Overclock.net banner

4781 - 4797 of 4797 Posts

·
Premium Member
Joined
·
5,017 Posts
I figured out if I set the below timings with SCL's at 4, reboot, enable Memory Fast Boot, reboot, set SCL's to 2, boot into Windows, I get a decent increase in AIDA64 bandwidth and TM5 passes which it doesn't if I don't enable Memory Fast Boot.

bbiab, see if I can do am hour of GSAT was well.
 

·
Premium Member
Joined
·
5,017 Posts
I figured out if I set the below timings with SCL's at 4, reboot, enable Memory Fast Boot, reboot, set SCL's to 2, boot into Windows, I get a decent increase in AIDA64 bandwidth and TM5 passes which it doesn't if I don't enable Memory Fast Boot.

bbiab, see if I can do am hour of GSAT was well.
GSAT passed, TM5 Anta Extreme passed.

2459070
 

·
Registered
Joined
·
882 Posts
I figured out if I set the below timings with SCL's at 4, reboot, enable Memory Fast Boot, reboot, set SCL's to 2, boot into Windows, I get a decent increase in AIDA64 bandwidth and TM5 passes which it doesn't if I don't enable Memory Fast Boot.

bbiab, see if I can do am hour of GSAT was well.
Glad to see other people are experimenting and seeing the "weirdness" with AIDA score dependent on enabling "Memory Fast Boot" and rebooting twice etc.

I wanted to take my investigations further so tried the following,

Disable read/write memory training, disable "Memory Fast Boot" and try to get something stable higher than 3533 mhz.

On my X370 it was not happening, pure carnage, was impossible to get anything stable.

As has been explained by Veii, there are many timings we do not have access to and without the read/write memory training we cant reach the frequency/timing combinations we are currently rocking.

However with the read/write memory training and "Memory Fast Boot" there is stuff going on that is not always producing the same results and we are never going to find out why .....

Would be interested to hear if others on X470/X570/B550 platforms experience the same issue when attempting to ramp up frequency/timings with both read/write memory training and "Memory Fast Boot" disabled.

:)
 

·
Registered
Joined
·
2,200 Posts
Glad to see other people are experimenting and seeing the "weirdness" with AIDA score dependent on enabling "Memory Fast Boot" and rebooting twice etc.

I wanted to take my investigations further so tried the following,

Disable read/write memory training, disable "Memory Fast Boot" and try to get something stable higher than 3533 mhz.

On my X370 it was not happening, pure carnage, was impossible to get anything stable.

As has been explained by Veii, there are many timings we do not have access to and without the read/write memory training we cant reach the frequency/timing combinations we are currently rocking.

However with the read/write memory training and "Memory Fast Boot" there is stuff going on that is not always producing the same results and we are never going to find out why .....

Would be interested to hear if others on X470/X570/B550 platforms experience the same issue when attempting to ramp up frequency/timings with both read/write memory training and "Memory Fast Boot" disabled.

:)
Of I definitely see better Aida scores after a reboot.
 

·
Registered
Joined
·
211 Posts
3800 CL16 with B-die subtimings is already better than everything else on the market.

Which kit do you own? Could you post a screenshot of the latest ZenTimings?
It crashed at round 20 so not stable yet, my entered timings is exactly the same ones as the ram calculator excel sheet gave me for 3800CL16, so not really sure why I'm unstable at 1.48v, higher voltages like 1.5 reduces stability.

EDIT: just swapped 2 of my sticks with my other PC, the same timings are now way more stable atm at tm5 cycle 25 of 30, possibly because all 4 sticks are now 3200CL14 instead of 2x 3200CL14, 2x 3600CL16.
 

·
Premium Member
Joined
·
5,017 Posts
Of I definitely see better Aida scores after a reboot.
I turn Memory Fast Boot off, set all my timings with SCLs at 4, reboot, it'll train my timings at what I set it at, then set Memory Fast Boot on, SCL's at 2, reboot again.

It'll keep the training at what I had it at, only with SCL's changed to 2. If I don't do it that way, both TM5 and GSAT will get errors.

And I still get decent bandwidth increase with having the SCL's at 2. :D
 

·
Premium Member
Joined
·
5,017 Posts
Hey peeps, I get much better timings and memory overclock with tRP and tRAS set at 19-21 respectively rather than the traditional settings like DRAM Calculator suggests.

I tried it on @Veii suggestions and improved my memory overclock and AIDA64 bandwidth quite a bit. Much easier to get stable with tighter timings.

2459136
 

·
Registered
Joined
·
17 Posts
I tried CAS 14 but my latency went higher than it is at CAS 16. Can someone tell me if this is good PCB ram for overclocking. I can't get near 62 ns like some.
2459145


2459146
2459147


If anyone see something that would help please tell me.
 

·
Registered
Joined
·
882 Posts
I tried CAS 14 but my latency went higher than it is at CAS 16. Can someone tell me if this is good PCB ram for overclocking. I can't get near 62 ns like some.

If anyone see something that would help please tell me.
You will not get near 62ns regularily unless you are running an all core overclock.

From your screenshots it looks like you are using PBO.

Dont fret about the couple of ns, it makes no difference .........
 

·
Registered
Joined
·
882 Posts
Hey peeps, I get much better timings and memory overclock with tRP and tRAS set at 19-21 respectively rather than the traditional settings like DRAM Calculator suggests.

I tried it on @Veii suggestions and improved my memory overclock and AIDA64 bandwidth quite a bit. Much easier to get stable with tighter timings.
Using those timings makes no difference for my setup with regards to AIDA64 bandwidth.

This could be because I am running 1T and only 16GB of RAM with a 3600.....

The only difference i saw was that the memory latency results were not consistent.

I.e. on my setup latency is usually between 62.5 to 62.9

However using the funky timings I have seen it jumping to as low as 61.8 which is unheard of for my setup and up to 63.4 while using [email protected], [email protected]

Me thinks AIDA is a benchmark that we should not rely on too much with regards to the results it throws out .
 

·
Registered
Joined
·
17 Posts
Are you capable of going faster because you run 16GB of RAM? I will never be able to run that fast. Fastest I've seen it get only 1 time was 63.4. Not really worried just like playing around with it, but I know the results don't really change much even if all the timings are lower.
 

·
Registered
Joined
·
17 Posts
I have disabled PBO, if I did an all core overclock does the voltage always stay high or can it scale up and down like it does currently?
 

·
Registered
Joined
·
882 Posts
I have disabled PBO, if I did an all core overclock does the voltage always stay high or can it scale up and down like it does currently?
You must sit and read a few of the countless threads that are at this forum and across the Internet with regards to your question.

I cannot give you a nice and simple answer for your question, im sure somebody else may do, just dont understand why some people dont read about their hardware first before asking elementary questions.

Because your questions depends on many factors .

Please do yourself a favour and read the basics

 

·
Registered
Joined
·
211 Posts
Managed to get this stable with 30 passes of tm5 V3 profile, anything i can do to tighten timings further without compromising stability, would 3800Cl14 2T be better than these timings on 1T if i manage to make it boot/stable?

Ram voltage is already at 1.48v but i don't think i can get it down to 3800Cl14 2T as i'm running 4xSR, maybe with 2xDR it would be possible...
 

Attachments

·
Computer Enthusiast
Joined
·
2,855 Posts
Still experimenting with bios 3004. Cannot do over 3733 on it whatever I try. Higher voltage, loser timings etc.

I did some tests with timings at 3733, and though I'm getting tighter, it doesn't seem to make much difference. I have got latency down to 61.x ns, but not able to get better throughput. 4266 and 4400 XMP profile just won't work any more.

This is after TM5 run while general stuff was running - which passed even though there were 2 Windows hardware bus errors. Any obvious problems anyone can see here?

3733fastExperiment.png
 

·
....
Joined
·
225 Posts
Anyone tried AGESA ComboV2 1.0.8.1?
 
4781 - 4797 of 4797 Posts
Top