Resolved To mix or not to mix (memory)....

I currently have a Sandy Bridge CPU and 4GB of 1600 MHz DDR3 memory, and am running Windows 8.

I plan to upgrade to a (faster) Ivy Bridge and at the same time invest in 8GB of 2133 MHz memory. The new memory-sticks will occupy two of the four DIMM slots. I plan to install a RAM-drive at the same time to use mainly as a scratch-disk when encoding video, probably of 4 GB.

Optimally I would of course invest now in 16GB of 2133 MHz, but my budget doesn't stretch to getting me there in one step. So what should I do in the meantime with my existing two 1600 MHz sticks - leave them in place?

12 GB of mixed-spec memory ought, surely, to offer better performance than 8 GB of (unmixed) superfast memory? But my mobo's manual says: "The system maps the total size of the lower-sized channel for the dual-channel configuration. Any excess memory from the higher-sized channel is then mapped for single-channel operation ". This sounds distinctly like throwing away the benefit of the super-fast memory's performance.

I simply don't know what is my best interim arrangement (until such time as I can afford to buy the other 8GB of 2133 MHz). Can someone who knows please guide me?

 

I take the point.

Nevertheless, won't it make a noticeable difference in the time taken to transcode , say, a 4 GB video file? Or, alternatively, won't it mean that higher quality is available for an equivalent time spent than with slower memory? Otherwise, why would people bother?

As I read what the manual says, all 8 GB of the faster memory would be mapped as single channel - ie its speed will be halved. Or have I got that wrong?

 

Hi Evan

Yes, that's the bit I follow. As I understand it I'd have 12 GB running at 1600 MHz (until such time as I can afford to replace the two !600 MHz sticks with two more 4 GB 2133 MHz sticks). I could live with that quite happily - a 300% increase in the size of my RAM compared with now, but with unchanged speed.

The bit I don't understand at all is what my mobo's manual means by saying that "any excess memory from the higher-sized channel" (ie the two 4GB 2133 MHz sticks) "is then mapped for single-channel operation ". What does "single-channel operation" mean, and what is its (by implication deleterious) effect upon performance? Will I lose on the roundabouts as much as or more than I'll gain on the swings?

 

Thanks for that clarification.

Conclusion?

Let's see if I've got this right. I would have my existing 4 GB of 1600 MHz memory operating - unchanged - in dual-channel configuration. Alongside that I'd have 8GB of 2133 MHz memory running at the same clock speed (1600 MHz) but at only half the rate (single- instead of dual-channel). Putting that together the net effect sounds to me to be exactly the same as if I had added another 4GB of 1600 MHz memory.

Is that correct?

If so, on the face of it I'd have done better to just double the existing memory. However, the purchase of the 8GB of 2133 MHz memory is only an interim step towards the eventual goal of 16GB at 2133 MHz, whereas to add another 4GB of 1600 MHz would be marking time at best - nor is the price difference that great.

I conclude that I should stick with my plan because at least the effect is to double my memory's capacity, but without adding to the amount which is due to become redundant at some - hopefully not too distant - point.

Do you see any flaw in that reasoning, may I ask?

 

And one more question, if I may.

Would I be likely to suffer any significant hit on latency through mixing memory in this way?

My existing 1600 MHz is shown by CPU-Z to be running at 8-8-8-24.

(I have XMP activated in the BIOS)

 

You don't mention Brand or type of RAM and yes the CAS Latency will probably be different unless you select matching RAM. Timing and Voltage that the RAM operate at should be matched to get best performance. Neil.

 

I have in mind G.Skill RipjawsX Gaming 2 x 4GB

 

That is a real show-stopper.

The only way, it seems, to avoid risking having wasted my money is to change nothing - at least not until I can afford to change completely.

In the meantime I could of course double my existing memory without running any risk. But then I'll have wasted whatever I spend on doing that once the time comes to replace it all.

Oh dear, why can't things be simpler! Is there really no way of knowing beforehand?

 

I say try it and see what happens. I would certainly install the faster pair on in the first channel and the slower in the second - if your particular memory management module is smart enough, it may allow your first pair to run at faster speeds - if voltage specs are the same. If not, as a general rule, adding more RAM provides a greater performance boost than adding faster RAM (of course, there are exceptions to every rule, but typically, more is better than faster when it comes to RAM.

As far as wasting money - that's just the cost of keeping up with technologies. I have a box full of old, but perfectly good memory that were removed to make room for bigger sticks.

I don't see any advantage to removing 4Gb from your RAM pool - especially for a scratch disk for video encoding - unless, maybe that was my job and I did that all day long and needed the isolation. Or maybe if I had 24 - 32Gb of RAM, instead of just 12Gb. Windows 8 is, by far, the best Windows yet for efficiently utilizing memory resources. I would rather buy a small SSD to add to my system and move the page file to it rather than steal RAM from my CPU and OS and dedicate it to a RAM drive that will see limited use.

Of course I am assuming you have 64-bit Windows 8. You have not specified which bit version you have. If 32-bit, you are stuck with your 4Gb and this discussion is moot.

 

You assume correctly.

And many thanks for the very helpful advice concerning RAM-drive. I got enthused by the speed aspects and was unaware of the downside - it never seems to get any mention.

 

RAM drives were great years ago when most users were limited to 4Gb of RAM due to a 32-bit OS, and when hard drives were very slow.

But today, 64-bit Windows outsells 32-bit by a long shot, and most computers are coming with more than 4Gb of RAM. Plus, hard drives come with much larger buffers (typically 32Mb, often 64Mb when back in the day, 2Mb was large).

One big advantage with a RAM disk is security. If working on a classified document, using it with a RAM disk helps ensure no copy of the document is saved on the hard drive - which might be recoverable by a badguy if he gets his hands on the drive. When using a RAM disk, all data is dumped out of RAM when the computer is powered off. But I highly suspect you are working with classified documents.