Gadgetory


All Cool Mind-blowing Gadgets You Love in One Place

RAM Benchmark: Dual-Channel vs. Single-Channel - Does it Matter?

2014-03-08
hey everyone this is Steve from gamers Nexus dotnet and today we're benchmarking single Channel versus dual channel configurations for memory on compatible platforms I wrote a very detailed 5,000 word article to accompany this test that includes a full detail breakdown of how these benchmarks work and how they apply to your uses in the real world but the video will focus more on Adobe After Effects performance and sort of skim over the other results what we're testing here today is the statement that you need to buy ram in pairs of two else you're significantly hampering the system that is what we're testing that's sort of the myth I into it for a very long time and I was personally invested in learning the results there are a couple tests out there but I wasn't satisfied with them my hypothesis going into this is that I'd have a significant impact on video editing but would be negligible elsewhere so we'll see if that was correct going forward the back story here is that we met with msi at their HQ in 2013 in july during our california tour and and this is also by the way where we show how ram was made so hit that link if you are interested they showed us some new laptops that all shipped with a single stick of 8 gigabytes of RAM and at the time I pressed and asked why they would opt for that because to me why would you sacrifice dual channel performance that was kind of my thinking so why not go with 2 x 4 gigabytes for instance they said that they first thought was a little cheaper and they found no noticeable difference between dual and single channel configurations when testing internally so I I returned home I did some lightweight testing and got results that appeared in line with this but it wasn't at all comprehensive I wasn't happy with leaving it there so we're returning now before diving in I want everyone to make note that there is no such thing as quote dual channel RAM or dual channel memory or a single channel Ram or single channel memory before you call me an idiot in the comments which happens quite frequently without any research I'm not bitter let me let me explain where I'm coming from there dual channel and single channel platforms but the memory itself isn't made in a special way nor does it contain some sort of special flippable bit that enables a so-called dual channel mode on the memory for this reason unless I slip up because I've been running test for 18 hours you will never hear me say the words dual channel memory without following it up with the words platform or configuration that's pretty important when you're talking about this dual channel memory is not a standalone item that exists it's just memory so hopefully that clears things up I want to get that out of the way first there are some special cases where you enter higher channel counts like quad channel or octave channel where the memory will actually exhibit fairly substantive differences as you increase channel count and that's because of a lot of things including and primarily the higher density like in the range of 64 gigabits for example so that's that's pretty impactful for your performance and isn't necessarily quite as easy as the dual channel stuff we're talking about here today when you're using a single stick of RAM the memory will communicate via a 64 bit wide data channel the channel can effectively be thought of as a lane on a road with 64 bits of total width in a single channel we're more limited in our maximum throughput then a wider channel or multiple channels will allow just like a multi-lane Road using that analogy dual channel platforms which all modern IMC's and chipsets support sans Kabini will utilize - 64 bit channels instead of one effectively doubling the maximum potential data rate of our memory as it communicates with the CPU assuming you've used a multiple of two sticks this sort of pain ponds our data down the channels so we've now introduced parallelism to the memory by multiplying the channel count this applies all the way up the line if you've got quad channel Ram multiply your channel count by four and so on this is easy to confuse with double data rate RAM or DDR so be careful to make that distinction it's it's a different thing they operate independently of each other and a dual channel configurations can only be utilized when the RAM is socketed and matching banks like Bank zero on most boards some boards will support dual channel configurations in both banks in our testing in order to retain a consistent memory capacity we used Bank zero for dual channel config regions that be slot 1 and 3 or 0 and 2 depending how you count it and we used mismatched banks for what we'll call single channel testing meaning slots 1 & 2 or 3 & 4 or whatever alright so enough of that if you want more information like this hit the article link below and that will tell you all the technical stuff it will answer almost all your questions and let's just dive into benchmarks right away here the test platform and complete test methodology are are pretty fleshed out again in the link if you'd like to run the test yourself and replicate my results I've tried to enable this and I encourage it please post a comment in the article if you do some your of your own testing because I'd love to include it in our results in the future we'll start with Euler 3d Euler 3d is a computational fluid dynamics benchmark shortened as CFD bench I'll read directly from my article here because it does a pretty good job at explaining it oiler 3d is a simulation benchmark simulation like this is a computation intensive task that tends to demand high-speed components and multi-threaded processing so CPU and RAM get hammered CFD uses a computer model to calculate fluid flow which is air in this case and my understanding I I haven't I didn't program this I'm not an expert on CFD my understanding is that the test uses a complex matrix to calculate fluid flow at multiple points on whatever we're testing and it then iterates over time with change in flow at surrounding points so this in mind it would mean that we have to store some values and then access them upon each iteration which becomes more and more RAM intensive in terms of speed as the test iterates the results here showed a 17.7% differential between single and dual channel platforms that were tested which is pretty damn substantial if you're doing anything with simulation or similar engineering like parametric modeling or something it is reasonable to presume that you could extrapolate the data to those applications then there's max mem this one benchmarks memory copy read/write latency and bandwidth and you can see here that our Delta's were upwards of 30% with the exception of memory latency which was a couple nanoseconds worse on the dual channel platform this is likely due to overhead and is substantial this is an entirely synthetic benchmark so it's tough to say how it relates to the real world but that's what I'm here for so let's move on to real-world stuff WinRAR is something that you're all familiar with we ran a WinRAR compression test to calculate the advantage of running two memory channels when archiving data and our test we see a difference of approximately two point eight seven percent which isn't really that impactful for consumers keep in mind that this sort of three percent advantage toward dual channel could be a seriously important boon to enterprise and server or database solutions where archiving massive log files or indexes on an hourly or faster basis would demand any small gains that it can get of course these systems will also be utilizing server architecture and are probably on the far more channels than two but you get the idea now we're on handbrake handbrake starts showing some advantages for power users on PCs that be enthusiasts rather than just developers and scientists like the previous tests handbrake transcodes video from one format to another and is often used for ripping or converting video files and is multi-threaded enabled making it good for this test and our test handbrake saw a four point four percent advantage with dual channel configurations which is starting to get somewhat noticeable now for gaming benchmarks and and then we'll move on to video editing I promise I didn't expect any noticeable difference here at all but I tested it for point of clarity to kind of make a point shogun 2 is benchmark tested for a load time advantage of only 0.78% for dual channel configurations while the FPS tested at a point to 1% increase on dual channel you generally won't see FPS increases in this regard with memory but that's another that's another topic for a much longer article both of these numbers are well within margin of error they're less than 1% and should be regarded as in substantial I also tested load times and fps in Skyrim and found similarly low numbers as in less than 1% Delta and I tested it in battlefield 4 but the test is yet inconclusive so far they are in substantial but I need to finish a few things so this gives us the conclusion that at least these two or three games will not see direct benefit of dual-channel platforms and we could probably extrapolate that to most games here's where you see some video from my test for Adobe Premiere I rendered out this very complex ten-second clip featuring for 1080p scenes at 60fps with a 28 megabit per second framerate I filmed all four of these and drop them into this window you can see that there is a rough 3% Delta favoring dual channel here it's noteworthy and if you're doing this sort of work you should probably be running multiple sticks anyway but even with the three percent difference that's only two minutes shaved off of every hour you spend rendering so it's completely up to you to determine whether that's beneficial or not for me it can actually pretty beneficial when I'm at a convention and rendering you know 20 hours a day but still we're looking at minutes here so now for After Effects Adobe's AE has what is called a ram preview that streams video data to memory for live playback with post-processing added in live that's the key in this demo I created another reel with four scenes at 1080p running 60fps with a rough 28 megabit per second framerate 24 in some cases this is the the dual channel test we're looking at the average frame rate of playback including color correction and exposure post-processing spits out at fifteen point one fps for dual channel configs and fourteen point two FPS for my single channel config the delta here is almost 6 percent which could absolutely be a big deal for those of you who regularly do this kind of work so what's our conclusion for all of this well here's the thing generally it's cost-effective to buy memory in kits however if you're on an ultra budget say less than $500 and you see a single stick of RAM that's maybe 8 gigabytes and you know something absurd like $50 FSM forbid that is a price available now then it's a don't freak out about losing the dual channel capabilities you can get by on one stake of your quote unquote just a gamer right it has a very little impact as we've seen the places it becomes relevant is largely going to be outside of the realm of this audience they're going to be server architects and enterprise and web solutions and databases and things that do really regular archiving or require redundancy or require very rapid accessing of data but on the enthusiast side ignoring all things aesthetic it is somewhat beneficial for a render rig if you're doing a lot of rendering if you are a scientist or you do a lot of simulations or you're an engineer and you work with parametric sand analyzers and signal integrity and electro magnetics it's definitely useful to have it there just judging by the Euler 3d test so that's kind of where I stand on all this just to really make things clearer if you are a gamer if it's out of your price range don't go out of the way to get a dual channel configuration going so I know we blasted through those numbers very fast and you've probably got a lot of questions as to how I validated everything and how I tested it but fear not the test case was very extensive and attempted to answer all those questions and my own concerns along the way I did have quite a few concerns going into the tests that I wanted to address so hit the link again in the description below for that not to not to verbally spam you with the link but it is very comprehensive and that's all for this time if you like this sort of content please like comment and subscribe let me know what else you want to see in terms of pc myths we'll call them and I will see you all next time peace
We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites.