F5F Stay Refreshed Hardware Desktop Consider upgrading if you want better performance.

Consider upgrading if you want better performance.

Consider upgrading if you want better performance.

I
209
06-13-2020, 07:27 AM
#1
Hello everyone. A few months ago, I switched from my i7 5820X (X299) configuration to a Ryzen 9 5900X with an AM4 motherboard (X570 Aorus Master). I decided to replace my quad channel memory setup, which is a Corsair CMU32GX4M4C3000C15. To keep things simple, it’s a 32GB kit rated at 3000 mHZ with a CAS latency of 15. Like my previous upgrade, I planned to rebuild the whole system gradually—memory, NVMe SSD, graphics card, etc. Now I’m thinking about moving to a 3600 mHZ kit that offers a CAS latency of 18 or 16 (both Trident Z Neo options, 32GB). Since lower CAS is usually preferable, I’m wondering if it’s better to stick with my 3000 mHZ unit or go for the faster 3600 mHZ version, even though it might have a higher latency. Given that I don’t run as many benchmarks anymore, I’m not sure how much those numbers really affect me. I guess I was thinking my current 3000 mHz quad would still work fine. Maybe I could just swap in a couple of Gen 4 NVMe drives instead. But I wanted to discuss this anyway. Thanks ahead!
I
IninhaGamer_BR
06-13-2020, 07:27 AM #1

Hello everyone. A few months ago, I switched from my i7 5820X (X299) configuration to a Ryzen 9 5900X with an AM4 motherboard (X570 Aorus Master). I decided to replace my quad channel memory setup, which is a Corsair CMU32GX4M4C3000C15. To keep things simple, it’s a 32GB kit rated at 3000 mHZ with a CAS latency of 15. Like my previous upgrade, I planned to rebuild the whole system gradually—memory, NVMe SSD, graphics card, etc. Now I’m thinking about moving to a 3600 mHZ kit that offers a CAS latency of 18 or 16 (both Trident Z Neo options, 32GB). Since lower CAS is usually preferable, I’m wondering if it’s better to stick with my 3000 mHZ unit or go for the faster 3600 mHZ version, even though it might have a higher latency. Given that I don’t run as many benchmarks anymore, I’m not sure how much those numbers really affect me. I guess I was thinking my current 3000 mHz quad would still work fine. Maybe I could just swap in a couple of Gen 4 NVMe drives instead. But I wanted to discuss this anyway. Thanks ahead!

M
mistercraft77
Posting Freak
900
06-14-2020, 01:59 AM
#2
Lower CAS latency improves mainly when measured at the same frequency, since it reflects clock cycles. A larger value at a higher frequency can actually mean reduced overall delay. A faster clock provides more cycles in the same period, meaning each cycle lasts shorter. For example, 16 cycles at 1800 MHz finish quicker than 15 cycles at 1500 MHz. At 3000 MT/s (1500 MHz), CL15 takes 10 ns; at 3600 MT/s (1800 MHz), CL18 also takes 10 ns. CL16 at 3600 MHz runs in 8.88 ns, making it both quicker and lower latency compared to slower options.
M
mistercraft77
06-14-2020, 01:59 AM #2

Lower CAS latency improves mainly when measured at the same frequency, since it reflects clock cycles. A larger value at a higher frequency can actually mean reduced overall delay. A faster clock provides more cycles in the same period, meaning each cycle lasts shorter. For example, 16 cycles at 1800 MHz finish quicker than 15 cycles at 1500 MHz. At 3000 MT/s (1500 MHz), CL15 takes 10 ns; at 3600 MT/s (1800 MHz), CL18 also takes 10 ns. CL16 at 3600 MHz runs in 8.88 ns, making it both quicker and lower latency compared to slower options.

I
IMayBeDead
Senior Member
696
06-14-2020, 08:19 AM
#3
Higher speeds can improve benchmark performance. However, you likely won’t see a difference in everyday use. With four sticks already active, dual-ranking is in place, and a 3000 CL15 chip isn’t extremely slow—so the gains are probably minimal.
I
IMayBeDead
06-14-2020, 08:19 AM #3

Higher speeds can improve benchmark performance. However, you likely won’t see a difference in everyday use. With four sticks already active, dual-ranking is in place, and a 3000 CL15 chip isn’t extremely slow—so the gains are probably minimal.

P
Pottsiee
Junior Member
17
06-15-2020, 09:29 PM
#4
Slightly boost the RAM speeds to 3200mhz or above. Trident ZX Neo is also trying to overclock the memory, which isn't a great idea if you want to keep things as they are. They're just trying to push performance beyond normal limits.
P
Pottsiee
06-15-2020, 09:29 PM #4

Slightly boost the RAM speeds to 3200mhz or above. Trident ZX Neo is also trying to overclock the memory, which isn't a great idea if you want to keep things as they are. They're just trying to push performance beyond normal limits.

T
The_D3mon
Senior Member
694
06-16-2020, 03:59 AM
#5
Consensus reached on securing the lock before proceeding. If successful... nice. Otherwise, reconsider.
T
The_D3mon
06-16-2020, 03:59 AM #5

Consensus reached on securing the lock before proceeding. If successful... nice. Otherwise, reconsider.

C
cooldude99889
Member
179
06-16-2020, 10:57 AM
#6
You're operating a CPU with much higher performance than 3600mt/s. Most users target at least 3800 units for a 1900 MHz setup. Exceeding 4000 would require a 1:2 ratio—2000mhz at 1000MHz. Performance loss will become noticeable around 4267MHz. It's like investing in faster cars; the upgrade costs more upfront. A premium B-Die package can handle whatever you need, though it won't exceed 4000mt/s. Boosting latency with a 3600 CL14-14-14 model is possible. My B-Die units with extra voltage can push that timing to 4000mt/s, but I only use 16GB kits. @freeagent has tips for 32GB options and some recommended kits you might want to look into. Just to really boost performance, it's there. Your 2700X reaches 3733MHz—your rig should easily surpass that. Good luck!
C
cooldude99889
06-16-2020, 10:57 AM #6

You're operating a CPU with much higher performance than 3600mt/s. Most users target at least 3800 units for a 1900 MHz setup. Exceeding 4000 would require a 1:2 ratio—2000mhz at 1000MHz. Performance loss will become noticeable around 4267MHz. It's like investing in faster cars; the upgrade costs more upfront. A premium B-Die package can handle whatever you need, though it won't exceed 4000mt/s. Boosting latency with a 3600 CL14-14-14 model is possible. My B-Die units with extra voltage can push that timing to 4000mt/s, but I only use 16GB kits. @freeagent has tips for 32GB options and some recommended kits you might want to look into. Just to really boost performance, it's there. Your 2700X reaches 3733MHz—your rig should easily surpass that. Good luck!

H
HuBaZX
Junior Member
18
06-16-2020, 07:40 PM
#7
I'm working with a 5950X and Trident-Z Neo RGB...ATM running a 64 GB 3600 set on my main PC. I also have a 5800X paired with Trident-Z Neo RGB modules, 3600, but only a 32 GB setup. From the start, these aren't Samsung "B" dies—they're Hynix "M" dies. Right from the beginning at installation, setting XMP got them running at 3600. After a fresh Windows install, I experimented with some overclocking (just increased the DRAM frequency in the BIOS), and they've stayed stable at 3800 for weeks now. I haven't pushed them higher. Check the photos for proof of my argument here. Reminder to myself: capture a clearer image of the system configuration.
H
HuBaZX
06-16-2020, 07:40 PM #7

I'm working with a 5950X and Trident-Z Neo RGB...ATM running a 64 GB 3600 set on my main PC. I also have a 5800X paired with Trident-Z Neo RGB modules, 3600, but only a 32 GB setup. From the start, these aren't Samsung "B" dies—they're Hynix "M" dies. Right from the beginning at installation, setting XMP got them running at 3600. After a fresh Windows install, I experimented with some overclocking (just increased the DRAM frequency in the BIOS), and they've stayed stable at 3800 for weeks now. I haven't pushed them higher. Check the photos for proof of my argument here. Reminder to myself: capture a clearer image of the system configuration.

K
kervinc
Posting Freak
804
06-16-2020, 09:15 PM
#8
Great! Based on the 3000C15 setting, we can begin. It looks like you're currently at 16-15-15, which is fine. You might be running at 15-15-15-15-36 now. Adjust the other timings manually to the numbers shown, so they don't auto-set. Once you reach the bottom, choose 2T for odd CL and turn GDM off. Reset memclk to 3200, fclk to 1600, set DRAM voltage to 1.45 for S-N-G save, then exit to check the results. Some secondary values may display two options—let them auto unless a value is zero, in which case keep it as 0.
K
kervinc
06-16-2020, 09:15 PM #8

Great! Based on the 3000C15 setting, we can begin. It looks like you're currently at 16-15-15, which is fine. You might be running at 15-15-15-15-36 now. Adjust the other timings manually to the numbers shown, so they don't auto-set. Once you reach the bottom, choose 2T for odd CL and turn GDM off. Reset memclk to 3200, fclk to 1600, set DRAM voltage to 1.45 for S-N-G save, then exit to check the results. Some secondary values may display two options—let them auto unless a value is zero, in which case keep it as 0.

W
WreckerSG
Member
201
06-17-2020, 05:54 AM
#9
I took the lead and changed your configuration significantly. It seems I couldn’t locate the memclk and fclk parameters, but managed to adjust things like Gigabyte Bios does, so everything stayed stable. I struggled with timing and memclock speeds, which led to no posting. I could fine-tune some settings in the XMP profile, but maxed out at 15-17-17-35-52 @ 1.35. Once I hit 3200, it stops posting and resets the BIOS automatically. I think the machine feels sluggish with this XMP setting, though I’m not sure if that’s the case. It does read the system at 15-15-15-15-36, but it doesn’t actually execute it. This is when auto-reset kicks in.
W
WreckerSG
06-17-2020, 05:54 AM #9

I took the lead and changed your configuration significantly. It seems I couldn’t locate the memclk and fclk parameters, but managed to adjust things like Gigabyte Bios does, so everything stayed stable. I struggled with timing and memclock speeds, which led to no posting. I could fine-tune some settings in the XMP profile, but maxed out at 15-17-17-35-52 @ 1.35. Once I hit 3200, it stops posting and resets the BIOS automatically. I think the machine feels sluggish with this XMP setting, though I’m not sure if that’s the case. It does read the system at 15-15-15-15-36, but it doesn’t actually execute it. This is when auto-reset kicks in.

S
SoyDash
Posting Freak
859
06-18-2020, 01:08 AM
#10
Sorry about that... it was just a guess for me.
S
SoyDash
06-18-2020, 01:08 AM #10

Sorry about that... it was just a guess for me.