Jump to content

Sign in to follow this  
BIGSKY

Question on 6 Core Affinity with 777

Recommended Posts

This is a P3D question that might best be served in their forum; however, i believe the P3D executable will use all available logical/physical cores by default.  I use the affinity setting simply to reserve core 0 for other applications that I run during a session.

Share this post


Link to post
Share on other sites

4 cores, HT disabled?

 

OP uses 6 cores with HT enabled. AF=1364 sounds just about right for him.

PMDG state in the manual the reason why. I dont have it handy here but to sum it up, it fixes freezing issues they cannot explain why

Share this post


Link to post
Share on other sites

PMDG state in the manual the reason why. I dont have it handy here but to sum it up, it fixes freezing issues they cannot explain why

Full names in the PMDG forums please.

 

It entirely depends on number of physical and logical cores, and it's simple binary counting. No black art.  However, note that P3D will use all available threads and the only value that using the affinity mask provides is blocking core 0, which I do.

Share this post


Link to post
Share on other sites

PMDG state in the manual the reason why. I dont have it handy here but to sum it up, it fixes freezing issues they cannot explain why

 

 

You're correct; PMDG state that AF=14 should even be used with 6 core processors, or with HT enabled, in order to eliminate freezes. With AF=14 FSX would use only 4 cores even in a 6, 8 or 12 core environment.

 

If freezes are no issue, different AF settings would make sense for performance reasons, though.

Share this post


Link to post
Share on other sites

1364 (5 Logical Processors) is a good call, keeping threads to single LPs per core. Can also try 2720 (4 LPs), or 2688 (3 LPs), and these reverse the order the LPs are loaded on the cores, worth a test. Remember that P3D and FSX need 3 cores at minimum for best results. Going for maximum threads can be a mistake, and an excess of threads synchronizing with the main thread can slow it down. Also each 32bit thread on a 64bit O/S consumes more memory. As PMDG state going for AM=14 generally will provide 3 LPs whatever CPU these days and get round any problems.

Share this post


Link to post
Share on other sites

Best performance with AM=244 here with i7 4core HT on.... (works better than 84 for me)

AM 14 is quite a performance roller coaster on my system...

Share this post


Link to post
Share on other sites

Best performance with AM=244 here with i7 4core HT on.... (works better than 84 for me)

AM 14 is quite a performance roller coaster on my system...

How about 170?

Share this post


Link to post
Share on other sites

How about 170?

just using 4 non physical processors sounds "odd" to me. Although 170 would also keep #0 clean... I will give it a try and see how intel turboboost tec handles it...

 

with AM 244 I know that the only degradation (9FPS) occurs on peaks caused by UT2 and ASN in extreme weather during dusk/dawn every 5 min for 1 second...

Share this post


Link to post
Share on other sites

just using 4 non physical processors sounds "odd" to me. Although 170 would also keep #0 clean... I will give it a try and see how intel turboboost tec handles it...

 

with AM 244 I know that the only degradation (9FPS) occurs on peaks caused by UT2 and ASN in extreme weather during dusk/dawn every 5 min for 1 second...

244 (11110100) demonstrates some good ideas and you probably found an optimum there. As you quite rightly intimate, the problem is not just one of spreading the load over the available cores but also it's keeping busy processes off the core with the first FS thread. With 244=(11,11,01,00) the two paired up cores (11,11) with the secondary FSX threads, mean they are always fully utilised. Although core 0 (00) is free for other threads starting up, core 1 (01) with the main FS thread, has a higher probability of being hit by other threads starting than the two pared up secondary FS threads. Which sounds like what you may be describing with ASN coming in with the weather updates. The two paired threads also use double the resources over one pair on two cores, and double the synchronizing with the main thread, and may be underutilised. Also worth trying on 4 core with HT=ON is 42 and 21, and maybe try 168 which is like 84 but with reversed LP allocations.

Share this post


Link to post
Share on other sites

Thanks,I read the 777 P3D introduction,but didn't recall seeing the affinity 14 suggestion. I have all my sliders at 1/2 settings,and cant even make a KBOS to KPHL flight without freezing up,that is with fresh install on an external solid state hard drive (/X:\Lockheed Martin\Prepar3D v2)with MTX 54.C ,Flytampa KBOS and SunSky Jet KPHL. I am using all the tweaks on this forum and whats in the P3D docs. Off to try AF14,thanks again, I dream of sloped runways!

Share this post


Link to post
Share on other sites

Remember if you have HT on (CPU hyperthreading) then the equivalent to 14 is 84, or 168.

Share this post


Link to post
Share on other sites

244=(11,11,01,00) the two paired up cores (11,11) with the secondary FSX threads, mean they are always fully utilised. Although core 0 (00) is free for other threads starting up, core 1 (01) with the main FS thread, has a higher probability of being hit by other threads starting than the two pared up secondary FS threads. Which sounds like what you may be describing with ASN coming in with the weather updates. The two paired threads also use double the resources over one pair on two cores, and double the synchronizing with the main thread, and may be underutilised.

 

Thank you Steve!

The AM is the only bastion that hampers full time fluidity in FSX for me.

 

Your thoughts (see above) about AM244 exactly represent my experience and I am looking for an answer to this frequently and for a long time...

So instead of AM244 I will give AM170, 168, 42 and 21 a try.

 

The latter 21 (10101) remembered me somehow of a NASA Apollo mission bit code. So I started a Google search and came across an interesting story about a wrongly set bit which caused the Lunar Excursion Module computer (by M.I.T.) to slow down during approach. This bit (set to "on" because of a failure in the procedure check list) turned the ascent radar to "AUTO mode" and by that made the computer wait for unnecessary ascent radar data. The performance hit was about 20%...

Hopefully finding the right AM bit combination will also give some performance boost in return  B)

https://www.hq.nasa.gov/alsj/a11/a11.1201-fm.html

 

I just downloaded P3D v2.5. Let's see if it goes off like a rocket...

 

Share this post


Link to post
Share on other sites

Here is a quote from referenced article:

 

The on-board Apollo Guidance Computer (AGC) was about 1 cubic foot with 2K of 16-bit RAM and 36K of hard-wired core-rope memory with copper wires threaded or not threaded through tiny magnetic cores. The 16-bit words were generally 14 bits of data (or two op-codes), 1 sign bit, and 1 parity bit. The cycle time was 11.7 micro-seconds. Programming was done in assembly language and in an interpretive language, in reverse Polish. Scaling was fixed point fractional. An assembly language ADD took about 23.4 micro-seconds. The operating system featured a multi-programmed, priority/event driven asynchronous executive packed into 2K of memory. The Mean Time to Failure (MTBF) of the machine in a space environment was calculated at 50,000 hours -- almost 6 years, and it never failed in flight operations. It was truly a marvel for its time, a tribute to M.I.T.'s designers, and it accomplished a most complex mission.

 

Keep things in perspective, Intel was only one year old and was working on an integrated chip (much to the chagrin of TI) for a Japanese watch maker. That chip became the 8008 a few years later.

Share this post


Link to post
Share on other sites

 

 


The Mean Time to Failure (MTBF) of the machine in a space environment was calculated at 50,000 hours -- almost 6 years, and it never failed in flight operations.

That was well before "some other folks" :ph34r:  started computing! :lol: 

It is really a fascinating excerpt of one of many NASA articles...

 

And to find and read all that great stuff today is made possible by the ongoing improvement of technology. One can actually go back to the invention of the reel and beyond. And in the other direction of the timeline e.g. "Startrek" gets reality and it still goes on...

B)

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Sign in to follow this  

×
×
  • Create New...