If there's heat, then since heat = work done on the CPU, I can't see why optimising the overclock for HT would not be a benefit overall on a P3D system. The only gain to HT off is applying a few MHz more OC for a few more fps (if that), but reduces the highly threaded base of resources to a greater extent.

Otherwise HT would not exist.

3 hours ago, tooting said:

What would you recommend for a 8700k with HT on?? 

Hi Pete,

I’m afraid I can’t make any recommendations that would guarantee the results you are hoping for on your system. Instead, I would encourage you to read through this thread which explains how I arrived at the optimum Affinity Mask that suited Prepar3D running on my 5960X:

https://www.avsim.com/forums/topic/541073-using-loading-times-to-determine-affinity-mask/

Pages 1 and 3 contain the procedures and results of my tests.

I can’t help but feel that we have been over complicating this business of finding the CPU Affinity Mask that best serves to deliver performance for Prepar3D on our systems. IMHO this method provides a logical and systematic approach to answering the question, and I have Steve to thank for that. Hitherto we were forced to try so many permutations that it all became very confusing and frequently unproductive.

All systems are different so one suggestion may not work as well on one as it does on another. There is really no substitute for experimenting with different Mask values and witnessing the outcomes. If this is done in a systematic way then you are more likely to arrive at a meaningful result.

Of course, you could simply take the shortcut and try the following:

8700K - 6 Core (12HT LPs) - assuming Hyperthreading = Enabled in the System BIOS.

Assume that the Prepar3D main thread and secondary tasks be given exclusivity of access to Physical Cores 0 and 1 by running on LP0 and LP2: 01,01

That leaves 4 Physical Cores (potentially up to 8HT LPs) for other tasks, predominantly texture loading. It may be that some of these LPs could be reserved for Addons without a significant hit on performance/loading times:

For example, try these Affinity Mask values:

1525 = 01,01,11,11,01,01

1530 = 01,01,11,11,10,10

2805 = 10,10,11,11,01,01

2810 = 10,10,11,11,10,10

1013 = 00,11,11,11,01,01

1018 = 00,11,11,11,10,10

Regards,

Mike

 

 

 

 

Edited January 20, 20197 yr by Cruachan

3 hours ago, Pete Dowson said:

Yes, but as I think I said, HT is not currently enabled.  This allows me to achieve 4.8GHz on all 10 cores. With HT enabled I'd need to reduce most of them if i'm to avoid generating too much heat (unless I changed the cooling system significantly. Currently it's a simple Corsair H100 arrangement).

And received wisdom used to be to avoid HT because of the clashes for each core. I have noted Steve's opposite advice, and would consider experimenting with HT re-enabled but in a few weeks I will have an expertly built, really well watercooled system, using a 9900X with 2 x 2080Ti GPUs and probably running at 5.3GHz at least (fingers crossed). Other parts will also be overclocked, and it wil be using 4500 MHz DDR4 memory. 😉

Pete

 

Hi Pete,

Boy, that’s some System - bet you’re salivating at the prospect!

Meantime, why not try enabling Hyperthreading and overclock on a per Core basis. For example, the 10-Cored 7900X has good single thread performance so you might consider clocking Cores 0 - 3 to 4.8GHz and keep the remaining 6 Cores at default (3.7GHz) or increase them to 4.4Ghz or 4.6Ghz.

As you see from my signature all 8 Cores on my 5960X (stock 3Ghz) are dynamically overclocked stable to 4.6Ghz (53% increase). I am experiencing no heating issues and I too use a water-cooling solution: Corsair H110i GT

Regards,

Mike

Edited January 20, 20197 yr by Cruachan

Actually, thanks for all the advice on this, but i really posted with the question about P3D4 performance changes.

Allowed ALL cores, in my test scenario, it hits core 0 at 100%, and 2 other cores at about 50% and one other at nearer 30%. This is with no addons running.

But then, restricted to 4 real cores (0-3) , it doesn't hit 100% on any of them, but all 4 at around 60-70%

This result simply seems illogical to me, hence my post way back when. I was hoping that the resident expert, Steve, would have an answer, or at least a theory.

All the other replies are fine, but i do actually know how to determine the best AM for my system, whether I allow P3D to do as it likes and simply schedule my add-ons around it, or restrict it and assign the others to addons.

Thanks,

Pete

 

22 minutes ago, Pete Dowson said:

All the other replies are fine, but i do actually know how to determine the best AM for my system, whether I allow P3D to do as it likes and simply schedule my add-ons around it, or restrict it and assign the others to addons.

Hi Pete,

I don’t doubt it for a moment. In fact, my last response surrounding the challenge of determining the most suitable Affinity Mask was in response to Pete King’s post.

To add to the overclocking topic, my inclination would be to purchase a MoBo that comes with an overclocking utility that spares me from all the time spent faffing about, can be relied upon and is sufficiently flexible to allow stuff like per Core overclocks.

Best regards,

Mike

23 minutes ago, Pete Dowson said:

Actually, thanks for all the advice on this, but i really posted with the question about P3D4 performance changes.

Allowed ALL cores, in my test scenario, it hits core 0 at 100%, and 2 other cores at about 50% and one other at nearer 30%. This is with no addons running.

But then, restricted to 4 real cores (0-3) , it doesn't hit 100% on any of them, but all 4 at around 60-70%

This result simply seems illogical to me, hence my post way back when. I was hoping that the resident expert, Steve, would have an answer, or at least a theory.

All the other replies are fine, but i do actually know how to determine the best AM for my system, whether I allow P3D to do as it likes and simply schedule my add-ons around it, or restrict it and assign the others to addons.

Thanks,

Pete

 

seen the same, what i did is run 8 real cores for P3D and 2 cores for addons , my strongest core is nr2 (1)  run like this 0,1,1,1,1,1,1,1,1,0 when i disable cores the hatd working one go down to 67-70% with all ten 100% and 8cores 95% with bursts 100%.

the heat on a 79XX series cpu is insane have a 5960X IT NOT 100MHZ more like 300mhz 

4 ht treads doing the same work as 1 real core , in my case i can enable 8cores in not bottleneck main core , with all 10 i proably bottelneck it.

run sync all cores no need to fiddle with per core as its one core that doing most of the work , i diont give as much heat as the main one and dont raise the temp on that .

my cpu 7920X 5.1ghz all cores( 12core CPU with 2 bad cores disabled in bios) and 4X8gb 3800mhz CL16 

 

 

Edited January 20, 20197 yr by westman

1 hour ago, Cruachan said:

Hi Pete,

I’m afraid I can’t make any recommendations that would guarantee the results you are hoping for on your system. Instead, I would encourage you to read through this thread which explains how I arrived at the optimum Affinity Mask that suited Prepar3D running on my 5960X:

https://www.avsim.com/forums/topic/541073-using-loading-times-to-determine-affinity-mask/

Pages 1 and 3 contain the procedures and results of my tests.

I can’t help but feel that we have been over complicating this business of finding the CPU Affinity Mask that best serves to deliver performance for Prepar3D on our systems. IMHO this method provides a logical and systematic approach to answering the question, and I have Steve to thank for that. Hitherto we were forced to try so many permutations that it all became very confusing and frequently unproductive.

All systems are different so one suggestion may not work as well on one as it does on another. There is really no substitute for experimenting with different Mask values and witnessing the outcomes. If this is done in a systematic way then you are more likely to arrive at a meaningful result.

Of course, you could simply take the shortcut and try the following:

8700K - 6 Core (12HT LPs) - assuming Hyperthreading = Enabled in the System BIOS.

Assume that the Prepar3D main thread and secondary tasks be given exclusivity of access to Physical Cores 0 and 1 by running on LP0 and LP2: 01,01

That leaves 4 Physical Cores (potentially up to 8HT LPs) for other tasks, predominantly texture loading. It may be that some of these LPs could be reserved for Addons without a significant hit on performance/loading times:

For example, try these Affinity Mask values:

1525 = 01,01,11,11,01,01

1530 = 01,01,11,11,10,10

2805 = 10,10,11,11,01,01

2810 = 10,10,11,11,10,10

1013 = 00,11,11,11,01,01

1018 = 00,11,11,11,10,10

Regards,

Mike

 

 

 

 

wow thank you for such as detailed answer. next time you need a cheap ticket on either a virgin atlantic or easyjet route give me a pm.  thanks again 

43 minutes ago, Pete Dowson said:

Actually, thanks for all the advice on this, but i really posted with the question about P3D4 performance changes.

Allowed ALL cores, in my test scenario, it hits core 0 at 100%, and 2 other cores at about 50% and one other at nearer 30%. This is with no addons running.

But then, restricted to 4 real cores (0-3) , it doesn't hit 100% on any of them, but all 4 at around 60-70%

This result simply seems illogical to me, hence my post way back when. I was hoping that the resident expert, Steve, would have an answer, or at least a theory.

All the other replies are fine, but i do actually know how to determine the best AM for my system, whether I allow P3D to do as it likes and simply schedule my add-ons around it, or restrict it and assign the others to addons.

Thanks,

Pete

 

Hi Pete,

When talking about % use of a core, we need to know where you are in the sim, loading a scenario, flying at altitude, parked at the gate. It is terribly important to ascertain the type of situation that is required for best performance aspects, maybe that's landing for me...

The reason for 100% on the main task is that it is given too much work to do.

We first limit the fps for those times when there's not much to do, or the fps climbs and frames are ignored with excess heat generated.

Parked in our fave. airport we set the system settings so the amount of work to do between frames can be handled without seeing 100% on that main task. If we set too high, maybe too much area or shadows whatever, the time to process that takes longer than a frame at 100% on that core. Part of that workload is the communication with that task from the other tasks. They will be seeing a harder time than need be.

Restricting to four cores and not seeing the main task reach 100%? Not enough information in there for me to ascertain what's going on. HT off saves working out a few things and also saves the processor from utilising spare cycles, that's a lot on a Windows platform.

 

 

 

1 hour ago, Pete Dowson said:

Actually, thanks for all the advice on this, but i really posted with the question about P3D4 performance changes.

Allowed ALL cores, in my test scenario, it hits core 0 at 100%, and 2 other cores at about 50% and one other at nearer 30%. This is with no addons running.

But then, restricted to 4 real cores (0-3) , it doesn't hit 100% on any of them, but all 4 at around 60-70%

This result simply seems illogical to me, hence my post way back when. 

Hi Pete,

I don’t really know but, if I was to hazard a guess, this may have something to do with the use of Level 2 cache on the CPU, the size of the cache and the competition between the Prepar3D threads to access/share that cache.

I don’t seem to be affected but would be very interested to read Steve’s take on this as well....whenever he has a moment! Goodness knows how he finds the time to cope with real world stuff when he is constantly chipping in to help us out!

Regards,

Mike

 

 

Edited January 20, 20197 yr by Cruachan

The shared cache is distributed on a per thread basis, rather than per core. So if we enable HT or not we have more or less the same number of threads and the same use of the cache. However, the HT mode applies better to that shared system - hence the cost increase for an HT mode CPU.

Edited January 20, 20197 yr by SteveW

I'm never that confident unless I can get hands on to a system.

But in Pete's case, I'm fairly sure that allowing all cores simply saturates the system. The main core reaches 100% through in some part the linkage to those repeat tasks on the extra cores. They would probably be robbing the system of fps as well.

Those extra tasks are non-beneficial if those cores do not represent an increase in scenario loading time.

Use only enough cores to improve the scenario loading time with the main airport. If only one or two seconds are gained by adding another core, that's probably excess to requirements. Too many cores allocated will eat into shared cache, slow down the entire process, and make more heat.

If we have to ask ourselves why 4 and 6 core systems work OK. And if so, then we should be able to use four or six cores of an 8 or 10 core without issues.

 

 

 

So can we see it like this :

- HT on : lower OC = less FPS, but less fluctuation in FPS = smoother

- No HT : higher OC = more FPS , but with larger fluctuation in FPS = less smooth

?

13 minutes ago, GSalden said:

So can we see it like this :

- HT on : lower OC = less FPS, but less fluctuation in FPS = smoother

- No HT : higher OC = more FPS , but with larger fluctuation in FPS = less smooth

?

Sounds about right to me.

 

5 hours ago, GSalden said:

So can we see it like this :

- HT on : lower OC = less FPS, but less fluctuation in FPS = smoother

Hi Gerard,

I went for a flight from KVPS last night in my recently purchased and updated Just Flight Avro Vulcan B Mk. 2. This is a very detailed simulation of the iconic aircraft  and I wasn’t expecting the performance to be all that great. However, I was wrong!

Scenario: Spring, Dawn, ORBX Global terrain and LC, Black Marble Vector NG and Traffic, Active Sky P3D4, REX SkyForce Cloud textures and structures, heavy overcast, 3 Layers, Rain, 4xSSAA

No synchronisation of frame rates with a 30Hz capable monitor here. Frames locked in Prepar3D at 30 (No VSync or TB).

Now, I would be the first to admit that the area around KVPS and beyond is not particularly demanding and performance is generally good. However, this was quite special! How can I convey the experience other than stating that, for once, I could not detect the joins between frames throughout the flight, and this impression was supported by an unvarying FPS variance of 0.2-0.3% both in and out of the VC. It was truly amazing.

Monitoring of core activity in Task Manager confirmed all cores, as defined by the Affinity Mask, were busy but none, including HT LP0, ever maxed out.

The JF Vulcan is a truly remarkable simulation and, I believe, underlines the importance of such coding finesse in aircraft Addons when it comes to reducing or eliminating the incidence of long frames/stuttering.

Regards,

Mike

 

8 hours ago, Cruachan said:

To add to the overclocking topic, my inclination would be to purchase a MoBo that comes with an overclocking utility that spares me from all the time spent faffing about, can be relied upon and is sufficiently flexible to allow stuff like per Core overclocks.

That's what I tried with my current 7900X system, despite it being supplied by the builders (Palicomp, UK) ready overlcocked on all cores to 4.8GHz (my request).  The AISuite software got up to 5.2 which failed, and retreated to 5.0. I tried that for a while but got blue screen crashes in P3D. Not immediately, but within an hour each time. I think P3D is more of a test than that carried out by the automatic software, which doesn't run long enough in any case.

I then tried 5.0 on core 0 and compensated by reducing two others to 4.6. That worked, but gave me no better performance. So I ended up restoring the original as-supplied settings.

For my new 9900K system next month I'm relying on a friend's reknowned expertise to get me a good and stable overclock.

Pete