RAM overclocking question

[Geriatrix]

Hi guys and girls,
I have an Asrock B450 Steel Legend motherboard with a Ryzen 9 3900X CPU.
My RAM is 2 x 16GB (so 32GB total) Kingston Fury Renegade DDR 4 3200MT/s.
My question is, can I safely push the RAM to 3600MT/s, given that both my CPU and Mobo can handle those speeds and is it beneficial in any way to do so?

 

[thecurto]

Hi guys and girls,
I have an Asrock B450 Steel Legend motherboard with a Ryzen 9 3900X CPU.
My RAM is 2 x 16GB (so 32GB total) Kingston Fury Renegade DDR 4 3200MT/s.
My question is, can I safely push the RAM to 3600MT/s, given that both my CPU and Mobo can handle those speeds and is it beneficial in any way to do so?

There generally isn't a giant performance leap from 3200 to 3600 but you will likely encounter some stability issues if you're tinkering above the rated XMP speeds. I'll leave the details for the more experienced Carbies but my suggestion would be to enable XMP and call it a day.

Here's an old YouTube video with your exact CPU and a 2080 Ti (3070 / 6750 XT equivalent). The 3200 results are with CL14 so that will be the best case scenario. The FPS increase is around 2-5 FPS so this may not be worth the risk of instability.

 

[Geriatrix]

There generally isn't a giant performance leap from 3200 to 3600 but you will likely encounter some stability issues if you're tinkering above the rated XMP speeds. I'll leave the details for the more experienced Carbies but my suggestion would be to enable XMP and call it a day.

Here's an old YouTube video with your exact CPU and a 2080 Ti (3070 / 6750 XT equivalent). The 3200 results are with CL14 so that will be the best case scenario. The FPS increase is around 2-5 FPS so this may not be worth the risk of instability.

Ah, thank you, forgot to mention, I have already enabled XMP, so it is running at 3200MT/s.

 

[Spaffy]

Hi guys and girls,
I have an Asrock B450 Steel Legend motherboard with a Ryzen 9 3900X CPU.
My RAM is 2 x 16GB (so 32GB total) Kingston Fury Renegade DDR 4 3200MT/s.
My question is, can I safely push the RAM to 3600MT/s, given that both my CPU and Mobo can handle those speeds and is it beneficial in any way to do so?

RAM timings make a bigger difference than RAM speed but if you're not that into overclocking then set "XMP" and forget.

 

[D-wired]

Hi guys and girls,
I have an Asrock B450 Steel Legend motherboard with a Ryzen 9 3900X CPU.
My RAM is 2 x 16GB (so 32GB total) Kingston Fury Renegade DDR 4 3200MT/s.
My question is, can I safely push the RAM to 3600MT/s, given that both my CPU and Mobo can handle those speeds and is it beneficial in any way to do so?

To define the word "Safely" would answer your question better.

Can push your ram as far as you want, but will get into stability issues. It is safe to do so, nothing will explode😂, but won't be stable. Will experience BSOD while busy on PC.

I would advise watching BuildZoid (HardcoreOverclocking) and Skatterbencher on YouTube to get an idea of what is necessary to tune your RAM. The advertised speeds are XMP or EXPO profiles, meaning that those speeds are what the modules could run at if your other hardware supports it. There are safety factors built in as not all hardware is equal, and could possibly see performance gains.

It is mainly for the tinkerers, as stability issues could ensue. You cant prove stability, only instability and would need to leave your pc while stress testing to ensure a stable overclock. Lots of effort for incremental gains above EXPO profiles.

I would advise first to educate yourself on the process and then to decide if you want to try your hand at RAM tuning.

And then go for it.

If you get stuck, ask chatGPT by sending all relevant timings and the error you observed. It will point you in the direct direction. Think critically about what it proposes, but it has helped me sort out a few issues after I tuned my RAM on another's guide.

Will increase PC performance, but be it however small, its fun to get it right and have something work that you tuned yourself.

 

[Couch_Verstappen]

@ThatGuyJD Know's RAM things.

 

[Geriatrix]

@D-wired and @Couch_Verstappen and the others, ya, no, I am not into tinkering THAT much. I was basically asking because with RAM prices the way are going, I don't think I would be able to afford DDR4 3600MT/s or 4000MT/s any time soon, so was just wondering if pushing my existing RAM would give any benefits and would be stable.

 

[NastyA55]

@D-wired and @Couch_Verstappen and the others, ya, no, I am not into tinkering THAT much. I was basically asking because with RAM prices the way are going, I don't think I would be able to afford DDR4 3600MT/s or 4000MT/s any time soon, so was just wondering if pushing my existing RAM would give any benefits and would be stable.

Pops, I like your reasoning.

I am in the boat that hopefully the DDR4 and DDR5 that I have will be sufficient for now... my only hang up might be the DDR4 in a bit.
Then I am kind of stuffed and hope that I can pick up a decently priced used kit on here.

 

[Geriatrix]

Pops, I like your reasoning.

I am in the boat that hopefully the DDR4 and DDR5 that I have will be sufficient for now... my only hang up might be the DDR4 in a bit.
Then I am kind of stuffed and hope that I can pick up a decently priced used kit on here.

LOL, thank you! This CUD is quite a beast, you upgrade one thing, then you think "oh, wait, now I will "need" need to upgrade the other thing", because speed difference or something and before you know it, you have a completely new system!! It's addictive, I tell you!!

 

[Spaffy]

@D-wired and @Couch_Verstappen and the others, ya, no, I am not into tinkering THAT much. I was basically asking because with RAM prices the way are going, I don't think I would be able to afford DDR4 3600MT/s or 4000MT/s any time soon, so was just wondering if pushing my existing RAM would give any benefits and would be stable.

The difference would be negligible anyways. Just make sure you keep that beast cpu cool and hopefully you have a decent gpu to play whatever it is you play.

 

[Geriatrix]

The difference would be negligible anyways. Just make sure you keep that beast cpu cool and hopefully you have a decent gpu to play whatever it is you play.

Ya, CPU is cooled by Scythe Mugen 6 (thanks Progenix!!) and I have an XFX SPEEDSTER SWFT 309 RX 6700XT which is an amzing GPU. Thank you @JustJean !!

 

[ThatGuyJD]

Yeah without dropping you into the deep end that is buildzoid and "glhf"

 

[Geriatrix]

Yeah without dropping you into the deep end that is buildzoid and "glhf"

OK, @ThatGuyJD, having now watched this, thank you, what do I now do? Do I look for the sub-timings SPECIFIC to my make of RAM, to the make of my motherboard, or to both, or just general "timing settings for B450 chipset with 32GB RAM"?

 

[ThatGuyJD]

OK, @ThatGuyJD, having now watched this, thank you, what do I now do? Do I look for the sub-timings SPECIFIC to my make of RAM, to the make of my motherboard, or to both, or just general "timing settings for B450 chipset with 32GB RAM"?

So you can totally do it sight unseen and with your XMP, it could literally be any DDR4 produced in the last couple years on that stick. So it's going to be more trial and error than normally. And obviously it's OC so your mileage may vary.

So what you'd do is leave it at XMP and then work your way up to 3400 and then 3600 on speed and see if it posts. When that is done, you trail and error each individual timing and then test to see if it's kinda stable... so change, boot into windows, run your selected test, if it passes, try even tighter, if it fails go back to what it was before and move onto the next timing. Starting with tRRD_S; tRRD_L & tFAW then tRFC and can try a lot of the other ones after that if you're so inclinced.

Yeah, it's not a a quickly with the wife before the kids wake up on a weekend.

 

[Geriatrix]

So you can totally do it sight unseen and with your XMP, it could literally be any DDR4 produced in the last couple years on that stick. So it's going to be more trial and error than normally. And obviously it's OC so your mileage may vary.

So what you'd do is leave it at XMP and then work your way up to 3400 and then 3600 on speed and see if it posts. When that is done, you trail and error each individual timing and then test to see if it's kinda stable... so change, boot into windows, run your selected test, if it passes, try even tighter, if it fails go back to what it was before and move onto the next timing. Starting with tRRD_S; tRRD_L & tFAW then tRFC and can try a lot of the other ones after that if you're so inclinced.

Yeah, it's not a a quickly with the wife before the kids wake up on a weekend.

Ok, I am not going to lie, this is not something I would contemplate or attempt. I have NEVER done overclocking in my life and at my age, simply put, I am too scared of doing something wrong. So thanks, however, I think I will just leave well enough alone. "if it ain't broke, don't fix it".

 

[ThatGuyJD]

Ok, I am not going to lie, this is not something I would contemplate or attempt. I have NEVER done overclocking in my life and at my age, simply put, I am too scared of doing something wrong. So thanks, however, I think I will just leave well enough alone. "if it ain't broke, don't fix it".

No judgement.

 

[D-wired]

Ok, I am not going to lie, this is not something I would contemplate or attempt. I have NEVER done overclocking in my life and at my age, simply put, I am too scared of doing something wrong. So thanks, however, I think I will just leave well enough alone. "if it ain't broke, don't fix it".

MythBusters proved that you can teach an old dog new tricks

 

[Geriatrix]

MythBusters proved that you can teach an old dog new tricks

Ya, I agree, however, in this case, this old dog is too scared to learn these new tricks!!

 

[D-wired]

OK, @ThatGuyJD, having now watched this, thank you, what do I now do? Do I look for the sub-timings SPECIFIC to my make of RAM, to the make of my motherboard, or to both, or just general "timing settings for B450 chipset with 32GB RAM"?

**NB** Just confirm the downloaded files are virus free (by scanning with an anti-virus or uploading the downloaded file to virustotal.com as that site uses something like 74 anti-virus scans on the file) 1 or 2 false positives is ok, but more than that and I wouldn't run the downloaded file on my pc **NB**


I would start by downloading HWiNFO 64, a hardware monitoring program to see temps, volts, speed and even your current RAM timings. Watch a few videos on the program, really just a reporting program - lots of numbers but really interesting to see what your computer is doing (or is set up as). Can see your RAM timings, BIOS Version, Motherboard Name, Graphics card specs, CPU specs, features, OS, drives etc. Can click on pictures to even get more info. Really cool tool for monitoring PC.

Then download TM5 (memory stability test program) watch a few videos on how it works, then let it run on your current PC just the way it is (takes 2 hours for 32GB on the extreme profile - which is what you want). And see if your current EXPO profile is stable - which it should be.

Then download Linpack Xtreme (another RAM test) can benchmark or stress test RAM to also see if stable. Then you can see a GFlops number for your RAM - this is important as now you can see from the video posted by @ThatGuyJD in the graph, where your set up lands and how it compares. If you get less than 280 GFlops I would suggest spending time to tune timings as you should be able to get close to 350 GFlops or more on tuned timings. And get closer to 330 GFlops if only tuning tRRD_S, tRRD_L and then tFAW.
I am assuming based on the video that you also have a 5800X, but I see now that you have a 3900X. So ignore the numbers I specified, just know that with simple tuning you can get an increase of about 15% by just tuning 3 parameters over EXPO profile. That would be a noticeable difference (anything over 10% is a noticeable difference). And even up to a 25% increase in GFlops if properly tuning the rest of the timings. So it is possible to increase performance by just changing numbers ;-)

So see where you currently are with the linpack benchmark and then google linpack results for your cpu and you will then see the possible improvements that can be achieved.

That is just the start. If you then feel comfortable changing timings in the BIOS, just start with tuning tRRD_S, tRRD_L and then tFAW.
Watch more videos of DDR4 Tuning to get the idea of what is happening and to get comfortable with what you are doing and not to just randomly enter numbers (well you could be lucky - you never know). It's learning, and learning is repetition (by watching more videos on the subject) and then you have to try your own hand at it - for practical understanding of the subject matter.

If you have time, definitely consider it, as problem solving keeps the brain young.

And just to end this off with a quote: You always miss 100% of the shots you don't take.

 

[D-wired]

**NB** Just confirm the downloaded files are virus free (by scanning with an anti-virus or uploading the downloaded file to virustotal.com as that site uses something like 74 anti-virus scans on the file) 1 or 2 false positives is ok, but more than that and I wouldn't run the downloaded file on my pc **NB**


I would start by downloading HWiNFO 64, a hardware monitoring program to see temps, volts, speed and even your current RAM timings. Watch a few videos on the program, really just a reporting program - lots of numbers but really interesting to see what your computer is doing (or is set up as). Can see your RAM timings, BIOS Version, Motherboard Name, Graphics card specs, CPU specs, features, OS, drives etc. Can click on pictures to even get more info. Really cool tool for monitoring PC.

Then download TM5 (memory stability test program) watch a few videos on how it works, then let it run on your current PC just the way it is (takes 2 hours for 32GB on the extreme profile - which is what you want). And see if your current EXPO profile is stable - which it should be.

Then download Linpack Xtreme (another RAM test) can benchmark or stress test RAM to also see if stable. Then you can see a GFlops number for your RAM - this is important as now you can see from the video posted by @ThatGuyJD in the graph, where your set up lands and how it compares. If you get less than 280 GFlops I would suggest spending time to tune timings as you should be able to get close to 350 GFlops or more on tuned timings. And get closer to 330 GFlops if only tuning tRRD_S, tRRD_L and then tFAW.
I am assuming based on the video that you also have a 5800X, but I see now that you have a 3900X. So ignore the numbers I specified, just know that with simple tuning you can get an increase of about 15% by just tuning 3 parameters over EXPO profile. That would be a noticeable difference (anything over 10% is a noticeable difference). And even up to a 25% increase in GFlops if properly tuning the rest of the timings. So it is possible to increase performance by just changing numbers ;-)

So see where you currently are with the linpack benchmark and then google linpack results for your cpu and you will then see the possible improvements that can be achieved.

That is just the start. If you then feel comfortable changing timings in the BIOS, just start with tuning tRRD_S, tRRD_L and then tFAW.
Watch more videos of DDR4 Tuning to get the idea of what is happening and to get comfortable with what you are doing and not to just randomly enter numbers (well you could be lucky - you never know). It's learning, and learning is repetition (by watching more videos on the subject) and then you have to try your own hand at it - for practical understanding of the subject matter.

If you have time, definitely consider it, as problem solving keeps the brain young.

And just to end this off with a quote: You always miss 100% of the shots you don't take.

Just a note all the programs are free to download, just to benchmark your current performance before even considering tuning.

 

[Geriatrix]

**NB** Just confirm the downloaded files are virus free (by scanning with an anti-virus or uploading the downloaded file to virustotal.com as that site uses something like 74 anti-virus scans on the file) 1 or 2 false positives is ok, but more than that and I wouldn't run the downloaded file on my pc **NB**


I would start by downloading HWiNFO 64, a hardware monitoring program to see temps, volts, speed and even your current RAM timings. Watch a few videos on the program, really just a reporting program - lots of numbers but really interesting to see what your computer is doing (or is set up as). Can see your RAM timings, BIOS Version, Motherboard Name, Graphics card specs, CPU specs, features, OS, drives etc. Can click on pictures to even get more info. Really cool tool for monitoring PC.

Then download TM5 (memory stability test program) watch a few videos on how it works, then let it run on your current PC just the way it is (takes 2 hours for 32GB on the extreme profile - which is what you want). And see if your current EXPO profile is stable - which it should be.

Then download Linpack Xtreme (another RAM test) can benchmark or stress test RAM to also see if stable. Then you can see a GFlops number for your RAM - this is important as now you can see from the video posted by @ThatGuyJD in the graph, where your set up lands and how it compares. If you get less than 280 GFlops I would suggest spending time to tune timings as you should be able to get close to 350 GFlops or more on tuned timings. And get closer to 330 GFlops if only tuning tRRD_S, tRRD_L and then tFAW.
I am assuming based on the video that you also have a 5800X, but I see now that you have a 3900X. So ignore the numbers I specified, just know that with simple tuning you can get an increase of about 15% by just tuning 3 parameters over EXPO profile. That would be a noticeable difference (anything over 10% is a noticeable difference). And even up to a 25% increase in GFlops if properly tuning the rest of the timings. So it is possible to increase performance by just changing numbers ;-)

So see where you currently are with the linpack benchmark and then google linpack results for your cpu and you will then see the possible improvements that can be achieved.

That is just the start. If you then feel comfortable changing timings in the BIOS, just start with tuning tRRD_S, tRRD_L and then tFAW.
Watch more videos of DDR4 Tuning to get the idea of what is happening and to get comfortable with what you are doing and not to just randomly enter numbers (well you could be lucky - you never know). It's learning, and learning is repetition (by watching more videos on the subject) and then you have to try your own hand at it - for practical understanding of the subject matter.

If you have time, definitely consider it, as problem solving keeps the brain young.

And just to end this off with a quote: You always miss 100% of the shots you don't take.

Thanks for this. I already have HWINFO64. I will look at the benchmark software you mentioned and I MAY give your suggestions a try. You must understand, I do not even understand or know what those "timing values" like RRD_S, tRRD_L and then tFAW even are, or what they do. The thing is, I prefer to know WHY I am changing something, not just do something parrot fashion. For instance, I looked up a LOT of info on the PBO settings for my CPU, before I even attempted to enable and customise those settings.

 

[D-wired]

Thanks for this. I already have HWINFO64. I will look at the benchmark software you mentioned and I MAY give your suggestions a try. You must understand, I do not even understand or know what those "timing values" like RRD_S, tRRD_L and then tFAW even are, or what they do. The thing is, I prefer to know WHY I am changing something, not just do something parrot fashion. For instance, I looked up a LOT of info on the PBO settings for my CPU, before I even attempted to enable and customise those settings.

Every expert was once a beginner too.

I am also still learning and trying to get better performance out of my build. If it didn't matter to you, you wouldn't have asked.

I google'd your question and this is what it said:

tRRD_S, tRRD_L, and tFAW are all DDR4 RAM timing parameters that control the speed of memory operations by setting minimum delays between ACTIVATE commands. tRRD_S (Activate to Activate Delay - Short) is the minimum time between activating different bank groups, while tRRD_L (Activate to Activate Delay - Long) is the minimum time between activating the same bank group. tFAW (Four Activate Window) is a time window during which only four ACTIVATE commands can be issued.

Not really helpful, better to watch a video on the topic:

Go to 10:32 timestamp where he starts talking about those 3.
Rather watch from the start. And even if you have more questions than answers, go to the beginning of the series of videos he talked about to completely understand DDR4 Timings

 

[Geriatrix]

Every expert was once a beginner too.

I am also still learning and trying to get better performance out of my build. If it didn't matter to you, you wouldn't have asked.

I google'd your question and this is what it said:

tRRD_S, tRRD_L, and tFAW are all DDR4 RAM timing parameters that control the speed of memory operations by setting minimum delays between ACTIVATE commands. tRRD_S (Activate to Activate Delay - Short) is the minimum time between activating different bank groups, while tRRD_L (Activate to Activate Delay - Long) is the minimum time between activating the same bank group. tFAW (Four Activate Window) is a time window during which only four ACTIVATE commands can be issued.

Not really helpful, better to watch a video on the topic:

Go to 10:32 timestamp where he starts talking about those 3.

Oh, wow, thank you, you did not have to do that, I would have done that later today. That is really kind of you. I will definitely take a look at the video. Much appreciated.

 

[D-wired]

Pleasure

 

[D-wired]

TLDW (Too Long Didnt Watch)

Set tFAW to 16
Then either start lowering tRRD_S 1 at a time in BIOS till you get to 4. Each time, you should run a linpack benchmark and check if it passes the 5 tests. If it passes the test, go back to BIOS and lower it again by 1. That's the scientific method - changing 1 variable and observing the outcome. When it doesnt pass, go back to BIOS and add 1 to the test where it did pass.

Then do the same thing for tRRD_L

If you want to save time, set (Or this is the goal you are working towards):
tFAW = 16
tRRD_S = 4
tRRD_L = 4

And check for stability. In the video example, he showed that this should be possible at 3600MHz, so first try this at your current 3200MHz - should be much easier on memory controller. But I start first by tuning primary timings - meaning for now just leave it at 3200MHz and see what improvements you get. Then you can go increase MHz - but will need to adjust primary timings to ensure the chip can handle the latencies. Will do another post later on the relationship between primary timings and MHz if you dont understand the first video I posted.

If your system is unstable (random BSODs or applications closing - check event viewer for errors - Google those. Should be related to the RAM timings you just changed.

But if your system can't run at those speeds, either you dont have the best chip die for DDR4, and would need to watch the first video in the series again.

If you dont have the specific chip die he is talking about, just Google typical tFAW values for your chip for example - should get a ballpark number to aim for.

Most important is to keep track of your changes. That's why you dont change 2 values at a time, as you won't know which one caused the instability.