• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

X299 motherboards

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
You are missing IMC and cache dividers/speed in this calculation as memory speed and also stability on new platforms is generally limited by these 2 factors. Even though there can be "idle cycles" then there is a way to improve performance. This is why memory overclocking not always gives higher performance. It gives when all factors are matching. List of timings that are affecting stability and performance is really long not only CL, CR and one more. There are also additional settings, signal strength and other things.

I'm not saying that I've set perfect timings but that at these settings memory is fully stable and there are no issues with IMC or cache. In real, results are about as good as 3600 16-16-16 because limiting factors are somewhere else than in memory.
On older chipsets perfect timings were improving stability and performance. On new chipsets and DDR4 generally is some range in which it runs stable. Also some timings have to match for full stability but in most cases when we leave settings at auto then motherboards can handle that.

Other thing is that barely anyone cares if settings are perfect as long as everything is stable and performs well. We are not making any lab projects here or send people into space on these computers. Even then it wouldn't matter much. If I wanted perfect settings then I wouldn't push memory at 4000+.
What more, you never know if what you set in BIOS is really how it works. Motherboards are adjusting many settings behind our back but most users don't know that. In real some settings that user set wrong can be correct.
 
You are missing IMC and cache dividers/speed in this calculation as memory speed and also stability on new platforms is generally limited by these 2 factors. Even though there can be "idle cycles" then there is a way to improve performance. This is why memory overclocking not always gives higher performance. It gives when all factors are matching. List of timings that are affecting stability and performance is really long not only CL, CR and one more. There are also additional settings, signal strength and other things...

Actually while there are about forty memory settings you can set in bios if you set them at all different numbers the cpu has to do more work because as you have pointed out the memory controller is built into the cpu not a separate north bridge and it is done so that when all the setting are set whily nilly with no bais on what the cpu needs to get the information to the stepping transformer which most people don't even understand how it work on the memory die to be able to change the memory state from zero to one and one to zero, there is only a couple timings you set that effect the ability to be able to write ones and zeros when the chips tries to write the information.

If you are drawing a red square in the center of the screen twice in a row, you can optimize how that memory location is prioritized to be read from first while the waiting banks clear from random ones and zeros to all zeros so they can be written with a CAS statement instead of finding out if you need a CAS or RAS statement, or if the memory needs to be cleared to all zeros to allow for write statements only, instead of read parse ifstatement result write||skip statement. that is where the other values can be tweaked but only if you are not actually getting new information and you know your not getting new information is there any value in prioritized those other parts of the ram. Random Access Modules are designed around each slot or sector holding a random value that is not important except when it is being used.

You ram is not stable if you can not prime test it, because by the very nature of testing every possible number you are in essence trying to write a number and compare it to every other slot or sector in memory, at least until you start stack overflowing the memory and older version of prime 95 crash due to invalid compares.

Most people are happy with good enough. It is simply truth as long as your games play correctly and you don't lose data no one cares if the ram is throwing errors that is why most people don't need ecc memory. I just pointed out that at the speeds you are running you are likley throwing more than good enough errors but likely have all kinds of weird bad data on the memory with that pull down as the memory's internal speeds to read and write are faster than it can get from the cpu to know if it is even writing to the proper sector. Plus you are generating far more waste heat doing it that way which means the chips if you take the time to do some research might actully have far more head room thinking smarter not simply using a brute force method.

And I have to say it, you might not be sending people to space but I am, and while it likely will not be on consumer motherboards and chips the tech base is the same. Once Intel and AMD figure out a way around the sheering people are going to see chips that binned parts based on those chips and likely people wanting lots of errors but a pretty picture of the operating system booting, will start screaming that the chips have too much over head. End result Intel and AMD ignore those people there is too much money in the new 37 series of craft being able to put people on an asteroid belt with no environment controls and not having to pay anyone to mine it for rare earth metals and gold and platinum. I am a test pilot for US Space Command and my day job is running a think tank for congress. So ya that may not effect you but it is impacting the entire industry, mostly as I get tired of waiting for a company to build parts the aerospace industry needs and give the designs to an industry group and the company I was trying to work with screams why did I give the information to their competitors. So every time I see someone scream things like they are not sending people to space on thousand dollar chips... I sigh and have to wonder what do people think are going into consumer level rockets and launch vehicles and ships? Magic beans? Fairy Dust? they use the same tech everyone else does except when it does not work at speeds faster than mach one. Tesla is using over the counter tech because the rockets go into space at mach one that is why it takes ten minutes to reach twenty five miles up. Seven hundred miles an hour and the smallest sonic boom downward they can manage. The 37 series are water based ships (ships so it does not become United States Aerospace Ship) they use the turbines in the centerline to act like hydrofoils up to one hundred forty nine miles and hour at while point they rock back on the stakes like a cigar boat doing stunts, only instead of spinning, they can simply spin the turbines faster achieving flight, then detonate a liquid fuel afterburner instead of solid fuel rocket, safely away from anyone getting hurt by the larger sonic boom, and achieve space flight at far higher speeds. The aerospace industry said they needed twenty years to be ready for it, well they had twenty years and now they are having to deal with far more competition, which is actually causing more people to be able to own space ship eventually so I am predicting those that get onboard are going to make more money than they are now but I keep seeing people scream they are going to lose everything.
 
There is something to be said about being stuck in the minutia...

there is truth in saying someone can not see the forest for the trees, but when you point to a stable build that likely crashes during memtest rather than admit you likely jumped the gun on the build by grabbing the first close to stable number before the chip burns out fall short of useful info. I have built a modern cpu from carbon steel substrate layers. I have built a modern ram bank. And I have destroyed enough equipment over the years because I planned to replace it in six months when the next latest and greatest came out because I was writing off the costs are part of testing out various hardware. Most cpu crumble under more than one gee of acceleration and are useless if they get too hot. As long as we tell the industry the stuff they are making is good enough even if we have to run it way out of spec they are not going to build nice toys just keep selling you the same snowmen they sold you last year. Responding first makes a board feel alive trying to walk up and say that guy is crazy talk there is now way we will ever need more than 256 Kilobytes of memory between the cpu and memory... reminds me of why we have two companies we can by retail chips from to play video games but fifteen to build commercial hardware and retail appliances.
 
....and ill take the blue pill and the story ends. Not sure what you are droning on about now..
 
Byórðæįr, sorry but it's hard to read your posts. People say that I write wall of text but really I doubt that most forum users are reading your posts at all. It's just my thought so I can be wrong but I just know how forum members are ignoring everything that is too long.
I don't really know what are you fighting for here. Many forum members are interested how memory can be overclocked on X299 motherboards so I posted one result which is stable for me ( never said it's 100% stable ). Even if it wasn't fully stable then this is overclockers.com ... people are sharing overclocking experience here. I didn't say it's 100% stable, I didn't say these settings are perfect. I don't know if you understand that.
I just see no point to continue that.
 
Byórðæįr;8033828 said:
there is truth in saying someone can not see the forest for the trees... blah, blah, blah, but when you point to a blah, blah, blah, grabbing the first blah, blah, blah, fall short of useful info blah, blah, blah cpu from carbon steel substrate layers. Blah, blah, blah, I have destroyed blah, blah, blah, too hot. Blah, blah, blah say that guy is crazy.

This is what I got from that last post of his.
 
Back