32GB (CMK16GX4M2B3000C15) x2 @ 3000 MHz?

In theory you will be ok, the Pro4 supports up to 64GB RAM. The biggest problems you will encounter is more strain on the IMC (integrated memory controller) by populating all DIMM banks which means you may need to run looser timings and/or higher voltage for full stability. Manufacturers (particularly Corsair) sometimes switch what ICs memory kits use so there is a chance you could get a kit with an identical part number that uses entirely different ICs to your existing kit.

As a whole however you should be fine worst case scenario is that you have to run the memory a bit slower than its rated speed due to added IMC strain or the possibility you get a kit using different ICs.

EDIT - Just seen you already found out that last part. Loosen timings a bit, increase vDIMM a touch and try again for 3066/3200. You're just going to have to do a bit of manual tuning. I'm not sure what the default TRC is for Micron but increasing to 65 or 75 should help a fair bit with stability, as will disabling Gear Down mode.


Edited by ket - 20 Oct 2017 at 7:28pm

donalgodon wrote: I used the XMP profile for 2933. I haven't received the second kit yet, and I'm not even sure if I'm going to install it when it arrives. I might sell it instead. I'm finding that 16GB is likely enough for my use case scenario. I was surprised to learn that they were Micron. I expected Hynix.

That's the problem sometimes with memory models made by manufactures that don't make their own memory chips, which is most memory manufactures. They may change the memory chips used in a model of their memory. Note how memory manufactures rarely specify the source of the memory chips in the product specifications.

The manufacture may get a better deal on a similar type of memory chip from another source. Or they cannot get the same memory chips from the original source. Or they cannot get them in the quantities they need, when they need them.

The basic specs of the memory chips may (should) be the same, but they may have more deeply technical differences in other ways that make them incompatible with another source of the "same" memory chips. Which is why mixing memory with different chips may not work, or work as well together when used at their limits, such as when over clocked, or really used beyond their stock/default guaranteed stable speed.

On the other hand, I don't want to scare you into not trying to use the second kit. When someone asks us a question like will two kits of memory model X work at Y speed, and we answer yes it will or it should work, and then it doesn't work, what do we tell that person? What is their reaction to that situation? Even if we had identical systems, with the "same" multiple memory kits in use, they could work for us over clocked, but not for someone else. We all wish these things were that certain and simple, but they aren't.

Add to this Ryzen's almost random memory compatibility situation, and we are reluctant to make definitive statements. The same is really true for any type of over clocking on any system, which always has the caveat, it is not guaranteed.

donalgodon wrote: Thanks for the info. What's the real-world performance difference between the default clock and the higher ones like 3200? Is it noticeable outside benchmark apps?

That's a very general question about something that depends on many details. AMD has said that the Infinity Fabric connection between the Ryzen processor core modules and memory performs better at higher memory speeds. But it depends upon what work or task you are doing if a higher memory speed will provide a tangible difference. What is a significant real world difference in your opinion, a few seconds or a few minutes, for example?

If you will really be using over 16GB of memory for the work you do, then having more memory at a somewhat slower speed is preferable to running out of memory and using the page file on a drive. If you want 32GB just to have it but are worried about performance, then stay with 16GB.

We are not saying that 32GB will need to run at its stock speed. Instead of 3000, use the settings you use now for 3000, but set the speed to 2667, and see if it completes POST and boots and runs your OS. Default clock speed for Ryzen, or really Ryzen specific memory, was 2400, but now new DDR4 memory with a default clock speed of 2666 is available. Your memory's default speed is 2133, but if you can use 16GB at 3000 now, 2666 with 32GB could be possible, or at 2400.

If you get 32GB total memory, when you install the new 16GB, you will need to clear the CMOS/BIOS anyway, since you are adding new hardware, and so it will start the first time at its default speed. What did you do to configure your memory at 3000 now?

donalgodon wrote: I'm new to all of this, so I'd appreciate any help the more experienced folks here could offer. I have two (2X8) kits of CMK16GX4M2B3000C15 and I was hoping to populate all four slots of my AsRock AB350M Pro4 Micro ATX motherboard for a total of 32GB. 16GB is currently installed and running on my system and running at 3000 MHz, but I'd like to add the second set (which I just ordered) to upgrade to 32GB. I didn't even realize there could be a problem running 32GB until AFTER I ordered. Is this going to work? Will I be able to run 32GB at 3000 MHz?

impossible to predict if it will work at 3000. Normally the more memory/capacity in use, the more difficult it is to reach higher speeds. Several factors related to this:

You are using two "unmatched" sets of memory. While the sets "should be the same", sometimes manufactures change to different memory chips in a memory model. Mixing two different sets may not work as well. Also, there are other true technical reasons why memory is sold in matched sets. That does not mean your two sets will definitely not work together, but is a potential factor.

The integrated memory controller (IMC, part of the CPU) has more work to do with more memory, so it may not be able to reach higher speeds when more memory is used. Similar to the way processors of the same model have different over clocking abilities, the same is true about the IMC, its quality varies between examples of each processor model. Since a speed of 3000 is a memory OC for the IMC, and not guaranteed by specification, you may need to reduce the speed one notch down or more to handle 32GB.

The board itself can be a factor regarding maximum memory speed with any amount of memory. More expensive boards are not simply boards with more features, they are built to a higher standard, which can include the quality of the DIMM/memory slots and their connections to the CPU's IMC. The idea that all boards are built in basically the same way and will perform identically in all ways is false. The higher price of high end boards goes into many things, including the amount of time put into coding the UEFI/BIOS for things like memory over clocking.

Put these thing together, and who can predict the outcome?

Now that I have scared you, this is simply an explanation of why it is more difficult for higher amounts of memory to reach high speeds. It's nothing new, and the reason CPU manufactures have memory speed specifications. I hope all your memory can reach 3000, but if it can't, these are the usual reasons why it can fail.

4 sticks @ 3200 will be an uphill push as far as voltages go.

Best would be 4 running something lower yet with tighter timings.

Then again, you may find it does boot with 4 @ 3200.

If you try, try with only two sticks, finding stability, then add the other two and be prepped to add some SoC and VTT_DDR voltage


Keep us posted.