Using Aorus Pro X570. Currently running SniperX 16gb x4 sticks at 3000Mhz.
I need 128GB in near future and looking at upgrade options.
The memory QVL list from Gigabyte doesn’t say much about four stick configurations or memory quantities above 16GB. There is only one mention of 2x sticks of 32GB.
Also the QVL list is dated 020/02/18 while the very recent BIOS upgrades do mention improved memory compatibility so wonder what that means in practice.
then, what kind of speeds should I expect? I remember someone saying that 64GB running at 3000Mhz is technically way out of spec for AMD processor, but is this thing of a past? Surely it cannot be as easy as dropping 128 memory to the board and enabling XMP and using those speeds?
Frankly for me I don’t care much about the memory speed if around 3000 but if it drops to base spec like 2133 this might start show up.
I speak unto correction, but AMD recommends that configurations beyond 64GB of RAM that the 16GB modules run at 2666MHz. It’s an expected state with the memory controller in quad channel configs or sticks beyond 16GBx4.
Again, I could be mistaken but that’ been my understanding.
Probablly somewhere inbetween 2400mhz and 2933mhz is what i kinda expect.
But this just depends on too many factors to put an exact number on it.
It’s just a matter for trail and error and see how far you get.
For what its worth I have struggled with Ryzen 4 stick memory speeds on 64 GB but I have mis-matched modules (4x16, all DDR4-3000 g-skill, but half are faster CAS latency so different - which I didn’t notice until after I ordered. none of the sticks are QVL).
I think they’re running at 1866 or 2133 at the moment on a 2700x on an asrock x470.
not ideal, but they were cheap and I needed the capacity more than the speed for work from home, figured I can move some of the RAM off to other DDR4 machines when I upgrade some of the other boxes at home.
in all honesty? I have noticed zero difference (not benchmarked, in terms of user experience) between that and 32 GB @ ddr4-2866 whilst gaming general use, etc.
benchmarks probably worse but I’d say don’t be too scared of the potential drop if you need capacity more than speed. its made no difference to me and I simply dropped clock speed as far as required to get stable without bothering to tune anything beyond that.
if you’re on a 3950 with a much better memory controller and much bigger/better cache, I’d say don’t stress about it - again if its capacity you want.
Yeah i kinda agree with that.
On workstations type of use cases where a 3950X is mostly used for,
the memory capacity is likely more important then the general speed.
Although with certain workloads memory speeds could make a significant difference.
However Ryzen is still a dual channel platform and in any case when your particular,
workloads get really bottle necked by the memory trough put.
Then you should likely be on HEDT instead.
