Processor is now faster, but memory is slower?

[beanzontoast]

The specs of base unit I'm looking at buying have recently changed - the processor has been 'improved' to a 4x3.3GHz i5 but the memory used in the system has changed from 4GB of 1666MHz to the same amount of 1333MHz.

To my simple way of thinking, would it not have been better with the 1666MHz? Or is there some interaction between the processor and memory to tkae into account which means it will work better with 1333Mhz?

 

[2Loose]

The specs of base unit I'm looking at buying have recently changed - the processor has been 'improved' to a 4x3.3GHz i5 but the memory used in the system has changed from 4GB of 1666MHz to the same amount of 1333MHz.

To my simple way of thinking, would it not have been better with the 1666MHz? Or is there some interaction between the processor and memory to tkae into account which means it will work better with 1333Mhz?

For optimal memory performance the memory speed must normally be a 'whole' divisor of the speed of the CPU. 3.9Ghz CPU \ 1.3GHz RAM (x3) would give similiar memory performance to a 3.3GHz CPU \ 1.6Ghz RAM (x2) setup.
IIRC the i5 uses triple ram channels rather than the more common twin, so I think that this calculation isn't quite as simple in reality.

A motoring analogy may be that a fuel pump can provide more petrol than an engine can burn at once (flooding), or alternatively that the engine could use more fuel than the pump can supply (fuel starvation). The parts should match for optimal performance.

 

[beanzontoast]

Thanks 2L. So, in this specific instance of a 4x3.3GHZ processor, do you think it will work out as an advantage having 1333 rather than 1666?

Edit - from a bit of reading I've now done on this, it appears that the processor only supports up to 1333 for this very reason.

 

[2Loose]

If you had 1666MHz RAM in the system it would be running at the slower 1333, so no advantage either way in this case.
(unless you count less cost as an advantage.)

 

[marinyork]

Memory tends to scale very well on all intel chips post Core (original). It's not really an issue. Was much more of an issue for a while on AMD stuff.

 

[GrasB]

There are many factors which define what ram is better. Normally though ram clock speed isn't a real bottle neck, it's usually the HDD/SSD & amount of ram which makes the computer feel fast or not. We recently evaluated new workstations for students at work, a 25% faster RAM & CPU clock speed was overwhelmed by an SSD system drive & 24GB instead of 6GB ram.

There is even the odd possibility the ram might actually work faster at the lower speed. The number ram is given tends to be bus speed, but only defines the transfer rate of the data once it's retrieved. There's this thing called latency, it's noted in clock cycles & defines how quickly you can access the data stored in ram. It gets complicated but thing is it's possible that at 1666MHz you have to to wait longer to access the data than at 1333MHz, so while the faster ram can transfer more data to the CPU you can't actually use that extra bandwidth effectively.

 

[marinyork]

Any overclocking site tells you that!

 

[Jezston]

For optimal memory performance the memory speed must normally be a 'whole' divisor of the speed of the CPU. 3.9Ghz CPU \ 1.3GHz RAM (x3) would give similiar memory performance to a 3.3GHz CPU \ 1.6Ghz RAM (x2) setup.
IIRC the i5 uses triple ram channels rather than the more common twin, so I think that this calculation isn't quite as simple in reality.

A motoring analogy may be that a fuel pump can provide more petrol than an engine can burn at once (flooding), or alternatively that the engine could use more fuel than the pump can supply (fuel starvation). The parts should match for optimal performance.

I disagree.

Memory 'speed' in terms of Hz is only relevant when doing continuous transfers of data between the RAM and other hardware (mostly the CPU). In practice, this means your PC doing a big 'job' like rendering video or suchlike where you'd walk away and leave it to it.

For day to day tasks the latency of the RAM, i.e. how quick to respond it is, has the greater effect. Cheaper higher speed RAM will typically have a higher latency than similarly priced lower speed RAM.

In practice, neither is likely to make a noticeable difference.

 

[2Loose]

Jezton, I agree that memory latency has a marked affect on performance especially for things such as gaming, video conversion / rendering etc.

For RAM to run at it's peak (manufactured) latency it also should be running at its optimal speed though, as overclocking memory normally increases latency, rather than decreasing it. Better quality memory will normally have less latency at the same bus speed (at a greater expense). Also, good quality 1333 ram will often have less latency than good quality 1666 ram when run at 1333...

In fact, I agree with all of your post...

especially the bit about not noticing unless running benchmarks!

which bits of mine did you disagree with?