Need advice on getting some more performance from 5950X

1Wolf

1Wolf

Limp Gawd
Joined
Jul 10, 2007
Messages
433
Built a 5950X/Asus Dark Hero/RTX3090 system on a 360mm AIO about a year and half ago. At the time, the performance provided by precision boost alone was all I needed.

However, these days I'm in Microsoft Flight Sim 2020 in VR with an HP Reverb G2 and need all the frames I can get. I'm CPU limited there. The in-game FPS counter shows that my 3090 has headroom left and is running pretty cool, while my CPU is limiting me.

At this point, when tweaking the in-game settings, even just a few FPS does make a difference.

So I'm investigating seeing if I can get a bit more performance with Ryzen Master, Precision Boost Overdrive, AutoOC, and Curve Optimizer. I don't have any experience with any of these. So, looking through YouTube and such I found some guides. Problem is that, after reviewing some videos, it looks like there is alot of bad information out there.

1) Looking for any help/advice on the best way to get some more performance out of my 5950X. I've read that simply using Curve Optimizer alone is best. I've also read that using Curve Optimizer + AutoOC is best. I've also read that Curve Optimizer + PBO is best....on and on. I don't know what to believe. I also don't know what order to approach them in. Should I take care of the CO optimizations first and then move on to...PBO...or should I move from there to AutoOC? Or should I start with PBO and then move to CO?

2) Looking for a good step-by-step guide from someone easy to follow, and knowledgeable on doing this, and what order to do it in. If you have any recommendations that would be greatly appreciated. I found some really nice, well produced guides on a smaller YouTube channel called the TheGrayingTech that covered CO, PBO, and AutoOC



..but the video wasn't too current (about a year and some old) and it looks like the channel has been largely inactive for some time so I'm not sure whether to trust his guide or whether it might be bad/outdated information leading me in the wrong direction. There is so much bad info out there that could even cause a person to unknowingly reduce the performance of their CPU or even damage it.

Any help, advice, or just pointers or links to get me in the right direction would be greatly appreciated. Thanks!
 
Quick advice if you go down the path of optimizing your core curve, you should check out CoreCycler in order to test and insure the stability of your cores after applying negative offsets.
 
Honestly first what are your specs, Mobo, cooling, PSU, etc...

IMO PBO and Curve optimizer is the best. OCCT is a great testing software. Use AMD Ryzen software to set starting values. These are usually to aggressive. OCCT cpu only test will find these issues quickly. Figure out which core spits out the errors and adjust till you can pass 1 hour on OCCT. Then if you get any other crashes try lowering the suspect core but very small amounts.
 
Draver said:
Quick advice if you go down the path of optimizing your core curve, you should check out CoreCycler in order to test and insure the stability of your cores after applying negative offsets.
Click to expand...

Thanks! I went out and snagged CoreCycler. I haven't had a chance to give it a try yet.

Vengance_01 said:
Honestly first what are your specs, Mobo, cooling, PSU, etc...

IMO PBO and Curve optimizer is the best. OCCT is a great testing software. Use AMD Ryzen software to set starting values. These are usually to aggressive. OCCT cpu only test will find these issues quickly. Figure out which core spits out the errors and adjust till you can pass 1 hour on OCCT. Then if you get any other crashes try lowering the suspect core but very small amounts.
Click to expand...

Thanks Vengance_01.

Specs are as follows:
CPU: AMD Ryzen 9 5950X
CPU COOLER: NZXT Kraken Z73 360mm AIO
MOBO: Asus Crosshair VIII Dark Hero X570
MEM: 64 Gigs. (4x16 GB) G.Skill Trident Z Neo DDR4 3600 16-16-16-36
GPU: EVGA RTX 3090 FTW3 Ultra Gaming
PSU: Seasonic Prime TX-1000 1000W
MAIN DRIVE: WD Black SN850 2TB NVMe M.2 SSD
STORAGE/SECONDARY: Sabrent Rocket Q 4 TB NVMe M.2 SSD
STORAGE: WD Black 10TB 7200 RPM HDD
OS: Windows 10 Pro 64-bit

I'm not familiar with OCCT. I've used Prime95 and Aida64. Is OCCT pretty much the same thing or is that the preferred method folks are using these days?
 
1Wolf said:
Thanks! I went out and snagged CoreCycler. I haven't had a chance to give it a try yet.



Thanks Vengance_01.

Specs are as follows:
CPU: AMD Ryzen 9 5950X
CPU COOLER: NZXT Kraken Z73 360mm AIO
MOBO: Asus Crosshair VIII Dark Hero X570
MEM: 64 Gigs. (4x16 GB) G.Skill Trident Z Neo DDR4 3600 16-16-16-36
GPU: EVGA RTX 3090 FTW3 Ultra Gaming
PSU: Seasonic Prime TX-1000 1000W
MAIN DRIVE: WD Black SN850 2TB NVMe M.2 SSD
STORAGE/SECONDARY: Sabrent Rocket Q 4 TB NVMe M.2 SSD
STORAGE: WD Black 10TB 7200 RPM HDD
OS: Windows 10 Pro 64-bit

I'm not familiar with OCCT. I've used Prime95 and Aida64. Is OCCT pretty much the same thing or is that the preferred method folks are using these days?
Click to expand...
Its like Prime95 with better details and more flexibility on testing

https://www.ocbase.com/download
 
I'm just curious what kinda settings you're using in game and what your performance is currently at. I have a 5950X and a 3080 12GB and it'll help me compare.
 
TheSlySyl said:
I'm just curious what kinda settings you're using in game and what your performance is currently at. I have a 5950X and a 3080 12GB and it'll help me compare.
Click to expand...

I'm pretty new to MSFS 2020 so am still tweaking and fiddling and trying to figure out what FPS feels "ok" and what is just too much of a slideshow to enjoy. That said, I've been working off the following settings:

HP Reverb G2:
Set Render Scale to 100% via OpenXR Tools for WMR in windows. I had motion projection on "Auto". I'm running OpenXR Toolkit in MSFS2020 and have the in-game panel for that set to upscale with FSR at 90% and 60% sharpness. Foveated Rendering Off. Shaking Reduction at 45%. World Scale usually set to 125% or thereabouts for GA aircraft depending on the cockpit.

MSFS2020 settings for VR:
Anti-Aliasing = TAA
Render Scale = 100
Terrain Level of Detail = 150
Off Screen Terrain Pre-Caching = Ultra
Terrain Vector Data = High
Buildings = Medium
Trees = Medium
Grass & Bushes = Medium
Object Level of Detail = 100
Volumetric Clouds = High
Texture Resolution = Ultra
Anisotropic Filtering = 16X
Texture Supersamping = 8x8
Texture Synthesis = High
Water Waves = High
Shadow Maps = 1024
Terrain Shadows = 512
Contact Shadows = Medium
Windshield Effects = High
Ambient Occlusion = Low
Cubemap Reflections = 192
Raymarched Reflections = Medium
Light Shafts = Medium
Bloom = On
Glass Cockpit Refresh Rate = High

As far as the data options go:
All of them are "On" (Photogrammetry is ON).
Running a 32GB Rolling Cache.
Unlimited Bandwidth on a Fiber connection.

Prior to Sim Update #10 I was doing "Ok" with the above.

Performing local flights in a C152 in portions of England outside of London, a C172 in the far west suburbs outside of Chicago (In the KDPA, KDKB area), and a Bonanza G36 around Telluride, CO....my FPS was high twenties and low thirties. Sometimes mid thirties depending on where I was looking.

After Sim Update #10 I was trying a short cross country in England from Shorestree to Biggin Hill, and then from Biggin Hill to Shoreham and I noticed my FPS was struggling more.

Then I tried my first flight in a more urban area. I was on the ground in a DA40NG at Seattle's Renton Municipal airport (KRNT) and my FPS was in the high teens and low twenties. Taxiing was awful. After takeoff in the Seattle area, it got to maybe the mid twenties and that was about it. Motion Reprojection made it worse. I tried Foveated Rendering and that helped...a little.

So I turned the Shadow maps down to 768, the Terrain Shadows down to 256, Light Shafts to Low, and Terrain Level of Detail to 135...then 125....then down to 100.

At that point I was in the low twenties on the ground, and the mid-twenties in the air. So thats when I started using the in-game FPS counter and saw I was terribly limited by my CPU and my GPU was sitting there twiddling its thumbs with temps in the high 60's.

I'd been following alot of great guides from some of the MSFS2020 YouTube channels online (VR Flight Sim Guy, Sim Hangar, Pie in the Sky Tours) as well as alot of tweaking info on bang-for-buck for the different graphics options but thats when I decided to go down the rabbit hole of seeing if I could squeeze some more out of my CPU safely. I'd been running the past year with PBO off, no OC, no Curve Optimizer...just at stock....because it was doing more than I needed it to. I also play DCS and for DCS my CPU frametimes were fine....and it was my GPU frame times that were hurting.

So anyhow, I hope that helps give you a frame of comparison.
 
Vengance_01 said:
Honestly first what are your specs, Mobo, cooling, PSU, etc...

IMO PBO and Curve optimizer is the best. OCCT is a great testing software. Use AMD Ryzen software to set starting values. These are usually to aggressive. OCCT cpu only test will find these issues quickly. Figure out which core spits out the errors and adjust till you can pass 1 hour on OCCT. Then if you get any other crashes try lowering the suspect core but very small amounts.
Click to expand...

Vengance_01

Just wondering exactly which type of CPU test and exactly which options you set in OCCT? I'm new to it and just watched a few videos and did some reading on the options available, but wanted to know which options to set to follow your advice and exactly what a 1-hour test consists of and how you set it up in the test options/advanced options.
 
1Wolf said:
I'm pretty new to MSFS 2020 so am still tweaking and fiddling and trying to figure out what FPS feels "ok" and what is just too much of a slideshow to enjoy. That said, I've been working off the following settings:

HP Reverb G2:
Set Render Scale to 100% via OpenXR Tools for WMR in windows. I had motion projection on "Auto". I'm running OpenXR Toolkit in MSFS2020 and have the in-game panel for that set to upscale with FSR at 90% and 60% sharpness. Foveated Rendering Off. Shaking Reduction at 45%. World Scale usually set to 125% or thereabouts for GA aircraft depending on the cockpit.

MSFS2020 settings for VR:
Anti-Aliasing = TAA
Render Scale = 100
Terrain Level of Detail = 150
Off Screen Terrain Pre-Caching = Ultra
Terrain Vector Data = High
Buildings = Medium
Trees = Medium
Grass & Bushes = Medium
Object Level of Detail = 100
Volumetric Clouds = High
Texture Resolution = Ultra
Anisotropic Filtering = 16X
Texture Supersamping = 8x8
Texture Synthesis = High
Water Waves = High
Shadow Maps = 1024
Terrain Shadows = 512
Contact Shadows = Medium
Windshield Effects = High
Ambient Occlusion = Low
Cubemap Reflections = 192
Raymarched Reflections = Medium
Light Shafts = Medium
Bloom = On
Glass Cockpit Refresh Rate = High

As far as the data options go:
All of them are "On" (Photogrammetry is ON).
Running a 32GB Rolling Cache.
Unlimited Bandwidth on a Fiber connection.

Prior to Sim Update #10 I was doing "Ok" with the above.

Performing local flights in a C152 in portions of England outside of London, a C172 in the far west suburbs outside of Chicago (In the KDPA, KDKB area), and a Bonanza G36 around Telluride, CO....my FPS was high twenties and low thirties. Sometimes mid thirties depending on where I was looking.

After Sim Update #10 I was trying a short cross country in England from Shorestree to Biggin Hill, and then from Biggin Hill to Shoreham and I noticed my FPS was struggling more.

Then I tried my first flight in a more urban area. I was on the ground in a DA40NG at Seattle's Renton Municipal airport (KRNT) and my FPS was in the high teens and low twenties. Taxiing was awful. After takeoff in the Seattle area, it got to maybe the mid twenties and that was about it. Motion Reprojection made it worse. I tried Foveated Rendering and that helped...a little.

So I turned the Shadow maps down to 768, the Terrain Shadows down to 256, Light Shafts to Low, and Terrain Level of Detail to 135...then 125....then down to 100.

At that point I was in the low twenties on the ground, and the mid-twenties in the air. So thats when I started using the in-game FPS counter and saw I was terribly limited by my CPU and my GPU was sitting there twiddling its thumbs with temps in the high 60's.

I'd been following alot of great guides from some of the MSFS2020 YouTube channels online (VR Flight Sim Guy, Sim Hangar, Pie in the Sky Tours) as well as alot of tweaking info on bang-for-buck for the different graphics options but thats when I decided to go down the rabbit hole of seeing if I could squeeze some more out of my CPU safely. I'd been running the past year with PBO off, no OC, no Curve Optimizer...just at stock....because it was doing more than I needed it to. I also play DCS and for DCS my CPU frametimes were fine....and it was my GPU frame times that were hurting.

So anyhow, I hope that helps give you a frame of comparison.
Click to expand...
I've been busy as hell lately, but I'll check out my own settings when I have time. I don't have a VR headset so I sadly can't see how much that'll constrain things.
 
1Wolf said:
Vengance_01

Just wondering exactly which type of CPU test and exactly which options you set in OCCT? I'm new to it and just watched a few videos and did some reading on the options available, but wanted to know which options to set to follow your advice and exactly what a 1-hour test consists of and how you set it up in the test options/advanced options.
Click to expand...
Think of it this way...

Use AVX2 for Instruction set.
Data Set: Small = all CPU (most heat) Medium = some ram and CPU and Large = Most ram and less CPU Heat
Thread = Advance and make sure you select all cores and threads
Mode I choose Normal when finding stable clocks and then when I think I have it do Extreme option
Load Type = Steady is best

Hopes this helps
 
Vengance_01 said:
Think of it this way...

Use AVX2 for Instruction set.
Data Set: Small = all CPU (most heat) Medium = some ram and CPU and Large = Most ram and less CPU Heat
Thread = Advance and make sure you select all cores and threads
Mode I choose Normal when finding stable clocks and then when I think I have it do Extreme option
Load Type = Steady is best

Hopes this helps
Click to expand...

Thanks Vengance_01

Got it. Just a couple follow ups and then I think I'm set to give it a go!

What duration do you set the test for? 1 Hour total? Or 1 Hour per Core (16 Hours on a 5950X)?

Are you selecting all physical cores in the advanced thread box and then testing them all at once? Or are you selecting "Core Cycle" and then specifying to cycle the active core every X seconds? If the later, how many seconds do you set it to test each core before it cycles?

The above two questions aside, just to read back the settings for the test:
Duration: 1 Hour?
Data Set: Small
Mode: Normal (and then later Extreme once I think I have a stable clock)
Load Type: Steady
Start At Cycle: 1
Instruction Set: AVX2
Threads: Advanced

Then in the "Advanced Thread Settings"
Physical Cores: Select all of them
Virtual Cores: Physical and Virtual
Core Cycle: Disabled or Cycle active core every X seconds???
Swap Active/Inactive Cores: Disabled.
 
1Wolf said:
Thanks Vengance_01

Got it. Just a couple follow ups and then I think I'm set to give it a go!

What duration do you set the test for? 1 Hour total? Or 1 Hour per Core (16 Hours on a 5950X)?

Are you selecting all physical cores in the advanced thread box and then testing them all at once? Or are you selecting "Core Cycle" and then specifying to cycle the active core every X seconds? If the later, how many seconds do you set it to test each core before it cycles?

The above two questions aside, just to read back the settings for the test:
Duration: 1 Hour?
Data Set: Small
Mode: Normal (and then later Extreme once I think I have a stable clock)
Load Type: Steady
Start At Cycle: 1
Instruction Set: AVX2
Threads: Advanced

Then in the "Advanced Thread Settings"
Physical Cores: Select all of them
Virtual Cores: Physical and Virtual
Core Cycle: Disabled or Cycle active core every X seconds???
Swap Active/Inactive Cores: Disabled.
Click to expand...
Disable core cycle and Duration is total time as all cores are used at once. 1H is a good baseline for testing stability :)
 
This in an interesting thread...

I too have a 5950x and hp reverb. Howver, I have just done the reverb stock but used the google earth imagery. I've been using a 3080 and have not noticed any problems. However I haven't run an update/playes in the past couple weeks...
 
You must log in or register to reply here.
Back
Top