F5F Stay Refreshed Hardware Desktop Performance degration lors du changement vers la mémoire DIMM de rang double sur l'Aorus X570 Elite

Performance degration lors du changement vers la mémoire DIMM de rang double sur l'Aorus X570 Elite

Performance degration lors du changement vers la mémoire DIMM de rang double sur l'Aorus X570 Elite

C
Crazy_Heaven
Posting Freak
811
06-13-2020, 03:11 AM
#1
Hey there... welcome to the group. I’ve got some thoughts on what’s been happening with my setup. Here’s what I know:

Recently, I built this computer about a month ago. The key parts involved are:
- Ryzen 5900X
- Gigabyte Aorus X570 Elite @F33 Firmware
- G.Skill 16x2 [F4-3600C18D-32GTZR] – not the AMD version

The machine ran smoothly with XMP enabled, and RAM was clocked at FSB3600 at 22-22-42 MHz. I tried PBO Auto and learned about dual vs single rank memory. From CPU-Z info, my first pair is single rank (Hynix), while the second is dual rank (Samsung).

After installing all DIMMs, the PC wouldn’t boot properly until I cleared CMOS. I attempted to load XMP, but it didn’t work. I then tried running at a lower speed (2133MHz) and managed to get into Windows with all RAM recognized. CPU-Z Bench showed slightly better performance than average.

Later, I removed the older pair and kept the newer one in. When I booted again, XMP worked fine. But when I checked the module details in CPU-Z, it said the second set was dual rank (Samsung) and the first was single rank (Hynix).

My questions:
1. Why did the two sets give such different results? Why did the second one perform worse despite similar settings?
2. Is dual rank really superior to single rank? Are there any BIOS tweaks I should try?
3. Can I combine both sets to reach 64GB, given that the first was Hynix single and the second was Samsung dual?
4. Why did using 64GB with 2133MHz give better results than 32GB at 3600MHz? Does RAM amount really matter in this test?

Any insights or experiences from others would be super helpful. Thanks!
C
Crazy_Heaven
06-13-2020, 03:11 AM #1

Hey there... welcome to the group. I’ve got some thoughts on what’s been happening with my setup. Here’s what I know:

Recently, I built this computer about a month ago. The key parts involved are:
- Ryzen 5900X
- Gigabyte Aorus X570 Elite @F33 Firmware
- G.Skill 16x2 [F4-3600C18D-32GTZR] – not the AMD version

The machine ran smoothly with XMP enabled, and RAM was clocked at FSB3600 at 22-22-42 MHz. I tried PBO Auto and learned about dual vs single rank memory. From CPU-Z info, my first pair is single rank (Hynix), while the second is dual rank (Samsung).

After installing all DIMMs, the PC wouldn’t boot properly until I cleared CMOS. I attempted to load XMP, but it didn’t work. I then tried running at a lower speed (2133MHz) and managed to get into Windows with all RAM recognized. CPU-Z Bench showed slightly better performance than average.

Later, I removed the older pair and kept the newer one in. When I booted again, XMP worked fine. But when I checked the module details in CPU-Z, it said the second set was dual rank (Samsung) and the first was single rank (Hynix).

My questions:
1. Why did the two sets give such different results? Why did the second one perform worse despite similar settings?
2. Is dual rank really superior to single rank? Are there any BIOS tweaks I should try?
3. Can I combine both sets to reach 64GB, given that the first was Hynix single and the second was Samsung dual?
4. Why did using 64GB with 2133MHz give better results than 32GB at 3600MHz? Does RAM amount really matter in this test?

Any insights or experiences from others would be super helpful. Thanks!

V
Vanna_Chan
Junior Member
17
06-13-2020, 08:38 AM
#2
Always purchase a kit since they come in the same packaging and are made by the same company. This isn't guaranteed to succeed, but it's likely not the best approach. Fingers crossed the CMOS trick functions!
V
Vanna_Chan
06-13-2020, 08:38 AM #2

Always purchase a kit since they come in the same packaging and are made by the same company. This isn't guaranteed to succeed, but it's likely not the best approach. Fingers crossed the CMOS trick functions!

B
Bartekdwarf
Posting Freak
791
06-21-2020, 02:54 AM
#3
You've already explored several configurations and found some useful insights. The BIOS reset and default setup didn’t fully resolve the performance issues, but switching to the Dual Rank DIMM helped. Your observations about Infinity Fabric settings and memory timings are accurate. Regarding your question, it seems the Dual Rank DIMM offers a slight edge in speed, while keeping the same timings. As for running both simultaneously, sticking with 2133 at 15/15/36 might be better than 3600 at 22/22/42, depending on stability and compatibility. Keep testing different settings to see what works best for your system.
B
Bartekdwarf
06-21-2020, 02:54 AM #3

You've already explored several configurations and found some useful insights. The BIOS reset and default setup didn’t fully resolve the performance issues, but switching to the Dual Rank DIMM helped. Your observations about Infinity Fabric settings and memory timings are accurate. Regarding your question, it seems the Dual Rank DIMM offers a slight edge in speed, while keeping the same timings. As for running both simultaneously, sticking with 2133 at 15/15/36 might be better than 3600 at 22/22/42, depending on stability and compatibility. Keep testing different settings to see what works best for your system.

M
Maximoreyrojo
Member
141
06-21-2020, 07:17 AM
#4
Consider testing your flat at 15s with 2T and GDM turned off. That should work well. Your 5600X handles 15-15-15-36 2T GDM effectively, capable of 2100 1:1 at 3200C14.
M
Maximoreyrojo
06-21-2020, 07:17 AM #4

Consider testing your flat at 15s with 2T and GDM turned off. That should work well. Your 5600X handles 15-15-15-36 2T GDM effectively, capable of 2100 1:1 at 3200C14.