How much RAM for Windows 95?
Jonathan Sachs
performance is pretty dismal. I'm seeking advice on whether to spend
about $400 and upgrade to 16 MB, or spent $700 and upgrade to 24 MB.
Since laptop memory is so expensive, I don't want to get more than I
need. On the other hand, the processor and disk are relatively slow to
start with, so I probably suffer more from excessive swapping than I
would on a desktop machine with the same amount of RAM.
I use the laptop mainly for writing, bookkeeping, and communication.
Due to screen size limitations I'm not likely to do much programming
on it no matter how fast it is.
- -
Jonathan Sachs
Sand River Software, Inc.
70047...@compuserve.com
Dean Cashen
>I'm running Windows 95 on a 486/75 laptop with 8 MB of RAM, and
>performance is pretty dismal. I'm seeking advice on whether to spend
>about $400 and upgrade to 16 MB, or spent $700 and upgrade to 24 MB.
.. . .
>I use the laptop mainly for writing, bookkeeping, and communication.
>Due to screen size limitations I'm not likely to do much programming
>on it no matter how fast it is.
For your use, 16MB should suffice, possibly just 12MB total will
alleviate the swapping. From personal experience on two laptops (A
P90 and a 486SX/25), it seems that 8MB is a terrible number for Win95
due to swapping, but moving to 16MB will do *wonders* for performance.
From there, adding more memory helps only incrementally, and any gain
you see will be dependent on the number and type of applications you
use. For the applications you've described, I don't believe you'll
benefit much in going to 24MB. In my case, I'm running a P90/8MB
system right now, but have ordered 16MB additional. I could get by
with 8MB more, but, in my case, I'll be doing work that will benefit
from the extra 8MB.
More is almost always better, but I think you'll be happy going with
the 8MB upgrade in place of 16MB.
Ken Wright
wrote:
>jona...@well.com (Jonathan Sachs) wrote:
>
>>I'm running Windows 95 on a 486/75 laptop with 8 MB of RAM, and
>>performance is pretty dismal. I'm seeking advice on whether to spend
>>about $400 and upgrade to 16 MB, or spent $700 and upgrade to 24 MB.
>>
>>Since laptop memory is so expensive, I don't want to get more than I
>>need. On the other hand, the processor and disk are relatively slow to
>>start with, so I probably suffer more from excessive swapping than I
>>would on a desktop machine with the same amount of RAM.
>>
>>I use the laptop mainly for writing, bookkeeping, and communication.
>>Due to screen size limitations I'm not likely to do much programming
>>on it no matter how fast it is.
I just upgraged my 486/75 to 16 megs and it made a bunch of
difference......so, that's what I would recommend.
--
Ken Wright
k...@primenet.com
http://www.primenet.com/~kw/
Paul Davison
>I'm running Windows 95 on a 486/75 laptop with 8 MB of RAM, and
>performance is pretty dismal. I'm seeking advice on whether to spend
>about $400 and upgrade to 16 MB, or spent $700 and upgrade to 24 MB.
>
>Since laptop memory is so expensive, I don't want to get more than I
>need. On the other hand, the processor and disk are relatively slow to
>start with, so I probably suffer more from excessive swapping than I
>would on a desktop machine with the same amount of RAM.
>
>I use the laptop mainly for writing, bookkeeping, and communication.
>Due to screen size limitations I'm not likely to do much programming
>on it no matter how fast it is.
All the reviews I've read show the 8M to 16M transition as being the most bang
for your buck.
I went from 8M to 24M just because when I bought RAM there was only 4M or 16M
chips available for my laptop, and I wasn't about to fill my last memory slot
with only a 4M chip.
I'd say you'd probably be quite happy with 16M.
---
Paul Davison
PDav...@SpryNet.com
Dean Cashen
>Dragon Lord wrote:
>>
>> On Thu, 07 Mar 1996 21:22:34 GMT, jona...@well.com (Jonathan Sachs) wrote:
>>
>> >I'm running Windows 95 on a 486/75 laptop with 8 MB of RAM, and
>> >performance is pretty dismal. I'm seeking advice on whether to spend
>> >about $400 and upgrade to 16 MB, or spent $700 and upgrade to 24 MB.
>> >
>> For the applications that you are doing I would suggest the 16MB
>> option. The only reason you would want memory above that is if you are
>> running large graphics programs or many applications at once. I have 32
>> MB in my laptop and it runs well with 6 big applications. If you plan on
>> doing something like this then go for the 24. Win 95 will be sure to suck
>> it all up.
>What I'd add to this is the caution that while you may not see
>"excessive" swapping in 16 MB you might see enough swapping to keep
>your hard drive spun up and thus cut into your battery life, if that
>matters to you.
This is a good point but, of course, it's still a matter of degrees;
the only way to minimize swapping *fully* is to have a no disk and
sufficient RAM (1.6GB) to store and support all apps. There are
inherent problems with that as well, of course.
> . . . More RAM cuts down on disk activity, as does making
>your swap file a fixed size, and setting a minimum disk cache size
>to leverage the extra memory you have.
I'm really confused on the practice of fixing swap size; how does this
affect, or minimize, Win95 swapping? I tried it myself, and found no
empirical gain in performance or swap minimization. Beyond that, it
just doesn't make sense to me; Win95 will still swap when it needs,
and if it runs out of SPACE, swaps will FAIL, but they won't be
avoided.
> . . . I just went from 12 to 20MB
>of RAM on a Toshiba T3400 where I usually run Netscape, Word and
>Agent (a newsreader)-- usually not all three, but at least two-- and
>I noticed a marked dropoff in disk activity.
Ditto. Except I went from 8MB to 24MB, and the difference is worth
every penny. Of course, Win95 gets along with 8MB like Lex Luthor
gets along with Superman; conflicts abound. You went from a much
better level of memory than 8MB -- 12MB, in my experience, helps a
great deal. AND, you went to a level that exceeds the Pretty Good
level of 16MB. In short, you went from Good to Grand. That differs
from going from Bad to Good, as it is when you move from 8MB to 16MB.
Your point is well-made, that "more is better" when it comes to RAM.
But, from a bang-per-buck perspective, I believe you get the most gain
going from 8 to 16MB.
>Your mileage, of course, may vary.
>-Antonio Romero aro...@ix.netcom.com
Antonio Romero
>
> > . . . More RAM cuts down on disk activity, as does making
> >your swap file a fixed size, and setting a minimum disk cache size
> >to leverage the extra memory you have.
>
> I'm really confused on the practice of fixing swap size; how does this
> affect, or minimize, Win95 swapping? I tried it myself, and found no
> empirical gain in performance or swap minimization. Beyond that, it
> just doesn't make sense to me; Win95 will still swap when it needs,
> and if it runs out of SPACE, swaps will FAIL, but they won't be
> avoided.
The trick is this: the swap file is ordinarily dynamically resized,
based on current/recent demand. Win95 increases swap file size
when you need it, which is great... but when you stop using it,
it may go back a little while later and shrink it-- which is also great
if you're low on space but which generates extra activity.
If Win95 sees your swap file is 32MB and you're only using half of it,
it will spin up the drive and reclaim some of the space while you're
not using it. If the size is fixed, it won't do this. You save
*some* (probably not that many) disk accesses. Likewise, fixing the
disk cache size up around 4-6 MB can help, though there's a point at
which you risk swapping more if you use too much RAM for disk cache.
Time permitting I'd love to figure out what's optimal for my application mix.
> Your point is well-made, that "more is better" when it comes to RAM.
> But, from a bang-per-buck perspective, I believe you get the most gain
> going from 8 to 16MB.
>
Agreed. (I'd have gone to 16 if I could, but my box has 4MB on the
motherboard, and a socket that takes a single 4, 8 or 16MB chip.
Thus 20MB.) 24MB would let you devote 8MB to disk cache, which
combined with a few other tricks should cut down disk activity
significantly. I have a coworker who's running a Thinkpad 760 with
40MB of RAM (which I think is excessive) and he says he can run
Netscape plus Word plus Excel and not hit disk... (though when I
look at what the machine costs, I figure I'd just as soon buy a
Kia Sportage...)
-Antonio Romero aro...@ix.netcom.com
Dean Cashen
: Dean Cashen wrote:
[deleted for brevity]
: > I'm really confused on the practice of fixing swap size; how does this
: > affect, or minimize, Win95 swapping? I tried it myself, and found no
: > empirical gain in performance or swap minimization. Beyond that, it
: > just doesn't make sense to me; Win95 will still swap when it needs,
: > and if it runs out of SPACE, swaps will FAIL, but they won't be
: > avoided.
: The trick is this: the swap file is ordinarily dynamically resized,
: based on current/recent demand. Win95 increases swap file size
: when you need it, which is great... but when you stop using it,
: it may go back a little while later and shrink it-- which is also great
: if you're low on space but which generates extra activity.
: If Win95 sees your swap file is 32MB and you're only using half of it,
: it will spin up the drive and reclaim some of the space while you're
: not using it. If the size is fixed, it won't do this. You save
: *some* (probably not that many) disk accesses. . . .
I can see that, but I agree it just can't be a heck-of-a-lot of savings
in disk access. Trivial, in fact. Have you (or anyone) actually
measured any improvement in performance when running with a fixed-
size swapfile instead of the dynamic file?
: . . . Likewise, fixing the
: disk cache size up around 4-6 MB can help, though there's a point at
: which you risk swapping more if you use too much RAM for disk cache.
: Time permitting I'd love to figure out what's optimal for my application mix.
Hey, when you do, e-mail me! Tuning caching parms is never that
much fun. . .
Thanks
Isaac Wong
In article <DoJos...@apollo.hp.com>, de...@boi.hp.com (Dean Cashen) writes:
|> Antonio Romero (rom...@genmagic.com) wrote:
|> : Dean Cashen wrote:
|> : If Win95 sees your swap file is 32MB and you're only using half of it,
|> : it will spin up the drive and reclaim some of the space while you're
|> : not using it. If the size is fixed, it won't do this. You save
|> : *some* (probably not that many) disk accesses. . . .
|>
|> I can see that, but I agree it just can't be a heck-of-a-lot of savings
|> in disk access. Trivial, in fact. Have you (or anyone) actually
|> measured any improvement in performance when running with a fixed-
|> size swapfile instead of the dynamic file?
|>
|> : . . . Likewise, fixing the
|> : disk cache size up around 4-6 MB can help, though there's a point at
|> : which you risk swapping more if you use too much RAM for disk cache.
|> : Time permitting I'd love to figure out what's optimal for my application mix.
|>
|> Hey, when you do, e-mail me! Tuning caching parms is never that
|> much fun. . .
Perhaps I would throw in my $0.02. When my hard drive in the notebook powers
down, it is really annoying to have Win95 momentarily tweaks the swap
file which causes the hard drive to spin up again for a few seconds. I'd
rather have a fixed-size swap file so that my hard drive can remain down
for a longer period of time just to save battery juice.
For a system with less than 32MB of RAM, a disk cache of size 4-6MB is way
too big. For 8MB systems, a size of 512k to 1024k is ideal. For 16MB systems,
around 1024k should be adequate for most work. A disk cache too large will
only eat up valuable memory which causes more swap file activities.
--
Isaac Wong (613) 763-6127 | Protel Compiler Development
iw...@chat.carleton.ca | NORTEL Switching, Ottawa, Canada
isc...@engsoc.carleton.ca | Dept. of Systems & Computer Engineering
http://chat.carleton.ca/~iwong | Carleton University, Ottawa, Canada
Kenneth Crudup
>Perhaps I would throw in my $0.02. When my hard drive in the notebook powers
>down, it is really annoying to have Win95 momentarily tweaks the swap
>file which causes the hard drive to spin up again for a few seconds.
How are so sure it's "swapping", and not the buffer cache being committed?
-Kenny
--
Kenneth R. Crudup, Unix & OS/2 Software Consultant, Scott County Consulting
ke...@panix.com CI$: 75032,3044 +1 617 524 5929/4949 Home/Office
16 Plainfield St, Boston, MA 02130-3633 +1 617 983 9410 Fax
OS/2 box: pkenny.tiac.net (when I'm online) Get Warp-ed! OS/2 3.0 is here NOW!