MAC Mini dual Core Memory-are there two slots on the new mac minis?

No that's not correct. Each config default comes with 512 at a minimum but that is 2x256. Look at the config when you buy a mini and it will say so right there.

-Seed

For best results always use matched pairs in the Intel mini. It uses PC2-5300, unbuffered nonparity, 200-pin, 667 MHz DDR2 SDRAM. Hope that helps.

Apple recommends using matched pairs in the new Intel Mac mini. However, several users have reported that it will work well with mixed configurations. 1GB in slot 1 and 512MB in Slot 2 will work. Or 256MB and 512MB.

Some tests indicate that video memory usage can change with different memory installs. See details below:

Memory Config. | Total Sys RAM | Shared VRAM
2 x 256MB | 512MB | 64MB
1GB + 256MB | 1.25GB | 256MB
2 x 1GB | 2GB | 64MB

You can see more on this at:
http://www.xlr8yourmac.com/

If I had the money to get matched pairs. I would do so but if I only had a 1GB chip I would not be concerned about using it.

Have fun,

Michael

I scored a couple of gigaram 1GB sticks from newegg and it changed the machine completely. It runs extremely smooth and this is on a Solo. Ran me about 200 for the pair.

I've read on these forums here that the matched pairs are in fact important. It has something to do with memory bandwidth. If the pairs are not matched, they are not treated as dual-channel memory and will run at DDR speeds (rather than DDR2 speeds they are designed for).
It will still work, but you'll only get half the effect speed of the FSB. The impact can be quite significant for work that is memory intensive.
I think the Mini's FSB is something like 667MHz... so that means with mismatched RAM will only perform at 333MHz.
Correct me if I am wrong.

"I've read on these forums here that the matched pairs are in fact important. It has something to do with memory bandwidth. "

I believe there is an Apple tech sheet on this. They said to always use matched pairs.

I just picked up the Core Duo mini and it does seem a little sluggish when I have multiple apps open. I'm thinking I may want to upgrade to at least 1GB of RAM later.

Do I have to go back to my local Apple store to have this done? If so, how much is the labor?

Mac mini Mac OS X (10.4.5)

Well I need to make a correction on my previous test posted here:

http://discussions.apple.com/thread.jspa?threadID=419810&tstart=0

I only ran it once, being a perfomance engineer that was a big no no, so I ran it 10 more times and the test results were in line with the original mac mini memory supposed "matched SODIMMs" or maybe I got lucky and got a matched pair. I don't know but I have yet to see any proof of a mismatched pair being of lesser performance than a matched pair.


Looking at PPC7410s post in that thread, there was a guy who had one 256 stick and one 1GB stick and his throughput was the same as everyone elses, certainly not a matched pair there.

Here is one with asymmetric RAM:
"Feri’s Mac mini"
Physical RAM 1280 MB
Model Macmini1,1
Memory Test 91.02
System 97.47
Allocate 102.16 375.18 Kalloc/sec
Fill 94.20 4580.02 MB/sec
Copy 96.40 1991.16 MB/sec
Stream 85.36
Copy 74.08 1530.07 MB/sec
Scale 73.37 1515.87 MB/sec
Add 101.27 2157.19 MB/sec
Triad 101.45 2170.17 MB/sec

Stock Mini:
Results 47.18
System Info
Xbench Version 1.2
System Version 10.4.5 (8H1619)
Physical RAM 512 MB
Model Macmini1,1
Drive Type FUJITSU MHV2060BHPL
Memory Test 93.42
System 92.52
Allocate 98.51 361.76 Kalloc/sec
Fill 85.09 4137.48 MB/sec
Copy 95.04 1962.97 MB/sec
Stream 94.33
Copy 79.33 1638.50 MB/sec
Scale 76.86 1587.92 MB/sec
Add 121.31 2584.26 MB/sec
Triad 117.03 2503.50 MB/sec


-Seed

This Mac novice Just bought a single core mini right before the dual cores came out. Is it upgradable to dual core?
Thanks for any clarification.

The only upgrade I'm aware of costs $799.00 (or as low as $19 a month).

Yes it is upgradable and in my opinion you novice made a smart decision. You'll be able to upgrade to a faster dual core that wasn't available from the factory and later on even consider going straight to the 64bit dual Merom chip.

You have options and you saved 200 bucks. Right now the 1.66 chip is about 250 from a retailer. Give it time and you can be ahead.

-Seed

I need to reply to my own post but I have to correct myself. I did not take enough samples and found the Gigaram to be on par with the factory SODIMMs. So I will say that I have yet to see any fact that matched dimms are faster than mismatched on these minis. These results are comparable:

Here is one with asymmetric RAM:
"Feri’s Mac mini"
Physical RAM 1280 MB
Model Macmini1,1
Memory Test 91.02
System 97.47
Allocate 102.16 375.18 Kalloc/sec
Fill 94.20 4580.02 MB/sec
Copy 96.40 1991.16 MB/sec
Stream 85.36
Copy 74.08 1530.07 MB/sec
Scale 73.37 1515.87 MB/sec
Add 101.27 2157.19 MB/sec
Triad 101.45 2170.17 MB/sec

2GB Core Solo
Physical RAM 2048 MB
Model Macmini1,1
Memory Test 98.21
System 95.52
Allocate 94.48 346.97 Kalloc/sec
Fill 89.98 4375.19 MB/sec
Copy 102.97 2126.87 MB/sec
Stream 101.06
Copy 84.26 1740.44 MB/sec
Scale 83.23 1719.51 MB/sec
Add 129.35 2755.45 MB/sec
Triad 125.52 2685.18 MB/sec



there is a guy who added a 1GB stick to his mini along side a 256 and saw minimal performance degradation. I would expect numbers to be far less
if you're running half the bandwidth.

-Seed

Seedling,
The memory doesn't need to be matched in any special way. It's not a specialized procedure of finding two modules which work together well. You just need to use two modules with identical specs, which is simplest if you just buy two identical modules. (Speculation: it might even work correctly if the two sizes are the same, but some of the timing specs differ, such as CAS latency. The system should run the RAM using the timing of the slower chip, Dual channel should work. This is speculation, would need to be tested.)

You're right, running a pair of identical DIMMS doesn't double RAM bandwidth, but your test results show a significant performance improvement. (though I'm curious about the Allocate result - might be an OS thing there)

My biggest concern about using two different-sized modules is the video RAM allocation mentioned a few posts up from here.

I'd use two identical modules. If money was tight, I'd install 2x512M rather than a 256M and a 1024M.

Thanks for the post tele but I'm not sure you're entirely correct. In the PC world, please don't beat me up for that, I am fluent in both, if one DIMM with same specs is of the same specs as a DIMM of a different size they will not run in dual channel mode. At least that's how I recall it mixing DIMMs and checking how the BIOS reads it. I can't imagine it's any different for OS X 'cause it's not really OS related.

Also, dual channel configuration should have twice the bandwidth, I mean that's the point of dual channel.



DDR2 5300 should be getting 5.3GBs per second in single channel mode and 10.6 or near in dual. I have not seen any numbers exceed 5GBs so I can't help to understand why, and I think I found the answer.



Actually it could be a limit of the FSB of the processor, from reading, the Yonah chip runs at 667 FSB, so the maximum memory speed that the chip could take advantage of is 667 x .8 or 5.3GBs per second, per core! So us Solo's can't take advantage of dual channel.

So with that said, with DDR2 667 even in single channel mode you won't see much of a difference because even in single channel the memory runs at the same speed that the CPU can take advantage of it.

Dual CPUs? Now that is a different story.

There I can see the perfect need for dual channel being needed as both cores can run at full efficiency accessing memory without any penalty of access from the other core, at least I think that's how it works. So with that said I guess Xbench can only run as one application and run results accessing only one of the cores so it will report the same as a single core performer.

Would love to see Xbench upgrade their software to run two benches at the same time one each core to see the memory bandwidth, then I bet the throughput would be cut by 50% if the system is running in single channel mode.

-Seed

Correct about matching.
You need to bear in mind that the RAM's bandwidth is being shared between the CPU and the GPU. So, the CPU (single or dual-core) alone won't be able to saturate the RAM's bus. That would starve the GPU, which would be bad.

My point was that running the mini's RAM in dual-channel mode has a significant performance benefit, at very little additional cost. Ignoring Xbench's cumulative RAM score, most of the individual RAM tests are about 15-25% faster in dual-channel mode, the allocate result skews the cumulative score down. It's probably not worthwhile to compare the Xbench results against the theoretical max speed of the RAM, since there are several other important factors - such as the CPU/GPU sharing, and the effect of the L1 and L2 caches.


Quad G5 2.5Ghz 4.5GB 2x250G, PB 15" 1.5Ghz,80G,1.5G Mac OS X (10.4.5)