X670 resource

[Shamino]

ill use this thread to collect some new test bioses for the boards, maybe also to explain some less understood options

to disable cores ccd go here and choose ccd xx bit map down core.
each ones stand for an enabled core
best to disable from the back, ie:
110000
instead of 0011000
after selection press downcore apply changes or discard if made mistake

ocpak/octools

FAQ:
7950x not boosting pass 5.5G -> check that CStates is not disabled
Detailed Explanation on CState Boot Limiter


Test BIOSes:

new:
X3D OC Preset for those MB with asynch BCLK Support: (for simple slight perf boost for X3D)


DOCP/EXPO Tweaked: (for simple timings tightening)


strixe-e 1515 

strixe-f 1515 

strix e a 1515 

crosshair hero 1515 

crosshair gene 1515 

crosshair extreme 1515 

creator 670 1515

creator b650 1515

strix 650E I

strix 670 itx

for crosshair and strix e-e:

explanation of segment2 Loadline:

customize a heterogenous loadline for a dual segment workload range.

example above shows loadline=L6 when current is in range of 0~40A, and Level4 when current is above 40A.

Adds for x3d

dynamic ccd priority switch with core flex, os / driver agnostic so win10 win11 ok





Algo as follows:
If condition reached and ccd0 specified, then check current mem/cache activity > threshold and hysteresis reached, if fulfilled then switch
If condition reached and ccd1 specified, then check current mem/cache activity <=threshold and hysteresis reached,, if fulfilled then switch
Default hysteresis =4

Can combine multiple algos for ccd priority so combinations are wide

works on non x3d too but of course senseless on it. detailed explanation here.

[ksenchy]

Reading all these memory problems, feels like AMD was a bit too optimistic with having super tight timings with AMD expo kit rams. Now it works now it doesn't.

I kind of had that a feeling this was going to happen so I bought an XMP 3.0 kits downclocked them and manually tuned every parameter of those modules. This takes a load of time and not a lot of people want to do this. They just want to enable EXPO profile and move on.

I hope this mess can be sorted out soon.

[kamyk155]

After longer tests I can still say this newest bios is still the slowest performance bios.
I did few times full bios reset using mobo button. No changes, but I saw one weird thing.
On latest bioses when I set expo tweaked it was great performance boost.
Now expo I / expo II / expo tweaked is marginal difference.
I think asus adding 24/48GB memory kits just ignored normal memory sets and something is slower now.

[kamyk155]

After 2h of BF2042 I saw this - readings bug or wow for 7950X3D ? 

 

[lordberti]

This is normal, same for me. It's only for a short time so much at max.

[SkyBeam]

Just a feedback on installing BIOS 1654 (AGESA 1.0.0.7c) on my ASUS PRIME X670-P: It went pretty smooth and I got a stable 4x16GB DDR5-6000 configuration but needed to tune back my previous settings very slightly. Overall seems to be stable. No significant performance changes observed.

System specs:

• ASUS PRIME X670-P
• AMD Ryzen 9 7950X
• 4x16GB G.Skill F5-6000J3038F16G DDR5-6000 CL30

Procedure:

• Just flashed from 1635 to 1654, no CMOS reset done, however all settings were lost during upgrade
• Dared even to load CMO settings from 1635 for laziness, no issues observed

I have noticed some occasional memory errors in Memtest86+ runs after hours (mainly test  #6 and #9) so I tuned back my memory slightly and changed my undervolt from 1.32 to 1.33v. Windows is now reading bang-on values of 1.35V.

Extract of the settings I use:

• EXPO Tweaked
• FCLK: 2033MHz
• AMD Eco  Mode cTDP 105W
• Curve Optimizer: -17 (YMMV)
• GFX Curve Optimizer -16 (YMMV)
• MC Voltage: Auto
• VDDP Voltage: Auto
• DRAM VDD: 1.33V
• DRAM VDDQ: 1.33V
• SoC Voltage Mode: Offset
• SoC Voltage: -0.05V
• PCIe ASPM: L0 And L1 Entry
• USB EHCI Handoff: Disabled
• POST Delay Time: 0s
• Fast Boot: Enabled
• Setup Mode: Advanced
• BIOS Image Rollback Support: Enabled
• Publish HII Resources: Enabled
• Download & INstall ARMOURSY CRATE app: Disabled
• Serial Port: Disabled

Memory Timings:

• Power Down Enable: Enabled
• UCLK DIV1 MODE: CLK=MEMCLK
• Context Restore: Enabled
• Timings see screenshot

CB23: ~35500, no errors on Y-Cruncher VST for about 30 minutes.

[Silent_Scone]

If errors occur hours into testing may be temperature related. Consider active cooling on the modules.

[Anigilator]

Hello, I have a Hynix HMCG78AGBUA081N BA and Rog strix b650e-f gaming, bios 1650, with hv mode enabled there is no startup, hv mode does not work as such. Maybe someone has similar memory sticks do you have hv mode working or is the problem in the board or bios?

[Falck]

So I have done a small test with EXPO and Nitro with my kit: https://www.gskill.com/product/165/393/1665020366/F5-6000J3040G32GX2-TZ5N

EXPO I:

EXPO II:

EXPO Tweaked:

EXPO Tweaked+Nitro:

TLDR: The only difference between EXPO I and II is TRC where EXPO II has "10 lower" than EXPO I. EXPO Tweaked and Nitro seems to default to EXPO I, but with *vastly* higher tREFI!.
So this to me seems like EXPO II is the fastest setting, Tweaked and Nitro does nothing to this kit?

I know the kit is NOT on QVL list, but is that a limitation to Nitro? Isn't Nitro "testing" the kit to find the "best" timings?

[SkyBeam]

---

@Falck wrote:

TLDR: The only difference between EXPO I and II is TRC where EXPO II has "10 lower" than EXPO I. EXPO Tweaked and Nitro seems to default to EXPO I, but with *vastly* higher tREFI!.
So this to me seems like EXPO II is the fastest setting, Tweaked and Nitro does nothing to this kit?

I know the kit is NOT on QVL list, but is that a limitation to Nitro? Isn't Nitro "testing" the kit to find the "best" timings?

---

I think there is lots of misunderstanding what those modes and specifically NITRO is.

NITRO only affecting memory training procedure. Memory training however does NOT have an impact on set memory timings. Buildzoid once mentioned this also in a video that it is fairly impossible for a mainboard to test all combinations of memory timings during training and this is NOT what training does. Essentially this is about other electrical parameters than timings.

You will see differences on timings when switching between EXPO I/II/Tweaked as of the way it reads and applies the timings:

• EXPO I: Primary timings read from SPD EEPROM of memory module. Some sub-timings are on "auto" and tweaked by ASUS. Remember that some timings also depend on the mainboard layout in terms of signal integrity or other parameters influenced by mainboard manufacturer - as well of course on the willingness of the mainboard manufacturer to sacrifice performance for stability.
• EXPO II: Also apply more sub-timings from SPD EEPROM of the memory module. This is closest to the recommended timings by the memory module manufacturer.
• EXPO Tweaked: Apply stronger optimizations by ASUS. Kind of Expo I plus some bias towards performance if in doubt. The description is kind of fuzzy though.

As you confirm above tREFI is an example where tweaked in your case seems to apply higher performance settings. Note that tREFI is the interval between refreshes. During refresh cycle the memory is unavailable. So longer cycles can improve memory performance at the risk of losing bits due to low charges in memory cells.

In your case none of the settings are likely to affect memory stability as the sub-timing changes are pretty subtle. NITRO however is not going to affect any of those parameters but might affect signal integrity and quality of memory transfer signals (between memory module and memory controller). In short: If your timings are not matching the memory chips also NITRO won't save you.

NITRO can help if your timings are alright and safe but you're facing communication problems (e.g. at higher frequencies).

Have a look at the G.Skill memory specifications. They provide tested and validated timings for their modules. If you exceed those specs also NITRO won't help you. Knowing that of course you might have won the silicon lottery and your memory modules are better than advertised. As manufacturers like G.Skill will not individually test all modules and apply the ideal and best performing timings in the EXPO profile of each single module they rather program values which are matched by every single module of this production batch. So typically those values are on the conservative side and you might be able to use faster timings on most of the modules at your own risk. But remember: Your timings need to be valid for the slowest module/chip in your setup (at least on the same memory channel). Tweaking for individual modules is typically not the effort you want to go through. This is also the reason why mix-and-match is essentially a no-go for memory modules at high speeds as you would have to run the optimization for the worst module in your setup and in worst case you can't find a setting being stable on all of the modules you use due to conflicting requirements.

As long as you stay within JEDEC timings (but this is not what we are here for in overclocking related forums) you should be fine as all the modules are bound to some standard timings and also modules of different manufacturers typically mix-and-match fairly well).

[Falck]

@SkyBeam , awesome explanation! Thank you!

Follow up questions/need clarification:

EXPO Tweaked raises the tREFI and that's good for performance?

EXPO Tweaked has the same timings as EXPO I, but it's still better performance than EXPO II?

EXPO Tweaked+Nitro is even better, potential, performance than EXPO Tweaked, but same timings as EXPO I?

The electrical signaling seems also unchanged if it's proc, rtt etc.?