F5F Stay Refreshed Hardware Desktop Timing specifications for RAM modules

Timing specifications for RAM modules

Timing specifications for RAM modules

B
Bananabut12
Member
64
01-17-2020, 10:08 PM
#1
I reviewed a Buildzoid video from earlier about RAM speeds for Ryzen processors. I changed my 5800X to an X3D IO setup and decided to revisit the timing settings. In the video, someone mentioned a tRFC above 1000, while mine was 1047—he warned it was problematic, advising against it. Eventually, I brought my RAM down to around 650. Unfortunately, I own two sets of RAM, both from Corsair, but the actual chips differ. One is SK Hynix and the other Micron. This makes it challenging to find better sub-timings because one type excels at certain speeds while the other is better in others. Overall, I’m curious about the impact of high tRFC values—so far, I haven’t seen any noticeable performance gains.
B
Bananabut12
01-17-2020, 10:08 PM #1

I reviewed a Buildzoid video from earlier about RAM speeds for Ryzen processors. I changed my 5800X to an X3D IO setup and decided to revisit the timing settings. In the video, someone mentioned a tRFC above 1000, while mine was 1047—he warned it was problematic, advising against it. Eventually, I brought my RAM down to around 650. Unfortunately, I own two sets of RAM, both from Corsair, but the actual chips differ. One is SK Hynix and the other Micron. This makes it challenging to find better sub-timings because one type excels at certain speeds while the other is better in others. Overall, I’m curious about the impact of high tRFC values—so far, I haven’t seen any noticeable performance gains.

_
_Vace
Junior Member
3
01-17-2020, 10:51 PM
#2
The key metric is the tRFC to tREFI ratio. This figure shows how often your RAM fails to support data operations. With a tRFC of 1000 and tREFI of 10000, you experience 10% downtime, refreshing enough to be usable. At 500 tRFC or below (like 250 on some DDR4 chips), only 5% or 2.5% of cycles are spent refreshing, allowing more operations per cycle. On AM4 boards, tREFI is fixed, so reducing tRFC is the main way to improve performance. For Intel platforms, the impact is smaller since you can adjust refresh intervals—even a low tRFC still results in minimal downtime (around 1.5%).
_
_Vace
01-17-2020, 10:51 PM #2

The key metric is the tRFC to tREFI ratio. This figure shows how often your RAM fails to support data operations. With a tRFC of 1000 and tREFI of 10000, you experience 10% downtime, refreshing enough to be usable. At 500 tRFC or below (like 250 on some DDR4 chips), only 5% or 2.5% of cycles are spent refreshing, allowing more operations per cycle. On AM4 boards, tREFI is fixed, so reducing tRFC is the main way to improve performance. For Intel platforms, the impact is smaller since you can adjust refresh intervals—even a low tRFC still results in minimal downtime (around 1.5%).

O
OhioGameBoyYT
Junior Member
1
01-17-2020, 11:41 PM
#3
generally...nothing... ram speed and timings are mostly overrated... i mean will ~5fps more really give you an edge? (in some cases maybe, but generally no) What's way more important for performance is having dual channel RAM (even at default speeds) as that gives huge benefits over single channel. *hypocrisy because i oc my 3200 ram to 3600 even though it shouldn't actually matter especially for my 5800x3D... but i did test it quite extensively and 3600 felt snappier, no fps boost or anything, just more responsive -- and im still not sure if its just placebo lol* Tldr: none issue, just make sure your RAM is actually stable and in dual channel mode
O
OhioGameBoyYT
01-17-2020, 11:41 PM #3

generally...nothing... ram speed and timings are mostly overrated... i mean will ~5fps more really give you an edge? (in some cases maybe, but generally no) What's way more important for performance is having dual channel RAM (even at default speeds) as that gives huge benefits over single channel. *hypocrisy because i oc my 3200 ram to 3600 even though it shouldn't actually matter especially for my 5800x3D... but i did test it quite extensively and 3600 felt snappier, no fps boost or anything, just more responsive -- and im still not sure if its just placebo lol* Tldr: none issue, just make sure your RAM is actually stable and in dual channel mode

J
Josi802
Member
100
01-19-2020, 05:19 PM
#4
If it's a Micron E-die, you might achieve performance around 545-560 at 3800MT/s, though this depends on your SOC voltage and DRAM voltage. At 3800MT/s you can reach 545 tRFC meeting RAM Test requirements for more than 10k% coverage, but there are challenges during cold boot. The advantage is clear in similar cases where the extra L3 cache provides significant gains, while outside those scenarios the improvements are minimal.
J
Josi802
01-19-2020, 05:19 PM #4

If it's a Micron E-die, you might achieve performance around 545-560 at 3800MT/s, though this depends on your SOC voltage and DRAM voltage. At 3800MT/s you can reach 545 tRFC meeting RAM Test requirements for more than 10k% coverage, but there are challenges during cold boot. The advantage is clear in similar cases where the extra L3 cache provides significant gains, while outside those scenarios the improvements are minimal.

Y
YT_SergioPvP
Junior Member
32
01-20-2020, 05:00 PM
#5
tRFC may influence outcomes more than strict primaries. I set mine at 255.
Y
YT_SergioPvP
01-20-2020, 05:00 PM #5

tRFC may influence outcomes more than strict primaries. I set mine at 255.

N
210
01-20-2020, 09:29 PM
#6
Check my RAM speeds right now. Is there anything that seems significantly off? I'm okay with what I see as long as it's acceptable.
N
NotTomZuper666
01-20-2020, 09:29 PM #6

Check my RAM speeds right now. Is there anything that seems significantly off? I'm okay with what I see as long as it's acceptable.

T
Teller
Junior Member
20
01-29-2020, 10:55 AM
#7
If its stable and running fine for you, then its good enough. Here is a shot of mine.
T
Teller
01-29-2020, 10:55 AM #7

If its stable and running fine for you, then its good enough. Here is a shot of mine.