F5F Stay Refreshed Hardware Desktop Memory configurations vary: 2x8GB, 4x8GB, or 2x16GB. Options include single rank or dual rank setups.

Memory configurations vary: 2x8GB, 4x8GB, or 2x16GB. Options include single rank or dual rank setups.

Memory configurations vary: 2x8GB, 4x8GB, or 2x16GB. Options include single rank or dual rank setups.

Pages (2): 1 2 Next
I
ImJaWe
Member
67
09-26-2020, 10:56 PM
#1
Gamersnexus covered memory settings for Ryzen 5000 processors. Steve from GN discussed performance variations between 2x8GB and 4x8GB configurations, referencing Wendell’s comments on single-rank memory. He noted that interleaving can affect results, which puzzled him. The key factors influencing this include memory rank and how data is accessed during gameplay. Without Wendell’s input on 2x16GB, it seems likely he believed dual-rank memory would outperform single-rank. For your setup—a 5950X with Aorus Master X570 and 2x G-skill Ripjaw 16GB RAM—consider whether sticking with 2x16GB keeps performance solid, especially if you plan to upgrade later. If you go for 4x16GB, a performance drop is possible, so a 2x32GB option might be safer now. Others have insights you can share.
I
ImJaWe
09-26-2020, 10:56 PM #1

Gamersnexus covered memory settings for Ryzen 5000 processors. Steve from GN discussed performance variations between 2x8GB and 4x8GB configurations, referencing Wendell’s comments on single-rank memory. He noted that interleaving can affect results, which puzzled him. The key factors influencing this include memory rank and how data is accessed during gameplay. Without Wendell’s input on 2x16GB, it seems likely he believed dual-rank memory would outperform single-rank. For your setup—a 5950X with Aorus Master X570 and 2x G-skill Ripjaw 16GB RAM—consider whether sticking with 2x16GB keeps performance solid, especially if you plan to upgrade later. If you go for 4x16GB, a performance drop is possible, so a 2x32GB option might be safer now. Others have insights you can share.

D
dutchmcsebas
Member
69
09-27-2020, 01:04 PM
#2
I anticipate a detailed video from Nexus gamers about this topic soon. It’s expected to significantly impact performance, and with the details emerging, they’ll likely act quickly. I plan to wait for the release to gather comprehensive insights before making any purchase.
D
dutchmcsebas
09-27-2020, 01:04 PM #2

I anticipate a detailed video from Nexus gamers about this topic soon. It’s expected to significantly impact performance, and with the details emerging, they’ll likely act quickly. I plan to wait for the release to gather comprehensive insights before making any purchase.

H
hrgriff
Senior Member
573
09-27-2020, 01:21 PM
#3
Thanks, the plan is confirmed. I was wondering if anyone else has already reported on this story from another source.
H
hrgriff
09-27-2020, 01:21 PM #3

Thanks, the plan is confirmed. I was wondering if anyone else has already reported on this story from another source.

D
dutchmcsebas
Member
69
09-27-2020, 10:08 PM
#4
Taking into account the I/O die remains consistent across generations, using very high-capacity dimms will have drawbacks such as reduced stable memory frequency and more flexible memory timing requirements. For workloads demanding substantial memory usage, beginning with a 2x16 GB configuration could be wise, allowing future upgrades to 64 GB if needed. If you're interested in detailed memory bandwidth analysis for various setups, Buildzoid offers a thorough guide on overclocking the 3000 series processors, which remains applicable even with the same memory controller in the 5000 edition. From what I observed, dual-rank performance plateaued as achieving stable quad-rank speeds became increasingly challenging, especially under tight timing constraints.
D
dutchmcsebas
09-27-2020, 10:08 PM #4

Taking into account the I/O die remains consistent across generations, using very high-capacity dimms will have drawbacks such as reduced stable memory frequency and more flexible memory timing requirements. For workloads demanding substantial memory usage, beginning with a 2x16 GB configuration could be wise, allowing future upgrades to 64 GB if needed. If you're interested in detailed memory bandwidth analysis for various setups, Buildzoid offers a thorough guide on overclocking the 3000 series processors, which remains applicable even with the same memory controller in the 5000 edition. From what I observed, dual-rank performance plateaued as achieving stable quad-rank speeds became increasingly challenging, especially under tight timing constraints.

T
techiseasy
Senior Member
688
10-14-2020, 01:13 AM
#5
The 2x16GB configuration will run in dual-rank mode. 16GB DDR4 modules also support dual-rank. I don’t understand the reasoning behind not using it.
T
techiseasy
10-14-2020, 01:13 AM #5

The 2x16GB configuration will run in dual-rank mode. 16GB DDR4 modules also support dual-rank. I don’t understand the reasoning behind not using it.

S
sifumind
Member
137
10-14-2020, 07:51 PM
#6
It might be worth revisiting GN... although I haven't been following the latest updates on Zen 3 testing. Generally, ram differences aren't huge in the bigger picture. Unless your project demands peak performance every second, you probably don't need to stress too much about it. The main scenarios where memory matters are heavy compute jobs (like running a system nonstop) and when you're doing detailed performance comparisons.
S
sifumind
10-14-2020, 07:51 PM #6

It might be worth revisiting GN... although I haven't been following the latest updates on Zen 3 testing. Generally, ram differences aren't huge in the bigger picture. Unless your project demands peak performance every second, you probably don't need to stress too much about it. The main scenarios where memory matters are heavy compute jobs (like running a system nonstop) and when you're doing detailed performance comparisons.

T
thingul
Member
136
10-14-2020, 08:50 PM
#7
Usually I'd support your view, though this scenario appears to create a 5-10% variation. Although doubling the RAM might not seem justified for 5-10%, when you're investing over $1500 in CPU, board, a 280mm AIO cooler and RAM, an additional $50 for that extra 5-10% makes sense.
T
thingul
10-14-2020, 08:50 PM #7

Usually I'd support your view, though this scenario appears to create a 5-10% variation. Although doubling the RAM might not seem justified for 5-10%, when you're investing over $1500 in CPU, board, a 280mm AIO cooler and RAM, an additional $50 for that extra 5-10% makes sense.

K
kalleboii
Senior Member
738
10-25-2020, 07:39 AM
#8
I realized why I stopped following GN. I glanced through the video and might have overlooked some details. In short, Steve mentioned in his chat with Wendell that 16GB single rank was optimal. His own experiments with 8GB modules revealed a 4x improvement over 2x8GB configurations. I think dual-rank 8GB units didn’t become common for a long time. Generally, the performance of 8GB modules was about 8% better than 2x8GB, possibly up to 10% more so in demanding tasks. This caught my attention since similar results appeared in Zen 2 and Intel’s compute lines, though I haven’t tested it in games much. During my tests, I noticed 2 DIMMs per channel (DPC) or using one dual-rank module per channel gave a boost over single-rank/1DPC, usually by around 8% to 10% in memory-heavy scenarios. This trade-off showed up in synthetic tests as higher effective bandwidth but with slightly more latency. The real impact depends on the application—rank interleaving helps unless rank isn’t a factor. Running 2DPC also uses a similar approach. Since I only have dual-rank 16GB modules, I lack experience with single-rank versions. Your original question remains: it could work depending on whether the modules are single or dual-rank, but we don’t have a direct comparison. In my experiments, 16GB dual-rank modules performed as expected for lower-capacity setups, so you’re likely seeing better results. If single-rank 16GB modules exist, they might deliver even more, but we’ll need more testing to confirm.
K
kalleboii
10-25-2020, 07:39 AM #8

I realized why I stopped following GN. I glanced through the video and might have overlooked some details. In short, Steve mentioned in his chat with Wendell that 16GB single rank was optimal. His own experiments with 8GB modules revealed a 4x improvement over 2x8GB configurations. I think dual-rank 8GB units didn’t become common for a long time. Generally, the performance of 8GB modules was about 8% better than 2x8GB, possibly up to 10% more so in demanding tasks. This caught my attention since similar results appeared in Zen 2 and Intel’s compute lines, though I haven’t tested it in games much. During my tests, I noticed 2 DIMMs per channel (DPC) or using one dual-rank module per channel gave a boost over single-rank/1DPC, usually by around 8% to 10% in memory-heavy scenarios. This trade-off showed up in synthetic tests as higher effective bandwidth but with slightly more latency. The real impact depends on the application—rank interleaving helps unless rank isn’t a factor. Running 2DPC also uses a similar approach. Since I only have dual-rank 16GB modules, I lack experience with single-rank versions. Your original question remains: it could work depending on whether the modules are single or dual-rank, but we don’t have a direct comparison. In my experiments, 16GB dual-rank modules performed as expected for lower-capacity setups, so you’re likely seeing better results. If single-rank 16GB modules exist, they might deliver even more, but we’ll need more testing to confirm.

X
xdpropelli
Junior Member
1
10-25-2020, 08:42 AM
#9
It seems single 16 GB modules perform better than dual-rank ones because using four of them lets you keep higher clock speeds even under tough conditions.
X
xdpropelli
10-25-2020, 08:42 AM #9

It seems single 16 GB modules perform better than dual-rank ones because using four of them lets you keep higher clock speeds even under tough conditions.

B
BadMemory
Member
206
10-25-2020, 09:46 AM
#10
I didn’t plan to rewatch, but I heard that a 2x16GB single rank setup was the top choice. This might differ from what I saw using only 8GB or less modules. Perhaps the high-capacity chips affect how they’re accessed, and this rank isn’t the main factor—something else matters. Also, when comparing setups, it’s important to keep speed and timing consistent. Those elements add extra considerations. We need the actual test data to grasp what’s happening for those aiming for peak performance. I should have noted earlier that the balance between CPU and RAM isn’t fixed, especially with Ryzen’s rapid CPU growth compared to RAM potential. Zen 3, being the most powerful core currently, will challenge the RAM system more, even with improvements like L3 cache to ease the load.
B
BadMemory
10-25-2020, 09:46 AM #10

I didn’t plan to rewatch, but I heard that a 2x16GB single rank setup was the top choice. This might differ from what I saw using only 8GB or less modules. Perhaps the high-capacity chips affect how they’re accessed, and this rank isn’t the main factor—something else matters. Also, when comparing setups, it’s important to keep speed and timing consistent. Those elements add extra considerations. We need the actual test data to grasp what’s happening for those aiming for peak performance. I should have noted earlier that the balance between CPU and RAM isn’t fixed, especially with Ryzen’s rapid CPU growth compared to RAM potential. Zen 3, being the most powerful core currently, will challenge the RAM system more, even with improvements like L3 cache to ease the load.

Pages (2): 1 2 Next