MacPro early 2009 memory - Speed vs. cas latency

[toddbe]

I am getting an early 2009 Two - 2.66GHz Quad-Core and looking at different possible solutions for 16GB of ram.

1- OWC - $290
Data Rate = 1066MHz
CAS 7-7-7-20

All the following from Tigerdirect-
2- Corsair - $300
Data Rate = 1333MHz
CAS 9-9-9-24

3- Corsair - $340
Data Rate = 1600MHz
CAS 9-9-9-24

4- Patriot Viper - $360
Data Rate = 1600MHz
CAS 7-7-7-20

I know the MacPro is set to use a Data Rate of 1066MHz. Would there be a benefit to using faster memory? I believe the faster cas latency is an advantage also.

 

[nanofrog]

I am getting an early 2009 Two - 2.66GHz Quad-Core and looking at different possible solutions for 16GB of ram.

1- OWC - $290
Data Rate = 1066MHz
CAS 7-7-7-20

All the following from Tigerdirect-
2- Corsair - $300
Data Rate = 1333MHz
CAS 9-9-9-24

3- Corsair - $340
Data Rate = 1600MHz
CAS 9-9-9-24

4- Patriot Viper - $360
Data Rate = 1600MHz
CAS 7-7-7-20

I know the MacPro is set to use a Data Rate of 1066MHz. Would there be a benefit to using faster memory? I believe the faster cas latency is an advantage also.

You'd get better performance from CL = 7. Since the MP is only rated for 1066MHz, it's the less expensive option, and recommended. It's also Unbuffered ECC, while the Patriot Viper is not.

However, CL = 7 Unbuffered ECC @ 1333MHz, 2 or 4GB sticks should be available soon, if not already. (You'd need to swap out the stock memory though).

 

[KBS756]

You'd get better performance from CL = 7. Since the MP is only rated for 1066MHz, it's the less expensive option, and recommended. It's also Unbuffered ECC, while the Patriot Viper is not.

However, CL = 7 Unbuffered ECC @ 1333MHz, 2 or 4GB sticks should be available soon, if not already. (You'd need to swap out the stock memory though).

would 1333MHz or 1600MHz have any speed benefit? or be killed by the fact its rated for 1066

 

[canonballs]

would 1333MHz or 1600MHz have any speed benefit? or be killed by the fact its rated for 1066

while you are pondering these, could there be any reliability benefit?

 

[nanofrog]

would 1333MHz or 1600MHz have any speed benefit? or be killed by the fact its rated for 1066

The 1333MHz, Yes in your case, as Intel designed the X5550 (2.66GHz) or higher clocked parts to work with 1333MHz clocked DDR3 (any variety; non ECC, Unbuffered, and Registered). On the lower, E5520 (2.26GHz), No, it won't. It was only designed for up to 1066MHz memory.

The 1600MHz variety won't do you any good unless you can overclock the MP. And none of the 1600MHz or higher is ECC yet. Only non ECC is currently available. I don't think this will change, as workstation and server parts aren't commonly OC'd systems. Not directly from system vendors at any rate.

while you are pondering these, could there be any reliability benefit?

Not really.

Specifics would depend on the grade of memory chips used. Just look for a Lifetime warranty, and you should be OK, keeping in mind, ECC isn't really intended for Overclocking. It may be possible, but this is when the grade (binning), really counts. No way to know without the actual memory IC part numbers.

 

[sidewinder]

The 1333MHz, Yes in your case, as Intel designed the X5550 (2.66GHz) or higher clocked parts to work with 1333MHz clocked DDR3 (any variety; non ECC, Unbuffered, and Registered).

Not likely true.

If Apple has hard wired the system to support a maximum of 1066MHz memory, it doesn't matter what the processor is capable of doing. Since the Apple specs say "1066MHz DDR3 ECC SDRAM", I suspect that buying 1333MHz RAM would be a waste of money.

S-

 

[nanofrog]

Not likely true.

If Apple has hard wired the system to support a maximum of 1066MHz memory, it doesn't matter what the processor is capable of doing. Since the Apple specs say "1066MHz DDR3 ECC SDRAM", I suspect that buying 1333MHz RAM would be a waste of money.

S-

If Apple did limit it, it wouldn't be the wiring, but rather a limit set in the EFI. Additional effort, for what gain?

 

[sidewinder]

If Apple did limit it, it wouldn't be the wiring, but rather a limit set in the EFI. Additional effort, for what gain?

nanofrog,

I used "hard wired" as a euphemism and not in the literal sense. Given the subject, that may not have been the best choice.

Regardless, why would Apple publish "1066MHz DDR3 ECC SDRAM" if that was not, in fact, as fast as the RAM would be clocked?

S-

 

[nanofrog]

nanofrog,

I used "hard wired" as a euphemism and not in the literal sense. Given the subject, that may not have been the best choice.

Regardless, why would Apple publish "1066MHz DDR3 ECC SDRAM" if that was not, in fact, as fast as the RAM would be clocked?

S-

My thinking for the 1066MHz limit is more of a cost/parts availability issue. Less expensive, and keeping a single/fewer quantity allows for simplified parts bins. Also has the ability to simplify assembly, particularly if the memory installation is by hand.

 

[sidewinder]

My thinking for the 1066MHz limit is more of a cost/parts availability issue. Less expensive, and keeping a single/fewer quantity allows for simplified parts bins. Also has the ability to simplify assembly, particularly if the memory installation is by hand.

Well, before you say what people should buy, wouldn't it be prudent know the actual answer?

S-

 

[eelmac]

Now I ain't too great with Mac related Hardware but from building computers in the past I have noticed some computers won't boot if the speed of the ram was higher than the motherboard would allow, in other cases it just slowed the ram down, I have no idea what it will do in a Mac, I am assuming it will just slow down. :/

Don't quote me tho, things have changed since i last had that issue

 

[nanofrog]

Well, before you say what people should buy, wouldn't it be prudent know the actual answer?

S-

It wouldn't be the first time specs have intentionally been under published.

Usually due to parts unavailable at the time of publishing. In this case, if they're waiting for 1333MHz @ CL = 7 timings, this is what I'd expect. (I happen to think this is likely).

I just don't necessarily agree with it in this case. The user should have the full knowledge of what the system can actually do, and make their own decisions. If they decide they can live with CL = 9 timings, it should be their call, not Apple's.

Now I ain't too great with Mac related Hardware but from building computers in the past I have noticed some computers won't boot if the speed of the ram was higher than the motherboard would allow, in other cases it just slowed the ram down, I have no idea what it will do in a Mac, I am assuming it will just slow down. :/

Don't quote me tho, things have changed since i last had that issue

Worst case, it would run at a lower clock (1066MHz). I seriously doubt the time and effort to intentionally disable the 1333MHz speed was done. Design phases are short as is, and the additional time and resources would be a waste for no gain. It would only serve to hinder users, not give them a performance boost, or hinder the competition in any manner.

I just can't see any real logic for doing so.

 

[sidewinder]

It wouldn't be the first time specs have intentionally been under published.

Usually due to parts unavailable at the time of publishing. In this case, if they're waiting for 1333MHz @ CL = 7 timings, this is what I'd expect. (I happen to think this is likely).

I just don't necessarily agree with it in this case. The user should have the full knowledge of what the system can actually do, and make their own decisions. If they decide they can live with CL = 9 timings, it should be their call, not Apple's.

Not this again. The system should do what Apple decides it wants it to do.

Regardless, you don't have any real idea if the new Mac Pro systems using the 5500 series Zeons will use 1333MHz memory at 1333MHz speeds. So you should not be telling anyone that they should buy 1333MHz memory until you know that answer.

S-

 

[VirtualRain]

It would be nice if the system just read the SPD from the memory modules and ran it at rated speeds up to the maximum supported by the memory and the memory controller on the CPU. Someone's going to need to try running some higher speed lower latency memory in a Mac Pro and then letting us know if it took to the tighter timings and/or higher frequency.

I look forward to hearing how this works out on the MP.

BTW, from my days working on PC's, just plugging in memory and assuming it's running at rated performance (speed and timings) is a bad idea. Often the RAM SPD is programmed for more "compatible" slack settings and needs to be manually tweaked to achieve maximum or even rated performance. If we don't have access to the EFI to adjust timings, I guess you need to ensure your memory SPD's are programmed for the best performance. I think OCZ makes an SPD programming tool.

Is there any utility that will tell you what your memory timings are under OSX?

 

[sidewinder]

Is there any utility that will tell you what your memory timings are under OSX?

Yes, it's called "System Profiler".

S-

 

[VirtualRain]

Yes, it's called "System Profiler".

S-

Thanks!

 

[Bartman01]

edited - not sure how this post showed up as a 'current' post for me since it is several months old.

Nothing to see here, move along...

 

[Gonk42]

Not this again. The system should do what Apple decides it wants it to do.

Regardless, you don't have any real idea if the new Mac Pro systems using the 5500 series Zeons will use 1333MHz memory at 1333MHz speeds. So you should not be telling anyone that they should buy 1333MHz memory until you know that answer.

S-

It would be interesting to know the answer to this but I think nanofrog is correct. My understanding is that the memory controller in the 5500 chips communicates with the RAM chips directly so for the 2.66GHz parts if it
was connected to 1333MHz parts (in a single bank) it would operate at
that speed. Apple would have to go to some lengths to sabotage what
Intel has built in.

As to Apple's specs, they specced memory that was common to all Mac Pros, both W3500 series and lower frequency W5500 series which is 1066MHz. It makes life a lot simpler - most users wouldn't notice any difference anyway especially given that the latency is the same (9 at 1333MHz corresponds to 7 at 1066MHz).

 

[nanofrog]

My understanding is that the memory controller in the 5500 chips communicates with the RAM chips directly so for the 2.66GHz parts if it
was connected to 1333MHz parts (in a single bank) it would operate at
that speed. Apple would have to go to some lengths to sabotage what
Intel has built in.

Yep. The Memory is attached directly to the CPU, which contains the IMC (Integrated Memory Controller). It is designed to handle both non ECC and ECC memory in both Unregistered and Registered versions. Speed is up to 1333MHz.

The person that tried 1600MHz non ECC got 1066MHz, and that is what would happen. Check out the various forums that are dedicated to Over Clocking. That's the factory default speed that would occur for that memory. You have to get into the board's firmware and set it to use the faster clock. Unfortunately, this isn't possible on the '09 MP's, and it's stuck at 1066MHz on that particular memory.

But I think that 1333MHz would run at the designated clock speed, as the controller would recognize it. That's what I've been hoping to see using UDIMMs.

As to Apple's specs, they specced memory that was common to all Mac Pros, both W3500 series and lower frequency W5500 series which is 1066MHz. It makes life a lot simpler - most users wouldn't notice any difference anyway especially given that the latency is the same (9 at 1333MHz corresponds to 7 at 1066MHz).

Intel's stock settings for DDR3 are in three flavors; 800, 1066, and 1333MHz. The latter is expensive, and is likely the primary reason it was skipped. If it only added a few bucks, they likely would have added it to begin with, as it would have set it apart from other systems, no matter if the clock speed would have improved performance or not (provided it wasn't a decrease, given the latency differences).