[kazablanka]

Although I can run nice and tight at 3733 CL14 IF @ 1866 I get quite nice results....

If I try to run 3800/1900 only way it boots so far is to let bios apply awful auto timings like 22 26 26 26.

Also, I can select DOCP 4266, which is meant to be 19 19 19 19, it boots, but with same awful timings, not the XMP timings. Currently trying to get it to boot with 20 20 20 20, but no luck yet. Trying various voltages and timings but just don't know what I'm doing yet.

Tried CL15, GearDown off, tried T2 as well, but not sure how T2 effects things like voltage on this chip.

Do others have the bios enter slower timings than XMP for them?

Think I can probably refine my 3733 with help from the last few pages.

Thanks for extra information @The Stilt - always helpful!

Αre you keeping the dram boot voltage same as the dram voltage every time?

 

[kazablanka]

All the R5 3600/R7 2700X data in CB is from C7H, the other Ryzen CPUs are from C6H.

PBO+200MHz 1669
PBO+150MHz 1665/1663
PBO+75MHz 1646
Stock is 1632

All R5 3600 results using RAM 3600MHz, 1:1:1, The Stilt's 3466MHz timings, 1T GDMD.

Spoiler

View attachment 282496

Upto PBO+100MHz I use 1x, PBO+150MHz just will not sustain MHz in Kahru RAM Test and bench in CB R15 with scaling if I do not use at least 4x, so I set 6x for over head in my mind, PBO+200MHz 10x was used, not tested it much. This post has some data as well.

Thanks I will check it.

 

[crakej]

Αre you keeping the dram boot voltage same as the dram voltage every time?

As far as I know, yes..... is this particularly sensitive with Matisse?

 

[gupsterg]

Αre you keeping the dram boot voltage same as the dram voltage every time?

I have always matched VBOOT & VDIMM on C6H/C7H/ZE/ZEA.

Thanks I will check it.

NP.

I found on UEFI 0068 that scalar change only applied via AMD menu in UEFI. So if the MSI UEFI has it in a differing section perhaps that does not work.

 

[eXteR]

And before you ask. No.
Matisse version of RTC will never be public, sorry

Why?

It's a copyright problem or something related to AMD?

 

[gupsterg]

Although I can run nice and tight at 3733 CL14 IF @ 1866 I get quite nice results....

If I try to run 3800/1900 only way it boots so far is to let bios apply awful auto timings like 22 26 26 26.

Also, I can select DOCP 4266, which is meant to be 19 19 19 19, it boots, but with same awful timings, not the XMP timings. Currently trying to get it to boot with 20 20 20 20, but no luck yet. Trying various voltages and timings but just don't know what I'm doing yet.

Tried CL15, GearDown off, tried T2 as well, but not sure how T2 effects things like voltage on this chip.

Do others have the bios enter slower timings than XMP for them?

Think I can probably refine my 3733 with help from the last few pages.

Thanks for extra information @The Stilt - always helpful!

This is where I am 3800v4.2, by about v4 I've got what I'd get I think unless hit primaries more, perhaps go 1T. Repeatable reruns, best so far ~4000% in RT.

 

[crakej]

This is where I am 3800v4.2, by about v4 I've got what I'd get I think unless hit primaries more, perhaps go 1T. Repeatable reruns, best so far ~4000% in RT.

Thanks for sharing those! Maybe I'll spot something i'm not doing..... rep+

Edit: @gupsterg what was your ram voltage for 3733?

 

[The Stilt]

Why?

It's a copyright problem or something related to AMD?

For multiple different reasons.

The biggest reason being that AMD doesn't give out any information, which is required to support their recent platforms.
In addition to that, the methods which are required to actually access the hardware are simply sick. Three different hardware access methods and two software "APIs" are required to access the information.
That's slow, expensive and unrealiable. A total hackjob, despite the fact that AMD themselves utilize it as well...

Due to the mush of all of these different methods required, there is no tolerance for errors or changes either.
A single change made to AGESA can break the existing methods. Naturally there is also the issue with AMD branching AGESA so heavily.
Because of the differences in the firmwares and physical configurations, separate workarounds / methods are required for each of the branches, CPU generations and sometimes even different SMU FW versions.
It is impossible to maintain, when you need at least 5-6 different code paths, due to the total lack of any uniformity from AMD side.

So in short; It would be insane to keep fighting the windmills.

When the tools stay in my own use, as originally intended, I don't need to verify the functionality on n+1 different platforms, and go through the hassle of getting the binaries EV signed each and every time.
Also keeping the tools to myself will prevent the work from getting stolen, which is an additional perk

 

[eXteR]

For multiple different reasons.

The biggest reason being that AMD doesn't give out any information, which is required to support their recent platforms.
In addition to that, the methods which are required to actually access the hardware are simply sick. Three different hardware access methods and two software "APIs" are required to access the information.
That's slow, expensive and unrealiable. A total hackjob, despite the fact that AMD themselves utilize it as well...

Due to the mush of all of these different methods required, there is no tolerance for errors or changes either.
A single change made to AGESA can break the existing methods. Naturally there is also the issue with AMD branching AGESA so heavily.
Because of the differences in the firmwares and physical configurations, separate workarounds / methods are required for each of the branches, CPU generations and sometimes even different SMU FW versions.
It is impossible to maintain, when you need at least 5-6 different code paths, due to the total lack of any uniformity from AMD side.

So in short; It would be insane to keep fighting the windmills.

When the tools stay in my own use, as originally intended, I don't need to verify the functionality on n+1 different platforms, and go through the hassle of getting the binaries EV signed each and every time.
Also keeping the tools to myself will prevent the work from getting stolen, which is an additional perk

Totally understandable.

Thanks for the detailed explanation.

You deserve a beer

 

[Nighthog]

For multiple different reasons.
snip*

Doesn't AMD Ryzen Master now give us all this information anyway this time around with the timings, CAD BUS values etc? We didn't have that before.
I do hope they don't break their own software in the future that will eventually happen but it's not as critical as before when your tool was the only way for us to present the info in a manageable manner.

Your work was really helpful, thanks for the work put in.

It is another matter with the awful design and manner which the AMD version presents the info but it's there nonetheless.

 

[majestynl]

This is where I am 3800v4.2, by about v4 I've got what I'd get I think unless hit primaries more, perhaps go 1T. Repeatable reruns, best so far ~4000% in RT.

Downloaded the files. Nice m8. Good luck pushing forward. Maybe you can try CL14?

Im pretty done with 3800 profile. Tighten the timings down. Cant see major improvements vs only base timings + TRFC! Just slightly better.
Below screenshot while running. That particular run passed 8K before i stopped!

@The Stilt
TRFC2 and TRFC4 are looking high. Is there any improvement if we "can" get them lower?

 

[gupsterg]

@kazablanka

Today I tested with lower room ambient. The PBO+200MHz result I showed was thermally limited. So are most of my PBO+150MHz, as here is 1676 in CB R15 .

Spoiler

Thanks for sharing those! Maybe I'll spot something i'm not doing..... rep+

Edit: @gupsterg what was your ram voltage for 3733?

Not really tried 3733MHz much TBH. I fought a few days with gaining 3666MHz using The Stilt 3466MHz timings (1T GDMD), always just before ~3000% or after I'd get an issue. I tried lots of things. Then I spent nearly I day testing 3600MHz with higher % pass in tests. Then I tried say all [Auto] timings just to validate could I get 3800MHz. From there on profile fell in to place as I stopped targeting SCL 2. I believe I maybe able to go back to 1T GDMD on 3800MHz, even if I can't seems from benches I did tonight I'm not losing anything, but gaining stability.

TBH even the SOC/VDDG/VDIMM I use currently are some ball park figures which I felt would work from all the meddling I have done with this R5 3600 since 9th July.

I too like Nighthog had seen on certain targeted setting using too much voltage (may that be SOC/VDDG/VDIMM or all) made a profile fail then using more conservative approach.

For example VDDG: ~1.005V had worked well for 3666MHz any more used kill the profile quicker, dunno why. Now 3800MHz is ~1.012V, which is nuts when I look at it. Now from how I can use SOC: 1.031 VDDG: 0.956V for 3600MHz with SCL2, etc, perhaps I can tweak down them for 3800MHz.

Time will tell .

Shocking how just ~2C lower room ambient got be nice scaling in CB R15 at same PBO+150MHz, also look at this WMV of RAM test. Last 6 screenies of this album are PBO+150MHz, 3800MHz v4.2, 1:1:1.

After so much time spent on this chip it still has had some more surprises .

Downloaded the files. Nice m8. Good luck pushing forward. Maybe you can try CL14?

Im pretty done with 3800 profile. Tighten the timings down. Cant see major improvements vs only base timings + TRFC! Just slightly better.
Below screenshot while running. That particular run passed 8K before i stopped!


@The Stilt
TRFC2 and TRFC4 are looking high. Is there any improvement if we "can" get them lower?

Cheers :thumb: .

You also had good results :thumb: .

C14 too much voltage spend for little gain IMO.

TRFC2/4 not used as before.

 

[crakej]

So, still stuck here at 3733. It was a breeze getting here - bit stumped as to why getting any further is so hard, guess I'm just not finding the right settings for my memory.

Many C5 and 07s that required the CLR CMos button, very time consuming, hopefully new bios won't be too long.

I wish AMD would order the ram timings in RM in the same order they appear in the bios.

 

[gupsterg]

So, still stuck here at 3733. It was a breeze getting here - bit stumped as to why getting any further is so hard, guess I'm just not finding the right settings for my memory.

Many C5 and 07s that required the CLR CMos button, very time consuming, hopefully new bios won't be too long.

I wish AMD would order the ram timings in RM in the same order they appear in the bios.

This is settings txt and UEFI menus screnies which I changed and may not be in txt.

View attachment R5PBO1503800v42.zip

 

[The Stilt]

Doesn't AMD Ryzen Master now give us all this information anyway this time around with the timings, CAD BUS values etc? We didn't have that before.
I do hope they don't break their own software in the future that will eventually happen but it's not as critical as before when your tool was the only way for us to present the info in a manageable manner.

Your work was really helpful, thanks for the work put in.

It is another matter with the awful design and manner which the AMD version presents the info but it's there nonetheless.

Ryzen Master has been able to show these timings for quite some time now.
IIRC since TR1 came to market, or slightly earlier.

 

[crakej]

This is settings txt and UEFI menus screnies which I changed and may not be in txt.

View attachment 282626

Thanks again!

Useful to see all the settings together. Of course I will not stop until I succeed!.....except for tonight!

 

[The Stilt]

@The Stilt
TRFC2 and TRFC4 are looking high. Is there any improvement if we "can" get them lower?

You can set tRFC2 & tRFC4 to 0 CLK if you like, makes no difference.

The refresh mode is 1x (7.8µs) and cannot be changed.
tRFC is for 1x refresh, tRFC2 for 2x refresh (3.9µs) and tRFC4 for 4x refresh (1.95µs).

 

[crakej]

This is settings txt and UEFI menus screnies which I changed and may not be in txt.

View attachment 282626

I just realized that I haven't been disabling c-states, can this help stability on Matisse? Seem to remember doing it when I had X370...

 

[Streetdragon]

So you go over 35XX Mhz on the Ram and still get a working boosting on the CPU? when i go over that ramspeed it stucks as 4.1Ghz single/multi or so

 

[gupsterg]

Thanks again!

Useful to see all the settings together. Of course I will not stop until I succeed!.....except for tonight!

NP.

I just realized that I haven't been disabling c-states, can this help stability on Matisse? Seem to remember doing it when I had X370...

I don't know. All the testing up til 3800MHz has been with them as whatever UEFI default is. I just wanted to have crack at 3800MHz and succeed and frankly the FW/HW did step up .

Be aware also there is setting in AMD Overclocking menu, SOC Overclock Mode. That is supposed to do the same for SOC, I didn't turn that on.

I do plan to set all C-States back, even with them off I saw CPU "Sleep" in RM. TBH didn't notice any differing operation of down clocking/volting, perhaps not working, dunno.

This ZIP has a ~18000% run on PBO+150MHz 3800MHz v4.2 and rerun on warm POST of ~5500%.

304 tRFC is ~160ns at profile MEMCLK, which is decent enough IMO. I may try SCL 3 at some point. tFAW lowered some more also, tWR as at 10, can't go lower. tRTP is 6, only 1 step down left, which really isn't gonna gain me much; just like the rest of the subtimings.

I may drop to 3800MHz C15 (~7.84ns), as 3600MHz C14 (~7.77ns) only needed ~1.425V and that was with Gear Down Mode: Off, SCL 2.