Macbook RAM Question

[Sysem]

I want to upgrade my 2012 13" Macbook Pro's RAM (currently 4GB) to 16GB. Being damn expensive, I'm thinking of doing it slowly. I want to buy 1 8GB stick and put it in (removing the 2x 4GB sticks). Then, say next month, buy another 8GB stick and put that in the last slot.

Is it going to be an issue have the one slot empty while I wait to get the other stick?

 

[bwana]

Nope. You don't need both slots filled.

EDIT: you can also fill one of the empty slots with a 4gb.

 

[Sysem]

Nope. You don't need both slots filled.

Great - thanks.

 

[bwana]

Great - thanks.

In case you missed my edit there's no reason to remove both 2gb sitcks… or is is 4gb sticks - your OP seems to have both.

 

[Sysem]

In case you missed my edit there's no reason to remove both 2gb sitcks

Oh, interesting. Thanks for the suggestion.

 

[Johnatan56]

Why can't you just leave one of the smaller RAM sticks in? As long as the clock-speed is the same there should not be a degrade of performance.

 

[Sysem]

Why can't you just leave one of the smaller RAM sticks in? As long as the clock-speed is the same there should not be a degrade of performance.

Already discussed, but thanks Will be doing that.

 

[koffiejunkie]

Allow me to voice of dissent.

These computers have dual channel memory controllers, which requires two identical DIMMs to operate (this is why they're available sold in pairs). Without it, you'll be running in single channel mode, so there will be a substantial reduction in memory bandwidth, i.e. performance. Whether or not the reduced performance will affect you negatively depends on what you do with your laptop. If it's general "desktoppy" computing, the gain in RAM quantity will more than outweigh the hit in speed.

If you decide to go ahead and buy one now and one later, make absolutely sure you buy the exact same thing, down to the part number. For this reason, I don't recommend doing one at a time, because even if you buy the same model number from the same manufacturer, there's still a chance the part might be different. Rather save up and be sure.

 

[bwana]

Allow me to voice of dissent.

These computers have dual channel memory controllers, which requires two identical DIMMs to operate (this is why they're available sold in pairs). Without it, you'll be running in single channel mode, so there will be a substantial reduction in memory bandwidth, i.e. performance. Whether or not the reduced performance will affect you negatively depends on what you do with your laptop. If it's general "desktoppy" computing, the gain in RAM quantity will more than outweigh the hit in speed.

If you decide to go ahead and buy one now and one later, make absolutely sure you buy the exact same thing, down to the part number. For this reason, I don't recommend doing one at a time, because even if you buy the same model number from the same manufacturer, there's still a chance the part might be different. Rather save up and be sure.

Would you disagree with this:

All Intel Core Macs support dual channel memory access if matching modules are installed. The customary estimate is that this gives a 6% - 8% real world performance benefit. The modules do not have to be the same brand. That means it is quite possible but not 100% guaranteed, that adding a 3rd party SODIMM to an Apple supplied SODIMM of the same size will make a matched pair.

If not don't you agree that he might be able to survive the month with a 6-8% performance hit between buying the first stick and the second.

TBC I'm only referring to the 6-8% bit - you're going to disagree with the next sentence.

 

[Sysem]

A bit late, went ahead and installed it. It's used for Android development, so nothing top hard-core. Just needed more memory for amount apps running with the VM open. Also, losing 6-8% is still better 4gb...

 

[koffiejunkie]

Would you disagree with this:

ll Intel Core Macs support dual channel memory access if matching modules are installed. The customary estimate is that this gives a 6% - 8% real world performance benefit. The modules do not have to be the same brand. That means it is quite possible but not 100% guaranteed, that adding a 3rd party SODIMM to an Apple supplied SODIMM of the same size will make a matched pair.

No, I agree with it. The bold bit is exacly why I suggested it might not affect him, depending on what he does. The real world in that sentence is important - the difference in memory bandwidth and real world preformance is not the same. The difference in memory bandwith wildly according to the memory controller. Going from single to dual channel, I've seen gains as low as 20% and over 100%.

TBC I'm only referring to the 6-8% bit - you're going to disagree with the next sentence.

I actually agree with the rest. Modules from different manufacturers *can* work in dual channel, and differnet modules from the same manufacturer might not. The thing is, it's difficult to guarantee, which is why I recommend buying in pairs, particularly since he's concerned about cost. There aren't that many manufacturers of memory dimms, and even fewer of the chips they contain, so it's entirely possible for two different branded dimms to be practically the same or even identical.

 

[Sebastian.s1994]

If I may ask, which RAM did you buy?

 

[sajunky]

The bold bit is exacly why I suggested it might not affect him, depending on what he does. The real world in that sentence is important - the difference in memory bandwidth and real world preformance is not the same. The difference in memory bandwith wildly according to the memory controller. Going from single to dual channel, I've seen gains as low as 20% and over 100%.

Difference between single channel and dual channel depends on number of ranks in the module, but more important on memory controller. It can be a high on the older DDR2 North Bridge memory controller, but it never goes over 100%. However with the shift to the memory controller integrated with CPU, a difference is much smaller. So bwana is right, you can expect 6 to 8% penalty in single channel mode.

If you decide on single channel add on, get a two-ranks module (commonly called double sided), it perfoms very close to dual channel configuration.

 

[koffiejunkie]

Difference between single channel and dual channel depends on number of ranks in the module, but more important on memory controller. It can be a high on the older DDR2 North Bridge memory controller, but it never goes over 100%. However with the shift to the memory controller integrated with CPU, a difference is much smaller. So bwana is right, you can expect 6 to 8% penalty in single channel mode.

I did say it depends on the memory controller - you even quoted me. And yes, it *can* be more than 100% difference, because the number of banks is not the only factor. The controller may use a more efficient algorhytm in dual channel mode than in single channel mode, or there may be other implementation factors on the motherboard. If you don't believe me, you can take two motherboards and benchmark their memory bandwith using the same CPU and dimms, and there's a chance you'll get a different result. It's reasonably small on modern CPUs, but it's still there. The system that I saw it on was an Athlon64_X2, so integrated memory controller.

Heck, I still encounter mismatches between motherbaords and memory dimms, i.e. Dimm fails memtest in one board - move dimm and CPU (so same memory controller) to another board and it passes. Try a different dimm (same model, same manufacturer) in the first board, it works. So I can't show that either dimm or motherboard is faulty. It's rare, but it happens.

 

[sajunky]

I did say it depends on the memory controller - you even quoted me.[...]
The system that I saw it on was an Athlon64_X2, so integrated memory controller.

I quoted, as it is important and if we can stick to Intel Core memory controller - not some early Athlon design - it will not confuse people who read it.

Heck, I still encounter mismatches between motherbaords and memory dimms, i.e. Dimm fails memtest in one board - move dimm and CPU (so same memory controller) to another board and it passes. Try a different dimm (same model, same manufacturer) in the first board, it works. So I can't show that either dimm or motherboard is faulty. It's rare, but it happens.

It is true, but completely off-topic.
If it is about new products, it is about reliability margins for memory chips selection, noisy PCB design and production quality control - programming SPD data with parameters which do not always give 100% error free results. So you talk about 'good' brands and not so good.
If it is about old memory sticks, it is understable. Parameters deteriorate over the time, components are exposed to various stress, including static electricity, bypass capacity deteriorate, etc. It is normal that margin will decrease over the time and eventually stop working.

 

[koffiejunkie]

It is true, but completely off-topic.

As was your response to me from the start. I already stated I agree with the 6-8% (give or take) and explained why, before you butted in.

So you talk about 'good' brands and not so good.

You make too many assumptions.

 

[sajunky]

As was your response to me from the start.

Come on, it is not true.