LRW

Hi Bryan, I installed 1.4 and there's a PDF that says: IMPORTANT ‘ULTIMATE GROUND CREW X’ INFO FOR NGX-U USERS! If using ULTIMATE GROUND CREW X for your pushback and ground services, please visit the FS2Crew Support Forum for important information related to the NGX-U! I can't find any topic under the NGX-U or UGCX about this. Can you point me to right place, please? Thanks.

I started getting these on my 1080 Ti in v5 so I fell back on the TdrDelay work-around that I had used on my previous PC. Microsoft documents it at https://docs.microsoft.com/en-us/windows-hardware/drivers/display/tdr-registry-keys. I don't remember what app I used it for before (FSX or P3D) and it didn't always prevent timeouts, but so-far-so-good this time. KeyPath : HKEY_LOCAL_MACHINE\System\CurrentControlSet\Control\GraphicsDrivers KeyValue : TdrDelay ValueType : REG_DWORD ValueData : Number of seconds to delay. 2 seconds is the default value. I set mine to 10 seconds. I wouldn't use too large a value or you might have to power cycle to recover.

Thanks. That did it. KMSP looked kinda barren without Delta.

This is from Rob's P3D Guide. This description was written from a 32-bit perspective, when OOM's were a major concern. Setting it to '1' in a 64-bit environment will result in higher VAS but keep textures available longer. ENABLE_MEMORY_OPTIMIZATION Section: [GRAPHICS] Source: LM – Beau Hollis Default: 1 (if no entry exists) Description: 1. Performs periodic cleanup of the geometric cache which stores buffers used for batched buildings. If you have high autogen settings and move spin the camera around, the geometric cache can get quite big. Looking strait up at the sky for a few seconds may get you back 100+ MB of VAS. This change doesn't really help with the worst case scenarios where 200+ MB of memory is needed for the current scene, but it could help a great deal if you were to start in london where autogen is very taxing and switch to FlyTampa Dubai, where autogen is mostly disabled and replaced with pre-batched placed models. 2. Clear textures out when changing airports. Going to a new airport would trigger loading of new texture before the current area's textures had been given their 10-seconds to page out. Clearing out textures on airport change, helps prevent a spike in texture memory related to this. Note that with MaintainSystemCopyOfDeviceTextures enabled, this is less important as the bulk of the memory associated with these textures should already be freed up before the switch. Benefits: Reduce VAS usage. Detriments: None.

Several people on the Flight1 UTL support forum have reported that they no longer see Delta mainline or AIr Baltic flights after the update. They still see Delta Connection flights. I'm holding off until I see that Delta is working.

Since the 32GB kit of the TridentZ 4400 is the only RAM at that speed on the QVL for the Maximus XI Formula, it will be interesting to see if you can get XMP to work or at least a higher speed. I know that it won't be a matched set of 4, but close enough maybe?

Rob initially wrote 32000 and later corrected it to 320000. I don't think the decimals matter.

It didn't eliminate 100% of autogen popping and also some black textures when I look perpendicular to the direction of flight, but it is a significant improvement for me.

Did you add the settings that Rob described in the v4.4 thread? That made a big difference for me. MAX_TEXTURE_REQUEST_DISTANCE=320000 ENABLE_MEMORY_OPTIMIZATION=0

Why? A mere 200 MHz will give you 1.2 frames per second at 30 frames per second! Not worth the extra strain on the CPU or the effort, or the extreme cost of a custom loop. However... if you are a competitive overclocker, or just curious as to what your CPU can do, or have an abundance of funds then fair enough. If I was on a budget I'd have gone with an 8086K. This is more in the category of trying something new for the fun of it.

I've watched that video a couple of times. It was helpful, although his conclusion was for most people to use AI overclocking, which I tried early on to see what it would do. I found that a run of Prime95 or even RealBench would lower the cooler score so much that AI would lower the multiplier to 47 or 48. Running only P3D would raise it to 51. For people who want to let AI overclocking do the work it's important to only run their normal day-to-day apps and not artificial loads, like benchmarks. I just set the mode to Adaptive and entered the OC voltage manually. When I booted into Windows the idle voltage showed 0.630V. I don't know what I did before when it was 1.15V. Maybe I'm confusing it with my tests in Offset mode. My spreadsheet with all of my benchmark runs doesn't include minimums, only max and avg. Oh well. I'll stick with Adaptive. I think today I'll try a long flight and stop tweaking for a bit.

What cooler are you using? Running Prime95 with my current settings would hit thermal throttling very quickly and individual threads would fail. I'm considering doing my first custom loop to see how much I could lower temps from the AIO and possibly go to 5.2 or 5.3 GHz.

I used Manual when I was pushing the voltage up to see what the speed/heat was, but I don't want the voltage to remain at that level when idle. The Auto setting allows the CPU to run with 0.630V at idle (800 Mhz) and, at 5.1 GHz, AI Suite shows an Adaptive OC Voltage of 1.315. I haven't been able to find an Adaptive setting of my own that drops below 1.15V at idle. My last 15 minute run of RealBench on Auto showed Vcore @ 1.279V under load with LLC 6 and AVX 0. My temps in P3D are mostly in the 50s and 60s, with some spikes in the 70s, so I'm not concerned about AVX. I'm still on this journey, so I don't know where I'm going to end up.

I just installed FLAi and AI2UTL and saw the same behavior. It appears to be an issue with UTL not resolving add-on.xml files, which is how my FLAi was configured. I manually added the FLAi directory path to simobjects.xml and then the FLAi repaints showed up in the UTL config tool. AI2UTIL is a major time-saver. Great job!

As Martin indicated I'm using the Auto setting for LLC on my Maximus XI Code. When I check the value it always shows either 6 or 7.

I'm going to have to run that test again. I found one of my hwinfo logs from that time (22 Nov) and it shows the AVX speed (4.7GHz) 67% of the time. Core 0 was AVX 48% of the time. The other 7 cores ranged from 65-74%. All 8 cores were at 4.7 24% of the time. On average 6 of the 8 cores would be at 4.7 and 2 cores would be 5.0 at any given time.

If memory serves (which is always iffy) when I used AVX offset in my standard test flight around the patch at KLAX I would see the frequency drop intermittently. Most of the time it was full speed. So I don't use AVX offset for normal flying because I would only see brief spikes in my temps.

I'm not concerned at this point. After running the RealBench stress test for 15 minutes at 5.0 with HT off, and a max power draw of 167W, the max VRM temp reported on my board was 51C. I have no direct cooling of the VRMs; just the 3 side-mounted 120mm fans.

It might depend on the silicon lottery for your specific 9900k. I can run at 5.0 with reasonable temps on all cores with HT off. I even had it running at 5.2 for a couple of days but had one BSOD so I backed it off. With HT on I can reasonably run at 4.8 without the temps getting out of hand.

I've been trying them out this morning with my standard test flight around KLAX using TrackIR. Take off from 25L, make a 20nm downwind over Long Beach and land 25R. When I turn onto the crosswind over the Pacific and look left to the city all the autgen buildings are black and the textures pop after a couple of seconds. But the FSDT KLAX is still fully visible, which is an improvement from before. My VRAM usage is just under 9GB vs. 5GB before. RAM is 10GB vs 8.5GB. [GRAPHICS] MAX_TEXTURE_REQUEST_DISTANCE=320000.000000 ENABLE_MEMORY_OPTIMIZATION=0 [TERRAIN] TEXTURE_SIZE_EXP=10 UseGlobalTerrainView=True My autogen settings were not maxed out for this test which was a leftover from previously having Orbx SoCal loaded, which is a frame killer and stutter inducer. I'm going to re-run the test with only OpenLC NA and max autogen. As soon as I look left: After a couple of seconds: Video showing autogen texture popping:

It's a different story with HT. My 9900k runs about 10C hotter with HT on than with it off. RealBench wll fail or BSOD at 5.0 with HT with some thermal throttling on 1 or 2 cores plus the CPU package. However, it easily runs at 5.0 with HT off in the 70's with peaks from 79-84.

Created the return on Amazon. It's packed and ready for UPS to pick up tomorrow. I ran comparisons between the Ryujin and the EVGA in Cinebench (which takes 20 seconds to run) and RealBench (15 minutes) . The Ryujin did better than the EVGA with HT off, but the EVGA did as well as, or better than, the Ryujin with HT on. The values are the average peak temps for the 8 cores. Cinebench R15 HT Off @ 5.0 HT On @ 5.0 Ryujin EVGA Diff Ryujin EVGA Diff 66 °C 73 °C 7 °C 91 °C 81 °C -10 °C RealBench 2.43 (15 min - 4GB) HT Off @ 5.0 HT On @ 5.0 HT On @ 4.9 Ryujin EVGA Diff Ryujin EVGA Diff Ryujin EVGA Diff 76 °C 88 °C 12 °C 94 °C 94 °C 0 °C 93 °C 94 °C 1 °C HT on @ 5.0 resulted in a BSOD for both coolers about half way through.

Ya. It appears that they put all of their effort into the bells and whistles of the OLED display and the actual functionality was a poorly implemented afterthought.

Big disappointment with the Asus Ryujin 360 AIO. I don't see that it cools any better than the EVGA CLC 280 and, even worse, it doesn't even have the basic features of the EVGA cooler, like showing the coolant temp and allowing setting the radiator fans based on it. And as for instructions on how to use the software: "Please visit the product support page at http://www.asus.com to download and install the control software." The download page for the Ryujin has AI Suite III, LiveDash, and firmware updates, but no additional documentation. The LiveDash is easy enough to use. It lets you select the following: Parameters you want to display on the OLED (although it doesn't remember previous choices - everything is unchecked when you open it) What image or animation to display, either predefined or custom A custom string to display with the ROG logo during boot Lighting control for the LED stripe, if you don't have an Asus mobo with Aura The first choice is the only one I used. Pump and fan control is done from AI Suite, which seems like it could be an issue if you bought this to use in a non-Asus mobo. Unfortunately for me AI Suite didn't see the radiator or VRM fans leaving me not knowing how I was supposed to control them, as there is no documentation or FAQ. I found out from a new review of the Ryujin AIOs by Tiny Tom Logan on OC3D's youtube channel. I found that my system doesn't see the USB connection on every boot, especially cold boot. Tom pointed out a non-trivial issue. Namely that the pump and fans all operate based on the CPU temp, but not the actual temp reported by the CPU; rather, it's from a mobo probe that has a definite lag and under reports the CPU temp by 5-10C. Although I found that it also spikes intermittently by 10-15C, which caused all the fans to suddenly run high for a couple of seconds before slowing down again. I adjusted the fan curves to eliminate those annoying run-ups. The best thing about the Ryujin 360 is the set of Noctua fans. I've never used them before and I really like them. The little VRM fan on the block is definitely the loudest thing in my system when it spools up. I don't know the speeds of any of the Ryujin fans because they don't appear in the fan list that shows the current speeds. The fans only show up in the profile graph that lets you customize the speed by percentage. I'll run some direct comparison tests in the morning with the Ryujin and the EVGA CLC with both HT on and off to get some hard numbers, but I'm leaning toward returning the Ryujin to Amazon. I think I can put the $279.99 ($301.06 w/tax) to better use. Maybe it's time I started looking at a custom loop.