F5F Stay Refreshed Hardware Desktop Activated the XMP profile for 4000 MHz RAM...

Activated the XMP profile for 4000 MHz RAM...

Activated the XMP profile for 4000 MHz RAM...

X
xXSuperNovaXx
Posting Freak
811
02-11-2019, 01:57 AM
#1
This doesn't look right. Also go ahead and laugh at my CAS 18 RAM. I would like to fix it, however: 1. How do I measure if a different set of settings is giving better performance? I didn't notice any difference at all in cinebench r20 between 2400 and 4000 mhz RAM, AIDA64 gave incorrect numbers for the write test, etc. 2. I tried boosting the fabric clock to 1900 and the memory clock down to 3800. That didn't go so well and I had to reset CMOS. Looks like I need to keep the fabric clock at 1800. So I could try downclocking the memory to 3600 and reducing latency..but how do I know if it is even helping?
X
xXSuperNovaXx
02-11-2019, 01:57 AM #1

This doesn't look right. Also go ahead and laugh at my CAS 18 RAM. I would like to fix it, however: 1. How do I measure if a different set of settings is giving better performance? I didn't notice any difference at all in cinebench r20 between 2400 and 4000 mhz RAM, AIDA64 gave incorrect numbers for the write test, etc. 2. I tried boosting the fabric clock to 1900 and the memory clock down to 3800. That didn't go so well and I had to reset CMOS. Looks like I need to keep the fabric clock at 1800. So I could try downclocking the memory to 3600 and reducing latency..but how do I know if it is even helping?

I
itacarambi
Member
189
02-13-2019, 12:03 PM
#2
2000Mhz is accurate; with double data rate it becomes 4000Mhz. Unless you're really into fine-tuning, keep it as is if it's performing well and stable. Doing RAM overclocking or tweaking latency settings requires significant effort and usually doesn't show much improvement in everyday situations.
I
itacarambi
02-13-2019, 12:03 PM #2

2000Mhz is accurate; with double data rate it becomes 4000Mhz. Unless you're really into fine-tuning, keep it as is if it's performing well and stable. Doing RAM overclocking or tweaking latency settings requires significant effort and usually doesn't show much improvement in everyday situations.

S
sniper3609
Member
59
02-18-2019, 04:21 AM
#3
The result isn't a perfect 1:1 match with the infinity fabric. The 1800:2000 range indicates inconsistencies, often involving occasional latency delays.
S
sniper3609
02-18-2019, 04:21 AM #3

The result isn't a perfect 1:1 match with the infinity fabric. The 1800:2000 range indicates inconsistencies, often involving occasional latency delays.

A
akjosh47
Member
190
02-18-2019, 05:44 AM
#4
Your system performance appears normal at 4000MT/s with a 1800MHz clock, but timing mismatches are causing latency issues. Consider reducing memory speed to 3600MHz and optimizing timing settings if needed. Updated November 16, 2020 by Mateyyy
A
akjosh47
02-18-2019, 05:44 AM #4

Your system performance appears normal at 4000MT/s with a 1800MHz clock, but timing mismatches are causing latency issues. Consider reducing memory speed to 3600MHz and optimizing timing settings if needed. Updated November 16, 2020 by Mateyyy

M
mondoelite
Member
154
02-18-2019, 07:15 AM
#5
Check for visible changes or effects after applying it. If unsure, test on a small area first to confirm its impact.
M
mondoelite
02-18-2019, 07:15 AM #5

Check for visible changes or effects after applying it. If unsure, test on a small area first to confirm its impact.

X
xxmastr123
Junior Member
7
02-18-2019, 02:11 PM
#6
You can assess memory latency and bandwidth using AIDA64's benchmarking tools. Synchronizing memory with FCLK typically reduces latency. With an 1800MHz FCLK, expect around 3600MHz (or MT/s) in memory performance. Most Zen 3 chips should handle 1900MHz on the FCLK, which would translate to about 3800MHz in memory.
X
xxmastr123
02-18-2019, 02:11 PM #6

You can assess memory latency and bandwidth using AIDA64's benchmarking tools. Synchronizing memory with FCLK typically reduces latency. With an 1800MHz FCLK, expect around 3600MHz (or MT/s) in memory performance. Most Zen 3 chips should handle 1900MHz on the FCLK, which would translate to about 3800MHz in memory.

B
BVD47
Junior Member
40
02-18-2019, 03:47 PM
#7
I succeeded finally. DDR4 3600, CL 16. I just installed new RAM (important ballistix). How do you achieve such low latency? I’m seeing 70 ns at 3600/CL16. Is that possible with 3800/CL14 if timing is adjusted?
B
BVD47
02-18-2019, 03:47 PM #7

I succeeded finally. DDR4 3600, CL 16. I just installed new RAM (important ballistix). How do you achieve such low latency? I’m seeing 70 ns at 3600/CL16. Is that possible with 3800/CL14 if timing is adjusted?

T
TNTJeb1
Junior Member
35
02-19-2019, 11:06 PM
#8
Great! Likely these are Microv Rev-E or Rev-B models that perform well under load. It was around 3600 CL14 dual-rank, though the latency will naturally be greater on Ryzen compared to Intel due to their differing architectures even with similar memory configurations.
T
TNTJeb1
02-19-2019, 11:06 PM #8

Great! Likely these are Microv Rev-E or Rev-B models that perform well under load. It was around 3600 CL14 dual-rank, though the latency will naturally be greater on Ryzen compared to Intel due to their differing architectures even with similar memory configurations.