30 minutes ago, V1ROTA7E said:

Oh no, I love PL, it’s a great tool. Just don’t force MSFS on to CCD0 with PL is what I’m saying. AMD made their drivers work in a very specific way to enable core parking automatically. It’d be likened to the difference between using native frame gen vs Lossless Scaling. Lossless has its uses and is easy to use, but native frame gen works better because it’s able to communicate with the driver and application. PL is great for (when core parking is working as intended) background tasks like BATC/SI.ai, Volanta, Discord, etc.you can put those programs onto CCD1 and free up CCD0 even more. 

Awesome thank you for the input.

I just started using PL as a way around the core parking. However I like the idea of moving other apps off of CCD0.

I will do as you suggested and let the drivers do their thing and just use PL to place the other apps on CCD1.

For what its worth, I've been using PL on my 7950X3D for years to keep MSFS on CCD0 all all other apps on CCD1. Works beautifully and increases MSFS performance (main thread) for me. Also means I can take the focus off MSFS and it remains on CCD0. YMMV. 

13 hours ago, Cpt_Piett said:

I've got core parking working. I'm on a Gigabyte X670 AORUS ELITE AX motherboard. My BIOS settings in case it's useful:

• CPPC dynamic preferred cores: auto
• PBO limits: motherboard
• CPU boost clock overdrive: enabled (positive)
• MAX CPU boost block override: 200
• Curve optimizer: Per CCD
• CCD 0: negative 20
• SMT on

Note: as I took the screenshot, MSFS2024 was in focus. As mentioned earlier in this thread - if it's not, core parking will not work properly.

Okay getting closer. With your settings the correct cores are being used for MSFS but here's the issue. When I am doing non-gaming stuff it still uses CCD 0. I want CCD 1 to be used, since those are the highest frequency cores. Still, doing better. Thanks for posting this!  

On 3/23/2025 at 8:54 AM, RobJC said:

Okay getting closer. With your settings the correct cores are being used for MSFS but here's the issue. When I am doing non-gaming stuff it still uses CCD 0. I want CCD 1 to be used, since those are the highest frequency cores. Still, doing better. Thanks for posting this!  

Not sure if it was mentioned here but the video JayzTwoCents did about the core parking on the 9950X3D he mentioned that for it to work correctly you need to have the power plan in Windows set to "Balanced" or the driver and BIOS updates that fix it do not work.  This also means that you can set to Max Performance and it will ignore those optimizations and prefer the CCD1 cores for faster speeds if you need that instead.

Edited March 24, 20251 yr by flyinion

Well I don't know exactly what fixed it, but now everything works as intended. CCD 1 is the default (the high frequency cores) and when I load any game it shifts to the CCD 0 (vcache cores). Thanks guys. 

On 3/14/2025 at 10:54 PM, mpo910 said:

Yes CCD 0 and CCD1. But the CCD0 Chip Cores are Overclockable. With Curve optimizer set to negatives it hits easily 5.7 GhZ on the Cache Cores, wich are all the 12 cores. So you are able to get this 128 L3 (2x64 MB) cache AND high clock speeds.

My 7900X3D had 64 MB cache on the second CCD chiplet, called CCD1, and flightsim but there was no cache on that chiplet. Now, with 2x64 MB cache spreaded over 2 Chiplet, CCD0 and CCD1, you always use cache cores and are able to use high clock speeds. 

I did some testing w/ the 9900X3D by enabling PBO.  I increased clock speed by 200mHz which was the maximum you could set, set the Curve Optimizer at -20.  SMT has been off and I left it off--have you proven it's a good thing?  How much temp increase?   When I increased mHz by 200 it did hit 5,560 or so but there was some instability in sim some lockups.  But the cost of doing this was crazy:  went from a core temp of 66C to 79C, but the frame rate was only about 5% greater for all of that temp increase.  So set it back to no PBO added mHz but set Curve Opt to -25 and back to around 66C and clock speed peaks at 5360 In the Asobo 787 + SLC + FSLTL + BATC using TAA and taking off out of MMMX Mexico City the frame rate after GSX finished pushback was 72-74FPS-native and it's GPU limited or close to balanced now.  I set it down to 60FPS-na and that just fine.  So nice to have perfect image stability once again with no artifact of any kind including judder from alternative focus between two displays, no blurring of characters, etc.

Looking forward to PMDG or I-Fly 737 for 2024.  Fenix A320 is something I should consider as I don't thing FBW's NX will work flawlessly in 2024 but it's running super well in 2020.

29 minutes ago, Noel said:

 went from a core temp of 66C to 79C

Sounds like you’re happy with what you’ve got, but if you do decide to tinker again in PBO, try the new “curve optimizer” mode with the latest bios. It essentially has 3 primary regions and 3 regions within each of the first three. Confused yet? Lol. 
 

The 3 primary regions are Low usage, Mid usage, and High usage scenarios for the CPU. Within each of those, there are temp regions: Low usage, low temp..low usage medium temp…low usage high temp. It continues in the same fashion for mid and high. The cool thing about this is, you can set an offset value (positive or negative) for each region. As an example, in my Low/Low, Low/mid, and Low/high setting I have -25 for each one. For the mid setting I have -20 in all 3, and in high I have -10. This allows the curve to adjust based on usage and temp..the end result is lower temps, and more stability. My CPU tops out at around 67C with the absolute highest being 71C. This is with the +200 boost enabled. 
 

The reason this method works with reasonable temps is because, compared to the blanket -25, when usage goes up, the CPU says, well I can squeeze *this much* more out because the temp is still within limits (with the +200 enabled)….but as you saw, that can lead to stability issues. Now take what I’m saying with a grain of salt, as I’m no genius….just been researching and trying out stuff with help from YouTube and ChatGPT. I will say, however that I’ve had no stability or temperature issues, and the chip (CCD0 Cores specifically) boosts to 5.6 on occasion, with CCD1 hitting 5.7 regularly. The cooling solution in use is a 360 AIO, btw. 

Edited March 26, 20251 yr by V1ROTA7E

18 minutes ago, V1ROTA7E said:

Sounds like you’re happy with what you’ve got, but if you do decide to tinker again in PBO, try the new “curve optimizer” mode with the latest bios. It essentially has 3 primary regions and 3 regions within each of the first three. Confused yet? Lol. 
 

The 3 primary regions are Low usage, Mid usage, and High usage scenarios for the CPU. Within each of those, there are temp regions: Low usage, low temp..low usage medium temp…low usage high temp. It continues in the same fashion for mid and high. The cool thing about this is, you can set an offset value (positive or negative) for each region. As an example, in my Low/Low, Low/mid, and Low/high setting I have -25 for each one. For the mid setting I have -20 in all 3, and in high I have -10. This allows the curve to adjust based on usage and temp..the end result is lower temps, and more stability. My CPU tops out at around 67C with the absolute highest being 71C. This is with the +200 boost enabled. 
 

The reason this method works with reasonable temps is because, compared to the blanket -25, when usage goes up, the CPU says, well I can squeeze *this much* more out because the temp is still within limits (with the +200 enabled)….but as you saw, that can lead to stability issues. Now take what I’m saying with a grain of salt, as I’m no genius….just been researching and trying out stuff with help from YouTube and ChatGPT. I will say, however that I’ve had no stability or temperature issues, and the chip (CCD0 Cores specifically) boosts to 5.6 on occasion, with CCD1 hitting 5.7 regularly. The cooling solution in use is a 360 AIO, btw. 

You have a different CPU and I don't see the new curve optimizer mode but can look again.  What I thought was weird was I wanted to apply the 200mHz only to CCD0 and use CCD1 for everything except MSFS, but it would not let me.  If I choose Enable 2 Cores on CCD1 it enables 2 cores on BOTH CCDs.   MSI latest BIOS for B650-P Wifi.  It's certainly working great now but the temp increase was prodigious for no sigificant benefit.  Even +200mHz is only a 3.7% increase which translated to a few frames difference and that difference would not be discernable by anyone consistently as the base was over 70FPS-native anyway.  Yep, my cooling soln is Noctua D15 which is great for an air cooler.

22 minutes ago, Noel said:

new curve optimizer mod

I miss spoke, it’s called “curve shaper.” And yes, the chips are slightly different, but not really. The 9950 has more cores and draws more power, so it should be running warmer, and it’s not  🤷‍♂️

 

9 hours ago, V1ROTA7E said:

I miss spoke, it’s called “curve shaper.” And yes, the chips are slightly different, but not really. The 9950 has more cores and draws more power, so it should be running warmer, and it’s not  🤷‍♂️

Ahh, okay I did see that but didn't mess with it.  Your CPU is running where it is because of your AIO water cooler and also it's supposed to run 200mHz faster as a top boost speed over 9900X3D.  When you're close to the bleeding edge power draw goes up sharply relative to increase in performance...but fun to tinker with I've certainly done plenty of it over the years.

Last evening tried the BIOS option of setting Curve Opt on a per CCD basis which kinda makes sense in that it's CCD0 that accesses the higher cache memory and as I'm using Process Lasso put other major CPU players on to CCD1.  I watched video explaining this and learned something interesting:  when the curve is moved to the right (negative volts) the other end of that curve, that address lower clock speed voltage needs will decrease and so can/will become the source of instability that is to say even when you're the higher clock speed needs met sufficiently the instability eventually comes from not having enough voltage on the lower end.  So I tried setting CCD0 at -25, but CCD1 at -10, and +100Mhz.  I had no stability problems and while I saw clock speed go over 5600 for the first time temps were not good, at least not as low as previous:  up to 82C at peak.   I know the CPU has a Tmax of 95C so should not be any real issues that I can tell.  Even so, this morning went back to the simple -20 on all cores and temp now flying to KDEN from KSFO in the 787X is 68-70C, and clock speed now is around 5360 when in demand situations.  The practical difference in frame rate isn't very meaningful as previous said.  Taxiing around KSFO + FSLTL + BATC + SLC the native rate dropped down as low as 42 briefly, where the lock was set at 50.  I probably should retry that exact sequence with the overclock (his guide said this:  set PBO to Advanced, set PBO scaler to 7x to 10x, PBO limited by motherboard, and add boost clock to whatever, I used 100mHz.  That is what got me over 5600 and also up to 82C.

On 3/21/2025 at 11:40 AM, booga said:

I upgraded from a 7800x3d to a 9950x3d yesterday. On this particular PC, I only have a 75hz monitor, and I use frame generation with a 4080, so I was happy with my fps in general, but I wanted to improve the lows when I'm on Vatsim at a busy complex airport. Unfortunately, things seem to be the same as before. In the Fenix at mkvy's Gatwick, I am still going down to ~55fps at times. I thought the non-v-cache ccd might run things like Simapppro, Stream Deck software and Vpilot, but it's completely parked when I'm in the sim. I therefore suspect that a 9800x3d would achieve pretty much the same results. 

Edit: forgot to mention: this is on MSFS 2020.

Second edit: All settings on Ultra

Update: I used Process Lasso to move all processes except MSFS and Fenix to the frequency CCD. The AMD chipset drivers and Windows Game Bar ensured that the sim and plane remained on the 3D v-cache CCD. On Vatsim, arriving at a busy Gatwick, things seemed smoother. I was maintaining 65-70 fps. I shifted the processes back and forth as best I could while taxiing, and sequestering the tasks on the second CCD gained me about 5 fps (=2-3 fps without frame gen). This might not seem like much, but the difference can be felt at such low fps.

3 hours ago, Noel said:

Taxiing around KSFO + FSLTL + BATC + SLC the native rate dropped down as low as 42 briefly, where the lock was set at 50.  I probably should retry that exact sequence with the overclock (his guide said this:  set PBO to Advanced, set PBO scaler to 7x to 10x, PBO limited by motherboard, and add boost clock to whatever, I used 100mHz.  That is what got me over 5600 and also up to 82C.

I implemented both the per CCD curve optimization w/ CCD0 at -25, and CCD1 at -10, as well as the above PBO scale at 7x, limited by motherboard, and this time +200mHz.  This time the taxi w/ 787X + BATC + FSLTL + SLC briefly dropped to 46 then back up to a solid locked 50FPS-native at KSFO.  So that is roughly a 10% increase which when trying to avoid Frame Generation matters.  I learned the Noctua NH-D15 is ~24% less effective than the newer NH-D15 G2 so I will install one tomorrow.  Temps currently aren't too bad at 80-82C when operating at 100% which is 5560mHz now up from 5360 but will benefit with a drop in temp and the new cooler is a simple install.

2 hours ago, Noel said:

 operating at 100% which is 5560mHz now up from 5360 but will benefit with a drop in temp and the new cooler is a simple install.

So in your post above are you saying that MSFS2024 *did* respond meaningfully to the 200 mhz clock increase in your tests?   

If so, pretty neat stuff...

35 minutes ago, Mace said:

So in your post above are you saying that MSFS2024 *did* respond meaningfully to the 200 mhz clock increase in your tests?   

If so, pretty neat stuff...

Yep, about 10% which matters here as I was close to edge of not being able to maintain at least 50FPS w/o Frame Gen, and that is the sole reason I upgraded the CPU.  45FPS is not terrible by any stretch*** but like most one gets used to greater frame rates and so it becomes easy to notice when that rate drops down.   I guess it's in part the interplay of all the players:  CPU, memory, VRAM, RAM, GPU etc as even though a 200mHz increase in rate *should* only have delivered 3.7% increase it turned out to be ~10%.  And in truth, I tried to duplicate the test flight but time of day varied though weather looked about the same. Def needs a better cooler so that happens tomorrow hopefully.

***And one can use FG when conditions mask its graphic artifacts and I am right now--low contrast weather at NZAA

Edited March 27, 20251 yr by Noel