Bob Scott

Per-core HT isn't possible on a 10th-gen CPU. I can do that on my 13900KS, but not on the 10900K in my MSFS box.

That's normal--P3D is showing how much VRAM is available to it...the physical VRAM minus what Windows reserves for itself.

I think you meant 45:38 into the video...

In a variation on this theme, I had been running MSFS on my 10-core 10900K with HT off. I've observed that the main thread runs on the last core (core 9 in my case). As an experiment I lowered my all-core overclock to 4.8 GHz from 5.1 GHz, turned HT back on, and since MSFS doesn't have a config option to manage processor affinity, I used Process Lasso to externally set an affinity mask that keeps MSFS off of the paired virtual CPU on the core running the main thread, so the main thread doesn't share its core with other threads. In other words, Process Lasso allows MSFS to use virtual CPUs 0-18, and blocks vCPU 19 so that the core running the main thread doesn't have to share its processor time with other threads. This is an alternative to the all-or-nothing HT on or HT off approach, and it seems to have significantly mitigated the stability issues I was having with memory access violation (0xC0000005) crashes, and without any real performance hit.

This sounds like a Simconnect problem. There have been issues where add-on software bombards the Simconnect interface and overwhelms it to where it gets way behind and essentially stops responding. I recently had a number of panel lockups in the Fenix A320, but it was on SU13 the week before SU14 released. The panel froze and stopped responding, but the sim kept flying, sounds continued to work etc.

Check the control assignments of all your hardware in MSFS to make sure you don't have another device assigned to the rudder or tiller and fighting you for control. MSFS service updates have a history of messing with control assignments.

I have a 4090, and it's been nearly an omnipresent thorn in my side since I returned to tinkering with MSFS a few weeks ago. If I switch to an outside view (using DX12), patches of blurry Bing photoscenery tile replace some of the scenery tiles in my add-on sceneries. Turning on dev mode, then deselecting and reselecting "Draw on Terrain" fixes it temporarily. Was an intermittent problem, but became a persistent pain in the butt. Have not tried DX12 with SU14 yet to see if it's still as pervasively underwhelming.

Sounds like you're likely right on the ragged edge of the voltage needed to run stable at the manufacturer's XMP settings. Im my experience, usually you can stabilize an iffy XMP profile by adding one or two hundredths of a volt to the manufacturer's voltage spec and set that Vdimm manually in the BIOS--motherboards aren't super-accurate w/r/t actual applied voltage vs what's sensed/reported by the mobo's hardware, so the voltage being applied via the XMP profile might be just a hair too low. You might try that, and then run Memtest86+ through a 4-8 hour test cycle to verify stability.

SU14 now released, so I'm closing this prerelease thread.

SU14 download size was 7.2 GB, disk space 11.2GB. Did see multiple retries on one of the files at about the 92% mark in the d/l.

My PFC yoke is the floor-mounted version, which uses rotary transducers rather than the linear ones. There used to be a couple of Honeywell Hall Effect sensor packages (a 90-deg and a 180-deg version) that were very close to drop-in replacements for the original rotary pots (same casing and shaft diameter). Linear pots are a very different animal--I had to custom engineer a solution to replace the linear pots in my PFC throttle quad, which involved fabricating a set of brass piers that hold rare earth magnets, and brass mounts that hold Allegro HE chips attached to the old pot sliders that move between the magnets. AFAIK, there still are no HE equivalents for those 100 KOhm sliders where they're used in the PFC hardware, which probably explains why PFC now uses HE transducers in its yokes, but still uses those pots in their throttle quad.

In the BIOS "Tools" page there's a profiles page that allows you to save up to 10 BIOS configurations...if you save what you have before you touch anything, then you can always go back there and reload that saved BIOS config if tweaking gets you to a bad place. You can also save those configs to a USB stick, so you can get back there if anything were ever to happen where you'd need to reset the BIOS to stock settings (e.g. a memory DIMM going south on you). The boost of two selected cores is for light, 1-2 core workloads...maybe something like a spreadsheet in a business setting. Not really much use for us.

@MattNischan On a related topic, is this the same with the WT Garmin 1000 NXi? It appears pretty dated--the description in the marketplace still says it's an early beta. It's not at all clear whether it should be installed for acft that use the G1000.

That's normal--the 13900K runs two cores at a higher clock when there's no/low load, but if using the default intel scheduling they will downclock when a significant multi-core load is present. I doubt you'll see those cores at 5.8 with P3D running, unless your builder manually configured the BIOS to lock those cores at those speeds. The JobScheduler entries of 0/2/4 are just default settings--P3D does not do any sort of intelligent assignment based on differential core performance. What I'd do is watch HWInfo with P3D running to determine if any of the cores are left to run faster than the others, and if so, first assign the main thread to that core and see if the core still runs faster. I had to force that in the BIOS on mine after running many hours of overclock drills to determine which core was the most robust under load.

The fuzziness is a result of using DLSS--with FG+TAA it works well. I'd find it a welcome addition to P3D if LM ever decided to support it.

No, it doesn't.

Honestly, I'm not well-enough versed in the current batch of TVs to make a recommendation...some preliminary research had me considering a 55" LG G3, but I'm still quite wary about long-term dimming and burn-in effects from prolonged PC use of OLED displays, and reviews of the previous (C3) generation of LG TVs did note some issues with flicker when used with a PC or game box. The ultimate goal is to replicate on my X-Plane and MSFS machine the sort of experience I have with a 55" display on my dedicated P3D setup, but with VRR instead of a 30Hz VSync lock to keep things fluid. I was really hoping to find a large format true G-Sync display. ASUS produced a 65" G-Sync monitor back in 2018--but alas, it's been discontinued and there looks to be nothing in that space out there now. Nothing is currently jumping up and grabbing my attention. So for now I'm gonna hold pat with my 34" G-Sync ultra-wide.

On the 13900KS, I'm using an AM of 262143 to run P3D on 18 cores (with the remaining two active e-cores used for ancillary stuff). To reduce heat I turned HT off and disabled one group/bank of four e-cores in the BIOS, so there are 20 active cores, each of which is also a VP. I set the MainThreadScheduler value to force P3D to run the main thread on the one core I've independently clocked up to 6.0GHz. @Ray Proudfoot I'm a bit sceptical that a 240mm AIO is enough cooling for the 13900K. With a radiator that small, I'd expect the fans to be screaming with the machine under a typical flight sim load. I've generally recommended a 360mm AIO as a minimum ever since the 10th-gen CPUs came out. Also, 1.44v and 48 deg C at idle is pretty high--I see 1.34v and 27 deg on mine. I have to wonder what your voltages look like under load...my gut feel is they're way high. I highly recommend you get a copy of HWInfo64 (it's free) and watch your core VIDs under load. @Luke The PL1/PL2 TDP limits went away with the 12th-gen Intel CPUs. There's a similar base power limit (125W) and a max turbo power limit (253W), and there's no longer a "tau" time limit at max TDP, and there's no hardware enforcement of those limits. The problem I see with setting those limits PL1/PL2-style via the BIOS is that, like PL1/PL2, it's an all-or-nothing proposition--if the power exceeds the base power limit (up to the max TDP) for a specified period (formerly tau) then it gets throttled all the way back down to the base limit. See https://pcper.com/2022/10/intel-core-i9-13900k-power-scaling-performance-explored/

@Ray Proudfoot How are you cooling the CPU? And how are the CPU voltages set? My first suspicions would be either the use of an auto-overclock which, as most auto-overclocking is prone to do, is jacking your voltages up way too high. Alternatively, it could be a problem with CPU cooling--something like a wimpy fan curve on an AIO or a bad application of TIM. But yeah, 100C at those loads looks aberrant to me. I also agree that the loading across the cores looks way too high for P3D. I'd use something like Process Lasso to rule out something else that might be running other than P3D that accounts for that load. I haven't played with PLs (power limits), as lowering them drives the CPU to clock down to stock after prolonged periods of high load, and prolonged periods of high load are normal flight sim behavior. I run HWInfo64 remotely via the LAN and watch power (and voltages) on a separate PC during my overclocking stress tests and when testing a flight sim config on a new or modified build, just to keep things sane. If it gets too high, I adjust it by reducing the clock mults and/or CPU core voltage.

Yikes--That's a really bad AM if you have HT off--that makes four of your eight p-cores unavailable to P3D, which is definitely not something you want to do. The pairs of 01 are what you might use with HT on, so that it blocks P3D from using one of the two VPs on each HT-enabled CPU core. With HT off all the VPs are physical cores, so that 01010101 sequence blocks every other physical core, and in this case the rightmost eight binary digits in the mask represent the p-cores, which are much faster--something on the order of 30% faster.

I have a 55" 4K TV on my P3D machine, and a 34" 3440x1440 ultra-wide on my XPlane/MSFS box. The lack of vertical real estate on the UW is an issue--it's not a showstopper, but it does detract somewhat. The large-format 4K display does as much as the UW w/r/t horizontal FOV. So I think a large-format 4K display is the best view solution overall, if you have room for the big display. One thing, though--if you want VRR, finding a large display with flicker-free G-Sync "compatible" VRR that can rival a true G-Sync equipped UW display is something of an undertaking.

On the 13900K with HT on, you have 32 virtual processors, not 24 (2 on each of the eight p-cores, and one on each of the 16 e-cores). With HT off, you have 24 VPs, all of which are physical processors. If you want to cool things down to get some thermal headroom, in addition to turning off HT, you can disable a bank of four e-cores, leaving you with 8p + 12e, or 20 VPs, which is more than enough. That's how I'm running my 13900KS--I have the CPU overclocked to 6.0 GHz on one core where P3D's main thread runs and 5.8 GHz on the remaining seven P-cores, but I also run very aggressive water cooling in an 18 deg C room -- two 360mm deep radiators in series plumbed with 0.5" ID tubing and a high flow Laing D5 pump. To disable e-cores (you need to do it in groups of four) you enable "Per E-Core Control" and you'll then see a list of the 16 e-cores...you can then disable e-cores individually. To confirm you have an entire bank shut down, you can go into Extreme Tweaker -> Specific Efficient Core, and you should see E-Core ratio limit and voltage for only three active banks, which is indication that the fourth and unlisted bank is shut down. If you don't disable all the e-cores in the bank (aka "Group"), you'll still see the ratio limit and voltage setting options listed, which indicates the bank is still powered-up and producing heat. In my case, I have the last four e-cores (12-15) disabled, and on the Specific Efficient Core page I see ratio limit and voltage setting options only for Group0 (e-cores 0-3), Group1 (e-cores 4-7) and Group2 (e-cores 8-11), and nothing for the now-depowered Group3. With HT off and a bank of e-cores shut down, if you use an AM it should be 20 bits, reflecting the 20 operating VPs.

Congrats. Take her on a nice simulated flight to Saint Maarten, perhaps? 😃

In the FSUIPC for Advanced Users document, see the section on traffic limiter ini parameters--should be in one of the appendices of the document. There's a parameter called Ground Preference that controls how much traffic gets deleted on the ground.

OK, but those are operator and/or acft manufacturer polices/procedures, which can be more (but not less) restrictive...the approach itself does not have published DH minima.