Jump to content

Does RAM speed (MHz and/or latency) affect gaming or anything else?

I'm starting to look for new RAM (2x8GB) to replace the 2x4GB sticks that are currently in my friend's desktop but idk what if any effect the MHz and latency have on Gaming or other performance.  

 

 

Link to comment
Share on other sites

Link to post
Share on other sites

Not really. Faster RAM can benefit iGPUs a little, since they use it as VRAM, but in general, it shouldn't make a huge difference. Most are at least 1600MHz, which is plenty.

Laptop: Asus GA502DU

RAM: 16GB DDR4 | CPU: Ryzen 3750H | GPU: GTX 1660ti

Link to comment
Share on other sites

Link to post
Share on other sites

1 minute ago, Rangaman42 said:

Not really. Faster RAM can benefit iGPUs a little, since they use it as VRAM, but in general, it shouldn't make a huge difference. Most are at least 1600MHz, which is plenty.

Wrong. What architecture, OP? 

RAM frequency increases minimum framerates and average framerates by decreasing the time it takes to unload and load stuff into memory. On pretty much every arch besides Haswell.

idk

Link to comment
Share on other sites

Link to post
Share on other sites

Before skylake it didnt matter at all, on haswell, sandy and other older architectures that used ddr3 the speed differences in performance were negligible.

 

skylake and kaby lake that uses ddr4 on the other hand finally shown significant improvement as you get higher memory frequencies, sure if you get just 2133mhz it will still be very good but there might be up to 10fps improvement in games going with 3000mhz speeds.

 

Keep in mind you usually need a z or x board to get support for higher memory frequency speeds than 2400mhz.

Personal Desktop":

CPU: Intel Core i7 10700K @5ghz |~| Cooling: bq! Dark Rock Pro 4 |~| MOBO: Gigabyte Z490UD ATX|~| RAM: 16gb DDR4 3333mhzCL16 G.Skill Trident Z |~| GPU: RX 6900XT Sapphire Nitro+ |~| PSU: Corsair TX650M 80Plus Gold |~| Boot:  SSD WD Green M.2 2280 240GB |~| Storage: 1x3TB HDD 7200rpm Seagate Barracuda + SanDisk Ultra 3D 1TB |~| Case: Fractal Design Meshify C Mini |~| Display: Toshiba UL7A 4K/60hz |~| OS: Windows 10 Pro.

Luna, the temporary Desktop:

CPU: AMD R9 7950XT  |~| Cooling: bq! Dark Rock 4 Pro |~| MOBO: Gigabyte Aorus Master |~| RAM: 32G Kingston HyperX |~| GPU: AMD Radeon RX 7900XTX (Reference) |~| PSU: Corsair HX1000 80+ Platinum |~| Windows Boot Drive: 2x 512GB (1TB total) Plextor SATA SSD (RAID0 volume) |~| Linux Boot Drive: 500GB Kingston A2000 |~| Storage: 4TB WD Black HDD |~| Case: Cooler Master Silencio S600 |~| Display 1 (leftmost): Eizo (unknown model) 1920x1080 IPS @ 60Hz|~| Display 2 (center): BenQ ZOWIE XL2540 1920x1080 TN @ 240Hz |~| Display 3 (rightmost): Wacom Cintiq Pro 24 3840x2160 IPS @ 60Hz 10-bit |~| OS: Windows 10 Pro (games / art) + Linux (distro: NixOS; programming and daily driver)
Link to comment
Share on other sites

Link to post
Share on other sites

3 minutes ago, Princess Cadence said:

Before skylake it didnt matter at all, on haswell, sandy and other architectures that used ddr3 the speed differences were negligible.

 

skylake and kaby lake that uses ddr4 on the other hand finally shown significant improvement as you get higher memory frequencies, sure if you get just 2133mhz it will still be very good but there might be up to 10fps improvement in games going with 3000mhz speeds.

 

Keep in mind you usually need a z or x board to get higher memory frequency speeds than 2400mhz.

 

That would be why I thought it made no difference, I don't have any experience with DDR4 or Skylake and newer.

Laptop: Asus GA502DU

RAM: 16GB DDR4 | CPU: Ryzen 3750H | GPU: GTX 1660ti

Link to comment
Share on other sites

Link to post
Share on other sites

In some cases, it doesn't matter:

http://techbuyersguru.com/gaming-ddr4-memory-2133-vs-26663200mhz-8gb-vs-16gb?page=1

 

In other cases it does:

http://www.techspot.com/article/1171-ddr4-4000-mhz-performance/page3.html

 

However you have to consider the cost of fast enough RAM to have an appreciable performance impact, so it only makes sense if you're budget is basically "I don't care how much I have to spend."

Link to comment
Share on other sites

Link to post
Share on other sites

1 minute ago, Droidbot said:

Wrong. What architecture, OP? 

RAM frequency increases minimum framerates and average framerates by decreasing the time it takes to unload and load stuff into memory. On pretty much every arch besides Haswell.

Wrong (only because you said it first, lol). 

Ram speed helps depending on the amount of CPU overhead present in the game. This can be due to having an extremely powerful GPU that the CPU can't fully drive, or a specific part of the game is poorly coded, causing the CPU to hold the GPU back. It works just as well on Haswell as it does on every single architecture to exist. I know this, because I have my brothers rig sitting behind me with a 4690k and 980 Ti with my manually overclocked 2000 C8 kit in it. 

 

1 minute ago, Princess Cadence said:

Before skylake it didnt matter at all, on haswell, sandy and other architectures that used ddr3 the speed differences were negligible.

 

skylake and kaby lake that uses ddr4 on the other hand finally shown significant improvement as you get higher memory frequencies, sure if you get just 2133mhz it will still be very good but there might be up to 10fps improvement in games going with 3000mhz speeds.

 

Keep in mind you usually need a z or x board to get higher memory frequency speeds than 2400mhz.

It still mattered on Haswell. Anyone can test this. Why is this information still spreading around? http://www.overclock.net/t/1487162/an-independent-study-does-the-speed-of-ram-directly-affect-fps-during-high-cpu-overhead-scenarios

 

This isn't something new. Just because Digital Foundry tested this with Skylake first, doesn't mean it did not exist pre-Skylake. 

 

This is my problem with this forum. People are quick to answer questions they themselves do not even know the answer to. It's causing a massive influx of misinformation, and it makes my job extremely difficult when it comes to actually helping people understand the questions they ask.

My (incomplete) memory overclocking guide: 

 

Does memory speed impact gaming performance? Click here to find out!

On 1/2/2017 at 9:32 PM, MageTank said:

Sometimes, we all need a little inspiration.

 

 

 

Link to comment
Share on other sites

Link to post
Share on other sites

11 minutes ago, MageTank said:

This is my problem with this forum. People are quick to answer questions they themselves do not even know the answer to. It's causing a massive influx of misinformation, and it makes my job extremely difficult when it comes to actually helping people understand the questions they ask.

And some people are quick to try judging other based on single sources as if they were experts on it and even faster in trying to put people down while they are giving legit help. If before going around looking people to offend you bothered to read the title, OP interest lays on game performance exclusively, which as stated has negligible different when using ddr3 and haswell.

 

You arent getting paid in here so if you find your ´´job´´ so hard already just quit lol

Personal Desktop":

CPU: Intel Core i7 10700K @5ghz |~| Cooling: bq! Dark Rock Pro 4 |~| MOBO: Gigabyte Z490UD ATX|~| RAM: 16gb DDR4 3333mhzCL16 G.Skill Trident Z |~| GPU: RX 6900XT Sapphire Nitro+ |~| PSU: Corsair TX650M 80Plus Gold |~| Boot:  SSD WD Green M.2 2280 240GB |~| Storage: 1x3TB HDD 7200rpm Seagate Barracuda + SanDisk Ultra 3D 1TB |~| Case: Fractal Design Meshify C Mini |~| Display: Toshiba UL7A 4K/60hz |~| OS: Windows 10 Pro.

Luna, the temporary Desktop:

CPU: AMD R9 7950XT  |~| Cooling: bq! Dark Rock 4 Pro |~| MOBO: Gigabyte Aorus Master |~| RAM: 32G Kingston HyperX |~| GPU: AMD Radeon RX 7900XTX (Reference) |~| PSU: Corsair HX1000 80+ Platinum |~| Windows Boot Drive: 2x 512GB (1TB total) Plextor SATA SSD (RAID0 volume) |~| Linux Boot Drive: 500GB Kingston A2000 |~| Storage: 4TB WD Black HDD |~| Case: Cooler Master Silencio S600 |~| Display 1 (leftmost): Eizo (unknown model) 1920x1080 IPS @ 60Hz|~| Display 2 (center): BenQ ZOWIE XL2540 1920x1080 TN @ 240Hz |~| Display 3 (rightmost): Wacom Cintiq Pro 24 3840x2160 IPS @ 60Hz 10-bit |~| OS: Windows 10 Pro (games / art) + Linux (distro: NixOS; programming and daily driver)
Link to comment
Share on other sites

Link to post
Share on other sites

RAM speed isn't an issue BUT you can have a substential bottleneck when you overclock your CPU (if the CPU is beefy enough in the first place). For example, I know my "old" i5 3570k at stock speed didn't show any improvement in gaming by going from 1600MHz RAM to 2000+. However, when overclocking it to 4.5-4.6GHz, things changed and some CPU intensive games such as Watch Dogs 2 or GTA V saw around 5 fps improvement at 1080p with a GTX 980. And in the case of Watch Dogs 2, stuttering disappeared completely after overclocking both the CPU and the RAM. It will not be like that everytime though but if your RAM can sustain higher clocks, why not ?

CPU : i7 8700k @5GHz, GPU : ASUS GTX 1080 Ti STRIX, RAM : 2x8Go 3000MHz Corsair Vengeance, MB : ASUS Prime Z370-A, PSU : CM V850, Case :  NZXT S340, CPU Cooler : NZXT Kraken x62, Monitor : Acer Predator XB271HU 27" 1440p 165Hz, OS : Windows 10 Home 64 bits  

 

Link to comment
Share on other sites

Link to post
Share on other sites

33 minutes ago, Bleedingyamato said:

I'm starting to look for new RAM (2x8GB) to replace the 2x4GB sticks that are currently in my friend's desktop but idk what if any effect the MHz and latency have on Gaming or other performance. 

Fast RAM is only really applicable if you have a fast computer. Like any other system component, it can and will affect your system performance if its not fast enough. and you won't get any performance benefit from it if the rest of your system isn't fast enough to keep up with how fast it wants to operate.

take the eternal GPU/CPU bottleneck struggle for example. Having a 7700k is undoubtedly the fastest gaming CPU you can have on the market today. Pair it with a GTX 750ti and you're going to get about the same amount of performance out of your i7-7700k as you will out of a i3-7100 because your GPU can't keep up with how fast your 7700k wants to run, therefore your system merely runs as fast as the 750ti is capable of handling. Similarly, DDR4 is already fast enough that most gaming systems will not see an appreciable benefit from faster RAM in most situations unless you're running a high end CPU/GPU as well. RAM speed performance gains show up best when in a CPU intensive game as well (or in other CPU intensive tasks), so if you are playing mostly CPU intensive games you will see your benefits of faster RAM more often. Of course this will only be shown if the GPU ALSO has performance headroom to take advantage of, since if your GPU is already bottlenecking your CPU, then you're not going to be going any faster anyway.

As far as speed/timings go, think of it as a "Speed VS Timing" sort of thing. For optimal performance you want your speed as high as possible and timings as low as possible. 3200MHz C14 trident Z memory is a good example of this as it offers much faster than the normal speeds, yet marginally slower latencies compared to standard 2133MHz C12/C13 RAM . Compared to some other RAM from G,Skill and other companies in the same frequency bracket, 3200MHz is often accompanied by C16 or even C17 latency, which would sort of sour the performance you could get out of your 3200MHz threshold.




ALL THAT BEING SAID! The tangible performance gains for most people in most scenarios often turns people away from investing in faster RAM as its usually not that much. If you do not have a monitor capable of displaying more than 60hz I would say don't even bother as 2133MHz will have no problem operating at that refresh rate just fine. However, faster RAM often isn't a whole lot more expensive than slower RAM (so long as you aren't looking in the tremendously high frequency range like 3600+) so for many people it falls into the "why not?" category of spending. Throwing out a few extra bucks on a 1500 dollar system for a couple extra frames  is a bit of a no-brainer for some people. All depends on who you are, what you're doing, and how much your money is worth to you ;)

Link to comment
Share on other sites

Link to post
Share on other sites

1 minute ago, Princess Cadence said:

And some people are quick to try judging other based on single sources as if they are experts on it and even faster in trying to put people down while they are giving legit help. If before going around looking people to offend you bothered to read the title, OP interest lays on game performance exclusively, which as stated has negligible different when using ddr3 and haswell.

 

 

I am not putting anyone down. I am merely questioning why you make claims that you don't know the answer to, as if they are facts. As for being an expert on the subject, people often consider me the memory guy on this forum. I spend far more time overclocking memory (not just DDR4, but all kinds) and I've documented it's impact long before anyone else would bat an eye at it. I was called crazy until after Digital Foundry came to be, and even to this day, people question the validity of my results without even bothering to test them for themselves. As for your video, his methodology is flawed. He is testing only shows you the average framerates, and neglects minimums entirely (the framerate that faster memory has the biggest impact on, hence the entire CPU overhead argument).

 

I read the title, which is why I called the two of you out in the first place. I even posted two different sources showing gaming performance being boosted by faster memory. Do you not consider the 10-20% difference in minimum framerates in Digital Foundry's tests to be significant? Do you find the boosts seen on the Ivy Bridge 3930k to be a non-factor even as it scales perfectly all the way from 1333 to 2400?

 

Even as I type this post to you, more and more people are replying, and are wrong in their answers. They still don't understand WHY faster memory is helping, and write it off as if it's a placebo, or as if it's such a small measure of performance that it doesn't matter. They fail to realize how easy it is to overclock memory for free (other than the initial cost of the Z/X platform). 

 

The price difference between 1333 and 2133 DDR3 isn't that big. The same can be said about DDR4 2133 and 2800. There is no reason to sacrifice 10-20% gains in minimum framerates just to save $5-$10 on a kit. If the two sources I've provided is inadequite for you, let me know. I have plenty of sources to back these claims up, as this is a subject I've studied immensely. I can even show you how to test it yourself if you want to see it with your own eyes.

 

As for the answer to the second half of OP's question: Latency helps more when it comes to frametime. It's mostly beneficial in multi-GPU scenarios where frametime is poor, so having lower latency ram helps reduce how big the frametime spikes are. There is a test both on [H] and OCN (still ongoing) that has provided a fair bit of evidence on this subject. If you are interested, i'll dig those up as well.

My (incomplete) memory overclocking guide: 

 

Does memory speed impact gaming performance? Click here to find out!

On 1/2/2017 at 9:32 PM, MageTank said:

Sometimes, we all need a little inspiration.

 

 

 

Link to comment
Share on other sites

Link to post
Share on other sites

Memory speed in tomb raider (own results)

 

 

Spoiler

memoryspeeds.jpg

Direct link: 

https://linustechtips.com/main/uploads/monthly_2017_02/memoryspeeds.jpg.7579d1b51e04667242e615ed31ad71b0.jpg

 

Especially look at the GPU saturation between 1333 and 2400.

 

Memory speeds in GTA 5

 

 

Spoiler

memoryspeeds_GTA.jpg

Direct Link: https://linustechtips.com/main/uploads/monthly_2017_02/memoryspeeds_GTA.jpg.8bde41c940e90a04412a2619e8108692.jpg

 

Yes it matters, in games with a lot of big texturestreaming. Like...any modern game at this point. Since the new console generation revolves heavily around textures to make it all pretty. 

Link to comment
Share on other sites

Link to post
Share on other sites

26 minutes ago, Droidbot said:

Wrong. What architecture, OP? 

RAM frequency increases minimum framerates and average framerates by decreasing the time it takes to unload and load stuff into memory. On pretty much every arch besides Haswell.

It's a 6700K at stock speed on a Z170 mb.  

 

24 minutes ago, Princess Cadence said:

Before skylake it didnt matter at all, on haswell, sandy and other older architectures that used ddr3 the speed differences in performance were negligible.

 

skylake and kaby lake that uses ddr4 on the other hand finally shown significant improvement as you get higher memory frequencies, sure if you get just 2133mhz it will still be very good but there might be up to 10fps improvement in games going with 3000mhz speeds.

 

Keep in mind you usually need a z or x board to get support for higher memory frequency speeds than 2400mhz.

It's using a 6700K with a Gigabyte Z170MX Gaming 5 mb (my friend is very nice but currently wouldn't be able to afford anything near this nice so I decided to be nice and get her something very good since this desktop was a birthday present.)

 

Here's the build log I did for it:  

 

 

There's pictures and all the specs listed.  

 

19 minutes ago, M.Yurizaki said:

In some cases, it doesn't matter:

http://techbuyersguru.com/gaming-ddr4-memory-2133-vs-26663200mhz-8gb-vs-16gb?page=1

 

In other cases it does:

http://www.techspot.com/article/1171-ddr4-4000-mhz-performance/page3.html

 

However you have to consider the cost of fast enough RAM to have an appreciable performance impact, so it only makes sense if you're budget is basically "I don't care how much I have to spend."

How fast would you say would make a difference?

 

 

So basically "it might help but either way it'll cost extra that might not be worth paying?"

 

 

7 minutes ago, roylapoutre said:

RAM speed isn't an issue BUT you can have a substential bottleneck when you overclock your CPU (if the CPU is beefy enough in the first place). For example, I know my "old" i5 3570k at stock speed didn't show any improvement in gaming by going from 1600MHz RAM to 2000+. However, when overclocking it to 4.5-4.6GHz, things changed and some CPU intensive games such as Watch Dogs 2 or GTA V saw around 5 fps improvement at 1080p with a GTX 980. And in the case of Watch Dogs 2, stuttering disappeared completely after overclocking both the CPU and the RAM. It will not be like that everytime though but if your RAM can sustain higher clocks, why not ?

I don't plan to OC the 6700K and my friend isn't interested in stuff like OCing.  She's just happy to have a working good computer.  

 

The issue is I'd be tempted to stick with Corsair Vengence LPX like I have that comes in higher MHz like 3000 like mine is but I get crashes on my desktop that like my friend's uses a Gigabyte Z170 board (Z170X Gaming 7) when I've tried to enable XMP so I had to set the RAM to a higher speed manually.  

 

I was thinking about getting a different brand for my friend like G. Skill maybe or another brand if there's other good ones to consider idk if there are.  

 

But at least with G. Skill looking at some blue RAM of their's on Newegg I think the fastest I've seen was 2400 that was in stock.  So idk if that would be good enough.  

 

5 minutes ago, Zyndo said:

Snip

The full specs are in the build log link above.  My friend is currently using a 60Hz tv as a monitor (to save space) so maybe crazy fast RAM wouldn't be as helpful as it could otherwise be?

Link to comment
Share on other sites

Link to post
Share on other sites

25 minutes ago, MageTank said:

-snip-

Now you made your word worth it, do not get me wrong you have greater knowledge on it than I do I recognize it, however regardless of how much you know of the subject when passing it to others there is always a more polished way that does not sound so offensive.

 

My field is software as I do college for information systems but I do not go bashing people for not being perfectly accurate on their believes in the field, that was the sole issue I had, your initial addressing.

 

I will give a read on your material and get myself better informed to be able to help people even more accurately, and I do appreciate how you make the information available for this end. But It could have been all more smooth if you were not as aggressive originally.

 

Knowing but exposing it poorly is as good as not knowing as you will end up leading people to a rather fight for pride instead of a healthy exchange of experience.

 

with this said I hope we can be in good terms now [:

Personal Desktop":

CPU: Intel Core i7 10700K @5ghz |~| Cooling: bq! Dark Rock Pro 4 |~| MOBO: Gigabyte Z490UD ATX|~| RAM: 16gb DDR4 3333mhzCL16 G.Skill Trident Z |~| GPU: RX 6900XT Sapphire Nitro+ |~| PSU: Corsair TX650M 80Plus Gold |~| Boot:  SSD WD Green M.2 2280 240GB |~| Storage: 1x3TB HDD 7200rpm Seagate Barracuda + SanDisk Ultra 3D 1TB |~| Case: Fractal Design Meshify C Mini |~| Display: Toshiba UL7A 4K/60hz |~| OS: Windows 10 Pro.

Luna, the temporary Desktop:

CPU: AMD R9 7950XT  |~| Cooling: bq! Dark Rock 4 Pro |~| MOBO: Gigabyte Aorus Master |~| RAM: 32G Kingston HyperX |~| GPU: AMD Radeon RX 7900XTX (Reference) |~| PSU: Corsair HX1000 80+ Platinum |~| Windows Boot Drive: 2x 512GB (1TB total) Plextor SATA SSD (RAID0 volume) |~| Linux Boot Drive: 500GB Kingston A2000 |~| Storage: 4TB WD Black HDD |~| Case: Cooler Master Silencio S600 |~| Display 1 (leftmost): Eizo (unknown model) 1920x1080 IPS @ 60Hz|~| Display 2 (center): BenQ ZOWIE XL2540 1920x1080 TN @ 240Hz |~| Display 3 (rightmost): Wacom Cintiq Pro 24 3840x2160 IPS @ 60Hz 10-bit |~| OS: Windows 10 Pro (games / art) + Linux (distro: NixOS; programming and daily driver)
Link to comment
Share on other sites

Link to post
Share on other sites

7 minutes ago, Bleedingyamato said:

How fast would you say would make a difference?

 

So basically "it might help but either way it'll cost extra that might not be worth paying?"

3ghz seems to be the sweetspot for DDR4.

 

Whether it will help depends on the games, but as of late it started mattering quite a bit. Which isn't weird considering the consoles are highly dependent on fast memory, and lots of it. 16GB of DDR4-3000 won't break the bank and it can matter a lot in newer, more memory dependent games. Like Watch Dogs 2, Witcher 3, Rise of the Tomb Raider. Etc. But also CPU-bound games like GTA 5.

 

DigitalFoundry did a test, and in Witcher 3 the bump from 2133 to 3ghz gave more of a performance boost than setting the CPU to 4,7ghz.

Link to comment
Share on other sites

Link to post
Share on other sites

3 minutes ago, Bleedingyamato said:

I don't plan to OC the 6700K and my friend isn't interested in stuff like OCing.  She's just happy to have a working good computer. 

putting RAM anything above 2133MHz on skylake is overclocking. even through XMP its overclocking. If you don't want to overclock, get 2133MHZ RAM (lower timings will still be helpful). If all she wants is a computer that works good and has no issues, then 2133mhz is also the way to go as it is the most stable DDR4 frequency for skylake. Extremely unlikely to have any sort of issue on 2133MHz.

There is no sense in investing in faster RAM if you do not plan to take advantage of it or want to deal with the occasional side effects of OC-ing.

Link to comment
Share on other sites

Link to post
Share on other sites

3 minutes ago, Zyndo said:

putting RAM anything above 2133MHz on skylake is overclocking. even through XMP its overclocking. If you don't want to overclock, get 2133MHZ RAM (lower timings will still be helpful). If all she wants is a computer that works good and has no issues, then 2133mhz is also the way to go as it is the most stable DDR4 frequency for skylake. Extremely unlikely to have any sort of issue on 2133MHz.

There is no sense in investing in faster RAM if you do not plan to take advantage of it or want to deal with the occasional side effects of OC-ing.

If you're not overclocking, getting fast ram is the smartest 'next-best' thing you can do. Setting an XMP is not difficult. You probably need to be in the BIOS anyway to set a few parameters, like fans etc. 

 

And on what are you basing the lack of stability from 3ghz modules?

Link to comment
Share on other sites

Link to post
Share on other sites

1 minute ago, Princess Cadence said:

Now you made your word worth it, do not get me wrong you have greater knowledge on it than I do I recognize it, however regardless of how much you know of the subject when passing it to others there is always a more polished way that does not sound so offensive.

 

My field is software as I do college for information systems but I do not go bashing people for not being perfectly accurate on their believes, that was the sole issue I had, your initial addressing.

 

I will give a read on your material and get myself better informed to be able to help people even more accurately, and I do appreciate how you make the information available for this end. But It could have been all more smooth if you were not as aggressive originally.

 

Knowing but exposing it poorly is as good as not knowing as you will end up leading people to a rather fight for pride instead of a healthy exchange of experience.

 

with this said I hope we can be in good terms now [:

I was never on bad terms with you, nor do I take any offense to anything anyone says on this forum. You just happened to be the second post in this thread to ignite my holy crusade for memory speed yet again, lol. 

 

I often see this question on a daily basis, and every single day, I fight the conclusion that is drawn from multiple youtubers and their poor testing methodologies. That LTT video gets brought up at least once per thread when it comes to memory speed (and trust me, it will eventually show up in this thread too). My biggest issue isn't even the dismissal of the evidence I present. It's people downright refusing to test it for themselves, and merely go on repeating what other people say to them. I would rather people test this for themselves, and prove me wrong, than just dismiss the information entirely. At the very least, evidence proving otherwise helps the community as a whole. If you test it, and I am right, then good, we can all help people get more performance with faster ram. If I am wrong, good, we can save people money by using cheaper kits. More information is normally great, as long as the methodology is open and can be tested by everyone.

 

At some point, I wish to make an all-encompassing guide on memory that can hopefully be stickied in this sub-forum, so that this question can be answered once and for all. Sadly, I need to wait until Zen releases so that I can test it's IMC and the impact memory has on it. For AMD's previous generations, memory overclocking was a massive pain and scaled quite poorly due to how the IMC handled memory in general. 

 

3 minutes ago, Zyndo said:

putting RAM anything above 2133MHz on skylake is overclocking. even through XMP its overclocking. If you don't want to overclock, get 2133MHZ RAM (lower timings will still be helpful). If all she wants is a computer that works good and has no issues, then 2133mhz is also the way to go as it is the most stable DDR4 frequency for skylake. Extremely unlikely to have any sort of issue on 2133MHz.

There is no sense in investing in faster RAM if you do not plan to take advantage of it or want to deal with the occasional side effects of OC-ing.

CPU overclocking isn't required to take advantage of faster memory. If anything, faster memory will show an even greater impact on a slower CPU (more CPU overhead holding the GPU back). That being said, if you are investing in a Z platform to take advantage of faster memory, one is better off overclocking the CPU as well to get the most bang for your buck.

 

Stability shouldn't be an issue, as long as the IC's themselves are capable of handing the speed, and signal on the board is great. Skylake's IMC is very robust according to both my personal tests and ASUS's tests. They themselves listed the IMC failure rate to be around 3200mhz (meaning most Skylake IMC's can handle up to 3200mhz before IMC quality becomes an issue). My personal tests made it to around 3400 on 8 samples, with only one failing to hit it (and one other scaling negatively from that speed, due to how tertiary timings trained afterwards). Board quality plays a very important factor in memory overclocking, but mostly on the high end. A lot of boards are rated to handle up to 3600, which is well beyond the realm of diminishing returns. For gaming, I cannot tell a difference between my 3200mhz C14 XMP ,and my 3600 C14 manual overclock, which has way tighter tertiary timings, and far more bandwidth and lower latency. Had I owned a faster GPU or ran at a lower resolution to introduce more CPU overhead, that might change. For now, with a GTX 1070 at 1440p, 3200mhz is more than enough to satisfy my stock 4.2ghz 6700k. 

My (incomplete) memory overclocking guide: 

 

Does memory speed impact gaming performance? Click here to find out!

On 1/2/2017 at 9:32 PM, MageTank said:

Sometimes, we all need a little inspiration.

 

 

 

Link to comment
Share on other sites

Link to post
Share on other sites

4 minutes ago, Majestic said:

If you're not overclocking, getting fast ram is the smartest 'next-best' thing you can do. Setting an XMP is not difficult. You probably need to be in the BIOS anyway to set a few parameters, like fans etc. 

 

And on what are you basing the lack of stability from 3ghz modules?

 

It's overclocked in the factory, they've tested it before shipping out for sure.. it's stable as houses running a 3000Mhz stick at 2133 for stability to diagnose problems.

XMP is literally as easy as "Find it. Hit Enter. Hit [Enabled]. Save. Reboot." 

 

Modern Intel architectures take advantage of faster memory. 3-3.2Ghz is the sweetspot for the Skylake memory controller, and the Skylake controller (like Intel's recent memory controllers) is very robust and up to the task. 

idk

Link to comment
Share on other sites

Link to post
Share on other sites

27 minutes ago, MageTank said:

As for the answer to the second half of OP's question: Latency helps more when it comes to frametime. It's mostly beneficial in multi-GPU scenarios where frametime is poor, so having lower latency ram helps reduce how big the frametime spikes are. There is a test both on [H] and OCN (still ongoing) that has provided a fair bit of evidence on this subject. If you are interested, i'll dig those up as well.

I was looking at a G Skill kit that I think was 2400.  Is that acceptable do you think of should I look for better?

 

29 minutes ago, Majestic said:

Memory speed in tomb raider (own results)

 

 

  Reveal hidden contents

memoryspeeds.jpg

Direct link: 

https://linustechtips.com/main/uploads/monthly_2017_02/memoryspeeds.jpg.7579d1b51e04667242e615ed31ad71b0.jpg

 

Especially look at the GPU saturation between 1333 and 2400.

 

Memory speeds in GTA 5

 

 

  Reveal hidden contents

memoryspeeds_GTA.jpg

Direct Link: https://linustechtips.com/main/uploads/monthly_2017_02/memoryspeeds_GTA.jpg.8bde41c940e90a04412a2619e8108692.jpg

 

Yes it matters, in games with a lot of big texturestreaming. Like...any modern game at this point. Since the new console generation revolves heavily around textures to make it all pretty. 

So it did make a difference for you.  My friend likes the reboot TR games so that's good to know.  

 

The GPU use does look more stable at the higher speed.  What would that help with?

 

Link to comment
Share on other sites

Link to post
Share on other sites

6 minutes ago, MageTank said:

I often see this question on a daily basis, and every single day, I fight the conclusion that is drawn from multiple youtubers and their poor testing methodologies. That LTT video gets brought up at least once per thread when it comes to memory speed (and trust me, it will eventually show up in this thread too). 

Wew lad, don't even get me started.

 

Linus has no technical background or substantial education to make proper assessments. Anything that isn't a subjective experience isn't worth listening to.

 

2 minutes ago, Bleedingyamato said:

I was looking at a G Skill kit that I think was 2400.  Is that acceptable do you think of should I look for better?

 

So it did make a difference for you.  My friend likes the reboot TR games so that's good to know.  

 

The GPU use does look more stable at the higher speed.  What would that help with?

 

No, get 3ghz sticks for Skylake. My 2400 sticks are DDR3 (CL11), it's a bit different.

 

Yup, in those specific games. It's not in every game. But TR was probably the biggest difference, can't test Witcher 3 because I don't own it.

 

A few things. You want to be GPU-bound mostly, CPU-bound usually involves more spiking and frametime irregularity. It also means your GPU is less restricted by the CPU and you can achieve higher framerates.

 

 

Link to comment
Share on other sites

Link to post
Share on other sites

1 minute ago, Bleedingyamato said:

I was looking at a G Skill kit that I think was 2400.  Is that acceptable do you think of should I look for better?

 

So it did make a difference for you.  My friend likes the reboot TR games so that's good to know.  

 

The GPU use does look more stable at the higher speed.  What would that help with?

 

2400mhz is on the low end. I would look for at least 2800mhz. That being said, you can manually overclock ANY kit. I've purchased the cheapest, slowest 2133 kit I could find, with the worst possible IC's (Micron) and still took that kit to 3000mhz C15. I didn't do this once either. I bought an EVGA kit from Microcenter that had the same IC's, and took that kit to 3000 C16. If you need to save money, I'll gladly help you manually overclock your kit.

 

DDR4 is far easier to overclock when compared to DDR3, and it only takes about a full day to test overall stability. Just be warned. Faster memory will increase your CPU package temp when using extreme stress tests (Prime95, Linpack, etc) due to how AVX uses memory bandwidth. Memory bandwidth/capacity directly scales with heat in AVX stress tests. For gaming, you won't notice much of a thermal difference, if any. Just something to keep in mind if you plan on benching/stressing later on, and wonder why you are running hotter.

My (incomplete) memory overclocking guide: 

 

Does memory speed impact gaming performance? Click here to find out!

On 1/2/2017 at 9:32 PM, MageTank said:

Sometimes, we all need a little inspiration.

 

 

 

Link to comment
Share on other sites

Link to post
Share on other sites

11 minutes ago, MageTank said:

CPU overclocking isn't required to take advantage of faster memory.

I wasn't talking about his CPU. If he doesn't want to deal with the hassles of overclocking [any component], then he shouldn't get overclockable RAM. Even XMP settings can be unstable (XMP IS overclocking, just with predetermined parameters). My Corsair 3000MHz RAM isn't 100% stable on my Z170-A motherboard in XMP mode and causes occasional crashes. If he doesn't want to deal with overclocking, or that potential risk, then he should not pay more in order to get something he isn't interested in... If stability and no-hassle are his priorities, then he should get 2133MHZ memory, save himself a bit of money, and call it a day. No need to confuse the matter any further than that.

 

7 minutes ago, Bleedingyamato said:

I was looking at a G Skill kit that I think was 2400.  Is that acceptable do you think of should I look for better?

I think for your desires and what you want out of your system you should get 2133MHz (or maybe choose your RAM for an aesthetic reason) and call it a day. For a user like you, you shouldn't really consider RAM speed as a buying factor.

Link to comment
Share on other sites

Link to post
Share on other sites

1 minute ago, Zyndo said:

Even XMP settings can be unstable (XMP IS overclocking, just with predetermined parameters). My Corsair 3000MHz RAM isn't 100% stable on my Z170-A motherboard in XMP mode and causes occasional crashes. 

That's your substantial evidence, your n=1 subjective (and perhaps incompetent) opinion? 

 

wew

lad

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×