Maximum PC

Does the dual channel mode make any difference?

Hey DiyGuy, welcome to the forums!

Having memory in dual-channel mode is concept which one would think would make a bigger difference than it does.

The basic idea is that memory is written across the two sticks, and thus the load on an individual stick and it's slot / interface is cut in half, allowing for a speedup.

It doesn't really seem to work as great as promised, the numbers I remember hearing are around 4% in standard configurations.

However, I've also heard that in use with integrated graphics, it can actually have a much more measurable difference. The downside of course is that most enthusiasts who would be inclined to use Dual-Channel configs are not inclined to use integrated graphics.


BTW, the above info on Dual-Channel doesn't necessarily reflect that of newer standards such as Triple-Channel, I haven't kept up enough with the recent changes so I'm unsure if anythings really changed.


Peace,
Dan O.

As someone with a setup just like that (4GB of DDR3 1333 RAM, 3.25GB usable in Win 7 Pro 32-bit and XP Home SP3) in dual channel mode (ASRock 880GM-LE mobo, AMD Athlon II x4 640) with integrated graphics (AMD HD4250, 128MB of RAM reserved in BIOS for graphics), I can tell you that I don't see any advantage. When my RAM starts getting consumed wholesale with folding and a ton of Firefox windows open, my screen slows down and my cursor starts lagging as if there isn't enough RAM or it's still too slow.

The problem with memory, and this affects all types, is that its performance scales logarithmically. The whole purpose of memory is to keep the CPU happy and not wait for anything. Thus if you double the performance of memory, it may improve overall performance by say 30%. But the next time you do it, it'll only improve 15%, then 7.5%, and so on. This is theoretical by the way.

Integrated graphics might be another story as people have pointed out. It likes faster memory more than the CPU does.

Yeah, it's not a game-changer either way, with or without integrated graphics. But the impact on performance on a non-integrated setup is barely measurable, on an integrated it can at least be measured, but I doubt anyone would really notice or "feel" an improvement.

Like LatiosXT said, it's one of those things where the law of diminishing returns has a huge influence.

Imagine you could make your memory 1,000x faster, but your CPU, the memory controller, the numerous pathways and circuits between it and the CPU, and everything else in the system stays the same, you would be lucky to see an overall improvement in system performance of 25%.

Dan O.

This reminds me of various tests that people have done with performance versus HDDs, SSDs, and RAM Disks and load times (which is the biggest factor on why you want faster memory). The overall consensus that I found was even though the RAM disk showed something like 40-800 times the performance over an HDD, loading times didn't decrease at that rate. What gives?

While I can't provide a definitive answer, one has to remember that Windows probably treats the RAM disk as any other disk. So on top of the file system operations that it must do (it makes software implementation easier if you unify everything), there some other things that people forget about. Loading isn't just putting the contents into RAM, it's also initializing data for program use. For instance, I'm fairly certain that loading an OS is not only memory intensive, but it's CPU intensive as well. In fact, Windows 8 shaved off a lot of time with the fast start up because it skips a lot of the initialization routines (their state gets stored).

To put this in another perspective, I work on microcontrollers and the initialization routines on them take about 91.5K cycles. And this may be a hard limit. So no matter how much or how fast the memory is, I have 91.5K cycles of boot time because the CPU can't go through it fast enough.