Hyperthreading off works great!

[mobiel 70]

My CPU: I7-7700k and I use P3D without HT and it works better than with HT set to on.

 

[mobiel 70]

Sorry guys I'm not sure now.

Yesterday I got with FSLABS A320 in EGLL 12 FPS.

Now, I will try to get best results: 

My System I7-7700 k (OC 4,7 GHz) and GTX 1070

Should I try HT on or off and which AM?

Edited February 3, 2019 by mobiel

[Rocky_53 1,579]

56 minutes ago, mobiel said:

Sorry guys I'm not sure now.

Yesterday I got with FSLABS A320 in EGLL 12 FPS.

Now, I will try to get best results: 

My System I7-7700 k (OC 4,7 GHz) and GTX 1070

Should I try HT on or off and which AM?

Why not turn it off and see! 😊

Edited February 3, 2019 by Rockliffe

[gsand 105]

35 minutes ago, Rockliffe said:

Why not turn it off and see! 😊

Agree with Howard on this, he convinced me, turned HT off on my 8700K and it's done a great job for me. Still get the occasional micro stutter when a big load of AI hits but it's minor and no way to avoid that around large airports that I know of.

Edited February 3, 2019 by gsand

[mobiel 70]

So I tested the following Situation:

EDDF Gate A16 FSLabs A320

HT ON:

AM 254

AM 255

AM 116

AM 85

AM 84 

Everytime I got 17-21 FPS

 

HT OFF:

AM 84

AM 85

AM 116

AM 254

AM 255

Also 17-21 FPS

[gsand 105]

Check your CPU heat, I got a substantially better thermal profile with HT off and same or better performance, rarely go above 70C now.

[GSalden 4,656]

22 minutes ago, mobiel said:

So I tested the following Situation:

EDDF Gate A16 FSLabs A320

HT ON:

AM 254

AM 255

AM 116

AM 85

AM 84 

Everytime I got 17-21 FPS

 

HT OFF:

AM 84

AM 85

AM 116

AM 254

AM 255

Also 17-21 FPS

Did you really test HT Off values with HT on and the other way round ?

[mobiel 70]

25 minutes ago, gsand said:

Check your CPU heat, I got a substantially better thermal profile with HT off and same or better performance, rarely go above 70C now.

My temperature is between 55-66 C

 

5 minutes ago, GSalden said:

Did you really test HT Off values with HT on and the other way round ?

Yes, same time, same airport, same position, same aircraft, same weather. And same AM

[GSalden 4,656]

14 minutes ago, mobiel said:

Yes, same time, same airport, same position, same aircraft, same weather. And same AM

Then you did not test it correctly.

AM values for HT on are different from AM values with HT off...

[mobiel 70]

Okay, so which values are right?

[Rocky_53 1,579]

2 hours ago, mobiel said:

So I tested the following Situation:

EDDF Gate A16 FSLabs A320

HT ON:

AM 254

AM 255

AM 116

AM 85

AM 84 

Everytime I got 17-21 FPS

 

HT OFF:

AM 84

AM 85

AM 116

AM 254

AM 255

Also 17-21 FPS

In my experience, you don't need an AM setting with HT off! (Ref LM) As for the FPS, HT isn't about FPS, it's about smoothness and texture loading.

Edited February 3, 2019 by Rockliffe

[SteveW 1,792]

The way to go about all this is simple:

HT disabled first

With HT disabled we find the number of cores that loads our major scenario the quickest. And I don't mean we will then fly to a more intense scenery package, we will use the worst case for the test that we expect to be flying usually to/from.

For example on an 8 core we compare the scenario loading time with 11110000, 11111000,11111100,11111110...and so on, 4,5,6,7 cores...and note the times.

From data around the world I have been receiving, we see around 5-7 cores are the maximum we can accelerate the scene building time with. We might only gain one or a half second from there. Since after that more cores simply thin-out the resource (the PC platform has limits) and other cores start to wait - reversing the process.

 

Since we know that HT enabled will use many more cycles that are otherwise wasted (otherwise it wouldn't exists), we need that HT to support the creaking platform we are trying to support too many cores with.

So we for example find that 11111100 gives us the loading performance, and 11111110 only provides maybe 1 second improvement, we don't need it.

The sim has a rendering system that relies on single core performance, and a scene building system that relies on paralleled operations to feed it.

So taking 11111100 we can now do 01,01,01,01,01,01,00,00.

The sim has no idea of the change and the performance should be not just identical, but better.

But at this point that's going to present problems possibly with other software running alongside the sim, but not for the sim itself.

If we enable HT and see a poor result, it is an indication that the system settings and or software that are at fault, not HT.

It may be simpler to turn off HT and avoid any problems. However, it still makes sense to only use the number of cores needed, else the sim will push out the very processes that it relies on for fast operation. That means the AM is still important. Maybe not if we don't have many cores.

 

 

 

 

 

[GSalden 4,656]

3 hours ago, Rockliffe said:

In my experience, you don't need an AM setting with HT off! (Ref LM) As for the FPS, HT isn't about FPS, it's about smoothness and texture loading.

Here I use an AM with HT OFF. But that is because I have a 14 core processor ( 7940X ).

 

[Dirk98 354]

5 hours ago, GSalden said:

Here I use an AM with HT OFF. But that is because I have a 14 core processor ( 7940X ).

 

and that is (use of AM) to exclude some number of unnecessary cores that would otherwise not provide any tangible loading improvement over previous tests (maybe just 1 second plus) and would "thin-out the resource" of the CPU instead.  😄 

My 1c.

Edited February 3, 2019 by Dirk98

[GSalden 4,656]

6 hours ago, Dirk98 said:

and that is (use of AM) to exclude some number of unnecessary cores that would otherwise not provide any tangible loading improvement over previous tests (maybe just 1 second plus) and would "thin-out the resource" of the CPU instead.  😄 

My 1c.

With just 1 view display I would have used 6 of the 14 cores for P3Db4, but as I have 3x 4K displays for the views I am using the 8 strongest cores.