I’m not familiar with Core Cycler. Does it include the adjusting of the PBO multiplier? I’ve found the PBO multiplier is key to dialing in the optimal undervolt as it increases the aggressiveness of the boost (both higher and longer).
No, it simply cycles through and stress tests the cores so you can identify the limits of core stability. I undervolted every core until it failed, then backed off a bit, monitoring boost clock as I went. This took several days, since you can’t really test stability unless you stress the CPU for hours at each iterative step.
I’m not sure if I did anything with PBO Multiplier, but I followed many of the recommendations in this video. Check it out @ 3:40. The ‘CPU vCore Loadline Calibration’ tool seemed to make a big difference in how long the cores remained in boost. I settled on ‘Turbo.’
1 Like
After setting the calibration to extreme (level line) I got a nice bump up in performance with my 7900x3d - thx!! 
. It resulted in the best Cinebench R23 score I’ve run to date. I’ll see over the course of time if my PC continues to run stable.
When I did my PBO exercise (which I think takes your process a step further), I did the following:
- PBO: Advanced
- PBO Limits: Motherboard limits
- PBO Scalar: 10x
- CPU Boost Clock Override:Enabled
- Max Boost Clock Override: started at the default 200
- Thermal Throttle limit: Auto
- Curve Optimizer: see below*
- I followed this iterative process:
1 Found the lowest undervolts in the curve optimizer
2 Benchmarked the settings (stop after fully optimized)
3 Lower Max Boost Clock Override by 25-50
4 Go to 1
I found that by lowering the max boost clock override value I could continue to lower my undervolts. This of course generates less heat and will allow the cpu to hold a higher boost. The scalar at 10x will allow for the most aggressive boost. Currently, I have my 7900x3d max boost clock set to -100 (minus 100). While -100 definitely lowered the boost headroom, it allowed my cpu to hold the highest boost value.
I did not win the silicon lottery with my 7900x3d (most likely a 7950x3d reject) however, I feel I’ve squeezed the most performance out of it (and your suggestion above helped squeeze a little more 
).
1 Like
I also upgraded from 32 to 64. With 32 you stay around 27-28 GB most of the time, but once you’ve put in 64 GB the SIM instantly utilizes a minimum of 32 GB all the time while not having ANY third party stuff installed right now, all stock. This is in 4K DLAA 95% maxed out with TLOD/LOD set to 400/170 and live weather/traffics and multiplayer active on a 4090 and 7800X3D, RC set to 256 GB.
All panning and mouse activity is totally smoothed out. Still it won’t save you from the famous VRAM attacks, even on a 4090 with 24 GB RAM.
Thanks for that. I’ll do some more tweaking, using your wizardry. 
1 Like
Hi @BegottenPoet228 - I suspect I have the same memory kit (purchased Oct 2024):
TeamGroup T-Force Xtreem Overclocking 48GB Kit (2 x 24GB) DDR5-8000 PC5-64000 CL38 Dual Channel Desktop Memory Kit FFXD548G8000HC38EDC01 - Black
You will enjoy the results of the DDR5-8000 memory as I have even been able to OC mine to 36-48-46-36:
A newer bios with AGESA PI 1.2.0.2b allowed me to drive my FCLK to 2200 and be stable which I was not able to do previously. Check out this video in which Buildzoid explains the benefits of FCLK=2200 in a way only Buidzoid could do 
. Good luck and have fun!
Edit: this is the performance from the memory kits stock XMP profile which you can compare to above to show much the OC helped:
1 Like
I saw a YT video where he benchmarked a bunch of games, including MSFS. The tests were all performed on a 9700X / 4090 combo, and he purposely ran the tests at 1080p with low graphics settings so as to force a CPU bottleneck. He tested 6000/CL30, 6400/CL32, and 8000/CL38.
The vast majority of the tests showed little performance delta between the three kits. Certainly not enough to justify replacing my current 6400/CL32 kit.
But the MSFS test of the 8000/CL38 RAM showed a 39.6% increase in 1% lows, relative to the 6000MT/s kit. 
Based on that result I just bought a 2x24GB DDR5-8000/CL38 kit that was on my motherboard QVL, and once I get it I’ll replace my DDR5-6400/CL32 kit and do some comparative tests.
I know I’ll get less benefit with my X3D chip versus his non-X3D chip.
I’ve been happy with the 6400 kit, but if I can get anywhere near a +25% delta for 1% lows it will definitely be worth the upgrade. Of course I will be hoping and praying that the new RAM will be stable at that speed. 
I’ll be going from 64GB down to 48GB, but that should be plenty for the sim.
1 Like
I remembered that someone had gotten some good results with 8000 MT/s RAM.
Thanks for reminding me it was you. I’m anxious to test out the new kit.
2 Likes
If you do any 64GB vs 48GB comparisons I’d be interested in seeing the results! I only had 32GB when I upgraded to 48GB so I was not able to benchmark it.
It doesn’t matter a this point. The game is not good enough to say what is better.
32 or 64
And the game plays differently every other day and time. It’s never the same
5 Likes
If you can afford it, go for 64GB. Even if 32 or 48 is OK today, with 64GB you’ll be better off when the requirements increase in the future.
What you really want to avoid is using your harddrive as a swap RAM memory. There will be a lot of read & writes to that disk and eventually it will wear out. It’s not happening soon though but eventually if will.
1 Like
Amen to that. I said a long time ago that testing MSFS 2024 is a waste of time.
Would you mind sharing the link to that benchmark video?
1 Like
I would have bought a 2x32GB kit of DDR5-8000 RAM, but there are no 32GB DIMMS of that speed ( none on my QVL anyway.)
As for your second point…
When I hear ‘hard drive’ I think of mostly obsolete spinning platter drives. There are still good uses for that old technology, like NAS. But modern SSD’s have no moving parts, and thus their MTBF is very high - on the order of 1.5 million hours.
Regardless, they are not infallible, and some do fail. Backing up critical data is still wise.
Yeah, ”hard drive” kinda gives those associations, doesn’t it? I guess a more modern word would have been SSD. My only excuse is that I’m old enough to have used tape back in the days… 
I had an SSD disk failing in my work laptop once. Really annoying and caused quite some headache until it was sorted out. That was when I learned that those disks can get worn out. I’ve been paranoid about backing up sensitive data ever since.
2 Likes
Same here. None of my gaming computer files are irreplaceable, but I still do daily incremental backups with full backups once a week of my C: drive (Windows) and my D: drive (pretty much everything else.) All my 3rd Party addons are in the cloud on Proton Drive as well.
Being paranoid about losing data is a good thing!
2 Likes
A post was merged into an existing topic: DX12 - Does Your Ram Max out?
RAM latency seems to be a bigger performance factor in some games rather than speed. Here’s what I’m getting with stock timings on two 32 GB G.Skill kits (4 modules totaling 64 GB).
BTW, these number vary considerably from test to test. I’ve seen read, write, and copy speeds higher than these and I’ve also seen latency as low as 60 ns without any system changes. It all comes down to whatever else the system is working on in the background when the test is being run. Hence my subscription to the KISS philosophy. A lean system is a fast system.
1 Like
My two centavos/kopeks/rupees:
I had my system set with 32gb RAM figuring that 32 gigs was enough RAM to park a Forrestal class carrier. And it was, except that MSFS (both versions), want a much larger drydock!
Example:
With a 19200x1200 display with things set reasonably high, (or even not so high), sitting at the end of the runway in a DV-20, looking right and left out of the cockpit, it would always stutter when changing views rapidly.
With 64 gigs, I can look around all day long without stuttering and flying is much smoother than before.
Note that when you add memory and then load the game, (especially '24), it will take much longer the first couple of times it loads and will likely stutter during the opening trailer (in '24), as (AFAIK), it is re-arranging memory allocation and various caches to take advantage of the increased memory available in the system.
Once you’ve loaded and run it a few times it smooths out and eventually the load times return to normal.
At least that’s my experience. If you can afford more, go for it!
I installed the 2x24GB DDR5-8000/CL38 RAM yesterday. I got an error during bootup and the BIOS reset itself to default values. I’m not sure what caused it, but I next tried the EXPO overclock setting (keeping everything else at BIOS defaults) and everything works fine.
- I ran MemTest86+ for an hour - no errors.
- The synthetic benchmarks are so far showing about 5% lower performance than with my custom 6000 MT/s setup (no doubt due to having to run the faster RAM with UCLK=MCLK/2 instead of UCLK=MCLK like I could with the 6000 MT/s RAM.) I was also running FCLK=2133 with the slower RAM, and it’s set to 2000 with the faster RAM. I’ll mess around with that later…
- The system has so far been 100% stable @ 8000 MT/s using the EXPO profile.
I’m continuing to waste time today running all the same benchmarks I’d gathered data for with my custom DDR5-6000 profile. Then I’ll do the same flights I did with CapFrameX to see the real-world comparison.
Finally, I’m going to start tweaking the BIOS back to where it was (PBO undervolt, etc.) and applying @WingWarper1 memory timing tweaks as well. I know I have a solid, stable system running the RAM at DDR5-8000 with the RAM using EXPO settings and all other BIOS settings at default, so I have a good baseline to return to.
Prepare yourselves for a big data dump in the near future. I’m putting everything into Excel and will post a plethora of graphs comparing the two RAM kits once I’ve got things humming.
3 Likes
It is possible that if you had enabled the BIOS feature for detecting newly installed RAM, BIOS might have reset the settings. Additionally, if you had an active EXPO profile or had manually tweaked RAM settings prior to installation, that could also have led to a BIOS reset.