[SOLVED] Will tuning my RAM from CL16 to CL14 make a difference?

[MxzsyXII]

DRAM Calculator Result
I have Corsair Vengeance RGB Pro 3200MHz CL16 DDR4 ( CMW16GX4M2C3200C16). I have been very bored during this lockdown so I thought I can tune my pc a little and get the most for my money.

I used thaiphoon burner to check the ram die, and it says its samsung B die, but im pretty sure its not a well binned one. which is why i dont want to push more than the 3200mhz rating, unless someone here thinks its very possible to go higher. but i thought lowering the latency might be worth it for my Ryzen 5 2600.

Will doing this be worth it? is it safe? do you think it will be stable?

also, is this a cl14 tune? Because I see 14 at the top and just assumed, sorry lol

I plan to game and stream, and some editing on premiere pro occasionally

 

[senseipanda123]

DRAM Calculator Result
I have Corsair Vengeance RGB Pro 3200MHz CL16 DDR4 ( CMW16GX4M2C3200C16). I have been very bored during this lockdown so I thought I can tune my pc a little and get the most for my money.

The difference between CL16 and 14 is 1.14285714. Multiply your clock speed by that factor to get 3657.14285. This suggests that 3200 CL14 will perform similar to 3600 CL16. Yet again, this is just a ballpark and different modules will perform differently depending on CPU+MB combo.

 

[MxzsyXII]

The difference between CL16 and 14 is 1.14285714. Multiply your clock speed by that factor to get 3657.14285. This suggests that 3200 CL14 will perform similar to 3600 CL16. Yet again, this is just a ballpark and different modules will perform differently depending on CPU+MB combo.

I have a Ryzen 5 2600 on B450 Tomahawk MAX. Would changing just timings likely be stable?

 

[kingbowcat]

I have a Ryzen 5 2600 on B450 Tomahawk MAX. Would changing just timings likely be stable?

I got a 2600x and Tomahawk max and tuned my 2400 ram to 3466 speed from 18 cl to 20cl. Just experiment with it and get as tight timings as you can. If it doesn't work it will default to the old speed and try again. The boost also helped out my ryzen quite a bit. showing like an 8% increase in performance userbenchmarks.com. And the ram showed the most improved ofc up to 116% from the average benchmark of 76% from the same ram.

 

[Deleted member 14196]

Benchmarks mean little or nothing at all regarding real world performance. It looks to me like you sre asking for trouble. If you want faster ram buy faster ram kit that’s compatible with your machine

 

[kingbowcat]

Well I noticed a big difference

 

[drea.drechsler]

The difference between CL16 and 14 is 1.14285714. Multiply your clock speed by that factor to get 3657.14285. This suggests that 3200 CL14 will perform similar to 3600 CL16.
...

That seems correct when talking about latency. But you should still get the benefit of the faster data transfer speed 3600 Mtps provides vs. 3200 Mtps.

That's why it's both wise (for stability) and not harmful (for performance) to go ahead and bump up the timings when pushing memory clocks. I think many people don't, it fails, and they give up.

 

[drea.drechsler]

Well I noticed a big difference

You probably would...but I'd still suggest doing some memory tests before declaring "all good". If nothing else, type Windows Memory Diagnostics in Windows search and run that. You can also download MemTest. Run any test through all the memory at least twice.

 

[kingbowcat]

Yeah you have to do that, memtest is very good I let mine run until it finished to be sure.

 

[zx128k]

Ryzen DRAM Calculator 1.7.0 builtin test at 400-600% works well. I run both, memtest 4 passes boot from USB or CD and then Ryzen DRAM Calculator 1.7.0 builtin test at 400%. Take approx. 4 hours. Then I run AIDA64 for an hour or two to make sure the IF etc is okay because I am on Ryzen 3800x. With the EDC 1 bug I hit 11700 time spy cpu with this RAM overclock. https://www.3dmark.com/3dm/45053767

 

[drea.drechsler]

Ryzen DRAM Calculator 1.7.0 builtin test at 400-600% works well. I run both, memtest 4 passes boot from USB or CD and then Ryzen DRAM Calculator 1.7.0 builtin test at 400%. Take approx. 4 hours. Then I run AIDA64 for an hour or two to make sure the IF etc is okay because I am on Ryzen 3800x. With the EDC 1 bug I hit 11700 time spy cpu with this RAM overclock. https://www.3dmark.com/3dm/45053767

That's the way to 'overclock' Ryzen systems! It's so safe, in my opinion, I think of it as 'optimizing', not really overclocking. But whatever you're doing, if it's stable in appropriate tests then you got it!

 

[zx128k]

That's the way to 'overclock' Ryzen systems! It's so safe, in my opinion, I think of it as 'optimizing', not really overclocking. But whatever you're doing, if it's stable in appropriate tests then you got it!

It's most likely a risk but I have the money saved for a replacement cpu if needed.

 

[drea.drechsler]

It's most likely a risk but I have the money saved for a replacement cpu if needed.

Not a risk at all...first, has anyone ever damaged a CPU by overclocking RAM?

Second...if you're exploiting the 'EDC 1 bug' that must mean you're using PBO, right? That's not really overclocking the processor since it's still observing the safe clock/voltage/temperature parameters fused in the processor's FIT tables at AMD if done right. All PBO does (even considering the bug exploit) is overrule the power and current limitations intended to protect motherboards with weak VRM's.

And at least in my case after I'd tweaked my BIOS PBO parameters to the 'EDC 1' bug, not only is it hitting max boost clocks more frequently on more cores and scoring over 100pts higher on Cinebench MT scores, but it's doing it with lower voltage. A win-win if I ever saw one!

 

[zx128k]

I would guess you are refering too this source,

You can safely maximise any or all of these.

Even when they are maxed out, there are other limiters still present (FIT), which prevent the chip from using voltages that are unsafe to the silicon.

You can use TDC and EDC to limit the frequencies in 256-bit workloads (AVX, AVX2, FMA) if you see the need to do so.
Scalar workloads generally are not as much current limited.

This applies to stock (or PBO) behavior, not to manual OC.

https://www.overclock.net/forum/10-...g-sizing-limits-ppt-tdc-edc.html#post28106432

It's still breaking the fused value for EDC. Also you hit 180 watts very quickly. EDC 1 appears to remove the auto detection of high current workload. So much higher temps.