More RAM necessary? (Kill the spinning wheel!)
 

I have a 600Mhz iBook with 384 MB (a 256 MB card on top of the built in 128MB). I see that the cost of a 512MB card is now quite cheap, but I'm wondering if it is worth it. Will I see much of a performance improvement by adding this additional RAM? Someone told me that each application rarely uses more than 200, but someone else says that it speeds up OSX to have as much RAM as possible. I feel the problem the most when I get a spinning wheel cursor a lot of the time, especially when switching to Mozilla, and I would like to never see the spinning wheel at all! I currently using 10.1.5 and plan to also upgrade to 10.2. I have also heard that this greatly reduces the spinning wheel, but I'm not sure if additional memory will help as well. TIA!

Try downloading MemoryStick from http://homepage.mac.com/WebObjects/F...cty=US&lang=en

Take a look at it, read about it (see Help menu:MemoryStick Help). Then, go to the Options menu, select "Signal Pageouts." This program might give you some idea of how much RAM you should have for your needs.

David

Kerim, I'm not sure if there is a speed hit for using a memory card/stick instead of a module you install into the machine...but I wouldn't operate Jaguar with less than 512MB of RAM.

Most people see consistent performance improvement up until the 700MB range and then it seems to taper off in my experience. Personally, I have a GB in my G4 - although I recognize that is a different situation in terms of ease of upgrade.

Also, more RAM seems to help people with machines running G3's or older G4's even more so than it does those with newer G4 hardware, so I think it would certainly be adviseable for you to consider investing in a memory upgrade.

Thanks. This explains a lot, and I am clearly having lots of pageouts. But what this doesn't tell me is whether OS 10.2 will solve this problem or whether I simply need more RAM - period. The only way to know may be to install 10.2 and then run this program again!!!???

Up to 700MB and still seeing significan performance improvements!? That is a lot. Anyone else confirm this?

BTW: MemoryStick is the name of a program, not really a memory stick. Memory sticks (like those from SONY) are completely unrelated and don't help iBook RAM.

Kerim: sorry for the confusion on the memory stick thing. I just know there are slot-loaded RAM devices for laptops, so I tried to encompass whatever might be out there.

The 700MB "rule" is just something I personally have noted over the last year or so. People seem to be very happy with the improvements they see up to that point, and then from there on, the improvemens in performance are more minor.

Keep in mind also that "significant" is a relative term. Very subjective. I'm mostly "paraphrasing" the collective comments I've heard about people upgrading their RAM and how it affected their OS X performance.

As far as page-outs / spinning beachballs due to heavy screen loads, how much 10.2 will help you depends not only on RAM but on whether or not your machine can utilize QuartzExtreme. I'm not familiar with the video cards in iBooks, but my guess would be not, since QE basically requires a 32MB card to operate efficiently.

I think RAM is your best bet for boosting OS X performance, regardless of which version you decide to use.

Just curious (I am thinking of installing OS X on an old G3 iMac with 384M of RAM), are you saying this because you are seeing nonzero pageouts with the top command, or are you concluding this from MemoryStick? It is not clear to me that MemoryStick distinguishes between pageins and pageouts, one reason I prefer Memory Monitor. Do you have multiple swapfiles in /var/vm?

Wish I could answer you, but I don't know ... I guess I don't pass the TOUFL!

Mainly I was asking what led you to say that you were "clearly having lots of pageouts."

To find out the number of swapfiles you have, type 'cd /var/vm' (without the quotes), then type 'ls'

on my machine, for example:

To find out the number of pageouts you have had since the last boot, just type 'vm_stat'

For example, I get:

Typing 'top' will give similar information. Normally, you prefer to see only one swapfile (I have three) and zero or only a few pageouts, whereas I have 77787.

I have over 1G of RAM installed, but I caused all these pageouts and swapfiles to be generated solely in the spirit of scientific enquiry by opening, all at once, a few hundred jpegs of, well, pictures. But if you generally see multiple swapfiles and pageouts with normal use, then you need more RAM.

I was just saying that because of what MemoryStick told me (it was chiming a lot).

I have 8 swapfiles!!!

swapfile0 swapfile2 swapfile4 swapfile6 swapfile8
swapfile1 swapfile3 swapfile5 swapfile7


And here are my VM Stats:

Mach Virtual Memory Statistics: (page size of 4096 bytes)
Pages free: 1287.
Pages active: 56900.
Pages inactive: 28417.
Pages wired down: 11700.
"Translation faults": 10300716.
Pages copy-on-write: 479535.
Pages zero filled: 5610486.
Pages reactivated: 884731.
Pageins: 285066.
Pageouts: 289431.
Object cache: 48909 hits of 616091 lookups (7% hit rate)


Again, this is in 10.1.5. Don't know what 10.2 will do. If it is true that most of the speed gain is from Quartz Extreem then it won't help me!

What do you think?

I think you should spring for more RAM. I don't believe 10.2 will have much effect on that. Adding 512M sounds like a good plan.

You will see an improvement with 10.2 - I'm doing OK, and due to some issues with OWC's returns dept., I'm only running with 64MB on my G4 Cube.

Well, Ok - until I run more than two apps at once, that is. The more RAM the better, is a good rule of thumb with OS 10.x - I agree with Aspiring G33K wholeheartedly. My experience has been the same as his. If you can fit a 512 in your iBook along side what's there, you'll definitely notice some differences.

All in all, I bet more RAM + 10.2 = a happier, faster iBook.

9KILLER, are you saying that 10.2 will affect the number of pageouts and the number of swapfiles generated?

What kind of issues, I was *just* about to hit the button to order a 512MB card from them (for $99) when I saw your post. I saw other sites offering the same thing for about $80, but I specifically was under the impression that OWC would have better support for Apple , and simply better support in general.

You folks are onto something that has been puzzling me.

Since upgrading to Jaguar on a TiBook w/512 mb of RAM I have noticed the vm swap files proliferating even without my activity on the computer. This is eating up half a gig to a gig of drive space a day. For example, last evening before going out, I restarted the computer and vm_stat and top showed the following.

[G4-PowerBook:~] dhough% vm_stat
Mach Virtual Memory Statistics: (page size of 4096 bytes)
Pages free: 49408.
Pages active: 21292.
Pages inactive: 46240.
Pages wired down: 14132.
"Translation faults": 1016819.
Pages copy-on-write: 109441.
Pages zero filled: 464599.
Pages reactivated: 0.
Pageins: 8939.
Pageouts: 0.
Object cache: 9639 hits of 26635 lookups (36% hit rate)


[G4-PowerBook:~] dhough% top

Processes: 57 total, 2 running, 55 sleeping... 173 threads 18:28:01
Load Avg: 1.30, 0.84, 0.69 CPU usage: 31.0% user, 23.9% sys, 45.1% idle
SharedLibs: num = 7, resident = 2.23M code, 172K data, 560K LinkEdit
MemRegions: num = 5906, resident = 108M + 15.1M private, 72.7M shared
PhysMem: 55.3M wired, 83.4M active, 184M inactive, 323M used, 189M free
VM: 2.00G + 3.62M 8954(0) pageins, 0(0) pageouts


First thing this morning I reran vm_stat and top and got the following.

Mach Virtual Memory Statistics: (page size of 4096 bytes)
Pages free: 1588.
Pages active: 77040.
Pages inactive: 38569.
Pages wired down: 13875.
"Translation faults": 17290457.
Pages copy-on-write: 2403894.
Pages zero filled: 7647042.
Pages reactivated: 94319.
Pageins: 10324.
Pageouts: 6859.
Object cache: 201001 hits of 218522 lookups (91% hit rate)


Processes: 57 total, 2 running, 55 sleeping... 174 threads 07:02:01
Load Avg: 1.18, 0.73, 0.54 CPU usage: 27.5% user, 20.0% sys, 52.5% idle
SharedLibs: num = 88, resident = 22.0M code, 1.77M data, 6.73M LinkEdit
MemRegions: num = 6835, resident = 317M + 15.0M private, 57.5M shared
PhysMem: 54.2M wired, 301M active, 150M inactive, 505M used, 6.53M free
VM: 2.21G + 61.1M 10341(1) pageins, 7048(23) pageouts

Doing nothing on the machine but leaving Mail and a browser open resulted in another swap file and a lot of paging. After just some web browsing the results were as follows.

Welcome to Darwin!
[G4-PowerBook:~] dhough% vm_stat
Mach Virtual Memory Statistics: (page size of 4096 bytes)
Pages free: 1588.
Pages active: 77040.
Pages inactive: 38569.
Pages wired down: 13875.
"Translation faults": 17290457.
Pages copy-on-write: 2403894.
Pages zero filled: 7647042.
Pages reactivated: 94319.
Pageins: 10324.
Pageouts: 6859.
Object cache: 201001 hits of 218522 lookups (91% hit rate)


Processes: 57 total, 2 running, 55 sleeping... 174 threads 07:02:01
Load Avg: 1.18, 0.73, 0.54 CPU usage: 27.5% user, 20.0% sys, 52.5% idle
SharedLibs: num = 88, resident = 22.0M code, 1.77M data, 6.73M LinkEdit
MemRegions: num = 6835, resident = 317M + 15.0M private, 57.5M shared
PhysMem: 54.2M wired, 301M active, 150M inactive, 505M used, 6.53M free
VM: 2.21G + 61.1M 10341(1) pageins, 7048(23) pageouts


Now I have six swap files with minimal work on the computer in less than one day.

Is this normal? I had been planning to increase RAM but will the same things be happening? Is this indicative of a bad install or other problems with the OS? I can't believe the OS is supposed to eat up gigabytes of drive space, or is it?

Appreciate any ideas. I did not see this type of paging via Memory Stick under previous versions of OS X. I’m wondering now if I should upgrade RAM or go for a larger internal disk drive to let the OS work efficiently.

dhough, my understanding is that one need not be alarmed by paging out from time to time nor even by the occasional generation of some extra swapfiles. After first reading up on it, I was under the mistaken impression that if you have an appropriate amount of RAM, paging out and extra swapfiles should never occur. I was assured, convincingly, that this is not the case, by, of all people, my plumber's adolescent assistant. I mentioned above that, even with more than 1 G or RAM, I will occasionally do something silly that generates lots of swapfiles. Tonight, by sheer coincidence, I did something exceptionally foolish and generated 32!

I am a bit surprised by what you describe. I would have thought that with 512 M of RAM you would not be paging out much and am very surprised that you would have six swapfiles from such ordinary use. Still, from what I gather, 512M seems to be regarded as enough RAM for OS X but probably less than an ideal amount. My guess is that it is more likely that you could use some more RAM than that something is actually wrong. Bear in mind, however, that even my plumber's assistant knows more about this stuff than I do.

Thanks everyone. I decided that no matter what it will be worth while to get the extra memory. I did some more research on OWC and they seem to have excellent reviews at Bizrate.com, so I went ahead and bought their 512MB memory upgrade for $99 ... witht that and 10.2 I hope things will improve! If they don't I'll be sure to let you all know.

Geez. Plumbing and UNIX - someone will never be out of work!

Smart fellow that kerim is.
;)

MemoryStick also seems to graphically display the unreleased "active" and especially "inactive" RAM that applications don't totally release when quit. My 1GB of RAM often grows very close to getting nearly fully maxed-out after using/quitting PhotoShop and especially VPC. I used to log-out to clear the RAM until discovering that simply launching OmniDisk Sweeper on the OSX volume clears inactive RAM useage to only running apps, in the backround, while still working. It seems the bar graph is accurate because I rarely get ANY pageout signals reported by either MemoryStick or Perfboard (0 page-in or page-outs after several days uptime) with my method. Additionally, I very rarely see any beachballs (outside of PS7) even with my lowly G3 400 running OS10.1.5.

Astonishing. I started up Memory Monitor, then OmniDisk Sweeper, swept the startup drive, and watched the graph of free memory increase, just like you said.

How does that work?!

bruce, it's true, there are several facilities that will "squish" the inactive pages.

in the shell, a find or du over a large section of the filesystem will peel off "expired" inactive pages.

seems like large disk cache i/o triggers virtual memory management of "expired" inactive pages.

Wouldn't this account for the incredible disparities of OSX performance perception/actual results where some like myself are very happy with OSX on a G3 400 and others complain about constant beachballs wih DP towers? Maybe they have little or no free RAM left and are actually running OSX from disk virtual memory! I have seen VPC gobble up +400MB of inactive RAM in use and release less than 100MB when quit. Relaunch again and your +1GB RAM is gone.

I always closely monitor free RAM and quit idle apps or scan the volume as needed to keep performance consistently high. I'm serious when I say I hardly ever see any beachballs.

I have seen it written many places that one does not need to quit idle apps on OSX. Are you saying that this is incorrect?

With my modest system, I can see RAM useage grow with SOME apps such as VPC (maybe an extreme example) simply sitting idle consuming huge amounts of free RAM. If I launch PhotoShop and use it, I can observe 1GB of RAM disappear because VPC is sitting idle. Play with MemoryStick, it takes up little desktop real estate and is VERY revealing about on-going all-important RAM useage. The promise of memory management in OSX and the reality (as least as of 10.1.5) are not the same apparently.

perception is a real oddity, influenced by mythology, hearsay and uncorroborated experiences. add to that the lack of knowledge of phenomena you find sprinkled around and it seems to me that a lot of people philosophically need a change of diapers.

you don't need to quit idle apps, but it sure affects memory management which improves overall performance over time; if you have a picture of memory and you can alter it by dispensing with idle apps, you are proactively managing your performance. idle they may be, but they have their fingers in resources that need managing by the kernel.

Merv,

Another example of cutting through the voodoo, I found was the simple concept that with VPC, the smaller the virtual hard drive image size, the faster Windows performed regardless of what version used. My 98SE flew when the drive was stripped down to an essential 350MB and was a pig when loaded with apps and became a 1.5GB drive image. Every app added cut speed just a little bit. The idea seemed rudimentary to me, after all doesn't a 300MB file take longer to open, scroll or alter than a 3MB file, which all a Windows OS is to the Mac, a file folder to open and search through? I thought, why not have multiple copies of a VPC Windows OS, each tailored to only the essential app needed for each drive image for the purpose of speed.

Well, you would think I committed heresy when I posted the concept at the Connectix VPC forum. One of their head engineers replied it was "one of the craziest ideas" he had heard of. I didn't bother to reply and suggest he simply try it and see for himself. Yet there are posters there who can't get Windows XP to even launch IE in VPC in less than 5 minutes on their DP towers, while I get IE pages to load in XP at about dial-up speed (via dsl) with my G3 400 and minimalist XP drive image. Logically, there has to be something going on to account for such huge performance extremes.

Now over at MacWindows a concept of separating the OS from applications (via networking) in Windows VPC was tested and is being proposed to create smaller multiple drive image files to use less disk i/o to compensate for the terrible speed hit that occurred with VPC in OS10.2.1.

hmm, very interesting. and i don't think a bit of it sounds crazy, except from a management overhead point of view. but, thanks for pioneering. let us know what 'develops'.

Amazing
 

The memory arrived and I installed it this afternoon. There is a truly remarkable difference in how responsive the machine simply feels. Here are the stats:

That's after working with several M$ memory hog apps for a few hours! As well as Mozilla.