Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
He kept suggesting the capacity difference was not the key variable.They tested 2x8 GB vs. 4x8 GB. A better test would have been 2x16 GB vs. 4x8 GB. I'm betting the advantage would disappear. Whatever they are testing obviously can take advantage of the extra RAM.
And IIRC the issue with four stick has never been about speed but stability at speed, and 3200 isn't that difficult a target.
He points out towards the end of the video Wendell said 2x16GB single rank performed best of all. I think he didnt go into detail to save it for a separate video. I agree about stability at speed being the real issue.They tested 2x8 GB vs. 4x8 GB. A better test would have been 2x16 GB vs. 4x8 GB. I'm betting the advantage would disappear. Whatever they are testing obviously can take advantage of the extra RAM.
And IIRC the issue with four stick has never been about speed but stability at speed, and 3200 isn't that difficult a target.
He kept suggesting the capacity difference was not the key variable.
He points out towards the end of the video Wendell said 2x16GB single rank performed best of all. I think he didnt go into detail to save it for a separate video. I agree about stability at speed being the real issue.
Haha I guess we will have to wait and see. He is putting that in a different video. His working thesis is Zen 3 performs better with four sticks rather than two. There was no evidence of 2x16GB performance in this video because he said it didn't make a difference but sounds like he will post findings in the next video about 2x16GB SR performance.I didn't get that far into the video, checked out when it appeared the test was flawed (varying RAM amounts). Was the 2x16 GB test at the same speeds as the others?
Huh? It's the only variable, unless he screwed up and missed something. The setup is otherwise identical, yet is faster with 32 GB. Whatever they're testing with obviously makes use of and benefits from that extra RAM. That it was spread over four sticks is irrelevant.
Cities skylines with mods would like to have a word with him....He said that no game even used close to 16GB, much less 32.
Haha I guess we will have to wait and see. He is putting that in a different video. His working thesis is Zen 3 performs better with four sticks rather than two. There was no evidence of 2x16GB performance in this video because he said it didn't make a difference but sounds like he will post findings in the next video about 2x16GB SR performance.
He said that no game even used close to 16GB, much less 32.
Cities skylines with mods would like to have a word with him....
Or galciv3...
There are actually several games out now that can easily push the 16gb limit.
His thesis appears flawed, and so is his testing to this point. Another stating that 2x16 GB performed best would seem to bear this out (again assuming the speeds were the same). The only way I can see four sticks inherently performing better is if the CPU had a quad-channel controller. However, like every modern mainstream desktop/laptop CPU, Zen 3 only dual-channel.
That he seeming rambled on for almost a half hour and couldn't find time to properly test 2x16 GB makes it seem like he's trying to push for views and follows.
What was the OS using? Other apps? Were the GPU drivers caching anything in RAM? What was the total RAM used? Is more being paged out if less RAM is installed?
There are tons of applications in general that can use all the RAM you have. Gamers may not "need" more than 16GB but you could use all of that in video projects, VFX like After Effects or inside of Resolve, 3D modelling/Rendering, amonst others.Cities skylines with mods would like to have a word with him....
Or galciv3...
There are actually several games out now that can easily push the 16gb limit.
That's not the end of the story. RAM fragmentation can also play a role. Your game might only use 8GB at any given time, but if suddenly there is a request to allocate new memory it matters wheter it can be done on completely free memory, or it has to use already used bits filled with cached assets.He said that no game even used close to 16GB, much less 32.
His thesis appears flawed, and so is his testing to this point. Another stating that 2x16 GB performed best would seem to bear this out (again assuming the speeds were the same). The only way I can see four sticks inherently performing better is if the CPU had a quad-channel controller. However, like every modern mainstream desktop/laptop CPU, Zen 3 only dual-channel.
Back in the day with single channel controllers, more sticks of ram improved performance with interleaving enabled.
Dual rank vs single rank. I'm surprised he missed that, considering it's been like that for a while now. 2x8GB usually just clocked higher so it was enough to overcome the benefits of the lower clocked 2x16GB/4x8GB kits.
Yes, which would turn it from dual rank to single rank. All things equal, dual rank configurations will outperform single rank. The performance difference varies, but it has been the case for a very long time. You would achieve the same performance with 2 dual rank dimms as 4 single rank dimms if the frequency and timings are equal.Steve used the same kit for testing, just pulled two sticks.
Yes, which would turn it from dual rank to single rank. All things equal, dual rank configurations will outperform single rank. The performance difference varies, but it has been the case for a very long time. You would achieve the same performance with 2 dual rank dimms as 4 single rank dimms if the frequency and timings are equal.
Depends on settings, but probably not that much. Keep in mind once you increase the resolution and settings to something you would actually use, the performance difference becomes even smaller.Yeah but never a 10% uplift right? I mean Ive never seen it
A 10 percent gain is in-line with going from single-channel to dual-channel, going all the way back. I doubt this has anything to do with the RAM capacity, just the RAM arrangement.Yeah but never a 10% uplift right? I mean Ive never seen it
Dual rank vs single rank. I'm surprised he missed that, considering it's been like that for a while now. 2x8GB usually just clocked higher so it was enough to overcome the benefits of the lower clocked 2x16GB/4x8GB kits.
Never mentioned it in the video and ninja edited the pinned comment.
This is SO SLOPPY! The FIRST rule when doing comparison tests is to have only ONE variable. Now in complex systems that can be difficult or even impossible so you do the best you can. Matching the total RAM used in both tests is EASY! If you do not have 2x16 GB you wait! No one was dying to see this test. I am very disappointed in their methodology.They tested 2x8 GB vs. 4x8 GB. A better test would have been 2x16 GB vs. 4x8 GB. I'm betting the advantage would disappear. Whatever they are testing obviously can take advantage of the extra RAM.
And IIRC the issue with four stick has never been about speed but stability at speed, and 3200 isn't that difficult a target.
In their testing suite. How does the message get so screwed up...Cities skylines with mods would like to have a word with him....
Or galciv3...
There are actually several games out now that can easily push the 16gb limit.
You will change ram density by stick there if do that no ?This is SO SLOPPY! The FIRST rule when doing comparison tests is to have only ONE variable. Now in complex systems that can be difficult or even impossible so you do the best you can. Matching the total RAM used in both tests is EASY! If you do not have 2x16 GB you wait! No one was dying to see this test. I am very disappointed in their methodology.
This is SO SLOPPY! The FIRST rule when doing comparison tests is to have only ONE variable. Now in complex systems that can be difficult or even impossible so you do the best you can. Matching the total RAM used in both tests is EASY! If you do not have 2x16 GB you wait! No one was dying to see this test. I am very disappointed in their methodology.
My guess is total RAM is more of a factor than RAM density.You will change ram density by stick there if do that no ?
If you are certain to use scenario that do not really use more than 8 gig, maybe the closest comparable is same sticks different number of the,.
You cannot have just one variable that change you have too chose which variable you change, the sticks used or the amount of memory used.
Ideally you then have to do multiple test where more than one variable change and trying to do some "regression analysis?" to isolate the one that interest you.
OK fair enough, but he should also test with total RAM the same to see if that makes a difference. It looks like he is going to do it.The "ONE variable" was the amount of RAM sticks. I don't think you quite realize how much difference there is between RAM kits.
It matters about 10 percent since it was invented. But the reason why it's kinda less known today is that nobody runs single-channel unless they have to, on like, a laptop with a single slot for RAM or soldered RAM.Maybe it was never or rarely focused on before because it did not matter much for games.
Just to clarify, I've heard about single vs dual channel memory for years. Its the single vs dual rank memory that is new to me.It matters about 10 percent since it was invented. But the reason why it's kinda less known today is that nobody runs single-channel unless they have to, on like, a laptop with a single slot for RAM or soldered RAM.
I'm 99 percent certain he just pulled the wrong sticks and ran all the tests, now he's in CYA mode since it's a pretty dumb mistake to make.
It matters about 10 percent since it was invented. But the reason why it's kinda less known today is that nobody runs single-channel unless they have to, on like, a laptop with a single slot for RAM or soldered RAM.
I'm 99 percent certain he just pulled the wrong sticks and ran all the tests, now he's in CYA mode since it's a pretty dumb mistake to make.
yay ..that's what I run!He points out towards the end of the video Wendell said 2x16GB single rank performed best of all. I think he didnt go into detail to save it for a separate video. I agree about stability at speed being the real issue.