If your workload requires 128gb... Even at base speeds (non overclocked) it'll perform better than not having enough ram. Most I've seen haven't had many issues getting 3000mhz on 4 (decent) sticks, but much more difficult to hit 3600 with 4 vs 2. Of course, some have been able to and others struggle, but I haven't heard to many issues besides slightly lower overclocks. If this is a worthy trade-off, then cool. If you don't need that much memory and prefer slightly better timings, 2 sticks would do better in general.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
ATX vs mATX vs ITX
- Thread starter 1337Goat
- Start date
True, but since Ryzen likes memory speeds, it's definitely a trade off you have to consider - unlike Intel which doesn't give a shit about speed, but takes whatever you throw at it (it seems). I'd benefit from speed, for VMs and video work, and so I'm torn - 64G is good, 32 would be really tight, but I want 3600Mhz.
It's not like doing it with 32, where it pretty much just "works" now.
Correct.
This has been an issue to varying degrees for years. It's not specifically an X570 issue, but rather a general issue that effects some platforms and chipsets more than others. As for why, it comes down to a lot of variables. Differences in CPU IMCs and motherboard design all factor in here. Let's also not forget that when pushing the envelope for every last MHz, your highest achievable frequency will be determined by the worst of the four installed DIMMs. Overclocking your CPU can impact what memory speeds you can achieve as well. That said, if you look at AMD's specifications for X570 and the Ryzen 3000 series, AMD states that speeds up to DDR4 3200MHz are supported with two DIMMs. In contrast, using four reduces supported speeds to DDR4 2933MHz for single ranked DIMMs and DDR4 2666MHz for dual ranked modules. Intel is now listing different speeds for two vs. four modules in its product specifications now too. Intel didn't used to do that, but it only recently started supporting memory speeds in excess of DDR4 2666MHz officially.
In short, your statement is accurate. All things being equal, using two DIMMs will allow for higher memory frequencies than using four modules will. As stated earlier, this is for a variety of reasons. Back in the day, enthusiasts and motherboard makers learned that the mini-ITX motherboards were actually capable of higher memory clocks than their ATX counterparts due to the reduced number of memory slots. Even if they aren't used, the trace layout is different for motherboards with fewer slots. This is why there are also full sized ATX and E-ATX motherboards with a reduced number of memory slots.
Achieving higher memory clocks is typically more trouble on X570 than it is on Intel chipsets. Specifically, Intel's IMC is simply better than AMD's when it comes to memory clocking and using four modules. It's as simple as that. However, the bulk of AMD motherboards being designed to use daisy chain vs. T-Topology comes down to AMD and motherboard manufacturers making a conscious design decision to do so. With a T-Topology setup, AMD had previously discovered that it was more limited in regard to memory clocks. With Daisy chain, AMD could achieve higher clock speeds than ever before with the added benefit that the Ryzen architecture benefits greatly from higher memory speeds. As a result, AMD traded a lower maximum memory clock and the ability to run four modules with greater ease for the ability to achieve clocks generally comparable to Intel while losing the ability to handle four DIMMs very well.
Now, keep in mind that the Ryzen 3000 series memory controller is vastly improved compared to its predecessors. You can easily run four modules at speeds greater than what you typically could on earlier Ryzen CPU's using X370 and X470 chipsets. Now, the exact speeds you can achieve under any conditions with any given processor vary so much that specific examples are hard to provide. There are cases where some people achieved decent speeds using four modules and the older chipsets or four modules on newer ones. But that doesn't change the fact that generally, daisy chain out clocks T-Topology and that's why AMD recommends it. Even if it makes using four DIMMs somewhat problematic.
Indeed I do. As someone who reviews motherboards professionally and have for over a decade, I've tested this and dealt with it on many platforms. Again, using four DIMMs on any modern platform limits you more than two DIMMs does. Whether or not you'll run into that limitation depends greatly on what your trying to achieve and what hardware you are using.
It's problematic for higher density DIMMs as well. This is for the simple reason that there aren't too many if any larger memory IC's that clock all that well comparatively to their lower capacity counterparts. In essence, average 16GB DIMMs will out clock good 32GB DIMMs all day. As larger IC's become more common and technology advances, this will change in time. With four DIMMs, motherboard design, IMC quality, motherboard BIOS, memory SPD programing, motherboard trace layout, motherboard power design, cooling, PCB quality etc. all factor in as well.
Again, Intel does have the same problem AMD does. The difference is the range of speeds where the limitations come into play. You won't see the problem on the Intel side until you are at the upper end of the spectrum of what's possible right now with current production memory modules. AMD hits the wall much sooner when using four DIMMs.
BTW, you can get 64GB of RAM at DDR4 3600MHz using four DIMMs. That's what I'm running in my personal machine.
How much fuss was it to get stable?
IdiotInCharge
NVIDIA SHILL
- Joined
- Jun 13, 2003
- Messages
- 14,675
Gotta ask, have you had much experience with 2x32GB setups?
I've been watching stock (and costs
) of these at 3600C16, and even with those two hurdles cleared, I'm not sure if such a kit would be functional on the average enthusiast AMD build. Pretty sure my 9900K would take them, just would like to know if it's even that common to get them running on AMD!Gotta ask, have you had much experience with 2x32GB setups?
I've been watching stock (and costs
I have no practical experience with 2x64GB setups. I've never purchased any and no one has sent them for review or anything.
The problem with higher density modules is as I stated above. The current crop of ICs don't clock as high as their 16GB counterparts. You'll also pay allot for decent clocks and timings for such modules. But, they should work fine as long as you aren't trying to overclock them past their ratings. The rule about downclocking isn't about the size of the modules, but rather they're number.
Last edited: