September 7, 20232 yr Author 8 hours ago, LRBS said: Andy, I just remembered a situation that I faced a while ago. I replaced a PSU and had an issue related to the built-in fan controller. I discovered that the sensor was malfunctioning and causing incorrect temperature readings, and overheating which led to the entire PC running hotter than normal. After checking the proper installation and airflow direction, I realized that an incorrect power supply combined with a faulty sensor could result in increased temperatures. Please see if that can be an issue. Hopefully not. I do have a huge PS (EVGA 1600w) but it seems to be fairly quiet. What I have noticed on my system is that Fan Control doesn't register any kind of PS fan, and I have been meaning to investigate that, but just from physical evidence (ears!) I can tell that it's not making any particular racket or blowing huge amounts of hot air up onto the card. The fact that the card temps are now a couple of degrees lower and the card fan speed reduced post undervolting sorta confirms that as well. Ryzen 9 7900X, Corsair H150 AIO cooler, 64 Gb DDR5, Asus X670E Hero m/b, 3090ti, 13Tb NVMe, 8Tb SSD, 16Tb HD, 55" Philips 4k HDR monitor, EVGA 1600w ps, all in Corsair 7000D airflow case. Sims in use - 2020, 2024, XP-12 and -11, FSX/SE, P3Dv4.5 and v5.4. DCS and AFS2 installed but rarely used
September 7, 20232 yr 13 minutes ago, andy1252 said: Things like the fans racing when sitting on the runway, or at the hangar screen or whatever you call it (the one with the world map on) would seem to indicate to me that MSFS is doing something it didn't oughta, and so maybe that will get resolved at some point as well. Not necessarily. Even if you are still and doing nothing at the parking, your hardware rendering n frames per second works n times harder than rendering just one frame and going idle. No matter how many things are apparently changing in the visible scene. Edited September 7, 20232 yr by MrFuzzy 7800X3D | 2x32 GB DDR5-6000 CL32 | RTX 5080 | Alienware OLED 34" | 1 Gbps fiber
September 7, 20232 yr 1 hour ago, andy1252 said: Things like the fans racing when sitting on the runway, or at the hangar screen or whatever you call it (the one with the world map on) would seem to indicate to me that MSFS is doing something it didn't oughta, and so maybe that will get resolved at some point as well. That was resolved, turn low power mode on in the settings...
September 7, 20232 yr It is not uncommon for under-volting to actually improve performance, particularly if you otherwise thermal throttle.
September 7, 20232 yr 5 hours ago, MrFuzzy said: LOL what? Undervolt usually increases sustained clock speeds and often results in the same or increased performance. The average clock speed of my 3090 is higher since I undervolted it, and the 3DMark score slightly higher. Heavy undervolt and power limiting can decrease the performance but some people opt for it anyway because the tradeoff for a -30% power consumption is merely a 5-10% performance decrease. Undervolt is a very well known practice for 3090 and 3080, there are many guides and videos online attesting its efficacy. I was thinking that this practice for amd ryzen CPU only Missing the PMDG DC6 in MSFS 2024 (she's here, but...).
September 7, 20232 yr 3 hours ago, Claudius_ said: I was thinking that this practice for amd ryzen CPU only The ability to undervolt is literally built into the supporting Radeon software for my GPU. So in my personal case the card in my sig is undervolted and overclocked past the already factory overclocked 2340 MHz boost clock to 2550 MHZ boost with headroom to further overclock past 2600 MHZ. AMD Ryzen 9800X3D/ Asus ROG Strix B650E F Gaming WiFi / Asrock Taichi 9070XT / 32GB G.Skill Trident Z5 Neo DDR5 6000 / 2x ADATA XPG 8200 Pro NVME / Arctic Liquid Freezer II 280 / Seasonic Vertex 1000w PSU / Lian Li LanCool II Mesh Performance / Asus VG34VQL3A / Topping E70 Velvet DAC & L70 Amp /Sennheiser HD660s2 Thrustmaster Boeing Yoke + TCA Sidestick + TFRP Rudders
September 7, 20232 yr 10 hours ago, Claudius_ said: I was thinking that this practice for amd ryzen CPU only I undervolted my Intel 9900K, which I run with stock clocks as it is in a SFX case the size of a shoebox, in the motherboard bios using an offset. The only issue if I undervolt the 9900K excessively is instability under no load. It runs fine under heavy load. Backing off slightly on the undervolt made it stable under all loads, gave the same performance and it ran about 5 degrees cooler (Arctic Liquid freezer II). Edited September 7, 20232 yr by Glenn Fitzpatrick
September 8, 20232 yr 1 hour ago, Glenn Fitzpatrick said: I undervolted my Intel 9900K, which I run with stock clocks as it is in a SFX case the size of a shoebox, in the motherboard bios using an offset. The only issue if I undervolt the 9900K excessively is instability under no load. It runs fine under heavy load. Backing off slightly on the undervolt made it stable under all loads, gave the same performance and it ran about 5 degrees cooler (Arctic Liquid freezer II). It isn't the same as the ryzens, if you undervolt following the optimization curve you'll have higher clocks and lower temps (but the same risk when under no load). It's called power boost overdrive. Missing the PMDG DC6 in MSFS 2024 (she's here, but...).
September 8, 20232 yr 2 hours ago, Claudius_ said: It isn't the same as the ryzens, if you undervolt following the optimization curve you'll have higher clocks and lower temps (but the same risk when under no load). It's called power boost overdrive. Yeah, the Intels, at least the older ones, you do not create a new curve you offset it downwards, which is fine under load as the default curve boosts voltage with increased load quite a bit, but you need to watch for instability at idle.
September 8, 20232 yr 2 hours ago, Glenn Fitzpatrick said: Yeah, the Intels, at least the older ones, you do not create a new curve you offset it downwards, which is fine under load as the default curve boosts voltage with increased load quite a bit, but you need to watch for instability at idle. As far as I remember, it is possible to create a sort of curve with Intel Xtreme tuning*. This would definitely resolve your instability at idle / light loads. Either this or fixed voltage of course. My old 10850K would boost at 4.8 GHz @ default but I managed to stabilize 5.0 GHz under all conditions once set 1.25V fixed voltage... saving a ton of power and heat too, because the beast VID requested 1.45 V at full load before! If anyone wonders why AMD, nVidia and Intel don't do that, it's because the voltage curves set by the factory must ensure stability for 100% of the chip production. In other words, the default values are the optimal ones only if you are very unlucky and end up with a chip produced from a silicone portion with many impurities. *Example of curve found on the Internet. Edited September 8, 20232 yr by MrFuzzy 7800X3D | 2x32 GB DDR5-6000 CL32 | RTX 5080 | Alienware OLED 34" | 1 Gbps fiber
January 13, 20242 yr You're spot on about the need to offset the curve downwards with some older Intel chips to avoid instability at idle. Using Intel Xtreme tuning or a fixed voltage can definitely help maintain stability while saving power and reducing heat.
Archived
This topic is now archived and is closed to further replies.