Archived

This topic is now archived and is closed to further replies.

Guest Nick_N

HyperThread - FSX - i7 -and YOU

Recommended Posts

There is a thread in the FSX forum where the following was suggested may have benefitAfter reviewing the docs on i7 and looking at the tech information at FSINSIDER I have come to the conclusion that there MAY BE some value in running HT with i7 and FSX if the affinity mask edit is used.I personally do not have time to drop what I am doing and go into this and test it correctly .. therefore I am going to post this for people using i7 to tryNOTE: IF you are running 4GHz and enabling HT may push the temp of the i7 processor over the 80c limit you should reduce the clock on the processor to ensure safe operation. You may wish to stress test with HT enabled and establish your clock first before running this test.Enter the BIOS and ENABLE this setting if it is not already enabled:Intel HT Technology In the FSX.cfg add the following if it is not already there[JOBSCHEDULER]AffinityMask=Nuse these settings for N as you test and you must EXIT the sim and relaunch between changes255 - all 8 in use 254 - 7 in use with the first remaining free252 - 6 in use with the first entire CORE remaining free127 - 7 in use with the last remaining free63 - 6 in use with the last entire CORE remaining free(edited by David per Nick's request)Verify the use and the perf then report back please.There are some unanswered questions about the JOBSCHEDULER and what I posted in affinity mask settings above with i7 would answer a lot of them. Please post back the findings.

Share this post


Link to post
Share on other sites
Help AVSIM continue to serve you!
Please donate today!

Tried it here... i7 920 on evga x58, FSX SP20) With HT OFF in the bios I noted 4 'cores' 0,2,4,6 being used by FSX.1) With HT on and NO AffinityMask set, I noted 4 'cores' 0,2,4,6 being used by FSX.Same result as having HT off in the bios.2) With HT on and AffinityMask = 255, I noted all 8 'cores' 0,1,2,3,4,5,6,7 being used by FSX.Loading times seemed dramatically fasted with HT ON and AffinityMask = 255maybe just a perception at this point could have been other reasons, like cache being used,but it was clearly faster for my quick test.3) With HT on and AffinityMask = 254, I noted cores 1,2,3,4,5,6,7 being used by FSX, core 0 was not used by FSX.I noticed no difference from AffinityMask = 255 as far as load times or stuttering.4) In areas were stutters happen, like Vancouver using Vancouver+, I noticed no difference HT on or HT off.Intel's i7 seems to be like a wild horse, that needs to be tamed.It has raw power, but it's misunderstood and tempermental :)The chips run very hot and need high voltage to get high clocks.I think it will take a couple chip batches and MB revisions to get it stable and predictable.FSX and i7 can be amazing at times but the stuttering when it happens seems to have more buck than the core setups did.I'm still not convinced it offers much more for FSX than a quad core/duo setup would at 4GHz.

Share this post


Link to post
Share on other sites
Tried it here... i7 920 on evga x58, FSX SP21) With HT on and NO AffinityMask set, I noted 4 'cores' 0,2,4,6 being used by FSX.2) With HT on and AffinityMask = 255, I noted all 8 'cores' 0,1,2,3,4,5,6,7 being used by FSX.Loading times seemed dramatically fasted with HT ON and AffinityMask = 255maybe just a perception at this point could have been other reasons, like cache being used,but it was clearly faster for my quick test.3) With HT on and AffinityMask = 254, I noted cores 1,2,3,4,5,6,7 being used by FSX, core 0 was not used by FSX.I noticed no difference from AffinityMask = 255 as far as load times or stuttering.4) In areas were stutters happen, like Vancouver using Vancouver+, I noticed no difference HT on or HT off.Intel's i7 seems to be like a wild horse, that needs to be tamed.It has raw power, but it's misunderstood and tempermental :)The chips run very hot and need high voltage to get high clocks.I think it will take a couple chip batches and MB revisions to get it stable and predictable.FSX and i7 can be amazing at times but the stuttering when it happens seems to have more buck than the core setups did.I'm still not convinced it offers much more for FSX than a quad core/duo setup would at 4GHz.
Jack, .. I do not see any of the stutters here you seem to be experiencing and I am certainly seeing better perf than a clocked 4GHz Qx9770 so I am not sure what is happening on your end at this pointI do wish to point out I did warn about 920 and that users of the 940/965 would not be as limited as 920 user who pushes the limit.. regardless of the internet stories, I posted that for a reason.. I have already burned (dead) both a 920 and 940 engineering sample here for discovery in voltage limits so I am aware of how hot the 920 runs in compare.the more info gathered about the use of HT from different users the better... thanks for looking at this!PS.. the P6T motherboard is now on revision 2 and out of all the boards on the market it is the best one for the typical user running i7 regardless of the revision. I use the Rampage Extreme II for the engineering access and for full DDR3 1800/2000 support which is the memory I am running with it now.

Share this post


Link to post
Share on other sites

Actually Nick, 252 runs FSX on the last six threads (last 3 cores) and 63 on the first six. I'm trying out different settings but I don't know how objective I'll be. But in Task M. it certainly pegs all six threads I had running so far.And I've got a Swiftech WC system on the way so I'm going to ditch air cooling for a while. If this all is for real then I'm gonna need it. Besides, I'd love to put my Cu TRUE up on the shelf.-jk

Share this post


Link to post
Share on other sites
Jack, .. I do not see any of the stutters here you seem to be experiencing and I am certainly seeing better perf than a clocked 4GHz Qx9770 so I am not sure what is happening on your end at this pointthe more infor gathered about the use of HT from different users the better... thanks for looking at this!
Don't know whats happening with your Qx9770 setup , but my 4G QX9650 is smooth as could be.I'd give you some tips but every one I used to get it smooth was a post from you :)My 4G i7 setup does have more raw power / framerate but when it jerks its jerks harder, if that makes sense.I've never seen an FSX setup flying the f18 full out that can not be made to stutter over some terrain?Great thread here, looking forward to more results.

Share this post


Link to post
Share on other sites

I have already burned (dead) both a 920 and 940 engineering sample here for discovery in voltage limits so I am aware of how hot the 920 runs in compare.Hey Nick, feel free to send out any chips my way for which I can help you discover those limits! :( It's still cool in San Diego :( -jk

Share this post


Link to post
Share on other sites
Actually Nick, 252 runs FSX on the last six threads (last 3 cores) and 63 on the first six. I'm trying out different settings but I don't know how objective I'll be. But in Task M. it certainly pegs all six threads I had running so far.And I've got a Swiftech WC system on the way so I'm going to ditch air cooling for a while. If this all is for real then I'm gonna need it. Besides, I'd love to put my Cu TRUE up on the shelf.-jk
you are correct.. I got those backwards becuse the binary is backwards from the cores in use.. will editEDIT: Time ran out on the edit... So anyone posting results, be aware252 - 6 in use with the first entire CORE remaining free63 - 6 in use with the last entire CORE remaining freesorry... :(Nick, I have edited your post.The info is now correct :--)Best,David Roch

Share this post


Link to post
Share on other sites
NickN - so what vcore did it take to burn out a 920?
It was not Vcore.. it was DRAM and that came in 10 days @ 1.80vI took out the 940 on VTT pushing a 4.37Ghz clock on what I consider (possibly) acceptable vcore @ 1.46I have 2 more slugs here that will go to the Vcore voltage discovery cause... first one is @ 1.55 now and has been fine for 12 days

Share this post


Link to post
Share on other sites
Don't know whats happening with your Qx9770 setup , but my 4G QX9650 is smooth as could be.I'd give you some tips but every one I used to get it smooth was a post from you :)My 4G i7 setup does have more raw power / framerate but when it jerks its jerks harder, if that makes sense.I've never seen an FSX setup flying the f18 full out that can not be made to stutter over some terrain?Great thread here, looking forward to more results.
F18 here pops mach1+ on the deck, no stutters and no terrain blurs.. love it especially canyon runsI will say this about the multithread and flight loading.. it makes sense based on what was posted about SP1 work=================================During loading, we run the DEM loader on threads. You'll see good balanced usage across all cores, as well as about 1/3 faster load times on average.=================================OK so that does not really mean chit to the game in progress although with Adams changes to how the 'world' and terrain with the thread work they added are used may, which is one of the primary reasons I want to check this out with as many i7 users as possible. It may load the flight like a rocket and that actually makes sense but I dont care about flight load.. I want to know about in-game changesToo many freakin MYTHS get posted about tweaks and the only way to know if JOBSCHEDULER really makes a difference is the RIGHT mindset in what is REALLY happening and of course reports from other so I can compare those reports to what I see here when I end up spending quality time testing all this.The forums are full of MIRACLE tweaks which are BS and I do not like posting something works unless I know they really have teeth

Share this post


Link to post
Share on other sites

NickAs i already posted my findings on the other thread, i'm not going to post them again here. The 252 setting i didn't test yet, though, but i will, just to see if it has any improvement over the 254 setting (which so far is the one to go, in my opinion).But i'm glad to see you slowly starting to accept you were wrong when you stated things like "absolutely nothing was gained or lost by enable/disable HT in i7" and "I normally find those who use the affinity mask tweak and say killing a core to FSX makes the system work better and run smoother in application, see that result because FSX is NOT correctly tuned". Just proves you are human like the rest of us. :( Looking forward for your own tests.Cheers.

Share this post


Link to post
Share on other sites
NickBut i'm glad to see you slowly starting to accept you were wrong when you stated things like "absolutely nothing was gained or lost by enable/disable HT in i7" and "I normally find those who use the affinity mask tweak and say killing a core to FSX makes the system work better and run smoother in application, see that result because FSX is NOT correctly tuned". Just proves you are human like the rest of us. :(
Is there a reason for this nonsense that was taken out of context in the same post I placed this?===========================================I will look at it again when I get time to see if I missed something.. which is possible but nothing I saw when I ran i7 tests back in Oct-Nov indicated HT was doing anything for FSX with i7 except heating up the proc in a high clock. All my test were run on SP2/Acceleration as I do not use RTM or SP1 at all..I am also on WindowsXP x64 although I did run FSX on Windows7 and saw no difference either so there are few different wildcards out there in configurations that may need to be looked at.The bottom line to all this is.. if you think you have found something that works for you. no matter the 'what' or 'how'.. use it :( =============================================Nothing was gained or lost in my tests, thanks for pointing out I wrote that correctlyIt is unknown at this point if this will provide anything across the board to FSX other than what appearsto be faster load time It may.. or it may turn out to be another "works for me" ... "I dont see any change" ... "Made mine run worse" ... tweak that ends up being a placebo for 1/2 and a small ratio of people may see something of benefit.. where others may find no change and still others find it makes their system run worse.So if it was your intent to come into this thread to insult someone with a twisted, out of context statement and by suggesting something provides benefit across the board without proper testing and research to confirm... you have indeed accomplished those goals!Multithread and Hyperthread are 2 different things http://forums1.avsim.net/index.php?s=&...t&p=1532293This may be taking advantage of the changes to how MULTITHREAD is applied on i7 but even if it is shown to have some benefit, that does not mean the HYPERTHREAD instuctions as defined by Intel are being used, only that hyperthread must be enabled to use the expanded multithread ability of the proc which was not available on C2/Q Intel procs.Until I get confirmation from Phil for whom I have a pending communication in progress, or, someone at Aces that says FSX takes full advantage of HYPERTHREAD instructions and not just MULTITHREAD, what I posted and why I posted it is quite true regardless of the outcome in testing, and, even if FSX is taking some advantage of HT in all its glory, until this gets looked at correctly for application with in-game use it does not represent any major benefit to users. Many of them may discover the 2, 3 or 4GHz loss in the clock they need for heavy iron aircraft due to heat generated by HT is not worth any minor gain the feature may allow.Until those unknowns are at the very least defined somewhat... none of this means anything other than what appears at this point to be faster FSX flight load timeLast, I have always posted in this forum and any other, if I find I am wrong about something.. I am the first to admit it and do not need the assistance of a anyone to do so.

Share this post


Link to post
Share on other sites
Is there a reason for this nonsense that was taken out of context in the same post I placed this?===========================================I will look at it again when I get time to see if I missed something.. which is possible but nothing I saw when I ran i7 tests back in Oct-Nov indicated HT was doing anything for FSX with i7 except heating up the proc in a high clock. All my test were run on SP2/Acceleration as I do not use RTM or SP1 at all..I am also on WindowsXP x64 although I did run FSX on Windows7 and saw no difference either so there are few different wildcards out there in configurations that may need to be looked at.The bottom line to all this is.. if you think you have found something that works for you. no matter the 'what' or 'how'.. use it :( =============================================Nothing was gained or lost in my tests, thanks for pointing out I wrote that correctlyIt is unknown at this point if this will provide anything across the board to FSX other than what appearsto be faster load time It may.. or it may turn out to be another "works for me" ... "I dont see any change" ... "Made mine run worse" ... tweak that ends up being a placebo for 1/2 and a small ratio of people may see something of benefit.. where others may find no change and still others find it makes their system run worse.So if it was your intent to come into this thread to insult someone with a twisted, out of context statement and by suggesting something provides benefit across the board without proper testing and research to confirm... you have indeed accomplished those goals!Multithread and Hyperthread are 2 different things http://forums1.avsim.net/index.php?s=&...t&p=1532293This may be taking advantage of the changes to how MULTITHREAD is applied on i7 but even if it is shown to have some benefit, that does not mean the HYPERTHREAD instuctions as defined by Intel are being used, only that hyperthread must be enabled to use the expanded multithread ability of the proc which was not available on C2/Q Intel procs.Until I get confirmation from Phil for whom I have a pending communication in progress, or, someone at Aces that says FSX takes full advantage of HYPERTHREAD instructions and not just MULTITHREAD, what I posted and why I posted it is quite true regardless of the outcome in testing, and, even if FSX is taking some advantage of HT in all its glory, until this gets looked at correctly for application with in-game use it does not represent any major benefit to users. Many of them may discover the 2, 3 or 4GHz loss in the clock they need for heavy iron aircraft due to heat generated by HT is not worth any minor gain the feature may allow.Until those unknowns are at the very least defined somewhat... none of this means anything other than what appears at this point to be faster FSX flight load timeLast, I have always posted in this forum and any other, if I find I am wrong about something.. I am the first to admit it and do not need the assistance of a anyone to do so.
It would be interesting to get the bottom of this because I'm pretty sure that MS said ages ago that FSX couldn't use hyper-threading. I wonder whether the i7 implementation somehow "dupes" FSX, in which case we may end up with the very collisions that made MS disable hyperthreading to begin with; or whether i7's implementation is such an improvement that there aren't any collisions any more.Tim

Share this post


Link to post
Share on other sites
It would be interesting to get the bottom of this because I'm pretty sure that MS said ages ago that FSX couldn't use hyper-threading. I wonder whether the i7 implementation somehow "dupes" FSX, in which case we may end up with the very collisions that made MS disable hyperthreading to begin with; or whether i7's implementation is such an improvement that there aren't any collisions any more.Tim
Hi TimI posted a reply here on what we 'know' for sure and why the affinitymask hook can help with procs prior to i7http://forums1.avsim.net/index.php?s=&...t&p=1532737I am waiting to hear back from Phil so when I do look at this I have it right in my mind what I am seeing on the screen and how what is presented on the screen is being translated by the application through the hardware.EDIT: and the key word used by Intel which may be allowing this to work is the new design they refer to as "uncore"It was the tech docs in the description of the processors in that they are treated as somewhat separate in use from the single 'core' for which they are contained, and, the self-contained cache for each which sparked my interest in this when applied to 'multicore' type threading.. in that what you said is very possible based on the i7 design change.

Share this post


Link to post
Share on other sites
Guest
This topic is now closed to further replies.