unconstrained by
CPU and HD; 865G chipset, Celeron 1.8G, 120G HD partitioned nicely.
I find slow PCs with heroic RAM something of a pain to maintain,
because RAM testing (MemTest86, etc.) takes forever. It's nice if you
want to push the OS envelope (e.g. Pentium 200 running WinME) or if
you want to run a contemporary NT, as that always needs more RAM (and
is better at using this more effectively).
>>>About using the ATA PCI Controller card, well for most systems
>>>in the last 5 years anyway, using one makes a dramatic difference
>>>in overall across the board system transfer speed, and that's a
>>>given as well and not even debatable really - it's a no contest.
>> Once again, I'd disagree; I can't see how that can work as a blanket
>> recommendadtion if you have no idea what the existing controller is,
>> and what "PCI" card you are supplanting it with.
>That's okay I respect your opinions always.
>I stand by what I said though and don't think it's a complicated issue and the card
>will pass all the criteria. The card takes over and does a so much better job that
>even the latest motherboards do in their area of function. Drivers and 80 wire cable
>for the card of course come with it!
Actually, the detail matters. If the card does exactly the same
thing, constrained as it is by other factors (e.g. the capabilities of
the HD) then there's likely to be no benefits whatsoever, just the
downsides of cost, lost PCI slot, extra current draw, and additional
fragility factor due to the driver dependence.
Perhaps you are assuming the context of adding a newer HD to an old
system? If so, then sure, there may be benefits such as faster
transfers or CPU relief (the latter being welcome with slow CPUs), or
even the necessity of supporting higher capacities.
Today, the risk is of finding the new PCI card doesn't have Win9x
driver support. Definitely something to check before buying.
There's nothing inherently magical about a HD controller that happens
to be a seperate card rather than part of the motherboard. There's a
disadvantage that BIOS and CMOS won't "know" the card, and you may
lose some UMB if the card's ROM has to be patched into the memory map.
But you'd want that if the intention is to fix a HD address limit.
Some technology boosts look good on paper, but don't translate well in
the real world. Examples include...
1) Faster RAM designs
In theory, boosting the transfer speed of RAM ('normal' to EDO, EDO to
SDRAM, SDRAM to DDR, DDR to DDR2) should make a big difference, but in
practice, most RAM access is handled in L1 and L2 caches (and the
first amount of these caches does most of the work) and what counts is
avoiding HD access by boosting the amount of RAM, not the speed.
So while a new RAM type is silly-priced, to the extent that you can
get double the capacity of "old" RAM for the same price, it's a
badidea if building to budget - though it makes sense if you really
need all the speedyou can get, and you've already maxed the capacity.
Once the price offset goes away, then it's a no-brainer - not because
the performance boost is compelling, but because future upgrades will
be cheaper and later repairs will be possible (while the old RAM may
no longer be available).
2) Faster HD transfer modes
A single HD is generally not constrained by the speed at which data
can be transferred from the HD unit to the RAM. The limiting factor
is the speed at which the data comes off the disk platters.
However, modern HDs will use on-board cache RAM to decouple data flow
from platters to transferring this from RAM, on a store-and-forward
basis much as printer drivers do between applications and printers. A
faster DMA mode can squirt the data through in a shorter time window,
and thus have less impact on processing - even if DMA reileves the CPU
of intimate involvement, the CPU may still be locked out of the RAM
being managed until the transfer is complete.
Once you place additional devices on the same channel, such as JABOD
or RAID, then the notional benefits of faster transfer modes become
more likely to apply in practice. I think HD manufacturers like
faster transfer modes too, because it can relieve pressure on the
on-drive cache; adopting a new mode may be cheaper (and a more
sellable checkbox feature) than boosting on-drive RAM capacity.
Historically, the same has sometimes unexpectedly applied to faster
busses and CPU details. We found that MMX and Local Bus were far less
effective at boosting SVGA and 3D speed in Windows, compared with
accelerator chipsets. We found that adopting VL-bus on a 386DX system
made little difference, as there was no 486 burst mode to "fill the
pipe". And initial VL-bus IDE cards didn't demonstrate the boost in
HD performance that we expected, as tested with Checkit.
Having said that, PCI may be the limiting factor with UDMA modes
beyond 133, as applied via Serial-ATA. This may well cause PCI
controller cards to be slower than integrated S-ATA that bypasses the
PCI at the motherboard chipset level.
I wonder when optical drives will get off IDE (ATAPI) and onto S-ATA?
Or will these use a different bus in the future, such as USB 2.0?
Having to juggle S-ATA vs. IDE is a PITA at times.
Rick Chauvin
Ah, I have the D865PERLL board myself, 800 FSB and all, actually have two of them one
in each computer, coupled with a 3.2c GHz and a 3.4e processor, and a pair of 1 GB of
DDR PC3200 RAM of course. Anyway, I find that compared to just using a 512 stick
verses a 1 GB stick there is a big difference when it comes to certain tasks. I know
you can't see my earlier post in this thread (I wish you could darn it) ...but I was
telling another that for me I have noticed in specific 'important to me instances'
where using 1GB performance wise it's quite noticeable as compared to using less. It
comes into play for one instance when I use my software reg/file trackers taking full
system file/registry comparison snapshots where on much less ram the procedure takes
much much longer....Once a first compare snapshot is done and retained in memory then
with installation/changes made - and for a second compare takes just 10 seconds since
the large RAM can hold and compare all the info 'in hand' and RaPiDlY process the
resulting differences within itself, whereas using much less ram does not have enough
room for that full process and has to rely on reprocessing all the information again
instead; and so I say 1GB of ram can make a big difference 'if' of course you have a
process that can utilize it.
> I find slow PCs with heroic RAM something of a pain to maintain,
> because RAM testing (MemTest86, etc.) takes forever. It's nice if you
> want to push the OS envelope (e.g. Pentium 200 running WinME) or if
> you want to run a contemporary NT, as that always needs more RAM (and
> is better at using this more effectively).
right
>>>>About using the ATA PCI Controller card, well for most systems
>>>>in the last 5 years anyway, using one makes a dramatic difference
>>>>in overall across the board system transfer speed, and that's a
>>>>given as well and not even debatable really - it's a no contest.
>
>>> Once again, I'd disagree; I can't see how that can work as a blanket
>>> recommendadtion if you have no idea what the existing controller is,
>>> and what "PCI" card you are supplanting it with.
>
>>That's okay I respect your opinions always.
>
>>I stand by what I said though and don't think it's a complicated issue
>>and the card will pass all the criteria. The card takes over and does a
>>so much better job that even the latest motherboards do in their area of
>>function. Drivers and 80 wire cable for the card of course come with it!
>
> Actually, the detail matters. If the card does exactly the same
I know details matter :) I smartly save the details for you though <g>
meaning since you have the brain power and hi iq aptitude for such things.
I may be slower, but slow but sure.... :)
> thing, constrained as it is by other factors (e.g. the capabilities of
> the HD) then there's likely to be no benefits whatsoever, just the
> downsides of cost, lost PCI slot, extra current draw, and additional
> fragility factor due to the driver dependence.
The hard drive does not really matter, well sure the old 5200 rpm half mb cache
drives are going to be slower, but it's all relevant, and as far the controller card
is concerned it does not care about the hard drive; however, yes if you have a modern
7200 rpm drive with a 2 mb cache, sure, it will be faster, and that will translate to
a faster turn around for the controller card in action too, but as I said before it's
all proportionally relevant, but the controller card does not care about the hard
drive and works it's magic proportionally irregardless and just steps up to the plate
with a faster hard drives and chimes in too......)
> Perhaps you are assuming the context of adding a newer HD to an old
> system? If so, then sure, there may be benefits such as faster
> transfers or CPU relief (the latter being welcome with slow CPUs), or
> even the necessity of supporting higher capacities.
No I'm not assuming that context, and will say again, even if you leave the old puny
HD and the small feeble RAM in place, there will Still be Significant gains in the
overall data transfer rate with the controller card irregardless, but it will be a
little less so and proportional as I said.
> Today, the risk is of finding the new PCI card doesn't have Win9x
> driver support. Definitely something to check before buying.
No risk, the link I gave previously to the PROMISE ULTRA133 TX2 PCI IDE Controller
Card ...is for all versions of Win x's :)
> There's nothing inherently magical about a HD controller that happens
> to be a seperate card rather than part of the motherboard. There's a
> disadvantage that BIOS and CMOS won't "know" the card, and you may
> lose some UMB if the card's ROM has to be patched into the memory map.
> But you'd want that if the intention is to fix a HD address limit.
I have never found that to be true though and never had any problems, the card is
designed to accommodate all and is backward compatable to meet the past.
> Some technology boosts look good on paper, but don't translate well in
> the real world. Examples include...
Just to jump in here about the ccard, I must say I personally don't care about the
specifications boast about the 133 ccard or even what the earlier 100 with what its
specs say about its performance, I know it's not 'real world' with those numbers,
but, what I do care about is that even at 70% of their numbers - the performance is
awesome indeed!
I like that mine runs Ultra DMA 6
> controller cards to be slower than integrated S-ATA that bypasses the
> PCI at the motherboard chipset level.
You are a good write Chris and that's for sure.
I can only tell you how good the controller card works, which I did.
Now I think you should go out and buy one for your setup to see for yourself <vbg>
> I wonder when optical drives will get off IDE (ATAPI) and onto S-ATA?
> Or will these use a different bus in the future, such as USB 2.0?
> Having to juggle S-ATA vs. IDE is a PITA at times.
I see lots of people run both PATA and SATA drives simultaneously on the same MB &
setup with no problem.
I specifically bought a couple of my D865PERLL boards I use now for that very purpose
because I already have SATA ports that came right on my MB (although here again I
would use the Promise SATA PCI Controller Cards in it's place, mostlikely ...but as
of yet I have not got into testing any SATA's yet, but the day is fast approaching.
gotta run,
supper is ready,
Rick