Intel Mac Mini SLOOOOOOOW...

[milo]

This thread reminds me of the threads that started and ran for days right after the Intel mini was released. The moment some people got ahold of the fact that they had shared vRAM, the argument was made that performance just had to stink on these systems. Then the benchmarks started coming out, and lo and behold, they didn't stink -- which lo and behold didn't stop some people from continuing to argue that performance just had to stink.

That's a funny comment, but I agree with it. But in this particular case, the people who have never used the machine are the ones insisting that it's fine with 512, while the people who actually have used the machines and experienced the performance firsthand know what the truth is: with 512 they bog down *very* easily, which is remedied by going to at least a gig.

The moral of the story? People end up looking foolish if they try and pretend to be an expert about technology that they've never used. Funny how some people can't seem to learn that.

This is a good question, and one I have pondered myself. One answer is obviously tha Rosetta has a large overhead, but even without Rosetta running (i.e. 100% Universal apps) Intel Macs appear to need more memory than PPC ones. This may be due to the size of Intel binaries (Intel binaries seem to be on e whole slightly larger than their PPC equivalents) or perhaps due to the still-somewhat-beta nature of the new Intel Mac OS X.

That said, I see you are running Panther on your mini. I have a 12" PB with 512Mb and I found it became unusable when upgrading to Tiger. I upgraded it to 768Mb and suddenly it was very happy again, my conclusion being that Tiger is a bigger RAM hog than Panther.

All true. Intel itself needs more, Rosetta needs more, Tiger needs more, and the graphics chip uses some as well.

I understand what you mean with UB's taking up more space, thusly more ram...but wouldn't OS X only read the part of the UB that it needs to run, and let the other half just "sit" there on the HDD and not do anything except take up space? at least, that's how I would think it would opperate..

I'm not sure what the reason is for it, but apps just use more memory running on intel. If you look in Activity Monitor and compare an intel mac with a PPC mac, the intel mac will have higher ram usage for each app.

Show me the evidence that they're slow at memory intensive tasks. Memory throughput benchmarks for my mini show it to be up their with the G5s.

I'm curious about this too. I've seen a bunch of benchmarks, and none have showed memory performance that was much different than any other intel mac.

 

[dr_lha]

I'm curious about this too. I've seen a bunch of benchmarks, and none have showed memory performance that was much different than any other intel mac.

I think he's basing this assumption that "shared" video RAM means crappy RAM performance. Even if this is true, the effect is minimal to say the least. The fact is that for speed my Core Solo mini gives my 1.8Ghz DP G5 a run for its money.

 

[IJ Reilly]

I'm sorry, but this is wrong. Unix (and hence OS X) does not work like that. I almost never reboot unless the power goes out, and my system does not see performance degradation, with 1.5GB of RAM (on a PPC machine, so no Rosetta). The only thing that could cause something like that is a memory leak. Rebooting would cure the problem temporarily, but you'd want to get rid of it by not running the buggy program that has the memory leak. It's certainly not a "feature" of OS X.

You can read up on vm file handling in OSX in any one of a dozen places. They are stored in private/var/vm, a normally invisible directory (to the Finder). Use TinkerTool to make the hidden directories visible. Look inside and you will see at least one 64 Mb file called "swapfile0". Add this directory to your sidebar and watch its behavior as you use the Mac for more RAM-intensive tasks. You'll see the swapfiles proliferate, and they won't go away until you reboot.

If you've got plenty of physical RAM, you may never notice degraded performance from all the disk flogging these files can cause (the argument in favor of adding RAM). If you don't have a lot of RAM, you will eventually notice the degraded performance. Rebooting will delete the swapfiles, and you'll be back to optimum performance for awhile at least (the argument in favor of at least rebooting the Mac before deciding that more RAM is an absolute necessity).

 

[IJ Reilly]

That's a funny comment, but I agree with it. But in this particular case, the people who have never used the machine are the ones insisting that it's fine with 512, while the people who actually have used the machines and experienced the performance firsthand know what the truth is: with 512 they bog down *very* easily, which is remedied by going to at least a gig.

The moral of the story? People end up looking foolish if they try and pretend to be an expert about technology that they've never used. Funny how some people can't seem to learn that.

You're misrepresenting my argument again, and it's getting pretty annoying. What I've been saying from the very start is that the original poster is complaining of serious performance issues which cannot be described as "normal" for this or any other Mac. Odd that simply suggesting that he reboot and see what that does, before telling him to spend more money, is seen as so controversial.

Watch that high horse -- you might fall off.

 

[milo]

You're misrepresenting my argument again, and it's getting pretty annoying. What I've been saying from the very start is that the original poster is complaining of serious performance issues which cannot be described as "normal" for this or any other Mac. Odd that simply suggesting that he reboot and see what that does, before telling him to spend more money, is seen as so controversial.

So as to avoid any accusations of misrepresntation, let me quote from your original post in this thread:

You don't need more RAM. This is not the problem.

I (and seemingly most of the other people posting in this thread) disagree.

He does need more ram. That is the problem. That's the part I disagree with. You haven't changed your position on that, have you? If you did, I missed it.

I've said multiple times, there's no harm in things like rebooting and basic software troubleshooting. I agree that these are simple and quick enough to be worth doing in most situations. So stop pretending that anyone has EVER said such a thing is controversial. That said, based on his description of the problem, the basic fixes are unlikely to solve the problem. (Under heavy use, the mini starts to bog almost immediately, you'd have to reboot before getting any work done) All symptoms point to ram. So do the basics...while you're on hold waiting to order more ram.

If you've never used a mini, how do you know that the performance he's seeing isn't normal? I've seen that behaviour on the mini. Others on this thread have confirmed seeing it. Others elsewhere on the net have confirmed it as well. Why are you in such denial? Do you think we're all lying? Do you think somehow the "exception" cases outnumber what you imagine to be the norm? What will it take for you to believe that other minis do the exact same thing? Does one of us have to take footage of the exact same performance as the OP describes and post it online?

What I really don't understand is, since you admit that performance will degrade when ram is exceeded and the system starts using swap files (which would be reduced by more ram), why are you so insistent that's not what's happening in this case?


Question for the austincolby, after a reboot, how quickly does performance degrade on your machine? In Activity Monitor, how quickly do the page counts go up? You mentioned page outs hitting a million. Is that after a month, a week, a day, an hour?

 

[asencif]

While this war continues, has the OP gotten more RAM or fixed the issue? Intel Macs do need more memory to perform well in comparison to the PPC Macs. This is probably due to Rosetta apps, but it is the case overall with all apps as was already said by another poster in this thread. The PPC Mini could function well with 512 Ram as I've used that with Safari and Word. My understanding was that the OP was having Safari beach ball for 10 sec after opening it up and using it for a few. Now is that the problem everyone is having with the Intel mini..Can't run more than two apps after rebooting? I know others have said wtih 512RAM they can't multitask much and start hanging up after a while of usage. So my question is for the Intel mini is it also supposed to hang up after just opening one app? Or Safari and then Word and that's it?

 

[milo]

While this war continues, has the OP gotten more RAM or fixed the issue?

I think this is what it will take to settle the whole thing.

 

[IJ Reilly]

I (and seemingly most of the other people posting in this thread) disagree.

He does need more ram. That is the problem. That's the part I disagree with. You haven't changed your position on that, have you? If you did, I missed it.

I didn't realize this was an election. Did I lose?

The absurdity of this debate is that our opinions are seemingly not very far apart. I thought I'd made my views clear already, but for the last time I hope: The kind of severe degradation in performance reported suggests the need for some basic diagnostics before deciding to throw any money at the problem. If what I've suggested is tried, and it simply fails to produce reasonable performance from this Mac, then I'll accept your argument that Apple has effectively released a defective product in its stock configuration, and that only installing more RAM can possibly help. Until then, I think it's reasonable to be skeptical about the absolute need to do so, especially given the bench tests for the Core Solo Mini, which indicate that it outperforms the G4 Mini in most tests.

 

[IJ Reilly]

I think this is what it will take to settle the whole thing.

Agreed. I think we might have scared him off.

 

[whooleytoo]

Agreed. I think we might have scared him off.

There was an OP?!?

(There I go again, me and my lack of memory.. )

 

[whooleytoo]

Show me the evidence that they're slow at memory intensive tasks. Memory throughput benchmarks for my mini show it to be up their with the G5s.

Honestly, I don't have benchmarks to back up that assertion, but that would be my expectation. What benchmarks show it to be similar in performance to the G5s? Are they from a test suite or real world (i.e. application) testing?

 

[quidire]

You can read up on vm file handling in OSX in any one of a dozen places. They are stored in private/var/vm, a normally invisible directory (to the Finder). Use TinkerTool to make the hidden directories visible. Look inside and you will see at least one 64 Mb file called "swapfile0". Add this directory to your sidebar and watch its behavior as you use the Mac for more RAM-intensive tasks. You'll see the swapfiles proliferate, and they won't go away until you reboot.

If you've got plenty of physical RAM, you may never notice degraded performance from all the disk flogging these files can cause (the argument in favor of adding RAM). If you don't have a lot of RAM, you will eventually notice the degraded performance. Rebooting will delete the swapfiles, and you'll be back to optimum performance for awhile at least (the argument in favor of at least rebooting the Mac before deciding that more RAM is an absolute necessity).

Okay, here you are really demonstrating your technological illiteracy.

These files don't "cause flogging". The files represent "software RAM"; the total size of addressable memory space is increased by the size of those files. When you've run large programs (or loaded large files) more of these swap files have been created. When you stopped using them, those files which were no longer necessary were left around (so that they wouldn't have to be created again) but aren't used until you exceed your available RAM again.

Understand - swap files cannot slow down your computer. The files existing don't "make" the OS swap out your active applications; only mathematical necessity will make the OS use your virtual memory for active processes. Deleting them is silly, they will just be created again unless you fix the root problem - insufficient RAM.

One last point: anyone who says that adding RAM won't increase speed simply doesn't understand the Unix memory management process. All unused RAM (save 20-30 MB) turns into disk cache. If you had 10 GB of RAM, 9 GB of it would be filled up with what portions of your harddrive your kernel thought would most likely get accessed next. As soon as you needed more RAM than the 1GB that was left over, however much extra was needed would cease to be disk cache and would be allocated to your applications.

Memory access is much much faster than hard drive access. Thus RAM will always speed up your computer, even after you stop needing active swap memory; you simply get much faster harddrive access beyond that point.

(There is another level of complexity; the Unix memory management algorithms will swap out memory contents that are not being used and "are unlikely to be used in the near future" and simultaneously disk cache some parts of the hard drive which are likely to be used shortly, and so swap files will always exist, as Mac OS X will optimise its memory handling no matter how much the results alarm the uneducated user who apparently has a panic attack at the sight of a swap file...)

 

[quidire]

Benchmarks for the mini

Bare Feats compares the iMac Core Duo, the Mac mini Core duo, and the Mac mini G4

Just booting up the mini and doing nothing else consumes 270MB, so two 256MB modules just doesn't cut it.

Indeed.

 

[milo]

The kind of severe degradation in performance reported suggests the need for some basic diagnostics before deciding to throw any money at the problem. If what I've suggested is tried, and it simply fails to produce reasonable performance from this Mac, then I'll accept your argument that Apple has effectively released a defective product in its stock configuration, and that only installing more RAM can possibly help. Until then, I think it's reasonable to be skeptical about the absolute need to do so, especially given the bench tests for the Core Solo Mini, which indicate that it outperforms the G4 Mini in most tests.

I absolutely feel that the stock configuration is just barely usable. It's OK if you're running few apps at a time, few documents or windows open. Or if you have low expectations and are willing to accept pokey performance. For some things like Front Row, performance is so slow with 512 that there were times when I was left staring at a black screen wondering if the machine had crashed (usually it would come back, sometimes after as long as 30 seconds). There are many reviews online that have said the exact same thing, that the machine is barely usable, unusable, unacceptable, lots of beach balls, etc with 512.

Benchmarks aren't an accurate reflection of this situation. They generally run one app and focus on one document at a time. This ram starvation is usually seen when multiple apps are open with multiple documents, and the user is switching back and forth, a situation rarely if ever duplicated by benchmarks. I'd love to see a benchmark that does such a thing, I'm sure it would show a major difference between intel macs with 512 and a gig or more.

 

[dr_lha]

Honestly, I don't have benchmarks to back up that assertion, but that would be my expectation.

Exactly. Your expectation is not a very good measure of performance, even worse than a benchmark!

What benchmarks show it to be similar in performance to the G5s? Are they from a test suite or real world (i.e. application) testing?

Here are the "geekbench" memory test results from my Core Solo mini and my 1.8Ghz DP G5 (both have 1Gb RAM):

Mini:

Memory Performance
                Latency   317  (1 thread,  32.99 nanoseconds/load)
        Read Sequential   252  (1 thread,  1.831 gigabytes/sec)
       Write Sequential   206  (1 thread,  1.186 gigabytes/sec)
        Stdlib Allocate    94  (1 thread,  73.9 kiloallocs/sec)
        Stdlib Allocate    98  (4 threads, 77 kiloallocs/sec)
           Stdlib Write   126  (1 thread,  1.992 gigabytes/sec)
            Stdlib Copy   143  (1 thread,  1.07 gigabytes/sec)

G5:

Memory Performance
                Latency   217  (1 thread,  48.1 nanoseconds/load)
        Read Sequential   200  (1 thread,  1.457 gigabytes/sec)
       Write Sequential   174  (1 thread,  1.001 gigabytes/sec)
        Stdlib Allocate    99  (1 thread,  77.55 kiloallocs/sec)
        Stdlib Allocate   108  (4 threads, 84.81 kiloallocs/sec)
           Stdlib Write   118  (1 thread,  1.858 gigabytes/sec)
            Stdlib Copy   102  (1 thread,  784 megabytes/sec)

As you can see, the results favor the mini in some, and the G5 in others, but in general they are very similar.

I know this isn't "application benchmarking", if you can find me an application benchmarking package then I'd be happy to run it. However I think this is a good representation.

 

[IJ Reilly]

Okay, here you are really demonstrating your technological illiteracy.

These files don't "cause flogging". The files represent "software RAM"; the total size of addressable memory space is increased by the size of those files. When you've run large programs (or loaded large files) more of these swap files have been created. When you stopped using them, those files which were no longer necessary were left around (so that they wouldn't have to be created again) but aren't used until you exceed your available RAM again.

Understand - swap files cannot slow down your computer. The files existing don't "make" the OS swap out your active applications; only mathematical necessity will make the OS use your virtual memory for active processes. Deleting them is silly, they will just be created again unless you fix the root problem - insufficient RAM.

One last point: anyone who says that adding RAM won't increase speed simply doesn't understand the Unix memory management process. All unused RAM (save 20-30 MB) turns into disk cache. If you had 10 GB of RAM, 9 GB of it would be filled up with what portions of your harddrive your kernel thought would most likely get accessed next. As soon as you needed more RAM than the 1GB that was left over, however much extra was needed would cease to be disk cache and would be allocated to your applications.

Memory access is much much faster than hard drive access. Thus RAM will always speed up your computer, even after you stop needing active swap memory; you simply get much faster harddrive access beyond that point.

(There is another level of complexity; the Unix memory management algorithms will swap out memory contents that are not being used and "are unlikely to be used in the near future" and simultaneously disk cache some parts of the hard drive which are likely to be used shortly, and so swap files will always exist, as Mac OS X will optimise its memory handling no matter how much the results alarm the uneducated user who apparently has a panic attack at the sight of a swap file...)

Nobody, as least not I, ever said that more RAM won't improve performance. Quite the opposite in fact. VM swapfiles absolutely do build up over time, the Mac will slow as they build, the swapfiles do get deleted when the Mac is rebooted, and the Mac will perform better after the reboot. An entire thread was just started on the performance benefits of occasional reboots. Since you're the technologically literate one here, perhaps you ought to explain the cause and effect, if the deletion of VM swapfiles isn't it.

 

[danny_w]

I still run Panther on my mini, I just can't justify the $100 for dashboard- the only feature I would use (and beta testing lightbox).

I understand what you mean with UB's taking up more space, thusly more ram...but wouldn't OS X only read the part of the UB that it needs to run, and let the other half just "sit" there on the HDD and not do anything except take up space? at least, that's how I would think it would opperate..

Maybe I'm just not getting why anyone who does basic tasks would need more than 512 megs of RAM since I'm still on Panther...

I have a G4 mini and it was perfectly useable under Panther with only 512MB (upped from the standard 256MB), but it got much slower under Tiger.

 

[IJ Reilly]

There was an OP?!?

(There I go again, me and my lack of memory.. )

I know, it was such a long time ago. The original poster has probably already died of old age.

 

[erikamsterdam]

never thought of that

posted way earlier in this loooong thread but see where I went wrong. I am still on Panther, and I remember now why. I noticed Tiger demands 256 Mb RAM on the box, and Panther 128.
So Tiger eats 128 extra, the video eats 80, the mini is already 200 Mb behind compared to a PPC with Panther. And then Rosetta is demanding too.
Explains everything.
BTW looks like Openoffice is getting close to Intel code for Mac. Might help those people cursing Micro$oft office on Intel Mac.

 

[IJ Reilly]

I absolutely feel that the stock configuration is just barely usable. It's OK if you're running few apps at a time, few documents or windows open. Or if you have low expectations and are willing to accept pokey performance. For some things like Front Row, performance is so slow with 512 that there were times when I was left staring at a black screen wondering if the machine had crashed (usually it would come back, sometimes after as long as 30 seconds). There are many reviews online that have said the exact same thing, that the machine is barely usable, unusable, unacceptable, lots of beach balls, etc with 512.

Benchmarks aren't an accurate reflection of this situation. They generally run one app and focus on one document at a time. This ram starvation is usually seen when multiple apps are open with multiple documents, and the user is switching back and forth, a situation rarely if ever duplicated by benchmarks. I'd love to see a benchmark that does such a thing, I'm sure it would show a major difference between intel macs with 512 and a gig or more.

Bench tests are always going to show improved performance on any given system with more RAM, so I don't think this an issue, or a matter of serious dispute.

I get how you feel about the Core Sole mini, but frankly I'm more interested in objective testing than feelings, and to date the objective tests indicate that it outperforms the G4 mini in just about every respect. It's a little difficult to gage people's performance expectations when they buy a computer, and even more difficult to know what they've done to create a performance issue, and/or attempted in an effort to remedy a perception of under-performance. Without at least this much information, it's a little like arguing about who's favorite color is better.

 

[milo]

I get how you feel about the Core Sole mini, but frankly I'm more interested in objective testing than feelings, and to date the objective tests indicate that it outperforms the G4 mini in just about every respect. It's a little difficult to gage people's performance expectations when they buy a computer, and even more difficult to know what they've done to create a performance issue, and/or attempted in an effort to remedy a perception of under-performance. Without at least this much information, it's a little like arguing about who's favorite color is better.

Taking ten seconds to change apps or documents, or fifteen to close a safari window is something objective that can be measured. When those times go down to near instantaneous when ram is increased, that IS objective and has nothing to do with opinions, perceptions, or feelings.

The tests you link to don't test the point at which the machine starts using swap files and the resulting slowdown. I explained this already, the results described in the OP are the result of multiple apps and docs open at once, and none of those tests simulate such a situation. I definitely think it's possible to come up with some sort of benchmark to test performance when ram is used up, but I haven't seen anyone actualy do it yet.

 

[dr_lha]

So Tiger eats 128 extra, the video eats 80Mb

The video doesn't eat 80Mb though, on my machine, running desktop apps at 1680x1050, the GFX card takes about 19Mb.

 

[Kreamy]

Even my G4s running Classic don't have this problem, and Classic is a major RAM and resource hog. Rosetta could hardly be any worse. When the Mac bogs down, I reboot. Problem solved.

More RAM is always better (assuming you don't get bad RAM, which is probably the commonest serious problem on the Mac), but I'm a big advocate for figuring out what is causing the actual issue before throwing money at it.

a) You just contradicted your previous post.

b) Running Rosetta is completely different to running Classic. Rosetta is an application that translates every single command from its native PPC format to one that is understood by the Intel x86 system - Classic ran on the same chips that are being used to run OS X.

to the original poster: If you want my advice (and this is what I do with photoshop) run that windows emulator application (the name evades me) and office for windows - it's much quicker and you have access to Access, visio and other Windows-only apps

 

[quidire]

Nobody, as least not I, ever said that more RAM won't improve performance. Quite the opposite in fact. VM swapfiles absolutely do build up over time, the Mac will slow as they build, the swapfiles do get deleted when the Mac is rebooted, and the Mac will perform better after the reboot. An entire thread was just started on the performance benefits of occasional reboots. Since you're the technologically literate one here, perhaps you ought to explain the cause and effect, if the deletion of VM swapfiles isn't it.

A reboot can reset many things that may have peformance implications:

1. Memory Fragmentation - If you have 100 units of RAN, and all get allocated, and then units 20-30 and 60-70 get deallocated, if the next time memory is requested 15 units are sought, that program will have a fragmented memory space. I'm not sure if this is a problem in OS X; it is in many Unix variants (and in almost all other operating systems; XP certainly can suffer from it).
2. Memory leaks in programs - No program that is at all complex can reasonably be said to have no memory leaks; so many different libraries are tied in, that verifying all dynamic memory allocations and deallocations is impossible. Over time, programs' memory footprints will expand.
3. When the Mac first boots up, the disk cache is being used to accelerate UI-related resources, and perhaps some application binaries; later on the disk cache has to cover many more bases, resulting in some UI-related disk accesses hitting the actual hard drive
4. on a related note, process priority of applications like SystemUIServer may be reniced by other privileged applications during the period the computer is active, leaving it at a lower priority than other processes even after the privileged applications have exited.

Mind you, I really don't notice a difference; all three of my Macs need a minute or so to settle down once I log in (with various applications setting up and making initial network connections) and then seem to run at the same speed for days; I reboot rarely and notice no benefit when I do. With more memory used up, more applications running things might be slightly slower, but this fixes itself after I close those programs down.

I do tend to have a surfeit of memory, so perhaps memory leaks and fragmentation don't impact me as much as others...

EDIT: I am correct in assuming that no user applications were running at either point of the various comparisons? Otherwise clearly the comparison is just silly. A real comparison would have the computer boot up w/o any user applications executing, benchmark run, the computer rebooted, no applications launched have the computer left on w/o auto-sleep for a few days, and then benchmarked again. I imagine would would not find much difference

 

[IJ Reilly]

a) You just contradicted your previous post.

b) Running Rosetta is completely different to running Classic. Rosetta is an application that translates every single command from its native PPC format to one that is understood by the Intel x86 system - Classic ran on the same chips that are being used to run OS X.

What contradiction?

Of course Rosetta is "different," but that's not the point. The point is, I can really load up a Mac with half the RAM and about 20% of the horsepower of a Core Solo Mini, and it doesn't perform unacceptably as a rule. And if it does get laggy, a restart works wonders.