Hyperthreading

[WarpD 824]

So... OC is very stable... unless you have HT on and run FSX (which isn't HT aware). Willing to bet if you removed your OC... FSX would be stable with HT on.

[vgbaron 3,608]

Patient: "Doc, it hurts when I do this."

Doc: "Don't do that. That will be $100."

 

:smile:

 

Vic
 

[angeli662 43]

So... OC is very stable... unless you have HT on and run FSX (which isn't HT aware). Willing to bet if you removed your OC... FSX would be stable with HT on.

Actually FSX will run with HT enable as long a set the AM to 8 cores, if I set to 16 will eventually freeze.  disabling OC on my system will defeat the purpose of having a system like mine, considering tha @ full load the temps never reaches beyond  50 c . Both of my PCs used to run FSX are OCed.

I also use my main PC for my work (AutoDesk Revit) and when is time to render the final I always enable HT as it takes several days to complete.

[WarpD 824]

Ah... yet another OC addict... it's so sad when we see them go before their time. :wink:

[SteveW 1,792]

Here I show how P3D v2.5 loses 8% throughput with no AM specified on a 6 core HT enabled system. In the lower half no AM (signified with AM=0) and a maxed LP 0 sharing core 0 with another P3D main thread on LP 1. In the upper half, LP 6 is hosting the main P3D thread with no other activity on LP 7 which shares throughput of core 3 with LP 6. In fact another process starts around 2/3 into the trace, whereby the jobscheduler efficiently distributes the work away from core 3.

[SteveW 1,792]

Also notable in the top half of the image we have the CPU partitioned off to three cores for all the threads spawned by the P3D process, and performs a few more fps better than the no AM setup. The AM 3392 was used as an example for test purposes, and a point of demonstration. In both examples above the maxed main thread can't go faster anyway, so allowing more secondary threads to interrupt it decreases performance still further. Therefore we prevent the ganging up of main threads on the primary core by masking one LP on the first core to be used, and when we have many LPs, we partition the CPU to disallow too many secondary threads forming which would stop the main thread too often to sync information.

[Nyxx 6,993]

Steve you have shown time and again that HT works with P3D if people want to quote a guy talking about old code in FSX then sadly we need to inform people so they to don't get misinformation drummed into them.

[SteveW 1,792]

Thanks David. We can always mould an app to a machine, that is to use that machine efficiently, to it's best advantage. In the case of FSX-SE, that can clearly be seen placing its primary threads carefully and always shows an AM=0. No matter what we put in the jobsheduler cfg, FSX-SE does things the way it wants. Even with this kind of app though, its core affinity can be controlled by a hosting app if need be.

[Murmur 3,941]

No matter what we put in the jobsheduler cfg, FSX-SE does things the way it wants.

 

I have a dual core with HT on.

Here's FSX:SE with no affinitymask added in fsx.cfg. You can see it's only using logical core 2:

 

 

Now using affinitymask=5 (0101). It's using logical cores 0 and 2:

 

 

Finally, using affinitymask=15 (1111). It's using all 4 logical cores. In this case FSX:SE becomes noticeably more stuttery:

 

[WarpD 824]

This is a discussion about Prepar3D, not FSX:SE.

[Murmur 3,941]

This is a discussion about Prepar3D, not FSX:SE.

 

But SteveW mentioned FSX:SE. You should have read it better:

 

In the case of FSX-SE, that can clearly be seen placing its primary threads carefully and always shows an AM=0. No matter what we put in the jobsheduler cfg, FSX-SE does things the way it wants. Even with this kind of app though, its core affinity can be controlled by a hosting app if need be.

 

Moreover, FSX has been mentioned multiple times in this thread, starting at reply #1. If someone gives technical informations about FSX:SE in a P3D forum, and I suspect those may be inaccurate informations, don't you think I'm entitled to reply to that?

[n4gix 4,947]

This is a discussion about Prepar3D, not FSX:SE.

Actually, the discussion topic is Hyper-threading, not P3D or FSX per se.

[WarpD 824]

I think that bringing up FSX:SE which is not FSX but rather a modified version that Dovetail has done who knows what... as well as bringing up FSX itself as a comparison to current versions of Prepar3D has no legitimacy at all.

 

This thread is a discussion of hyperthreading for Prepar3D:

 

Does P3D use hyperthreading, in other words would an i7 run P3D better than an i% because of the hyperthreading capability?

I'm trying to get it back 'on track'.

[SteveW 1,792]

I have a dual core with HT on...

I've not tried it with a dual core. Unfortunately I can't make a conclusion about FSX-SE on a dual core from your post.

[SteveW 1,792]

...Checking the latest FSX-SE version on my 6 core Win10 system reveals it does indeed pay attention to the jobscheduler section, and appears to behave indistinguishably from FSX. Certainly working differently now since I tested that function. Maybe they had trouble with it somehow but it seemed to work OK.