F5F Stay Refreshed Hardware Desktop Hi! Have some questions about ram overclocking

Hi! Have some questions about ram overclocking

Hi! Have some questions about ram overclocking

M
mistercraft77
Posting Freak
900
05-25-2019, 01:33 AM
#1
I'm using an i5-9600k with a gigabyte z390 Aorus Master and trying to boost the RAM speed. Right now I'm stuck at 3600 with 1.35V, which seems to be a 16-18-18 setup, though it was cheaper back then. After some trial and error I'm at 3600 with 1.35V and it's stable. I'm curious about the timing performance—what tests would be best? I've used basic error checking tools and haven't seen any issues. The system isn't crashing, which is odd since I think timing problems could cause crashes if the numbers change. If I try 18-19-19-38-58 at 3600, the RAM only runs at 3466, even though I set it higher. My notes suggest this might drop to 2400 or 2666 and reset the timings back to auto.
M
mistercraft77
05-25-2019, 01:33 AM #1

I'm using an i5-9600k with a gigabyte z390 Aorus Master and trying to boost the RAM speed. Right now I'm stuck at 3600 with 1.35V, which seems to be a 16-18-18 setup, though it was cheaper back then. After some trial and error I'm at 3600 with 1.35V and it's stable. I'm curious about the timing performance—what tests would be best? I've used basic error checking tools and haven't seen any issues. The system isn't crashing, which is odd since I think timing problems could cause crashes if the numbers change. If I try 18-19-19-38-58 at 3600, the RAM only runs at 3466, even though I set it higher. My notes suggest this might drop to 2400 or 2666 and reset the timings back to auto.

C
cecedabro
Member
182
05-26-2019, 01:59 AM
#2
It varies based on the memory IC you're using. Most are decent, but some can be extremely good or very poor. I usually aim for 3600 CL16, though it might handle 3733 CL17 or 3800 CL18 better—consider trying a higher frequency. They're not bad, but they're also not outstanding. I prefer lower tCL values, but your system likely can't reach them. I typically run three memory tests: one IMC stress test, two different stress tests. For controller testing, Y-Cruncher VST is reliable; it often fails quickly if memory speed is too high. For stress tests, there are many options—HCI Memtest, TestMem5, Karhu, MemTestPro, MemTest86 are popular choices. If you run two for 2–3 hours without issues, the setup seems stable. There are some oddities, possibly due to board settings at CL18, but it's probably a BIOS problem. With memory overclocking, these issues can appear randomly (some boards failed to POST at CL14 even on rated kits). You might be misreading timings as 16-19-19; likely your board can't reach 18-19-19 with the voltage you're using. If you invest time, tweaking subtimings can boost performance in memory-heavy tasks—though results are usually clear-cut. Updated June 16, 2023 by RONOTHAN##
C
cecedabro
05-26-2019, 01:59 AM #2

It varies based on the memory IC you're using. Most are decent, but some can be extremely good or very poor. I usually aim for 3600 CL16, though it might handle 3733 CL17 or 3800 CL18 better—consider trying a higher frequency. They're not bad, but they're also not outstanding. I prefer lower tCL values, but your system likely can't reach them. I typically run three memory tests: one IMC stress test, two different stress tests. For controller testing, Y-Cruncher VST is reliable; it often fails quickly if memory speed is too high. For stress tests, there are many options—HCI Memtest, TestMem5, Karhu, MemTestPro, MemTest86 are popular choices. If you run two for 2–3 hours without issues, the setup seems stable. There are some oddities, possibly due to board settings at CL18, but it's probably a BIOS problem. With memory overclocking, these issues can appear randomly (some boards failed to POST at CL14 even on rated kits). You might be misreading timings as 16-19-19; likely your board can't reach 18-19-19 with the voltage you're using. If you invest time, tweaking subtimings can boost performance in memory-heavy tasks—though results are usually clear-cut. Updated June 16, 2023 by RONOTHAN##

L
Larkn
Junior Member
38
05-27-2019, 11:01 AM
#3
Thanks to that RAM you're pleased you can reach 3600mhz at cl19. The 18-19-19 setting is tighter than 19-19, which limits you to 3466. You might need to increase the voltage. Around 1.55v should work on a Z390 chip. Keep in mind that 9600k only supports 2666mhz RAM by default, so you may need to adjust VCCIO and VCCSA. On most boards you won't have to apply equal voltages to both lines when overclocking; typical values are VCCIO: 0.950 and VCCSA: 1.050. With your lower overclock, you should be able to use the standard settings. You could also add a small boost—like 0.15v—to both lines for VCCIO (1.100v) and VCCSA (1,200v).
L
Larkn
05-27-2019, 11:01 AM #3

Thanks to that RAM you're pleased you can reach 3600mhz at cl19. The 18-19-19 setting is tighter than 19-19, which limits you to 3466. You might need to increase the voltage. Around 1.55v should work on a Z390 chip. Keep in mind that 9600k only supports 2666mhz RAM by default, so you may need to adjust VCCIO and VCCSA. On most boards you won't have to apply equal voltages to both lines when overclocking; typical values are VCCIO: 0.950 and VCCSA: 1.050. With your lower overclock, you should be able to use the standard settings. You could also add a small boost—like 0.15v—to both lines for VCCIO (1.100v) and VCCSA (1,200v).

D
DriveIn
Senior Member
739
06-06-2019, 07:21 AM
#4
Primaries aren't helpful; concentrate on frequency and timing details. Share a burner image and your current voltage settings. You should pick an IC to ease overclocking (use other people's examples). For stress testing, I use Prime95 LargeFFTS and others like HCI Memtest, TM5 (anta777), etc. When optimizing for overclocking, set voltage and frequency goals, stabilize the frequency first, then tweak timing, after that run primaries (save all voltages except for TLM). You can skip voltage tuning if the RAM IC handles scaling and cooling well. Aim for 1.5V before RAM overheats or crashes due to heat. With a fan, higher voltages are possible but depend on the IC and board. Generally, aim for 1.45V for stability, then adjust later. Vdimm depends on the RAM IC, motherboard, and case. Usually you're limited by the IC's capabilities or your board's limits—especially if it's a low-end chip like 8Gb, AFR, or C-DIE. Make sure your board supports the needed voltages and cooling. Good RAM ICs, like Bdie, RevE, DJR, etc., help, but cooling is the main bottleneck at higher voltages.
D
DriveIn
06-06-2019, 07:21 AM #4

Primaries aren't helpful; concentrate on frequency and timing details. Share a burner image and your current voltage settings. You should pick an IC to ease overclocking (use other people's examples). For stress testing, I use Prime95 LargeFFTS and others like HCI Memtest, TM5 (anta777), etc. When optimizing for overclocking, set voltage and frequency goals, stabilize the frequency first, then tweak timing, after that run primaries (save all voltages except for TLM). You can skip voltage tuning if the RAM IC handles scaling and cooling well. Aim for 1.5V before RAM overheats or crashes due to heat. With a fan, higher voltages are possible but depend on the IC and board. Generally, aim for 1.45V for stability, then adjust later. Vdimm depends on the RAM IC, motherboard, and case. Usually you're limited by the IC's capabilities or your board's limits—especially if it's a low-end chip like 8Gb, AFR, or C-DIE. Make sure your board supports the needed voltages and cooling. Good RAM ICs, like Bdie, RevE, DJR, etc., help, but cooling is the main bottleneck at higher voltages.

J
jordi1218
Member
105
06-16-2019, 09:20 PM
#5
The replacement part is essentially the same 3200 16-20-20 RAM, just from a standalone unit. Whether it impacts overclocking depends on the specific model and compatibility, but since it's not a kit, you should verify its specifications before proceeding.
J
jordi1218
06-16-2019, 09:20 PM #5

The replacement part is essentially the same 3200 16-20-20 RAM, just from a standalone unit. Whether it impacts overclocking depends on the specific model and compatibility, but since it's not a kit, you should verify its specifications before proceeding.

I
inconitar
Junior Member
6
06-17-2019, 02:27 AM
#6
Sure, that works. Just because the model number matches doesn't guarantee identical RAM. It's possible the new stock uses a different chip than the original, and they might not be compatible.
I
inconitar
06-17-2019, 02:27 AM #6

Sure, that works. Just because the model number matches doesn't guarantee identical RAM. It's possible the new stock uses a different chip than the original, and they might not be compatible.