Single channel VS Dual Channel RAM (8700K)

[cucker tarlson]

I asked the reviewer about the testing place and conditions in which he obtained these results:

I never got an answer.

sorry to say that,but TPU's memory testing methodology for gaming is just weak. No way are you able to obtain such results (no difference between 2133 and 4000) unless you're 100% gpu bound, and if you test memory in GPU bound places then you don't know what you are doing. That 3466 CL14 kit should run circles around 2133 cl15 in witcher 3 if you know where to test.

 

[silkstone]

I asked the reviewer about the testing place and conditions in which he obtained these results:

I never got an answer.

sorry to say that,but TPU's memory testing methodology for gaming is just weak. No way are you able to obtain such results (no difference between 2133 and 4000) unless you're 100% gpu bound, and if you test memory in GPU bound places then you don't know what you are doing. That 3466 CL14 kit should run circles around 2133 cl15 in witcher 3 if you know where to test.

Have a read through the comments. Even on the 'old' 6700K, there is little difference going from slower to faster ram compared with single vs. dual channel.

 

[therealmeep]

In my benchmark video, that second number in RAM section, that is about 12000+ is actually my pagefile that game is using, not a second RAM. I only have one RAM stick with 8GB (2400mhz). So you think that RAM is the problem?

The fact that you are getting adequate pagefile usage while gaming does mean that more RAM is necessary. Ideally you should not have any pagefile usage, however at absolute worst, moving the paging file to an SSD could give you some performance boost, although this is going to shorten the ssd lifespan.

 

[Bill_Bright]

Ideally you should not have any pagefile usage,

Ummm, not true. You can have 64GB of RAM and Windows will still use a PF - but that's a good thing, not bad! That's also why you should never disable the PF. Best to just let Windows manage it. Contrary to what some think and want others to believe, Windows really does know how to manage virtual memory (RAM + the PF) very well.

What you don't want is for the OS to be constantly banging on the PF because there is too little RAM.

 

[cucker tarlson]

Have a read through the comments. Even on the 'old' 6700K, there is little difference going from slower to faster ram compared with single vs. dual channel.

You are generalizing too much. Cases of too slow ram will be different across the board. You can't take 6700K in arma 3 (and unknown gpu) and draw a conclusion about 8700k in witcher 3 on 1080ti.

btw the difference is almost the same anjd pretty huge in both cases, did you read the chart correctly ?

If you are referring to the top of the chart, then look at latencies. 3000MHz at cl12 is absolutely killer for gaming since latency is very,very important in single thread heavy scenarios. That's why 3000 cl12 is up there with 4000 at cl17.
Look at this, 2400 CL9 ddr3 can outperform 3333 cl16 in some cases

What I find especially interesting is stock (4.0GHz) 6700K with 3200 ram is 10% faster than oc'd (4700MHz, +18%) with 2133 ram.

 

[therealmeep]

Ummm, not true. You can have 64GB of RAM and Windows will still use a PF - but that's a good thing, not bad! That's also why you should never disable the PF. Best to just let Windows manage it. Contrary to what some think and want others to believe, Windows really does know how to manage virtual memory (RAM + the PF) very well.

What you don't want is for the OS to be constantly banging on the PF because there is too little RAM.

Agreed, I'm amazed at how well Windows is able to manage the pagefile, my only complaint with it would be the space it takes up (somewhere in the 60 gig range for my 64GB of RAM), but in most conceivible builds with this much ram it isn't an issue. That and the performance is a decent bit better than Linux swap.

You are generalizing too much. Cases of too slow ram will be different across the board. You can't take 6700K in arma 3 (and unknown gpu) and draw a conclusion about 8700k in witcher 3 on 1080ti.

btw the difference is almost the same anjd pretty huge in both cases, did you read the chart correctly ?

View attachment 105001

Unfortunately this is probably always going to be subjective, A3 has been something where I have yet to notice a difference between my 3770 and my 6800k. I have yet to see a noticeable difference between single, dual, and even quad channel in anything other than benchmarks.

 

[Devon68]

I have a motherboard with only 2 slots and was planing on getting 2x4GB sticks of ram but now I found a single 8GB stick for 50$ so I'm considering it. The 2x4GB option would be around 60$. Which option would make more sense?. I'm buying the ram for a pc that I'm planing to sell as soon as I get all the parts.

 

[Bill_Bright]

Agreed, I'm amazed at how well Windows is able to manage the pagefile, my only complaint with it would be the space it takes up (somewhere in the 60 gig range for my 64GB of RAM),

I am not amazed. Microsoft has an army of world class experts in memory management and more than 20 years and many exabytes of empirical data to draw from to learn how to make it right. After all, they want our computers to run optimally too.

As far as your 60 gig of disk space issue, that is not really true either. That maximum size is not automatically set, nor is the size set in stone and never changes. It is called a dynamically management PF for a reason. It will expand and contract as needed and available disk space is one the parameters used to determine the size. I note right now, I have 16GB of RAM installed yet my Windows managed PF is set to 2938MB recommended with 2432MB "Currently allocated".

 

[therealmeep]

I have a motherboard with only 2 slots and was planing on getting 2x4GB sticks of ram but now I found a single 8GB stick for 50$ so I'm considering it. The 2x4GB option would be around 60$. Which option would make more sense?. I'm buying the ram for a pc that I'm planing to sell as soon as I get all the parts.

I'd say to get a single 8GB stick, as it shouldn't handicap too much performance, and the gains you would get from the 2x4 really shouldn't be noticable unless you are doing ram heavy work. This also gives you the option to go to either 12,16, or 24 gigs of RAM down the road.

 

[Vayra86]

The fact that you are getting adequate pagefile usage while gaming does mean that more RAM is necessary. Ideally you should not have any pagefile usage, however at absolute worst, moving the paging file to an SSD could give you some performance boost, although this is going to shorten the ssd lifespan.

This is nonsense. Pagefile usage is a normal, documented Windows functionality that only improves performance, rather than decrease it. And it also does not necessarily rely on the amount of RAM you have at all. I have 16 GB and still see Pagefile over 10 GB quite often; with more data in there than what resides in RAM. What matters is how fast the system can access and use the data from pagefile, and that is where RAM bandwidth + single/dual channel comes in.

Don't spread BS pls

As for the video, you can clearly see that 6.7 GB of RAM is in use, which is quite fine as far as capacity is concerned. The problem here is single channel and speed. 8 > 16 GB is just bonus (for this game at least)

 

[RedeyeLegend]

This is nonsense. Pagefile usage is a normal, documented Windows functionality that only improves performance, rather than decrease it. And it also does not necessarily rely on the amount of RAM you have at all. I have 16 GB and still see Pagefile over 10 GB quite often; with more data in there than what resides in RAM. What matters is how fast the system can access and use the data from pagefile, and that is where RAM bandwidth + single/dual channel comes in.

Don't spread BS pls

As for the video, you can clearly see that 6.7 GB of RAM is in use, which is quite fine as far as capacity is concerned. The problem here is single channel and speed. 8 > 16 GB is just bonus (for this game at least)

So final conclusion is that I get more ram to solve my problem?

 

[Vayra86]

So final conclusion is that I get more ram to solve my problem?

Faster > 2 sticks > more

In that order of importance (and performance benefit)

 

[cucker tarlson]

Faster > 2 sticks > more

In that order of importance (and performance benefit)

I think all 3 are equally important.

 

[Vayra86]

I think all 3 are equally important.

Obviously, but the guy says 'more' as if its the takeaway here and it really isnt. 16 GB is the least immediate issue. He could start with one fast stick and already enjoy improved perf, then add another and boom problem solved.

It would ALSO be a bad route to stick another slow one next to what he has.

 

[RedeyeLegend]

Obviously, but the guy says 'more' as if its the takeaway here and it really isnt. 16 GB is the least immediate issue. He could start with one fast stick and already enjoy improved perf, then add another and boom problem solved.

It would ALSO be a bad route to stick another slow one next to what he has.

Here are screenshot with detailed benchmarks
https://www.mediafire.com/file/pvl2eveppgljpkx/Benchmarks.rar/file

 

[cucker tarlson]

Yup, 2x4 3000 cl15 would perfrom better than 2x8 2400.

Here are screenshot with detailed benchmarks
https://www.mediafire.com/file/pvl2eveppgljpkx/Benchmarks.rar/file

synthetic benchmarks are worthless since they are very multithreaded, unlike real games which prefer fast single thread and low latency. In 3d mark 2700x is faster than 8700k, in real games it's the other way around.

 

[RedeyeLegend]

Yup, 2x4 3000 cl15 would perfrom better than 2x8 2400.

Here are screenshot with detailed benchmarks
https://www.mediafire.com/file/pvl2eveppgljpkx/Benchmarks.rar/file

 

[silkstone]

I asked the reviewer about the testing place and conditions in which he obtained these results:

I never got an answer.

sorry to say that,but TPU's memory testing methodology for gaming is just weak. No way are you able to obtain such results (no difference between 2133 and 4000) unless you're 100% gpu bound, and if you test memory in GPU bound places then you don't know what you are doing. That 3466 CL14 kit should run circles around 2133 cl15 in witcher 3 if you know where to test.

There is no way that any of those games are GPU bound on 1080p with a GTX 1080. There is also lots of technical analysis as to why the results are the way they were for the coffee lake chips. I agree that faster ram would be better, and with faster ram, you'd likely see an increase in the minimum FPS, I just think you are overestimating the effect and having more memory that is dual channel is much more important than having faster RAM. The best situation would be to buy as fast a RAM stick as possible from the outset, but as the OP already has RAM, junking it in favour of a faster stick would not be a wise investment.

 

[RedeyeLegend]

Don't waste your money on faster ram for gaming, just get a second stick.

Here is the evidence: https://www.techpowerup.com/reviews..._Memory_Performance_Benchmark_Analysis/9.html

Do you think it is possible that my 60hz monitor is also causing low FPS? I read on internet that my GPU is in full power when I play 1440p on 144hz monitor. Do you think my GPU usage is low because of ram and also low refresh rate?

 

[c0de]

Do you think it is possible that my 60hz monitor is also causing low FPS? I read on internet that my GPU is in full power when I play 1440p on 144hz monitor. Do you think my GPU usage is low because of ram and also low refresh rate?

No, it wouldn't cause low FPS
Monitors don't make low FPS
They only display pictures at different refresh rates and times

Because 1440p on 144hz is your graphics card limit
It's the moment when your GPU hit full potential of power
Your GPU has got very good amount of power

Unless you have v-sync Enabled or capped frame limiter for example 60 FPS your GPU will use only power to generate and deliver only 60 FPS !
The rest power of GPU will be in IDLE mode for better cooling solution

So when your GPU have enough power capacity it will not hit 100% utilization when you have v-sync Enabled on monitor 60 Hz or frame limiter @ 60 FPS
Frame limiter is good choice when you don't need more than 60 FPS and want cooler GPU for example in very hot summer days

Frame limiter is better solution than v-sync Enbaled, because v-sync produce unnecessary input lags on devices

You can set frame limiter in RivaTuner Statistics, which comes with MSI Afterburner software at different FPS

I use frame limiter on notebooks which have bad cooling solutions on GPU chip to generate only 60 FPS especially for better cooling solutions

I don't need more than 60 FPS, because my monitor can display only 60 FPS@ 60 Hz
You use more FPS when you play shooter FPS
More FPS decrease inputs lags from devices like.: wheels controler, gamepad, keyboard and mouse
But you will suffer from tearing problem unless you have monitor that can display more than 60 FPS
For example 140 Hz can produce max 140 FPS

You will have tearing problem always above refresh rate of your monitor

 

[silkstone]

Do you think it is possible that my 60hz monitor is also causing low FPS? I read on internet that my GPU is in full power when I play 1440p on 144hz monitor. Do you think my GPU usage is low because of ram and also low refresh rate?

Nope. Refresh rate and FPS aren't linked unless you turn on VSync, even then it would just make your FPS a constant 60 fps for 60 Hz.

 

[RedeyeLegend]

This is on 7820x but you get the picture

https://www.purepc.pl/pamieci_ram/t...bach_single_vs_dual_vs_quad_channel?page=0,22
https://www.purepc.pl/pamieci_ram/t...bach_single_vs_dual_vs_quad_channel?page=0,21
https://www.purepc.pl/pamieci_ram/t...bach_single_vs_dual_vs_quad_channel?page=0,20
https://www.purepc.pl/pamieci_ram/t...bach_single_vs_dual_vs_quad_channel?page=0,19
https://www.purepc.pl/pamieci_ram/t...bach_single_vs_dual_vs_quad_channel?page=0,18
https://www.purepc.pl/pamieci_ram/t...bach_single_vs_dual_vs_quad_channel?page=0,17
https://www.purepc.pl/pamieci_ram/t...bach_single_vs_dual_vs_quad_channel?page=0,16
https://www.purepc.pl/pamieci_ram/t...bach_single_vs_dual_vs_quad_channel?page=0,15
https://www.purepc.pl/pamieci_ram/t...bach_single_vs_dual_vs_quad_channel?page=0,14
https://www.purepc.pl/pamieci_ram/t...bach_single_vs_dual_vs_quad_channel?page=0,13


2400mhz is slow to begin with, and if it's in single channel then your cpu is suffering a huge bottleneck.

get 2x4gb 3000 cl15 at least. 16gb 3200 cl14 is the sweet spot perf/price. 16gb 3600 cl15 is the fastest you can get without breaking the bank, +4000 cl17 sticks are faster but cost ridiculous amounts of money.

Man you saved me and you were right. It was about RAM. Today I bought 2x8 3200 MHZ and now my FPS is 140-160 in Novigrad. Thank you very much

 

[Bill_Bright]

I bought 2x8 3200 MHZ and now my FPS is 140-160 in Novigrad.

Sweet! And thanks for posting your followup!

 

[cucker tarlson]

Man you saved me and you were right. It was about RAM. Today I bought 2x8 3200 MHZ and now my FPS is 140-160 in Novigrad. Thank you very much

What was it before ? Like 90 ? That's huge but no wonder the improvement is so big if your gpu was at 50% utilization.

 

[RedeyeLegend]

What was it before ? Like 90 ? That's huge but no wonder the improvement is so big if your gpu was at 50% utilization.

It was 65-85 Novigrad. Now is 135+. Outside is even 175