Since when is 4x4gb better than 2x8gb for gaming

[AntiTrust]

These videos are showing a 4x4gb ram layout sometimes producing a 10-20% performance increase over 2x8gb layout of the same speed ram / Can anyone explain why this is? I always thought that the difference was negligible, or that 2x8 was slightly better for gaming. 

 

 

 

[Origami Cactus]

Well in the first video he didn't say if he ran the 2x8 in dual channel, so he probably ran them in single channel, reducing performance.

On the second video he uses a platform that supports quad channel.

 

But generally consumer platforms support dual channel, so if you set up the ram correctly then 2x8gb should work good too.

[CiBi]

4 slots to transfer data over i guess. But I always thought that 2 sticks in a dual channel config allowed you to use up the full bandwith, could be wrong though (and apparantly i am...)

[BleachedFur]

It's because of quad channel. 4 channels has twice the bandwidth of dual channel, and 4x the bandwidth of single channel.

 

Memory channeling is no joke, which is one thing GamersNexus freaked out about with the walmart PC. 16gb...in single channel.

 

I have quad channel DDR3 which outperforms DDR4 at less than half the frequency strictly because of the amount of bandwidth that I have, and one reason I'm still using X79.

[Fasauceome]

1 minute ago, CiBi said:

But I always thought that 2 sticks in a dual channel config allowed you to use up the full bandwith

Yes that is the max bandwidth, unless you're using a high end desktop motherboard and an i9 or threadripper and have quad channel, which gives more bandwidth.

 

@AntiTrustthe videos you linked are the unreliable type of benchmarks, they don't even show a picture or video of themselves with the hardware, it's often very likely that this is either simulated performance or just fabricated. When you see benchmark vids that don't have any introduction and voice over, don't take bother with them.

[BleachedFur]

2 minutes ago, fasauceome said:

just fabricated.

those figures are 100% fabricated, all the videos on that channel are fake.

[Fasauceome]

1 minute ago, BleachedFur said:

those figures are 100% fabricated, all the videos on that channel are fake.

At least it's not as bad as the vids that just have EDM over a bunch of colorful "performance graphs"

[AntiTrust]

7 minutes ago, fasauceome said:

fabricated

Well at least that would make sense. I'm not sure why you would fake it though its not like coming by 4 sticks of ram is very difficult if you run a channel on benchmarking pc components. If anyone has a link to a more reliable source doing this test I'd appreciate it, 

[kdawwgg1221]

9 minutes ago, fasauceome said:

At least it's not as bad as the vids that just have EDM over a bunch of colorful "performance graphs"

or as informative as The Verge's PC build guide...(sarcasm)

[emosun]

triple channel ftw

 

[Fasauceome]

4 minutes ago, AntiTrust said:

Well at least that would make sense. I'm not sure why you would fake it though its not like coming by 4 sticks of ram is very difficult if you run a channel on benchmarking pc components. If anyone has a link to a more reliable source doing this test I'd appreciate it, 

You can just assume 4x4GB is the same as 2x8 because it's just dual channel.

[porina]

Haven't watched the videos, which based on earlier comments, there may be other explanations than what I'm about to write.

 

Rank per channel can affect performance. I don't entirely understand it myself, but I have seen significant gains in other uses from having 2 rank per channel instead of 1.

 

For DDR4, 4GB modules are single rank. 8GB modules are almost all single rank, but some older ones might be dual rank. 16GB modules probably are dual rank but I never owned any. 

 

On a dual channel system, when you have 2x8GB, you probably only have one rank per channel. If you have 4x4GB, you will have 2 rank per channel. In synthetic tests, 2R per channel results in slightly higher bandwidth, at slightly higher latency. In my uses which are heavily ram bandwidth limited, and are mixed reads and writes, there can be tens of % difference in performance from having 2R per channel at otherwise exactly the same speed and timings.

[Princess Luna]

Honestly both these channels, I seen people link them in the past and I personally call it *fake news*.

 

To begin with, if they have money to test RTX TITAN, TITAN V, 2080 Ti's Radeon 7 on every CPU you can think of even i9 9980XE's then why don't they ever show the hardware? Why they never shown a video host?

 

Put together charts and forge Rivaturner numbers with pre-recorded gameplay and call it w/e is fairly doable and if done right it's easy to fool lots of people, so I personally feel it's all fake and their numbers are educated guesses based on other people's finding.

[RoseLuck462]

Quad Channel homie 

[BleachedFur]

54 minutes ago, AntiTrust said:

not sure why you would fake it though

getting views from people who had the same question as you.

 

54 minutes ago, emosun said:

triple channel ftw

 

lemme guess, X58?

 

43 minutes ago, porina said:

On a dual channel system, when you have 2x8GB, you probably only have one rank per channel. If you have 4x4GB, you will have 2 rank per channel. In synthetic tests, 2R per channel results in slightly higher bandwidth, at slightly higher latency. In my uses which are heavily ram bandwidth limited, and are mixed reads and writes, there can be tens of % difference in performance from having 2R per channel at otherwise exactly the same speed and timings.

Well, afaik all 4 of my 8GB DIMMS (DDR3, mind you) are dual rank with 8GB per channel, in quad channel.

 

30 minutes ago, RoseLuck462 said:

Quad Channel homie 

quad channel is amazing. Never leaving HEDT unless I'm forced to.

[porina]

8 minutes ago, BleachedFur said:

Well, afaik all 4 of my 8GB DIMMS (DDR3, mind you) are dual rank with 8GB per channel, in quad channel.

Yup, that's why I was specifically talking DDR4 in my reply. In part it is related to the density of the individual ram chips. As density increases, there is less need to go dual rank for a given module capacity.

 

I think I even have dual rank 4GB DDR3 modules.

[AngryBeaver]

4 hours ago, BleachedFur said:

It's because of quad channel. 4 channels has twice the bandwidth of dual channel, and 4x the bandwidth of single channel.

 

Memory channeling is no joke, which is one thing GamersNexus freaked out about with the walmart PC. 16gb...in single channel.

 

I have quad channel DDR3 which outperforms DDR4 at less than half the frequency strictly because of the amount of bandwidth that I have, and one reason I'm still using X79.

Single channel vs dual channel only makes a big difference if you are benchmarking your memory.

 

In real world usage the difference is imperceptible. You might gain a 1 or 2 fps in a game, so if you are seeing more than that then the reviewer has a flawed test setup.

[porina]

1 hour ago, AngryBeaver said:

Single channel vs dual channel only makes a big difference if you are benchmarking your memory.

 

In real world usage the difference is imperceptible. You might gain a 1 or 2 fps in a game, so if you are seeing more than that then the reviewer has a flawed test setup.

There is no question that ram bandwidth will impact performance. The only question is if it is worth paying for the difference. I've even been testing rank differences as the next layer of optimisation.

 

There are many variables but it simply shouldn't be a question in the first place in any but the most budget sensitive builds, where 2x4GB is slightly more expensive than 1x8GB for example. Having dual channel from the start will mean it works as best it can.

 

2 minutes ago, comander said:

On balance, I do believe that enthusiasts overemphasize memory performance. 

Personally I think it is over-stated on Ryzen. The Zen CPUs so far just aren't as bandwidth demanding as Intel ones, although many point towards differences from the tied infinity fabric. Having multiple systems, I'd prioritise faster ram to go in my Intel systems where it makes more impact.

[BleachedFur]

20 hours ago, AngryBeaver said:

Single channel vs dual channel only makes a big difference if you are benchmarking your memory.

 

In real world usage the difference is imperceptible. You might gain a 1 or 2 fps in a game, so if you are seeing more than that then the reviewer has a flawed test setup.

wrong. You'll see better 1% and 0.1% lows. Especially in CPU intensive games. GPU intensive games like Skyrim (20:1 GPU/CPU usage) RAM speed won't make a huge impact.