Jump to content

Recommended Posts

Posted

OK, I have just done a few test, but the results don't make sense to me.  Looking for conformation that they are correct, and some explanation, or what I might have wrong.

First test indicated little improvement in fps, but some in consistency.  Second test indicated about a 10 - 12 fps increase (from 30 - 39 fps to 42 - 51 fps).  Does MT require/use "learning"?

 

What doesn't make sense to me is that in both cases it says I am GPU bound.  My system is a i7 8700K (6 cores/12 threads), 3080 GPU (10 GB VRAM), 64 GB 3600 DDR$, HP Reverb G2 and DCS on a SSD.

With a 3080 GPU and a i7 8700K I don't see how this would be GPPU bound.

What am I missing?

----------------

AKA_Clutter

 

Win 10 Pro, Intel i7 12700k @4.6 GHz, EVGA RTX 3080  FTW, Ultra 64 GB G.Skill DDR4 3600 RAM, Acer 27" flat screen, HP Reverb G2, TM Warthog HOTAS with Virpil warBRD base, MFG Rudder Pedals, Virpil TCS Rotor Base with AH-64Dcollective, TrackIR 5 Pro w/Vector Expansion, PointCTRL.

Posted

Any idea why?  I thought the VRAM might be a thing, but with you having a 4090, that isn't it.

----------------

AKA_Clutter

 

Win 10 Pro, Intel i7 12700k @4.6 GHz, EVGA RTX 3080  FTW, Ultra 64 GB G.Skill DDR4 3600 RAM, Acer 27" flat screen, HP Reverb G2, TM Warthog HOTAS with Virpil warBRD base, MFG Rudder Pedals, Virpil TCS Rotor Base with AH-64Dcollective, TrackIR 5 Pro w/Vector Expansion, PointCTRL.

Posted
2 minutes ago, AKA_Clutter said:

Any idea why?  I thought the VRAM might be a thing, but with you having a 4090, that isn't it.

It's enough to use lower graphics settings

 

Posted (edited)
1 hour ago, AKA_Clutter said:

Any idea why?  I thought the VRAM might be a thing, but with you having a 4090, that isn't it.

 You will literally always have a bottleneck. That's not a ''problem'' or something to be fixed, there's always some part of your PC that's ''in the way'' of the others. The question is ''how much''. If you're trying to jam a typhoon through a 2'' pipe, that's a problem, and that was the issue we had on CPU side. We had extremely powerful processors that were being choked down to two cores, and it was causing a back up in other areas. Now that those areas are freed up, your primary restriction is the GPU itself, which is the ideal scenario.

Again, there is no such thing as ''no bottleneck''. A G2 is capable of extremely high resolutions (several times 2d 4k in pixel count) and a 3080 cannot handle that many pixels easily, meaning it's going all out, which is normal and to be expected.

Edited by Mars Exulte
  • Like 1
  • Thanks 1

Де вороги, знайдуться козаки їх перемогти.

5800x3d * 3090 * 64gb * Reverb G2

Posted
10 minutes ago, Mars Exulte said:

 You will literally always have a bottleneck. That's not a ''problem'' or something to be fixed, there's always some part of your PC that's ''in the way'' of the others. The question is ''how much''. If you're trying to jam a typhoon through a 2'' pipe, that's a problem, and that was the issue we had on CPU side. We had extremely powerful processors that were being choked down to two cores, and it was causing a back up in other areas. Now that those areas are freed up, your primary restriction is the GPU itself, which is the ideal scenario.

Again, there is no such thing as ''no bottleneck''. A G2 is capable of extremely high resolutions (several times 2d 4k in pixel count) and a 3080 cannot handle that many pixels easily, meaning it's going all out, which is normal and to be expected.

 

 

Yeah, I'm finally pinning my 2080Ti, which used to be difficult to do as the CPU rarely let it get there.

I see the GPU @ 100% as an extremely positive thing.  Your GPU is finally getting to use it's full performance, whereas, to this point, the CPU has been the limiter.

Rex's Rig

Intel i9-14900K | Nvidia RTX 4090 | 64GB DDR5 | 3x4TB 990 Pro M2 SSDs | HP Reverb 2 | 49" Samsung 5120x1440 @ 120Mhz

TM Warthog Stick + Throttle | TM Pendulum Pedals | MS Sidewinder 2 FFB | Track IR |  Cougar MFD x 2 

 

Posted (edited)

I do believe DLSS will be more important for 3000 and 2000 series owners than MT.

Not everyone can afford a 4000 series card.

I run 3080 + 13600k, and no extra FPS. I might see a bit more clarity now though.

Edited by KCOOL
  • Like 1
Posted

yea did what someone stated and run from the MT folder and the shader box is gone. 

"any failure you meet, is never a defeat; merely a set up for a greater come back",  W Forbes.

"Success is not final, failure is not fatal, it is the courage to continue that counts",
"He who never changes his mind, never changes anything," Winston Churchill.

MSI z690 MPG DDR4 || i9-14900k|| ddr4-64gb PC3200 |zotac RTX 5080|Game max 1300w|Win11| |turtle beach elite pro 5.1|| ViRpiL,T50cm2||MFG Crosswinds|| VT50CM-plus rotor Throttle || Z10 RGB EVGA Keyboard/ G502LogiMouse || PiMax Crystal VR || 32 Asus||

Posted

All good information.  Always learning.

I know that either the CPU or GPU will be the bottleneck, one has to be.  I guess what surprised me the first run was no FPS increase, and with the GPU at the max, I don't think I will gain anything on the setting side.  Time will tell if I did actually gain anything in fps.

And as someone said, not all can afford a new GPU.   I am (was?) thinking of upgrading my system to a 13 gen CPU/Mb/ram but gains are uncertain.

Again, thanks for all the good information.

 

----------------

AKA_Clutter

 

Win 10 Pro, Intel i7 12700k @4.6 GHz, EVGA RTX 3080  FTW, Ultra 64 GB G.Skill DDR4 3600 RAM, Acer 27" flat screen, HP Reverb G2, TM Warthog HOTAS with Virpil warBRD base, MFG Rudder Pedals, Virpil TCS Rotor Base with AH-64Dcollective, TrackIR 5 Pro w/Vector Expansion, PointCTRL.

Posted
6 hours ago, AKA_Clutter said:

All good information.  Always learning.

I know that either the CPU or GPU will be the bottleneck, one has to be.  I guess what surprised me the first run was no FPS increase, and with the GPU at the max, I don't think I will gain anything on the setting side.  Time will tell if I did actually gain anything in fps.

And as someone said, not all can afford a new GPU.   I am (was?) thinking of upgrading my system to a 13 gen CPU/Mb/ram but gains are uncertain.

Again, thanks for all the good information.

 

I can also confirm that I am now GPU bound (with a 4090) while my CPU (7950x3D) is performing in the 3-4ms range in initial testing in low level Persian Gulf flights. So it's a great update overall but still sucks for the golden goose that is 90fps without any settings concessions. 

PC: ASUS TUF 4090oc - Ryzen 7950X3D - 32gb DDR5 6000 - Quest Pro

Sims: DCS, IL2, MSFS

Pilot Skill: Drunk guy from Independence Day

RIO Skill: Goose (post neck-break) 

Posted (edited)

I see that people with higher end can add their info, and see about CPU or GPU bound. I just didn't have enough info to give any answers. Also seeing what others can get with the 2080-ti. I am doing better with MT, with my Intel 7700k (4.2 GHZ cpu, currently running at 4.5 GHZ CPU). But old Asus Z170 motherboard. The VR was getting killed, now it is at least consistent 20 + FPS and even more sometimes. Depending on the map and aircraft. It was easily getting half of those FPS numbers. I just say getting more than 5 FPS gain. But it is more consistent FPS.

Edited by Ramstein

ASUS Strix Z790-H, i9-13900, WartHog HOTAS and MFG Crosswind

G.Skill 64 GB Ram, 2TB SSD

EVGA Nvidia RTX 2080-TI (trying to hang on for a bit longer)

55" Sony OLED TV, Oculus VR

 

Posted
12 hours ago, AKA_Clutter said:

My system is a i7 8700K (6 cores/12 threads), 3080 GPU (10 GB VRAM), 64 GB 3600 DDR$, HP Reverb G2 and DCS on a SSD.

If you did not turn of hyperthreading yet, turn it off, it should result in more efficient CPU usage and CPU might be able to feed the 3070 better.

I had same issue, and got the same advise from another user. In addition I can confirm that MT + HT do not produce better results than only MT.

 

Posted
2 hours ago, okopanja said:

If you did not turn of hyperthreading yet, turn it off, it should result in more efficient CPU usage and CPU might be able to feed the 3070 better.

I had same issue, and got the same advise from another user. In addition I can confirm that MT + HT do not produce better results than only MT.

 

That seems strange, but worth a try.  Where did you turn off hyperthreading?  BIOS or somewhere else?

 

----------------

AKA_Clutter

 

Win 10 Pro, Intel i7 12700k @4.6 GHz, EVGA RTX 3080  FTW, Ultra 64 GB G.Skill DDR4 3600 RAM, Acer 27" flat screen, HP Reverb G2, TM Warthog HOTAS with Virpil warBRD base, MFG Rudder Pedals, Virpil TCS Rotor Base with AH-64Dcollective, TrackIR 5 Pro w/Vector Expansion, PointCTRL.

Posted
7 minutes ago, AKA_Clutter said:

That seems strange, but worth a try.  Where did you turn off hyperthreading?  BIOS or somewhere else?

 

 

I got little better results with HT on running MT.

Don B

EVGA Z390 Dark MB | i9 9900k CPU @ 5.1 GHz | Gigabyte 4090 OC | 64 GB Corsair Vengeance 3200 MHz CL16 | Corsair H150i Pro Cooler |Virpil CM3 Stick w/ Alpha Prime Grip 200mm ext| Virpil CM3 Throttle | VPC Rotor TCS Base w/ Alpha-L Grip| Point Control V2|Varjo Aero|

Posted

LOL  Like most things in life ... ... "Your mileage may vary!"

----------------

AKA_Clutter

 

Win 10 Pro, Intel i7 12700k @4.6 GHz, EVGA RTX 3080  FTW, Ultra 64 GB G.Skill DDR4 3600 RAM, Acer 27" flat screen, HP Reverb G2, TM Warthog HOTAS with Virpil warBRD base, MFG Rudder Pedals, Virpil TCS Rotor Base with AH-64Dcollective, TrackIR 5 Pro w/Vector Expansion, PointCTRL.

Posted (edited)
1 hour ago, AKA_Clutter said:

That seems strange, but worth a try.  Where did you turn off hyperthreading?  BIOS or somewhere else?

 

BIOS. Check your PC documentation, but typically F5, F8, F10 or F12 during early startup. Generally, DCS is CPU intensive, hence hyperthreading will usually cause worse performance.

There should be setting under Performance.

Edited by okopanja
  • 2 weeks later...
Posted (edited)
On 3/11/2023 at 4:54 AM, okopanja said:

If you did not turn of hyperthreading yet, turn it off, it should result in more efficient CPU usage and CPU might be able to feed the 3070 better.

I had same issue, and got the same advise from another user. In addition I can confirm that MT + HT do not produce better results than only MT.

 

 

Per DCS, it seems to depend on how many cores/threads you have.

From the MT FAQ: (emphasis mine)

Quote

 

* If you have 32 or less cores, please ensure that HT/SMT is enabled. If you 33 or more cores, please disable this.

* Currently the rendering pool scales up to 16 threads on P-cores (1 core = 2 threads). The other P-cores are used by the main logic thread, sound engine threads, and the auxiliary thread pool that occupy all available space. E-cores are used only by the resources streaming pool that has no limits (1 core = 1 thread).

 

I imagine most people have fewer than 33 cores, and thus should leave HT on?

It looks like the logic threads will use all available cores, but the rendering threads will use a max of 8 P-Cores if Hyper-threaded.

At least in theory, I supposed everyone's MMV.

 

Edited by Rex

Rex's Rig

Intel i9-14900K | Nvidia RTX 4090 | 64GB DDR5 | 3x4TB 990 Pro M2 SSDs | HP Reverb 2 | 49" Samsung 5120x1440 @ 120Mhz

TM Warthog Stick + Throttle | TM Pendulum Pedals | MS Sidewinder 2 FFB | Track IR |  Cougar MFD x 2 

 

Posted (edited)
On 3/11/2023 at 8:21 AM, okopanja said:

BIOS. Check your PC documentation, but typically F5, F8, F10 or F12 during early startup. Generally, DCS is CPU intensive, hence hyperthreading will usually cause worse performance.

There should be setting under Performance.

 

I don't think this is true.  At least not always.

Highly-CPU-intensive tasks usually benefit from SMT to some degree.

Handbrake is one of the most CPU-intensive tasks you can run, and even it benefits from SMT a little bit. Some CPU-bottlenecked apps run worse, but most run a bit better. I may be off, but if I had to take a complete stab in the dark, the DCS render threads are probably somewhere around Cinebench or Blender, with the logic cores probably in the Geekbench neighborhood.

I think DCS has to pull from RAM quite a bit, and it needs to talk to the GPU, network cards, disk drive, etc.  This would yield enough idle cycles to make SMT beneficial.

 

hyperthreading.png

Edited by Rex

Rex's Rig

Intel i9-14900K | Nvidia RTX 4090 | 64GB DDR5 | 3x4TB 990 Pro M2 SSDs | HP Reverb 2 | 49" Samsung 5120x1440 @ 120Mhz

TM Warthog Stick + Throttle | TM Pendulum Pedals | MS Sidewinder 2 FFB | Track IR |  Cougar MFD x 2 

 

Posted (edited)

With that being said, I found another set of benchmarks which would support your conclusions in some gaming titles:

 

http://www.redgamingtech.com/does-intel-hyper-threading-hurt-gaming-performance-i9-9900k-analysis/

 

BF5 is about 5 frames faster across the board with HT disabled, and with DX11 it is massively faster with Max settings ... 163FPS with HT off vs 131 with it on. 🤯

Edited by Rex

Rex's Rig

Intel i9-14900K | Nvidia RTX 4090 | 64GB DDR5 | 3x4TB 990 Pro M2 SSDs | HP Reverb 2 | 49" Samsung 5120x1440 @ 120Mhz

TM Warthog Stick + Throttle | TM Pendulum Pedals | MS Sidewinder 2 FFB | Track IR |  Cougar MFD x 2 

 

  • 3 months later...
Posted

I could not longer fly with my system, I ordered a new Motherboard, cpu and ram,  and next week it should be running.

new:

Motherboard: Asus Z790 , CPU: Intel 13900F , and RAM: G.Skill DDR5 ram. For now I am going to see my Nvidia RTX 2080-ti video will be enough...

 

ASUS Strix Z790-H, i9-13900, WartHog HOTAS and MFG Crosswind

G.Skill 64 GB Ram, 2TB SSD

EVGA Nvidia RTX 2080-TI (trying to hang on for a bit longer)

55" Sony OLED TV, Oculus VR

 

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...