Can someone explain why 2133MHz RAM is working on 2011 MBPs that should only take 1600MHz?

[JTToft]

So can someone please explain the reports that 2133MHz modules are working AND running at that speed in Late 2011 MacBook Pros (both 15" and 17") AND providing a visible performance increase?

See this thread (Geekbench scores included): RAM-upgrade-question in a 17" 2011 Macbook Pro
And this thread: late 2011 mbp ram

According to Intel, the Core i7 in these machines should only accept up to 1600MHz.
http://ark.intel.com/products/53476/Intel-Core-i7-2860QM-Processor-8M-Cache-up-to-3_60-GHz

Colour me confused. What's going on here?

 

[simonsi]

Because RAM is designed to run at the clock speed of the device, its rated speed is a maximum. RAM is normally fine 1 or 2 clock speeds down from its rated speed.

The reason they see a small performance increase under test conditions (you are unlikely to notice anything in real-life), is because faster RAM has a faster settle time, ie once the CPU selects a location the RAM controller on the modules will flag that the data is ready to be read by the CPU. Faster RAM has a shorter/quicker settle time before flagging data ready to the CPU, on a slower machine this can easily mean the data is ready quicker and the CPU takes advantage of it by reading it faster after requesting it (ie 1 or more CPU clock cycles before it would have read slower RAM), hence a small speed increase.

 

[JTToft]

Because RAM is designed to run at the clock speed of the device, its rated speed is a maximum. RAM is normally fine 1 or 2 clock speeds down from its rated speed.

- I'm with you that far. So it would be your contention (as well as mine initially in those threads) that the 2133 RAM isn't actually running at 2133 but has been downclocked to the 1600 maximum that the machine can handle, and, by extension, that About This Mac/System Information as well as Geekbench doesn't report the speed that installed RAM is running at but instead the maximum speed that the modules are capable of running at independently of the machine they're installed in?

If this is the case, I have a question: How can you know which speed the modules you have installed are actually running at?

 

[AZhappyjack]

- I'm with you that far. So it would be your contention (as well as mine initially in those threads) that the 2133 RAM isn't actually running at 2133 but has been downclocked to the 1600 maximum that the machine can handle, and, by extension, that About This Mac/System Information as well as Geekbench doesn't report the speed that installed RAM is running at but instead the maximum speed that the modules are capable of running at independently of the machine they're installed in?

If this is the case, I have a question: How can you know which speed the modules you have installed are actually running at?

I am not sure where you might see that information (actual speed), but clearly, the machine (processor/bus/internals) cannot run faster than they can go, regardless of what memory is installed. So installing faster memory will not allow your computer to run faster than it does with memory that matches the machine specs.

I also have experience with some older machines and some memory brands that don't do the down-throttling... and those combinations will give you memory errors, memory faults and kernel panics.

 

[simonsi]

- I'm with you that far. So it would be your contention (as well as mine initially in those threads) that the 2133 RAM isn't actually running at 2133 but has been downclocked to the 1600 maximum that the machine can handle, and, by extension, that About This Mac/System Information as well as Geekbench doesn't report the speed that installed RAM is running at but instead the maximum speed that the modules are capable of running at independently of the machine they're installed in?

If this is the case, I have a question: How can you know which speed the modules you have installed are actually running at?

RAM doesn't run at its own speed, it runs at the Bus speed set by the CPU, so if the CPU supports 1600MHz, that is the speed it runs at, slower RAM will cause an error, faster RAM will be downclocked. Hence there is no need to measure RAM actual speed. The RAM modules provide ID info that enable the CPU to tell if they are suitable, I'd expect About this MAC reports and decodes that ID info.

 

[leman]

I wouldn't be surprised if the RAM actually runs at 2133MHz. It is quite possible that the CPU "supports" it, but simply isn't rated for it (e.g. because those RAM speeds were considered experimental when the CPU was tested or maybe because there was some instability, or maybe because it simply wasn't tested at all). Another explanation, as already mentioned by others, could be improved RAM timings (but I doubt that would show much difference).

 

[JTToft]

RAM doesn't run at its own speed, it runs at the Bus speed set by the CPU, so if the CPU supports 1600MHz, that is the speed it runs at, slower RAM will cause an error, faster RAM will be downclocked. Hence there is no need to measure RAM actual speed. The RAM modules provide ID info that enable the CPU to tell if they are suitable, I'd expect About this MAC reports and decodes that ID info.

- I'm not so sure. For there not to be any need to measure it, we would need to know for sure what the maximum speed that the CPU supports is. And who's to say that spec page I linked to isn't wrong or at least outdated? We all know manufacturer specs can be inaccurate or only report the supported features at the time of sale, as in the case of Apple reporting a maximum of 1333MHz on these machines, or reporting a maximum of 8GB on many of their machines, even though they can take 16GB just fine.

I'd like also to hear your thoughts on leman's idea above.

 

[Samuelsan2001]

I it works, it's not making anything worse, what more do you need to know....

 

[JTToft]

I it works, it's not making anything worse, what more do you need to know....

- Have you ever heard of acquiring knowledge for the sake of the knowledge itself (pretty much the definition of higher education)? I needn't have a practical application for the knowledge I'm attempting to acquire in order for it to be sensible for me to acquire that knowledge.
That said, there is a practical application: Knowing whether there is any sense in spending extra money on 2133MHz modules rather than 1600MHz modules for these machines.

 

[ron1004]

I it works, it's not making anything worse, what more do you need to know....

There's a marked difference in price between the 1600 and the 2133 - I'd not pay the extra for the 2133 if it was confirmed that it runs at the same speed as the 1600 in my late 2011 17".
[doublepost=1454613785][/doublepost]I'll be receiving the below 1600 RAM in a couple of days.

The latency is 9 on this RAM vs 11 on the 2133 - my guess is that the latency plays a role in performance, but have no idea how much.

G.SKILL Ripjaws Series 16GB (2 x 8G) 204-Pin DDR3 SO-DIMM DDR3L 1600 (PC3L 12800) Laptop Memory Model F3-1600C9D-16GRSL