Are More RAM Modules Better for Gaming? 4 x 4GB vs. 2 x 8GB

I have four 8GB Hyper X modules installed in both my machines.

I see no need for more than 32GB even if I could have 64.

So far, everything runs perfectly.
 

Attachments

  • images.jpg
    images.jpg
    13.5 KB · Views: 1
  • 82710871_2270149796628235_4788462122096918528_o.jpg
    82710871_2270149796628235_4788462122096918528_o.jpg
    94.5 KB · Views: 1
An interesting concept that I never thought about, of course I'm running 2x16 and not having any issues at all ..... seems a bit like 'name brand' vs. 'generic' .....
 
More people will start opting for 32GB configurations so this is useful information to have.

The gains in tuning your memory frequencies and latencies seem to be more beneficial to me across a wide range of applications though. Not least because that's free. You don't need any more hardware to spend some time fiddling the memory you have to optimise performance.

This particularly applies to AMD platforms as we know. When AMD launch Zen 3 another in depth RAM test with it would definitely be a worthy article.
 
I just want to explain the reason for these results. Even though there is a single 64-bit data bus per memory channel, having more than 1 rank allows to issue different set of commands (interleaved commands) to different set of 64-bit of memory modules thus parallelizing memory command execution. Modern memory controllers take full advantage of this as can be seen in these results.
 
Lol I paid through the nose for the Corsair Vengeance 4x8GB 3466mhz CL16 kit back in Sep 2017 when 8700K released (470usd). Glad it still considered the best RAM money can buy to this day, maybe short of 4x16GB 3600mhz CL16 Kit that sell for almost the same price but there is no performance uplift.
 
Last edited:
I payed £200 for 32 gig of 3600 cas14 memory cant be more happy.
I remember spending weeks checking for a board with certified memory but in the end I ignored that.
 
They should include the 1440p results too. Let's get real, you're NOT going to notice the difference between 165fps and 175fps in a game. Lots of people pretend they do just so they can justify spending the extra cash on 32GB ram, a super-high refresh monitor (20 blade razors anyone?), or so-called 'faster' ram, which in the tests I've seen is a wash once you get up around 3000 (yes, I know, AMD systems NEED 3200 or they'll spontaneously combust and kill all your neighbours).

I run one of the most demanding games known to mankind (DCS World) at 1440p with 16GB ram and a 1070 and still never use more than 14GB, even in populated servers. That's another point...how about some benchmarks with ACTUAL demanding games, not these same ones loaded with fluffy FPS games all the time. My son can run COD on his i5 4690 and 1060 3GB at almost 100fps. Throw DCS, ANNO 1800 and the Division 2 at 1440p into the mix for once, then you'll see some of these ridiculously high frames come back down to earth.
 
Last edited:
Everyone I know who has a i9 games @ 4k.
doing 1080p tests with such a cpu is just PR for Intel .
should be 1440 minimum res to test
 
Everyone I know who has a i9 games @ 4k.
doing 1080p tests with such a cpu is just PR for Intel .
should be 1440 minimum res to test

You don't need an i9 to game at 4K (look at the articles on this site). You do however need at least a 2070 Super vid card if you want decent frames.
 
Interesting stuff. I've always used two DIMMs because I wanted the option to add more capacity in the future. I've always wondered if there would be a performance difference between 2 and 4 sticks and now I have an answer.
 
Interesting comparison, I would be curious if even the same size modules from the same vendor and same type but different batch would be different.
the comparison is missing a single 16GB module performance though
 
While potentially, possible, and I'm not saying I believe the results to be an accurate representation of what you'd see across the board for every increase to four DIMMs vs two because it's very likely that it's not, or that it's not for every generational or manufacturer platform, it also needs to be pointed out that this was on a clean lab system with a specific set of sticks.

Notably, that they weren't even the same sticks as what was used in the 2 DIMM configuration which is evident by the fact that they were 4GB DIMMs instead of 8GB DIMMs. That alone almost negates the study, but we can overlook that for now.

The bigger concern is whether stability and long term endurance of the CPU and motherboard are affected. It is well documented, by Intel as well as a good many OTHER engineers, that running four DIMMs vs two introduces a much higher stress, double in fact, on the IMC and to some degree on the motherboard as well. Normally, we see this translate into a direct increase in CPU temperature AND usually an overall increase in DIMM temperatures as well since the DIMM sockets become more thoroughly heat saturated.

Realistically, in some cases, this could easily translate into an accelerated rate of electromigration and voltage threshold shift, or at least an increase in potential for it. Electromigration is a direct result of increased or excessive voltage and heat, and adding two DIMMs doubles the amount of power used and increases the memory related heat that will impact the IMC as well as the package. Long term, that COULD translate into a reduction of longevity and the chances of that are greatly increased on systems where the voltage or thermal compliance are already at an elevated level compared to the stock condition.

Whether that is a problem or not will depend greatly on the specifics of the built and the overall cooling picture, not to mention that it is likely GREATLY affected by the quality of the board being used. Let's fact it, an 80 dollar motherboard is not going to handle or withstand the same thermal and configurational stress as a 165+ dollar board will in practically every case. A cheaper board almost universally will not only not handle higher speed memory AS well, but will not handle a four DIMM configuration AS well either.

I think this study is skewed on multiple levels and while it shows some things, I'm not confident that WHAT it shows is what we think it shows. The evidence would seem to be a lot more "this case" than "widely empirical". I 100% do not believe that if every person playing a CPU limited game runs out and adds two DIMMs to their configuration, or swaps in a four DIMM kit, they are going to see ANY kind of an increase in performance.

Some might, as with the configuration in this study, but I think many others not only won't, but might even see a decrease in performance if they are running a system that is already toeing the line in terms of thermal compliance, and they might also shave several years off the longevity of their hardware in some cases.
 
Last edited:
You should have done 2x8 vs 4x8 because if the 2x8 showed even the slightest benefit (due to using two sticks) you would have the 32GB users livid and screaming at you saying how flawed your test was...

Agreed, that is what Tom's Hardware did and they found that was the biggest jump in performance, They said using 4x8GB modules at 3200MHz was better than 2x8GB at 3600MHz for AMD CPU's. I have to assume the Techspot authors read that article before doing this article.
 
Great article and well done in the testing. When configuring my old 4700K I picked 4x4GB DDR3 modules, it was the better choice for years over 2x8GB.
 
Agreed, that is what Tom's Hardware did and they found that was the biggest jump in performance, They said using 4x8GB modules at 3200MHz was better than 2x8GB at 3600MHz for AMD CPU's. I have to assume the Techspot authors read that article before doing this article.


Agreed.

And probably the reason why I will move up to 4x8 32GB memory configuration in my build sometime this year.

https://www.tomshardware.com/reviews/amd-ryzen-3000-best-memory-timings,6310-2.html
 
Lol I paid through the nose for the Corsair Vengeance 4x8GB 3466mhz CL16 kit back in Sep 2017 when 8700K released (470usd). Glad it still considered the best RAM money can buy to this day, maybe short of 4x16GB 3600mhz CL16 Kit that sell for almost the same price but there is no performance uplift.

No, DDR4 3733 CL14 1R is the best RAM money can buy today and a good chunk faster then what you have. With increased ranks comes a trade-off in frequency and/or timings. This is stated in the article.

Great article and well done in the testing. When configuring my old 4700K I picked 4x4GB DDR3 modules, it was the better choice for years over 2x8GB.

People seem to be misunderstanding the article.

"It’s also worth noting that four modules can also limit memory performance in the sense that you might not be able to achieve the same frequency and timings that you could with just two modules."

There is a reason the best RAM you can buy is single rank. While having more ranks gives you a small performance boost, the maximum frequency you can obtain is reduced and often times you cannot obtain as tight timings.

If you are not spending a ton on RAM, a 4 stick or 2 stick dual rank kit is a good choice. If you have the money to spend though, ultimately all the best kits are single rank 2 module kits.

I'm actually surprised using more modules ended up being consistently faster. I would have figured the reverse would be true.

This test was conducted at a low frequency. All the fastest DDR4 kits are single rank and you cannot achieve those frequencies with dual rank.

This test is not surprising, it just illustrates that for budget RAM, dual rank or 4 piece kits are better then single rank assuming all other factors are equal.

An interesting concept that I never thought about, of course I'm running 2x16 and not having any issues at all ..... seems a bit like 'name brand' vs. 'generic' .....

Ranks are just one factor to consider. Frequency and timings are equally important. Dual rank sacrifices timings and frequency. If you are debating on whether to get a 4x8GB 1R CL16 vs a 2x16GB CL15 kit, they will be almost equal in performance. That said, you typically always want the kit with higher density as the resale value is higher and it allows you to add more in the future.
 
Last edited:
I thought this argument ended during the old socket 775 days.
Spreading RAM across the channels gives greater bandwidth to the RAM to CPU rather than trying to pigeon hole the greatest amount in the least amount of dimm channels.....
 
While potentially, possible, and I'm not saying I believe the results to be an accurate representation of what you'd see across the board for every increase to four DIMMs vs two because it's very likely that it's not, or that it's not for every generational or manufacturer platform, it also needs to be pointed out that this was on a clean lab system with a specific set of sticks.

Notably, that they weren't even the same sticks as what was used in the 2 DIMM configuration which is evident by the fact that they were 4GB DIMMs instead of 8GB DIMMs. That alone almost negates the study, but we can overlook that for now.

The bigger concern is whether stability and long term endurance of the CPU and motherboard are affected. It is well documented, by Intel as well as a good many OTHER engineers, that running four DIMMs vs two introduces a much higher stress, double in fact, on the IMC and to some degree on the motherboard as well. Normally, we see this translate into a direct increase in CPU temperature AND usually an overall increase in DIMM temperatures as well since the DIMM sockets become more thoroughly heat saturated.

Realistically, in some cases, this could easily translate into an accelerated rate of electromigration and voltage threshold shift, or at least an increase in potential for it. Electromigration is a direct result of increased or excessive voltage and heat, and adding two DIMMs doubles the amount of voltage and memory related heat that will impact the IMC as well as the package. Long term, that COULD translate into a reduction of longevity and the chances of that are greatly increased on systems where the voltage or thermal compliance are already at an elevated level compared to the stock condition.

Whether that is a problem or not will depend greatly on the specifics of the built and the overall cooling picture, not to mention that it is likely GREATLY affected by the quality of the board being used. Let's fact it, an 80 dollar motherboard is not going to handle or withstand the same thermal and configurational stress as a 165+ dollar board will in practically every case. A cheaper board almost universally will not only not handle higher speed memory AS well, but will not handle a four DIMM configuration AS well either.

I think this study is skewed on multiple levels and while it shows some things, I'm not confident that WHAT it shows is what we think it shows. The evidence would seem to be a lot more "this case" than "widely empirical". I 100% do not believe that if every person playing a CPU limited game runs out and adds two DIMMs to their configuration, or swaps in a four DIMM kit, they are going to see ANY kind of an increase in performance.

Some might, as with the configuration in this study, but I think many others not only won't, but might even see a decrease in performance if they are running a system that is already toeing the line in terms of thermal compliance, and they might also shave several years off the longevity of their hardware in some cases.
I'm sorry, but that is a full load of BS. running 4 RAM sticks doesnt increase voltage to the IMC, it poses no more risk then running your CPU at 100% load does, any properly cooled CPU will be able to handle that for incredible lengths of time. This "multiple sticks per channel" design goes back to the earliest x86 computers, literally any old computer with multiple memory sticks does this, and memory controller failures are extremely rare.

As for the rest of your claims, where are you getting this data? Some might see worse performance? Cheaper mobos may perform worse? Maybes, mights, I thinks, could, possibly, your conjecture is worth as much as used toilet paper. Post data to back up your claims, there is no reason to spread fear mongering of running 4 sticks frying your CPU, the additional power use from the IMC managing 4 sticks is negledgeable compared to having a 6 or 8 core CPU.
 
Back