Test your iMac G5 for a 64 or a 128 bit memory bus (and why is mine 64 bit?)

[Elan0204]

I ordered my iMac G5 with 1 512MB DIMM from Apple, and a bought a second 512MB DIMM from Crucial (Crucial brand, but bought for less from someone else).

Anyways, I opened up my iMac and installed the new chip, and also took a look at the Apple provided chip. I was happy to discover that the chip already in my iMac had the same Micron sticker on it my new chip from Crucial did. So both chips were from the same manufacturer and had the same CAS latency of 3. However, I noticed that my new chip from Crucial had all its little memory chips on the same side, whereas the Apple provided chip had memory chips on both sides.

So now my iMac is up and running again, and recognizes that I have 1GB of memory. However, when I run this Applescript provided by MacBidouille, it reports that my memory bus is still only 64 bit. Is there anything I can do to get it switch to 128 bit?

I've attached a picture of the memory section of my System Profiler, and also attached the Applescript so that others can test their iMacs. I'd appreciate reports from others who have run the test.

 

[varmit]

I ordered my iMac G5 with 1 512MB DIMM from Apple, and a bought a second 512MB DIMM from Crucial (Crucial brand, but bought for less from someone else).

Anyways, I opened up my iMac and installed the new chip, and also took a look at the Apple provided chip. I was happy to discover that the chip already in my iMac had the same Micron sticker on it my new chip from Crucial did. So both chips were from the same manufacturer and had the same CAS latency of 3. However, I noticed that my new chip from Crucial had all its little memory chips on the same side, whereas the Apple provided chip had memory chips on both sides.

So now my iMac is up and running again, and recognizes that I have 1GB of memory. However, when I run this Applescript provided by MacBidouille, it reports that my memory bus is still only 64 bit. Is there anything I can do to get it switch to 128 bit?

I've attached a picture of the memory section of my System Profiler, and also attached the Applescript so that others can test their iMacs. I'd appreciate reports from others who have run the test.

Um, did you ever think that because the chips are setup differently, they are different. Thus the idea that same chips produces 128bit, does not apply. English Translation: "We thus had the confirmation which it is not enough to put two of the same bars cuts to be entitled to the 128 bits."

 

[Elan0204]

Um, did you ever think that because the chips are setup differently, they are different.

Of course I did, that is why I mentioned it. If I hadn't noticed the difference myself, then we wouldn't even have a starting place for why I'm still stuck at 64 bits. However if the memory chip layout is the true reason for the problem, this is going to become a very serious issue for a lot people. When you buy RAM you don't get to specify how the chip is constucted, and you don't get to ask Apple to give you a chip with a specific construction when you order your iMac G5. This would make getting a 128 bit memory bus very very difficult, unless you throw away whatever chip Apple gives you, and install 2 new chips. Part of the reason I posted this in the first place is that this information is going to be very important to anyone looking at getting (or who already has) an iMac G5.

 

[jeremy.king]

1. Sell Apple's DIMM
2. Buy second DIMM from crucial

I think people are well aware you should either buy both chips from the same vendor or order it with your BTO, not rely on some false hope that mixing chips will achieve 128bit despite same size and CAS latency. Your situation is proof that this won't work *most* of the time.

Anyways, does it matter that much? Its an iMac for betty's sake.

 

[edesignuk]

There's no way you have to have two EXACTLY the same DIMMS for this to work, that's just ludicrous. All you need is matching pairs (as in same size, basic specs).

 

[psycho bob]

I'm putting on my dunce novice hat hear so be nice. I don't see how putting in another DIMM would change a reading from 64bits to 128bits or vice versa. What readings do you get just by putting in one of the DIMMs at a time. The Powermac G5 supports dual channel RAM and the memory controller provides 128bit data path for RAM, that equates to 64bit per channel. The imac only uses single channel memory wouldn't that mean you would only get a reading of 64bits maximum?

 

[edesignuk]

I'm putting on my dunce novice hat hear so be nice. I don't see how putting in another DIMM would change a reading from 64bits to 128bits or vice versa. What readings do you get just by putting in one of the DIMMs at a time. The Powermac G5 supports dual channel RAM and the memory controller provides 128bit data path for RAM, that equates to 64bit per channel. The imac only uses single channel memory wouldn't that mean you would only get a reading of 64bits maximum?

Apparently not! There was an article posted a few days ago that says although the iMac does not support dual channel, it can increase the memory BUS to 128-bit if you use two matching sized DIMMs.

 

[psycho bob]

Apparently not! There was an article posted a few days ago that says although the iMac does not support dual channel, it can increase the memory BUS to 128-bit if you use two matching sized DIMMs.

Indeed I've just searched Apple's developer pages and it states that 'when two 'identical in size and composition' DIMMs are installed the memory bus is 128bit'. Quite how this works I don't know I assume it can't be the same as dual channel as we know it on the Powermac yet the two modules must work together in some way. If the two RAM modules look different with one having chips on both sides then they cannot be of the same composition. Looks like its a money spinner for Apple for those buying without knowing any better or a cash cow for Crucial for those of us that do
We need to see some xbench memory comparisons between a G5 powermac and an imac which has both RAM slots filled with identical RAM. The Powermac on paper should be quicker I don't know how much of a difference in RAM terms 64/128 bit makes. I would have thought for most things throughput was more important.

 

[pckilla]

I'm putting on my dunce novice hat hear so be nice. I don't see how putting in another DIMM would change a reading from 64bits to 128bits or vice versa. What readings do you get just by putting in one of the DIMMs at a time. The Powermac G5 supports dual channel RAM and the memory controller provides 128bit data path for RAM, that equates to 64bit per channel. The imac only uses single channel memory wouldn't that mean you would only get a reading of 64bits maximum?

I ran the same test that the first post said to and it said(128bit)

 

[psycho bob]

I ran the same test that the first post said to and it said(128bit)

That I assume was with two DIMMs in?

 

[FuzzyBallz]

Gonna get the 20" iMac from the school bookstore on Wednesday, will put in the pair of Crucial PC3200 C3 (single sided) from my Dual G I bought from newegg. I'll let you know what happens.

 

[pckilla]

That I assume was with two DIMMs in?

Yes I ordered the imac with 2 512DIMMs from apple

 

[Jovian9]

Are 2 identical dimms that yield 128bit worth it? Will they provide that much of a difference b/t 1GB RAM at 128bit vs. 1GB RAM at 64bit? If so, do you think it would provide more performance than 1.5GB RAM at 64bit?
I ordered my iMac with 512MB RAM and was thinking about just buying a 1GB chip to put in the other slot. But I guess I'll have to re-evaluate if the memory bus running at 128bit is that much better.

 

[psycho bob]

Are 2 identical dimms that yield 128bit worth it? Will they provide that much of a difference b/t 1GB RAM at 128bit vs. 1GB RAM at 64bit? If so, do you think it would provide more performance than 1.5GB RAM at 64bit?
I ordered my iMac with 512MB RAM and was thinking about just buying a 1GB chip to put in the other slot. But I guess I'll have to re-evaluate if the memory bus running at 128bit is that much better.

In my opinion and from briefly looking at some of the xbench results the difference between having 2 identical DIMMs (128bit) and 2 of different sizes (64bit) is minimal. In real application performance I doubt anyone could tell the difference and for those that use applications that can they would probably be better off with a Powermac anyway. The quantity of RAM is more important for applications and remember the difference between RAM settings gains thousandths of a second at best in these terms. Until we know more I'd just put in as much RAM in as possible.

 

[Elan0204]

MacFixit has some xbench results comparing an iMac G5 with matched 512MB chips (1GB total) versus unmatched chips totaling 1.25GB, and I think everyone will find the results pretty interesting. Here is the article, but I'll put the contents below.

Matched RAM Reader Doug Eldred also sent us the following question: "Does anyone have any actual data for how much difference 'matched RAM' makes on the G5 iMac? Is this a mostly theoretical boost, or does it really provide a user-noticeable improvement?"

Doug is referring to Apple's documentation stating that:

"If only one DIMM is installed, the memory bus is 64-bit. If two non-identical DIMMs are installed, there are two 64–bit memory buses. If two identical DIMMs are installed, the memory bus is 128-bit. Identical DIMM pairs have the same size and composition and provide the fastest and most efficient throughput."

Although we haven't yet received any real-world reports from readers, the German site PPCNUX provides benchmarks for a 1.8GHz iMac G5 using XBench 1.1.3. Specifically, the benchmarks compare an iMac G5 with two unmatched DIMMs totaling 1.25GB of memory to the same iMac with two matched 512MB DIMMs (for a total of 1GB). Even though the former system had more RAM, the "matched" system returned significantly better results:

System Fill:
Unmatched: 1625.25
Matched: 2231.49

System Copy:
Unmatched: 861.25
Matched: 1041.32

Stream Copy:
Unmatched: 1255.89
Matched: 1615.10

Stream Scale:
Unmatched: 1251.27
Matched: 1657.52

Stream Add:
Unmatched: 1423.78
Matched: 1781.15

Stream Triad:
Unmatched: 1441.01
Matched: 1794.59

Personally, I don't think that xbench is a very good benchmark, but there is no denying that those speed differences are pretty significant.

 

[slooksterPSV]

Would having the same exact type of RAM upgrade performance on a PC. We have so many different kinds... what the heck am I talking about, screw the PC, I'm moving to Mac ppl.

 

[Jovian9]

If anyone does happen to buy 3rd party RAM and it matches with the Apple RAM to create a 128bit memory bus........please post where and what RAM you purchased.
I think there may be quite a few people, including myself, who may be making RAM purchase decisions for the G5 iMacs based on this.

 

[thedoc1111]

It is quite simple: Apple RAM is overpriced, so buy as little as you can i.e. the basic 256MB.

Then go to Crucial and buy 2 of exactly the same module (2x standard 512MB DDR400).

Sell your old 256MB on eBay.

Simple (and the cheapest way to get a gig of RAM if you do the maths)

There are all sorts of hidden timing issues in RAM that will proclude different RAM dimms from becoming 'matched pairs'

Mike

 

[coconn06]

It is quite simple: Apple RAM is overpriced, so buy as little as you can i.e. the basic 256MB.

Then go to Crucial and buy 2 of exactly the same module (2x standard 512MB DDR400).

Sell your old 256MB on eBay.

Simple (and the cheapest way to get a gig of RAM if you do the maths)

There are all sorts of hidden timing issues in RAM that will proclude different RAM dimms from becoming 'matched pairs'

Mike

Except also don't buy Crucial, because it's nearly as expensive as Apple's RAM. Check out ramseeker. See this thread for a discussion of Crucial RAM versus others.

 

[coconn06]

Personally, I don't think that xbench is a very good benchmark, but there is no denying that those speed differences are pretty significant.

You're right, XBench isn't a very good benchmark. Benchmarks in general are a bad way to compare performance in two different machines. However, comparing performance of the same machine with only one variable different (RAM, in this case) seems to me to be a very reliable way to benchmark. Therefore these results look great to me.

 

[thedoc1111]

Except also don't buy Crucial, because it's nearly as expensive as Apple's RAM. Check out ramseeker. See this thread for a discussion of Crucial RAM versus others.

Sorry, I am talking UK prices where Crucial RAM is around two-thirds of the price of Apple RAM. That is a big difference!

I don't know about other manufacturers but I have never had a stick of Crucial RAM fail on me ever and I have used about 8 from time to time.

Also Kingston RAM is throwing up issues with getting a proper matched pair (ppl buy 2 sticks of 'identical' Kingston RAM - iMac runs 64 bit). No-name RAM would likely cause even more trouble.

 

[VooDooPope]

Hmmm based on those numbers I wonder what I should do.

My 20" with 512 just came in today. I was going to order another 512 stick from crucial but now I'm not sure based on other peoples experience doing this and still running at 64.

I wonder if I'd be better off buying 2 matching 512's (aprox $200) for 1gb total running at 128 or buying a 1gb stick to go with my 512 I already have for 1.5gb total running at 64.

Ideas, thoughts, anyone?

 

[Elan0204]

I wonder if I'd be better off buying 2 matching 512's (aprox $200) for 1gb total running at 128 or buying a 1gb stick to go with my 512 I already have for 1.5gb total running at 64.

Ideas, thoughts, anyone?

Do you think you will be running anything that will need 1.5GB of memory? Will you be working with large video and image files? Do you do work in Final Cut Pro and Photoshop a lot?

Also, the test results above compare 1.25GB of memory to 1GB of matched memory, and the 1GB is significantly faster. It's possible that it still may be faster than 1.5GB of unmatched memory, but there are no numbers available yet to find out.

 

[VooDooPope]

Do you think you will be running anything that will need 1.5GB of memory? Will you be working with large video and image files? Do you do work in Final Cut Pro and Photoshop a lot?

Also, the test results above compare 1.25GB of memory to 1GB of matched memory, and the 1GB is significantly faster. It's possible that it still may be faster than 1.5GB of unmatched memory, but there are no numbers available yet to find out.

Most of the time the iMac will be used for surfing the web and email etc. But about once a week I bring work home (print and/or web design) and I'll use the iMac for that instead of using my powerbook like I did last night. I do work on large photoshop files from time to time as well as some minor video editing using final cut.

I wish I could drop 2 matching 1gb sticks in but the wife doesn't like that $500 dollar idea. Imagine that.

 

[Elan0204]

Most of the time the iMac will be used for surfing the web and email etc. But about once a week I bring work home (print and/or web design) and I'll use the iMac for that instead of using my powerbook like I did last night. I do work on large photoshop files from time to time as well as some minor video editing using final cut.

Based on that I would say that 1GB should be fine for you. I don't think you would benefit enough from the extra RAM to justify the added expense. Plus, it seems like there is a pretty big benefit to matched RAM. You already spent $75 to have Apple upgrade you to 512MB, now you could spend another $85 to get a second 512MB chip. 1GB chips go for over $200, and I don't think you'll benefit that much from the added cost.