Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Raunien

macrumors 6502a
Original poster
Aug 3, 2011
540
63
Hi everyone,

I had a 3x8 gb RAM set up with a 6core - 3.33 ghz 5,1. The ram was running at 1333 mhz. The actual ram modules were 1600 mhz non ecc.

I went and bought a 4th 8gb ram module today that was matched the others and put it in my machine and now the RAM is running at 1066.

I've seen others on this forum have 4x8 1333 mhz ram on a single CPU machine.

Anyone have any ideas?
 
Do some digging, this has been explained before but I don't remember the reason, only know that it is here.

I've pretty much read every single thread on this issue. But, basically it seems that the people who can get it to run at 1333 are people who have exactly 1333 - ECC ram.

I have 1600 non-ecc RAM so that could be the reason, but I'm wondering if there's anything to be done. I don't want to buy a whole new ram kit.

Basically, will I see a noticeable difference between 32gb-1333 and 32gb-1066? My main usage is massive multitasking with huge pdf's, VM's and millions of chrome tabs
 
Did you buy quad ranked RAM? I think quad-ranked RAM downclocks in certain circumstances but I cannot remember exactly.
 
I could be wrong, but I think as soon as you dont use identical memory all the way through, you'll lose 1333GHz speed.

I have 128GB and it still does 1333, despite using the 4th and 8th bank, but all the memory is identical.
 
All my memory is essentially identical. The 4th ram stick that I bought is the same as the other 3
 
I think it is. This is what I have deduced from everything that's been posted.

Now the question is: Is the difference between 1066 and 1333 big enough to justify going out to buy new ECC ram? (I think the answer is no, but someone else prove me wrong)
 
All my memory is essentially identical. The 4th ram stick that I bought is the same as the other 3
Being "essentially identical" and bing identical are not the same thing. It's been a long time sense I fooled with Ram, but from what I remember Ram sticks have an SPD that sets the timing parameters of the Ram during the boot sequence of the PC.

It is posable that you have an SPD conflict between the EFI firmware of your Mac and your new Ram stick, or two or more of your Ram sticks.

Try removing all the Ram, but the new Ram stick, if they don't need to be installed in pairs, and booting your Mac with only that one and see if it reports 1333Mhz or 1066Mhz.

There are a few tools for Windows for reading, editing, and flashing SPD's, but it's advanced stuff and can kill your Ram stick.
 
Being "essentially identical" and bing identical are not the same thing. It's been a long time sense I fooled with Ram, but from what I remember Ram sticks have an SPD that sets the timing parameters of the Ram during the boot sequence of the PC.

It is posable that you have an SPD conflict between the EFI firmware of your Mac and your new Ram stick, or two or more of your Ram sticks.

Try removing all the Ram, but the new Ram stick, if they don't need to be installed in pairs, and booting your Mac with only that one and see if it reports 1333Mhz or 1066Mhz.

There are a few tools for Windows for reading, editing, and flashing SPD's, but it's advanced stuff and can kill your Ram stick.

I removed all the ram and booted with 2 old ram sticks +1 new ram stick. It registered as 1333 mhz.

I then booted with only the new ram stick and it registered as 1333 mhz.
 
Would you do it? Get ECC to increase the speed to 1333 mhz bearing in mind that the latency is higher with 1333 mhz and ECC is also slower

I suppose only if I needed ECC Ram, but my computing needs aren't that high. I compile code, and I've never had a compile fail due to a Ram error.

It's $200 bucks over at OWC for 4x8GB sticks of ECC so not a great deal of money.
 
I will be very very surprise if you can tell the speed difference without look at the system info / benchmark. So, in real world, there should be no performance penalty for your usage. And more RAM = more cache, in fact it may speed up your system rather than slow down, even the RAM down clock from 1333 CL9 to 1066 CL7.

THB, your workflow doesn't look like memory bandwidth sensitive. Even if it does, the main speed difference will because you didn't optimise the triple channel architecture, but not the RAM running at 1066.

For most users, the main performance hit comes when running out of RAM, VM can use up lots of RAM easily. However, VM is memory size demanding, but not that memory speed sensitive. So, in your particular case, I assume 32G 1066 is actually a better choice than 24G 1333. And it can speed up rather than slow down your Mac under heavy multitasking enviournment.

If you really want more RAM with max speed, you have to go for 3x16G RAM. But again, I will be very surprise if you can really tell the difference (before your system run out of RAM).

For your info. A simple way to calculate the time between each clock tick in the RAM is

(CL / Frequency in MHZ) * 1000) = speed in ns

If you put all the numbers in, then you will realise that the theoretical speed difference between 1066 CL7 and 1333 CL9 is just about 2.7%. And guess what? ECC RAM is about 2% slower than non ECC RAM (reliability won't come for free, in both money and speed). So, if you "upgrade" from 1066 CL7 non ECC to 1333 CL9 ECC RAM, even in theory, you can't get more than 1% speed gain. However, you will have a more reliable system (if that's a matter for your workflow).
 
Last edited:
  • Like
Reactions: Raunien
I will be very very surprise if you can tell the speed difference without look at the system info / benchmark. So, in real world, there should be no performance penalty for your usage. And more RAM = more cache, in fact it may speed up your system rather than slow down, even the RAM down clock from 1333 CL9 to 1066 CL7.

THB, your workflow doesn't look like memory bandwidth sensitive. Even if it does, the main speed difference will because you didn't optimise the triple channel architecture, but not the RAM running at 1066.

For most users, the main performance hit comes when running out of RAM, VM can use up lots of RAM easily. However, VM is memory size demanding, but not that memory speed sensitive. So, in your particular case, I assume 32G 1066 is actually a better choice than 24G 1333. And it can speed up rather than slow down your Mac under heavy multitasking enviournment.

If you really want more RAM with max speed, you have to go for 3x16G RAM. But again, I will be very surprise if you can really tell the difference (before your system run out of RAM).

For your info. A simple way to calculate the time between each clock tick in the RAM is

(CL / Frequency in MHZ) * 1000) = speed in ns

If you put all the numbers in, then you will realise that the theoretical speed difference between 1066 CL7 and 1333 CL9 is just about 2.7%. And guess what? ECC RAM is about 2% slower than non ECC RAM (reliability won't come for free, in both money and speed). So, if you "upgrade" from 1066 CL7 non ECC to 1333 CL9 ECC RAM, even in theory, you can't get more than 1% speed gain. However, you will have a more reliable system (if that's a matter for your workflow).

Thanks alot H. I really appreciate your insight. I guess I will hold off on purchasing anything for now.
 
There's no point in using the Xeon without ECC memory.

I know several people who use Xeon CPUs with non ECC RAM, and reason is very simple: cost effective way to get more CPU cores/threads, because you can't use multiple i5/i7s (single QPI).
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.