Geekbench score went down after doubling ram

[amoda]

Okay,

So I was having some performance issues. I kept having tons and tons of page outs, going down to less than 50mb free ram, etc.

So I decided to upgrade the ram. I bought 4Gb of Nuimpact ram 1066 DDR3, etc. I bought it from a store that caters only to Macs so it has to be 100% compatible.

Anywho, I decided to run Geekbench (64-bit) before and after. My score prior to the upgrade was 3580, post upgrade it varies from 3548-3567. 99% of the drop was in the two ram tests.

I know the score didn't go down much, but really I was expecting it to go up. Like scores like that I would expect as normal variance of running the test on the same machine several times. But to see a drop when going from 2 to 4GB caught me by surprise.

The ram seems to be working fine in general. Not paging out any more, no hang-ups or sleep problems.

Just kinda bummed me out I guess.

So, any guesses as to why the score went down? Could it be shoddy ram despite no other problems?

 

[ihabime]

The simple answer is that your geekbench score didn't go down. Scores will vary by that small an amount even on the exact same machine.

The reason your score didn't go up is because Page Outs and having only 50MB FREE ram wasn't causing your machine to slow down.

Page Outs happen all the time, OSX will copy something to disk cache for a variety of reasons most of which aren't because it needs the memory at that moment.

It may cache some memory to disk but still leave it allocated just in case you need the memory quickly. It could be caching in order to reorganize memory into larger free blocks.

Page Outs are not a problem, they happen in the background and don't slow down your computer.

Page Ins can be a problem, that can mean OSX has swapped out active memory because another program needed it.

 

[amoda]

Actually ihabime, you're mistaken. Thanks for trying thought


First off,

"Scores will vary by that small an amount even on the exact same machine."

Yes, I know that. Thing is this isn't the same machine...it's a better machine. So the fact that the score improve, or even stay the same, but actually went down is worrisome to me.

Secondly,

Page Outs are not a problem, they happen in the background and don't slow down your computer.

Page Ins can be a problem, that can mean OSX has swapped out active memory because another program needed it

That isn't quite true.

Page INs aren't a problem. The system is putting and getting information off the super quick ram.

Page OUTs on the other hand, occur when there is insufficient memory. The computer has to send and and retrieve information from the HD instead, which runs slower than ram; therefore slowing the computer.

 

[localoid]

Although it is "100% compatible", perhaps the CAS latency of the new RAM is higher than the original Apple RAM(?)

 

[ihabime]

Actually ihabime, you're mistaken. Thanks for trying thought
Yes, I know that. Thing is this isn't the same machine...it's a better machine. So the fact that the score improve, or even stay the same, but actually went down is worrisome to me.

That was my point, if geekbench scores vary that much on the same machine then the difference between your machine before and after upgrade is insignificant.

The score didn't go down, the worst variance you had was from 3580 to 3548 a difference of less than 1% (.00893). To compare I just ran Geekbench 10 times on the same machine, restarting each time. There was a 1.89% difference between the best and worst score.

Page INs aren't a problem. The system is putting and getting information off the super quick ram.

Page Ins aren't getting information off of super quick ram, they are copying information from the disk to ram.

Page OUTs on the other hand, occur when there is insufficient memory. The computer has to send and and retrieve information from the HD instead, which runs slower than ram; therefore slowing the computer.

That's a vast oversimplification.
Page Outs sometimes occur because there is insufficient memory, but they also occur for the reasons I previously stated plus a few more.

The OSX memory model is more complex than that, it doesn't just sit around waiting for an App to run out of memory, it keeps a list of all memory allocated by most recent use and slowly ages out unused areas in order to make the most memory available to more recently used processes. It tries to predict your needs by your behavior.

Say you are working in Pages and Keynote with safari in the background, if you haven't switched to Safari in a while OSX will eventually swap it's memory out in order to give the most possible ram to Pages and Keynote. This will happen even when you haven't used up all your ram, because OSX is trying to make your most used apps run as quickly as possible.

It doesn't deallocate the safari memory, it marks it as available if necessary. If you switch back to safari before its ram is used OSX will simply mark that ram as allocated to safari again without needing to do a Page In.

If your system had been ram starved then you would have seen speed increases by adding the ram, whether it was slower ram or not. If it was shoddy ram you'd see freezes and kernel panics.

 

[obeygiant]

^^^

I don't know what "page outs" or "page ins" are but my machine was super laggy on Lightroom with about 121000 images so i thought I'd double the ram from 4gb to 8gb. I can't tell yet if its helped but my geekbench score did go down a little.

This helps something right? Otherwise why did I do this?

 

[zachsilvey]

^^^

I don't know what "page outs" or "page ins" are but my machine was super laggy on Lightroom with about 121000 images so i thought I'd double the ram from 4gb to 8gb. I can't tell yet if its helped but my geekbench score did go down a little.

This helps something right? Otherwise why did I do this?

Keeping that large of a library is dumb. You should really consider splitting that up unless you want significant slow down.

 

[obeygiant]

Keeping that large of a library is dumb. You should really consider splitting that up unless you want significant slow down.

Those images represent my current client roster, I have to keep them handy.

 

[Cboss]

Geekbench measures the speed of the RAM. So, your score won't improve by adding more RAM, it will only increase if you add faster RAM, which you most likely can't do without replacing the whole motherboard. This is why benchmarks aren't usually the best way to test a computer.

Your scores are so close that it doesn't mean anything. I'll bet that if you test it a few more times it will vary by much more than it did in your first two tests.

 

[obeygiant]

Geekbench measures the speed of the RAM. So, your score won't improve by adding more RAM, it will only increase if you add faster RAM, which you most likely can't do without replacing the whole motherboard. This is why benchmarks aren't usually the best way to test a computer.

Your scores are so close that it doesn't mean anything. I'll bet that if you test it a few more times it will vary by much more than it did in your first two tests.

Of course, you're right. I was hoping to improve the performance of Lightroom, which has a tendency to lag after I import images from a CF card. That or it will just stop responding, like there is no beachball I have a mouse and I can click on things but nothing happens. Then all of the sudden all the commands/clicking I just did will come flying back at me. Only time will tell if this extra ram will help me I guess.