MSI reaches DDR4 frequency of 4032MHz
Click here to post a comment for MSI reaches DDR4 frequency of 4032MHz on our message forum
tigermoth
im running 16 gig @ 2133 MHz @ 10,11,10,27 @ 1.55v but I thought memory speed today is pointless ?
I`ve here`d people say anything faster than 1866 MHz is pointless for performance during gaming yet i`ve allways argued with those who say it.
But with mantle around the corner and dx12 that both use the graphics ram to store what was usually stored in system memory in dx 11 makes memory speed less important today than yesterday.
So pls whats the point of memory running at this speed ?
Seketh
^
I agree and I ask the same. I have my memory at 2000Mhz 9-9-9-24 (still haven't checked memtest, honestly).
I love that DDR4 is coming, but so far it looks like it's less significant than the jump from DDR2 to DDR3.
fry178
the fact that most systems like that will not be used for gaming.
photo/video editing/rendering/server, will make use of all the ram and its speed.
eXXon
Not really, it is used for benchmarks. That's about it.
As for gaming, 2133 is the sweet spot now.
Servers or workstations would benefit a lot more going with ECC memory rather than pure speed.
Black_ice_Spain
there are apps which are heavily memory bound in the world, and some are very important, of course not executed that maaaaaaany times compared to games maybe.
Some in medicine and oil AFAIK
eXXon
For sure, but these apps require huge amounts of memory, not crazy speeds.
That's why server-grade ECC memory is rated at much, much lower speeds than gaming-grade memory.
Primey0
As for gaming it doesn't matter what speed. There is no sweet spot.
Cyberdyne
Yeah, I don't even think that 2133 or 1866 is the 'sweet spot'. I've seen test run DDR3 at 800mhz for games, and the most you see is a 1-3 frame drop compared to overpriced speedy RAM.
Nurmi
Because all sites seem to report average frames, I've been wondering myself that which end does it affect that frame or two.
ruiner13
Looks like they are using a 125 MHz BCLK. I'm of the opinion that having a faster FSB is more important than pure RAM speed. I bench higher when I run my CPU at 125 BCLK (vs 100MHz standard) resulting in slightly slower RAM (2000MHz vs 2133 at same timings) and CPU speed (4250MHz vs 4300). But, as usual, I'm sure it varies based on your components and where your bottleneck is in your system.
eXXon
Depends on the game. A quick Google search for BF4 shows this:
http://cwsmgmt.corsair.com/media/Blogs/DustinImages/BF41920.png
ApexiTT
That's a game a stock game taking advantage, I play my games at insanely high resolutions, maxed in game settings plus other profile hacks. When you are gaming sli/crossfire faster ram matters the most. I can tell a major difference between 1866 & just 2133 it can clear microstutter.
I always say only poor people say stupid **** that defies logic like You can't tell the difference between 1866 and higher ram or just don't believe everything you read.
Reddoguk
Linus tech tips has a video on YT showing tests of ram from 1333 -3000.
There is no difference really, if you could get DDR4 down to 9-9-9-24 then maybe a small increase.
https://www.youtube.com/watch?v=dWgzA2C61z4
ScoobyDooby
It was always looking to be that way.. certainly shouldn't be any revelations here about it.
That said, sellers gon' sell. They will push DDR4 every bit as hard as they did before with pretty shiny numbers on the package to get everyone to buy 32gb (still don't understand by anyone would buy over 16gb unless they are working in Unity/Autodesk/Heavy Adobe).
And here I am sitting with 8gb of DDR3 feelin fine about myself 🙂
tigermoth
I strongly disagree with the test done on that linus tech tips vid above and I will tell you why that is.
His using the same memory for all the tests so the quality of the memory people pay a premium for did not change and the performance of 1866 settings will be running @ a premium performance point. That's actually a fake result because 1866 memory isn't as good as 2400 so he cant do the test in this manner.
The performance of 1866 will be greater on this 2400 MHz memory kit of a memory kit designed to run @ 1866.
Its like me testing how good my car can stop using tires designed for a ferarri.
Its daft and he should not of done the test that way
All his showing is how good will 2400 MHz memory run @ 1866 mhz.
He should of used memory rated @ 1866 because that's real life.
No one runs 2400 MHz memory @ 1866
Make sense have I put my point across for you to understand ?
eXXon
A little harsh, but completely agree.
Once you go a little high-end with multi-GPUs & trying to eliminate stutters, Ram speed gets overlooked a lot.
And for the person rocking an 800MHz DDR2 claiming it is fine: if it works for you, then good for "you", not everyone else.
For me (and many other users), there's no way we could even enjoy a game with such low speeds.
And as for Linus, he's one of the worst with his inconsistent/weird testing methodology and holds no credibility with experienced users/gamers.
I once saw him review the Zalman CNPS12X (my cooler of choice) and he claimed it was not good at all and a cheap Evo212 is better.
yeah right.
(he should've checked out the review here regarding the washer mix up, which was stated in the comments on his video, but he didn't get it, which didn't surprise me)
Cyberdyne
No, it does not make sense. Because that is not how reality works. There is no magical secret sauce on higher speed RAM, either it is capable of higher speeds and lower timing or it's not.
In order to quell confusion, as you obviously are, we have a handy organization called JEDEC to create standards for us. All RAM must follow these standards, and regardless if it has a fancy heatsink or uses delicious gold plated tits, a piece of RAM sold at 2133mhz following JEDEC will perform identical to every other piece sold at the same speed. If we didn't do this companies could sell crap RAM with horrible timings but advertise as being something like 2400mhz.
So yes, an awesome piece of 2400mhz RAM running at 1600mhz using JEDEC (your BIOS does this automatically unless you manually change it) will be identical in performance to a cheapo 1600mhz stick.
Linus' video, while annoying (his voice, it's just nails on chalkboard), is accurate.
Also as a side note, some people will underclock their faster RAM at slower spec, a RAM stick binned at 1866mhz is likely to run stable for longer will less errors at 1600mhz.
CPC_RedDawn
Nope, its much like how CPU's and GPU's are made. They don't make different SKU's for each variant of CPU and its speed. They create the architecture and aim for a clock speed with it. This is where yields come into account, some of these chips will be binned and not able to reach said speeds. They are not actually thrown in the bin, they are then clocked at lower speeds and sold as lower versions of the same CPU. Or in a GPU's case some of the CU's and other features are turned off in order to make them stable and then released as lower end models. So in reality those 5960X CPU's more than likely came from the same wafer as the lower models from its range.
The same applies the memory chips, some can or can't reach said speeds so are binned as lower end models. But people like us like to alter settings and change voltages and multipliers to get our lower end chips stable at higher speeds 😉
What is wrong with Linus? Sure he can be a bit arrogant sometimes but he can be, hes probably the biggest tech youtuber around that actually knows his stuff.
Cyberdyne
I did say the video is accurate. I don't like his voice.
Also your explanation is incomplete. You don't explain how fast RAM can be underclocked to perform exactly like slower RAM due to standards.
CPUs are a flawed example, there are aspects of CPU's specification that cannot be controlled or changed easily from CPU to CPU (certainly not from the BIOS) like for example cache amounts. With RAM there is even less uncertainty, and because of the simplistic nature of RAM and the specifications there can be many many manufacturers of RAM, where CPUs only have two.
Corrupt^
I'm just abiding my time, watching as DDR4 matures...
Seems like we're going to have to push serious speeds to get any serious gain over DDR3. Though when that happens it'll finally be time to upgrade my 2600K and get a system with DDR4.
(not that I'm upgrading specifically for DDR4, but I like to be up to date when I buy a new motherboard/etc)
