Jump to content

How to lower the Latency of Memory

Net3

I have Ryzen 3600. It supports 3200MHz/s, usually unstable so i have it at 3000MHz 

Mobo: Asus TUF GAMING PLUS B450

RAM: Patriot Viper Steel 3733MHz CL17

Is it possible for me to lower the latency down as i lower the frequency of the ram? 

image.png.1edf3d105780e88e6df773ee9a984111.png

 

Am I doing something wrong that there is CAS 22? I think i have XMP enabled but manually set freq to 3000MHz e.g. 1500MHz

Link to comment
Share on other sites

Link to post
Share on other sites

For AM4 don't have RAM speeds lower than 3200MHz, anything lower than that's a pretty steep drop in performance.

Desktop: Ryzen 7 5800X3D - Kraken X62 Rev 2 - STRIX X470-I - 3600MHz 32GB Kingston Fury - 250GB 970 Evo boot - 2x 500GB 860 Evo - 1TB P3 - 4TB HDD - RX6800 - RMx 750 W 80+ Gold - Manta - Silent Wings Pro 4's enjoyer

SetupZowie XL2740 27.0" 240hz - Roccat Burt Pro Corsair K70 LUX browns - PC38X - Mackie CR5X's

Current build on PCPartPicker

 

Link to comment
Share on other sites

Link to post
Share on other sites

Just now, venomtail said:

For AM4 don't have RAM speeds lower than 3200MHz, anything lower than that's a pretty steep drop in performance.

But what can i do if it is unstable? Literally everything out of the box more than 3200 or even at that speed crashes and i have no idea why. at3200 it takes some time but it crashes eventaully.

Link to comment
Share on other sites

Link to post
Share on other sites

16 minutes ago, Net3 said:

But what can i do if it is unstable? Literally everything out of the box more than 3200 or even at that speed crashes and i have no idea why. at3200 it takes some time but it crashes eventaully.

Hello. The XMP/DOCP profile might not work on your motherboard. I suggest downloading Aida64, in Aida64 go to motherboard -> SPD. Find 3733mhz or 3200mhz and write down the full timings table like this:

image.png.cb1665e58596834d586430bbf7b3f080.png

 

Go to bios and set dram frequency, dram voltage and timings manually.

CPU: Ryzen 5800X3D | Motherboard: Gigabyte B550 Elite V2 | RAM: G.Skill Aegis 2x16gb 3200 @3600mhz | PSU: EVGA SuperNova 750 G3 | Monitor: LG 27GL850-B , Samsung C27HG70 | 
GPU: Red Devil RX 7900XT | Sound: Odac + Fiio E09K | Case: Fractal Design R6 TG Blackout |Storage: MP510 960gb and 860 Evo 500gb | Cooling: CPU: Noctua NH-D15 with one fan

FS in Denmark/EU:

Asus Dual GTX 1060 3GB. Used maximum 4 months total. Looks like new. Card never opened. Give me a price. 

Link to comment
Share on other sites

Link to post
Share on other sites

1 hour ago, DoctorNick said:

so far the only stable preset i've managed to get was  @1646MHz and ran it at FCLK 1600MHz /3200MHz

Tried the preset for 1755MHz and ran it at 1733MHz / 3466MHz and got BSOD. tried 3600MHz and got black screen and had to clear cmos. 

I think the 3600 cpu is garbage when it comes to that. I ran same ram sticks in another system with x570 board but with another 3600 cpu and got the same results, friend also had different ram sticks, but was also running at max 3200mhz.

 

But this at least works for now, i've used DRAM calculator before for Ryzen and it kinda sucks, absolutely everything crashes or is in need of cmos reset. 

but the only thing that i do not know is since i typed in CL15, it shows on CPUZ CL16, could you explain why is that? 

Link to comment
Share on other sites

Link to post
Share on other sites

5 minutes ago, Net3 said:

 

Yes I would choose either 3733 or 3200mhz - You can see the names of the timings on the right side: image.png.c14c6aba7259d65c9f1606eaa63b0e5d.png

CPU: Ryzen 5800X3D | Motherboard: Gigabyte B550 Elite V2 | RAM: G.Skill Aegis 2x16gb 3200 @3600mhz | PSU: EVGA SuperNova 750 G3 | Monitor: LG 27GL850-B , Samsung C27HG70 | 
GPU: Red Devil RX 7900XT | Sound: Odac + Fiio E09K | Case: Fractal Design R6 TG Blackout |Storage: MP510 960gb and 860 Evo 500gb | Cooling: CPU: Noctua NH-D15 with one fan

FS in Denmark/EU:

Asus Dual GTX 1060 3GB. Used maximum 4 months total. Looks like new. Card never opened. Give me a price. 

Link to comment
Share on other sites

Link to post
Share on other sites

2 hours ago, DoctorNick said:

Yes I would choose either 3733 or 3200mhz - You can see the names of the timings on the right side: image.png.c14c6aba7259d65c9f1606eaa63b0e5d.png

only 3200 works:)

Link to comment
Share on other sites

Link to post
Share on other sites

3 hours ago, Net3 said:

but the only thing that i do not know is since i typed in CL15, it shows on CPUZ CL16, could you explain why is that? 

To run an odd CL you need to turn Gear Down Mode off, and possibly switch to 2T. 1T with GDM on is like 1.5T so it’s doubtful you would notice. But it is hard to get a low latency with Zen 2. If you are under 60ns you are doing well.

AMD R7 5800X3D | Thermalright Frost Commander 140, TY-143
Asus Crosshair VIII Dark Hero | 32GB G.Skill Trident Z @ 3733C14
Zotac 4070 Ti Trinity OC @ 3045/1495 | WD SN850, SN850X, SN770
Seasonic Vertex GX-1000 | Fractal Torrent Compact RGB, Many CFM's

Link to comment
Share on other sites

Link to post
Share on other sites

  • 2 weeks later...
On 11/29/2022 at 2:45 PM, freeagent said:

To run an odd CL you need to turn Gear Down Mode off, and possibly switch to 2T. 1T with GDM on is like 1.5T so it’s doubtful you would notice. But it is hard to get a low latency with Zen 2. If you are under 60ns you are doing well.

so what you are saying is that i should favor the the latency in an actual bechmark test, rather than the CAS on the ram? 

I've watched today one of the buildzoid's videos and he had better latency (decreased latency) with cl40 than cl28 

kinda crazy. btw I have 78/79ns 😕

Link to comment
Share on other sites

Link to post
Share on other sites

2 minutes ago, Net3 said:

so what you are saying is that i should favor the the latency in an actual bechmark test, rather than the CAS on the ram? 

Yes. Lower is usually better of course, but tweaking certain sub timings can bring lower latency than simply switching CAS latencies depending on your sticks.

AMD R7 5800X3D | Thermalright Frost Commander 140, TY-143
Asus Crosshair VIII Dark Hero | 32GB G.Skill Trident Z @ 3733C14
Zotac 4070 Ti Trinity OC @ 3045/1495 | WD SN850, SN850X, SN770
Seasonic Vertex GX-1000 | Fractal Torrent Compact RGB, Many CFM's

Link to comment
Share on other sites

Link to post
Share on other sites

8 hours ago, freeagent said:

Yes. Lower is usually better of course, but tweaking certain sub timings can bring lower latency than simply switching CAS latencies depending on your sticks.

Yeah Trfc drop in 20’s till it won’t boot then put it back up 20

trefi 35000 - 66666

Tfaw 16 with the read and write activations at 4 work in most dies according to Buildzoid 

just those 5 settings get you a lot of performance 

-13600kf 

- 4000 32gb ram 

-4070ti super duper 

Link to comment
Share on other sites

Link to post
Share on other sites

8 hours ago, Net3 said:

so what you are saying is that i should favor the the latency in an actual bechmark test, rather than the CAS on the ram? 

I've watched today one of the buildzoid's videos and he had better latency (decreased latency) with cl40 than cl28 

kinda crazy. btw I have 78/79ns 😕

Yeah intel mlc gui/ aida64 

pyprime score 

 

-13600kf 

- 4000 32gb ram 

-4070ti super duper 

Link to comment
Share on other sites

Link to post
Share on other sites

  • 3 weeks later...
On 12/8/2022 at 3:06 AM, Ebony Falcon said:

Yeah Trfc drop in 20’s till it won’t boot then put it back up 20

trefi 35000 - 66666

Tfaw 16 with the read and write activations at 4 work in most dies according to Buildzoid 

just those 5 settings get you a lot of performance 

i will try to play around with them then. got a new Corsair Vengeance 32GB 3200MT kit, the timings are a bit higher in sub range than on the 16GB 3733MHz kit which i dont really understand. 

Is that like kit / die dependant? or the higher the maximum speed on the kit, and the lower you can go for your own needs, the better lows you can get on sub range timings?  

Let's take my example, viper kit with 3733mhz max speed, but underclocked at 3200mhz vs vengeance 3200mhz kit at the actual speed.

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×