Hello everyone, I'm new here but I have a big question about the memory band performance using Ryzen 5 and Ryzen 7. Taking the tests on AIDA64, I obtained the following results, with my memory Corsair Vengeance 2666 16-18-18-36
Ryzen 7 1700 Writing and Reading: 42,000
Ryzen 5 1600 Writing and Reading: 33,000
Now comes the question: Does this memory band depend on which CPU I am using? I ask this because I saw Rygen 2400G using a frequency of 2933 (which is not much above 2666 that I use) and the result is giving almost 45,000 being that it is a quad core processor weaker than the R5 1600.
How to explain this factor? I got to test my platform R5 1600 with the memory located at 2800 Mhz but not even close to the 42,000 I got with the R7 1700.
If anyone can explain this to me, thank you.
Memory controller is the same, bandwidth depends -heavily- on RAM speed due to the way Infinity Fabric works, but remember these numbers are fairly meaningless
even though the 2400g memory controller is more up-to-date compared to the first generation and explains the fact that my old R7 1700 hit around 42,000 and the current R5 1600 with the same motherboard configuration and Memory have much less memory bandwidth ? What is puzzling me ...
Your memory is much better than mine. I have 64gb and overclocar 4 modules is much more difficult for me but after much research I discovered the problem .. It was the last AGESA. I made the reversion to the previous version and now everything runs as it should and I do not have low memory band .. Even so .. Running at 2800 mhz only at 16-16-36 CL16 because my memories are 2400 mhz CL14 @
Simply because regardless of the CPU used if it is R5 or R7 and being the memory controller of the same generation and the memory clock being equal in both systems was not to have given that difference but as I explained above the problem was precisely in the new AGESA 1.0. 0.1a ... With AGESA 184.108.40.206 everything is OK but until I find out it was the new BIOS I lost almost 2 days.
Because these are not servers or professional workstations running half a dozen memory intensive applications simultaneously. Ryzen runs the Infinity Fabric at the same speed as the RAM, which is why the faster the RAM can run the better, but if bandwidth were that important to home programs and gaming, Threadripper 1900X would show clear advantages to Ryzen 1800X because it has a quad channel controller vs dual channel (amdmatt you have a Threadripper system if I remember correctly, run the AIDA64 cache and memory benchmark and post the results if you would) , but they perform within the margin of error between each other, showing that it is not a factor, while Ryzen RAM benchmarks show 8-10 FPS difference between DDR4 2400 and DDR4-3200. That's why those numbers are meaningless, home applications aren't RAM bound, haven't been for ages.
Your posts are double talk, to me anyway. Maybe I'm missing something. Maybe your talking about Ryzen with iGPUs on them versus vanilla Ryzen CPUs? I don't know.
Then you talk about 8-10 FPS differences which would absolutely have to be related to gaming. Then you switch back to talking about home applications in your last sentence. For God sake, please make some sense. Please be clear about what you are saying.
Memory bandwidth makes the difference between a game being fast or slow when measuring FPS. So, the more memory bandwidth performance you can get out of your memory chips all the better. I'm talking about "games" here now. Not, "home applications".
8-10 FPS difference when comparing DDR4-2400 and DDR4-3200, which is due to the way Infinity Fabric works, it runs at the same speed as the RAM.
Here is an absolutely stellar example of how games are not memory bound: Intel Coffee Lake. They don't use the Mesh technology, they're traditional chips. Guess what it shows? There is ZERO difference between DDR4-2600 and DDR4-4000, even on highly detailed games such as The Witcher 3, despite having more bandwidth due to a higher operating speed
So I stand by my statement that the numbers AIDA64 Cache & Memory Benchmark give you are pretty much meaningless.
Also, games are home applications. They're not VMWare running 20 virtual machines for multiple remote clients, they're not MySQL databases, they're applications for the home user.
I tried to clear this up in my last post when I said "Maybe your talking about Ryzen with iGPUs on them versus vanilla Ryzen CPUs? I don't know."
You keep making generalized statements which apply only to systems using discrete graphics cards. On APU's the higher the maximum memory bandwidth the better because the iGPU uses all that memory bandwidth.
Of course, when your using a discrete graphics card the game relies much less on system memory bandwidth because it's using it's own dedicated GDDR5 memory with it's own memory bandwidth at an effective memory bandwidth in GB/s that is probably at least 3x or more faster than the system memory.
Who seriously calls games "home applications"? I've never heard that one before. I guess I have now. Anyway, there is not really a disagreement between us at this point. Your posting of benchmarks in a review using a GeForce GTX 1080 pretty much cleared up the question I was asking. I just think you make a lot of blanket statements without specifying exactly what your talking about.