Jump to content

edmuss

Members
 View Profile  See their activity

  • Posts

    1740

  • Joined

  • Last visited

 Content Type 

Everything posted by edmuss

  1. edmuss
    Slightly off topic, but why on earth are people with 4090s getting such low min fps on heaven? I get 51fps minimum with my 3080ti, granted my average is only about 285 and high about 590. For sharpening I would recommend using CAS from the openxr toolkit, it doesn't upscale at all but the sharpening is really nice; not sure how it might compare to 3Dmigoto performance wise though.
  2. edmuss
    Weirdo I hang mine up on a hook, the cable is already coiled and cleated to the wall with enough free slack to reach both my head and the hook.
  3. edmuss
    I've just finished running some back to back testing specifically on the three terrain shadows settings for corroboration. Rolled back to 2.7.18.30765 and recorded a short track flying over Beirut in the A10. Replayed the track with each terrain shadow setting and all others as per my normal config. Updated to 2.8.0.33006 and repeated the three tests. There is some performance lost between OB's which is to be expected with the new lighting. The main issue is with the flat shadows now incurring a much larger performance penalty than they did, 1ms of lost frametime, presumably due to the soft edges? @BIGNEWY are the soft edges intentional on the current flat shadows or are they meant to be hard as per 2.7? My personal preference would be to retain the option to use the same hard edged shadows as 2.7 or perhaps be able to seperate the default/flat/hard/soft into different options?
  4. edmuss
    I haven't had a chance to clean out with DDU and retest yet, been struck down with flu for the last couple three days so not even booted the computer for anything meaningful. Will hopefully be able to corroborate in the very near future
  5. edmuss
    Mine starts to throttle above 76°C, only by 10-15mhz but by the time it's got to 90°C it's wound itself back to about 4ghz. Obviously the cooler you keep it the more consistently it'll keep the boost clocks up. In DCS mine sits at 55-65°C.
  6. edmuss
    I have approximately 2 foot from the front of my shoulder to the closest point of the desk and that's enough gap to be able to reach all of the forward controls on the A10 and KA50. I'm unfortunately constrained on the left hand side (to the outside and rear of the A10 throttle and some of the centre console/overheads of the Huey) simply because I'm flying in a cupboard Similarly the rear right panel on the KA50 can be difficult to get. Everything in the forward arc is dead easy though.
  7. edmuss
    If I remember correctly, the 5800x3D is a single chiplet whereas the 5900x is a double chiplet, as such the heat density on the x3D is much higher; add the extra layer of cache on there as well and you'll get far more heat building.
  8. edmuss
    MSFS is simply using more cores so it will generate more heat. Use PBO2 Tuner to undervolt the cpu and the temps should drop significantly. Mine sits at 55°C or there about in DCS and after a 10 minute cinebench R23 run it gets up to 75°, that is with all 8 cores pegged at 4.45Ghz @ 100% load. Using an old thermalright truespirit 140 that I made an AM4 bracket for.
  9. edmuss
    In the latest open beta (2.8), ED moved the setting for the laser pointers to the VR tab, I think perhaps the same functionality is shared to the VR controllers? I don't think you need to copy the leapC.dll any more, I haven't done so for months now and it's all worked fine. There is still some flakiness with the leap not initialising when DCS starts or the laser pointer settings being retained, frustrating that you need to reset it each time but not the end of the world!
  10. edmuss
    Another issue on the VRAM front is memory bandwidth, this is comprised of the VRAM speed and the bus width in bits. Whilst 8gb is actually plenty to fly DCS in VR (it isn't too bad at handling it's VRAM, particularly now @mbucchia fixed the reprojection tanking bug with the toolkit) those cards with the lower amounts tend to also have lower bandwidth. This results in an overall GPU slowdown because it's waiting for the textures to copy. The 3070 has 8gb 14gbps VRAM with a 256bit bus and 448GB/s bandwidth. The 3080ti has 12gb 19gbps VRAM with a 384bit bus and 912GB/S bandwidth. The 3070 will run high textures all day long as but it will get to saturation much faster and start to slow down. The 4080 actually has only has a 256bit bus but the VRAM is running at 22gbps, resulting bandwidth is 716GB/S which is less than the 3080ti. In theory in DCS vr where there is a huge amount of memory bandwidth required it may actually be slower at the very high resolution, high memory usage scenarios!
  11. edmuss
    It's on the top surface, but set back a few mm from the front face so the edge of the headset occludes the hands. I use the hands use stick option enabled because it saves needing to set a deadzone in the axis. Also as I mostly fly helos I'm developing a mechanical force trim system that will hold the stick in one place, without the hands use stick option this would permanently disable the right hand.
  12. edmuss
    The trick I've found with leap motion is to have the sensor positioned such that it occludes your hands when your head up and they only appear when you're looking down at controls. The laser pointers on the leap implementation is now a LOT better than it ever has been, but to me there is no better realism than turning off the pointers and reaching out to hit the control that you want. Obviously limitations aside, it's getting better all the time. I got my leap for 35 quid on eBay and it's the best value immersion booster I have
  13. edmuss
    Freely available from scan.... But then he'll not get his profit margin on the CPU.
  14. edmuss
    Well all us Europeans can whinge equally then Well 650 euro is 560 quid so yes, we're being fleeced
  15. edmuss
    Meanwhile we get the bargain brexit price of £682.80......
  16. edmuss
    No I didn't, that could be linked to it, is it possible to save the profile out and re-import it to save having to set it all again?
  17. edmuss
    Have recorded a track from the Syria A10C free flight and run through both drivers: - 526.98 Average appCPU 6.64ms Average appGPU 11.21ms 512.95 Average appCPU 7.16ms Average appGPU 11.41ms Both with head held still staring through the hud, the track was recorded on the 526.98 driver, not sure if that could contribute to the different appCPU numbers? The track comprised of around 2700 data points and included low level city flying, rolling hills, trees and some light clouds. edit: I am seeing the same reduction in appCPU with the newer driver, I don't see any appGPU change though. Could it be linked to settings? Did you have any terrain shadows on for your tests? I didn't.
  18. edmuss
    I shall have a test later on tonight in general flight and see if there is any improvement with the older driver, a 1ms jump should be pretty easy to spot.
  19. edmuss
    As an alternative quick and dirty test, I've just run both 512.95 and 526.98 back to back in both 2D and VR; I didn't reset any profiles. KA50 instant acces Haifa to Megiddo tour (ramp hot start on Syria), no touching controls, look forwards, then left, then right; granted certainly not a sample that incorporates a large range of different graphical variations. In 2D it pegged at 180fps (5.5ms frametime) and in VR it pegged at 14.1ms (71fps), the same result with both drivers to within 0.1ms appGPU.
  20. edmuss
    It's a very handy resource to have on tap, for the cost of a few ciders I can get stuff made For the simpler turning and drilling jobs I can just jump on a lathe on a lunchtime. I will likely be manufacturing the structural parts from aluminium and stainless and use the printer for the odd shape bits, who knows how well it will work though!
  21. edmuss
    As I have free access to solidworks and a 3D printer (or I will by the time I design it), along with a tame CNC man I can probably design a specific adjustable tension clamp. Good heads up though!
  22. edmuss
    Turbo mode will never improve your appGPU frametime. To the best of my understanding, when below refresh rate, WMR will force the the FPS to a lower value (relative to appGPU) in certain circumstances. For example you might be getting 17ms appGPU which equates to 58fps, yet the FPS counter could be reading 45fps (22ms appGPU). Sometimes it will manifest, sometimes it won't.
  23. edmuss
    Love this I'm looking to start designing my collective soon, mind if I steal your brake design? I had been racking my brains about how to get a cable pulled brake to work and never considered a sprung clamp; too set on how a bike disk brake works to consider inverting the caliper!
  24. edmuss
    This was my findings aswell, I get better image quality and equal performance by reducing resolution to 80% than FSR upscaling 80% up to 100%. There is also a computational GPU overhead associated with upscaling so there is a point of dimishing returns, on my machine it takes about 0.8-1ms to upscale at 0.77; as your framerates get higher, this 1ms hit equates in more cost. 11ms = 90fps vs 12ms = 83fps is a 7fps delta 19ms = 52fps vs 20ms = 50 fps is a 2fps delta I'm not sure if the computational load increases if you increase the amount of upscaling (by reducing the value), it's not something that I've dedicated much time to.
  25. edmuss
    The turbo mode will do nothing to improve frametimes, only to get the FPS to match the frametimes in certain circumstances when the WMR runtime throttles the frames to reduce latency. Openxr toolit does the same FSR upscaling as openvr_FSR so you should see similar results with equal settings. With FSR I can get about 3-5 ms reduction which could put me very close to 90fps all the time, however I don't care for the aliasing introduced by the upscaling so prefer to run without it and get 65-80 fps reliably @60hz headset refresh rate.
×
×
  • Create New...