To take advantage of Infinity Cache, you will want 4 modules of ram.
There are several reasons why that statement is wrong. One of them is that AMD's "Infinity Cache" is only on their GPUs, which neither know nor care how many RAM modules you have attached to your CPU.
To take advantage of Infinity Cache, you will want 4 modules of ram.
No.
The idea or theory which will be explored further in his next video on this is not that you need 4 stick of ram instead of 2, it's that you need 4 full ranks instead of two ranks.
The test were done with all the same 8gb sticks of "single rank" ram but in single channel vs dual channel or 2 stick vs 4 stick. The 4 sticks configuration won out because they think it was a full 4 ranks.
Steve will be doing another video on this as the theory is the sweet spot will be 2 sticks of dual rank ram which will equal 4 ranks. The belief is it's the ranks that matter not how many physical sticks you use.
This is something they are testing as it's new behavior being shown by zen 3 specific cpu's.
It's that using "4-Ranks" rather than 2 is more optimal because of the memory cycles. It's something about allowing more cycle to start while more are finishing.
The "Sweet Spot" is 4 ranks of memory. This in theory should mean 2 sticks of dual rank will be the best as to OC 2 sticks is invariably easier than 4 and should net you this increased performance shown in the video since it will be 4 ranks just like his 4 stick of single rank used in the video.
People are getting channels and ranks confused, they are not the same thing.
Read the description of the video and the first comment from GN:
This didn't used to matter (much) on previous AMD platforms; in fact, running 2 sticks of RAM was sometimes better because it'd be easier to stabilize the tighter timings and higher frequencies than running 4 sticks of RAM. Now, with Zen 3 and the AMD Ryzen 9 5950X, R9 5900X, R7 5800X, and Ryzen 5 5600X, it's become clear that 2x8GB vs. 4x8GB can impact performance, although this may not carry to other capacities as you exit single-rank configurations. In this benchmarking we're looking at several kits of memory in the beginnings of a "best memory" performance round-up for Zen 3, but we broke it into an initial 2x vs. 4x stick test.
Yes, we know it's to do with ranks. As stated, we have another video coming up with more detail. Everything is rushed with a launch. Re: Speaking with Wendell, we'll hopefully be doing something next week! He seems to indicate 2x16GB is a sweet spot, so we'll talk with him more about that soon and hope to feature him in a video!
Brenics ~ Just to point out I do believe Chris Roberts is going down as the man who cheated backers and took down crowdfunding for gaming.
What Linus was pointing out is the dependency on the memory speeds.If you are using slower memory you will basically be wasting your money. Also pointed out and something I figure as well is that the average gamer just sees what he wants to see,WOW this new chip is AWESOME but in reality never does any benchmarking to notice that he just spent money to get a 4-6% increase.
It all depends on weather you truly NEED an upgrade.I would say that I could use an upgrade because I play Atlas and the game is a super hog to run but lucky for me even at lower settings the game looks great.
I should make note that I did play FFXIV at around 40 frames and FFXIV looks as good as any mmorpg so if that is all your playing,there is no need for upgrades.How many gamers are playing games like Fortnite,lol's,ARPG's,WOW,Aion you do NOT need upgrades to play those games.
Never forget 3 mile Island and never trust a government official or company spokesman.
I believe you are making something more difficult then it is. I am sticking with my original analysis.
That's funny considering your analysis is different from the what the peeps in the video you posted came to.
Maybe you should read up on what ranks are and what channels, the physical slots on the mobo, are then learn the difference and then rewatch the video?
To make my statement less..... difficult ...... I think your analysis is flawed due to lack of understanding of what was said and talked about in the video.
However the next video they make should help clear things up as it will take more testing to be sure of what exactly is going on.
Brenics ~ Just to point out I do believe Chris Roberts is going down as the man who cheated backers and took down crowdfunding for gaming.
You don't know what the cause of the problem is until you've found it. I'm extremely skeptical that AMD would design their memory controllers to choke if you only use two modules. Given that their memory controllers are literally exactly the same as in their previous generation, it would be very strange if it's a design problem unique to the 5000 series. It would also be very strange if it was also true of the 3000 series and no one noticed it until now.
One simple cause would be that when he's using two modules, he's putting them in the wrong slots. Normally, there are two slots for each memory channel, and you can have either a 64-bit connection to one module or a 32-bit connection to each of two modules. If you have two memory modules and put them in the two slots for the same channel, you've just cut your memory bandwidth in half.
Even if you put one module in each channel, there's no guarantee that the motherboard is wired to be able to offer a 64-bit connection from either of the slots in a given channel. Some motherboard manuals will tell you that if you're only going to use two modules, use these two particular slots. If the other two slots are only wired to offer a 32-bit connection, using the wrong slot can cut your memory bandwidth in half.
It's also possible that it's just a BIOS bug. Maybe tweaking memory timings so much caused something to break. Maybe it's an AGESA problem that AMD will need to fix. But again, you don't know what the problem is until you've found it.
It's a completely new memory controller, this is why I am not going to jump down peoples throats for having a difference of interpretation, if it is wrong on "either or".
This shouting down phenom in media is pretty much what I'm writing my blog about, though specific to social media, the same theory can be applied to a host media types. .
AMD says it's the same memory controller. They just increased the max speed a bit
It's a completely new memory controller, this is why I am not going to jump down peoples throats for having a difference of interpretation, if it is wrong on "either or".
This shouting down phenom in media is pretty much what I'm writing my blog about, though specific to social media, the same theory can be applied to a host media types. .
The memory controllers are part of the I/O die, and that is literally the same die as the Ryzen 3000 series used.
People pointing out that your claims are objectively, factually wrong is a rather poor example of "shouting down". No one is calling your employer demanding that you be fired. No one is even demanding that you be kicked off of this forum.
Comments
Brenics ~ Just to point out I do believe Chris Roberts is going down as the man who cheated backers and took down crowdfunding for gaming.
Also pointed out and something I figure as well is that the average gamer just sees what he wants to see,WOW this new chip is AWESOME but in reality never does any benchmarking to notice that he just spent money to get a 4-6% increase.
It all depends on weather you truly NEED an upgrade.I would say that I could use an upgrade because I play Atlas and the game is a super hog to run but lucky for me even at lower settings the game looks great.
I should make note that I did play FFXIV at around 40 frames and FFXIV looks as good as any mmorpg so if that is all your playing,there is no need for upgrades.How many gamers are playing games like Fortnite,lol's,ARPG's,WOW,Aion you do NOT need upgrades to play those games.
Never forget 3 mile Island and never trust a government official or company spokesman.
Brenics ~ Just to point out I do believe Chris Roberts is going down as the man who cheated backers and took down crowdfunding for gaming.
One simple cause would be that when he's using two modules, he's putting them in the wrong slots. Normally, there are two slots for each memory channel, and you can have either a 64-bit connection to one module or a 32-bit connection to each of two modules. If you have two memory modules and put them in the two slots for the same channel, you've just cut your memory bandwidth in half.
Even if you put one module in each channel, there's no guarantee that the motherboard is wired to be able to offer a 64-bit connection from either of the slots in a given channel. Some motherboard manuals will tell you that if you're only going to use two modules, use these two particular slots. If the other two slots are only wired to offer a 32-bit connection, using the wrong slot can cut your memory bandwidth in half.
It's also possible that it's just a BIOS bug. Maybe tweaking memory timings so much caused something to break. Maybe it's an AGESA problem that AMD will need to fix. But again, you don't know what the problem is until you've found it.
Brenics ~ Just to point out I do believe Chris Roberts is going down as the man who cheated backers and took down crowdfunding for gaming.
Source:
https://www.reddit.com/r/Amd/comments/jomorg/conclusion_on_the_best_ram_mhz_for_ryzen_5000/
People pointing out that your claims are objectively, factually wrong is a rather poor example of "shouting down". No one is calling your employer demanding that you be fired. No one is even demanding that you be kicked off of this forum.