HyperThread - FSX - i7 -and YOU

[jwenham 1]

Hi Nick, I have gone thru the steps of reducing the USERVA amount down 4x (64) and still get out of memory errors when flying complex aircraft into large airports. I dont want to reduce sliders as when its not out of memory everything is great so I have decided to go 64 bit. This is what I have ordered and just want to make sure it is the right version.http://www.newegg.com/Product/Product.aspx...N82E16832116378Thanks, Jim

[Guest Nick_N]

Hi Nick, I have gone thru the steps of reducing the USERVA amount down 4x (64) and still get out of memory errors when flying complex aircraft into large airports. I dont want to reduce sliders as when its not out of memory everything is great so I have decided to go 64 bit. This is what I have ordered and just want to make sure it is the right version.http://www.newegg.com/Product/Product.aspx...N82E16832116378Thanks, Jim

yep... that'al do it Still getting them? hmmm .. well I can say there have been a few circumstances where no matter what the user does in a 32bit OS, OOM graphic spikes and errors just will not let up with the heavy hitting aircraft and the 64bit OS is the only solution. We had several of those over at Simforums and all but one were solved with the move to 64bit. One turned out to be a defective video card.If you have the setting 'MEMORY REMAP FEATURE' in your BIOS under the advanced chipset settings... set it to ENABLED for 64bit OS use. Some BIOSes have that setting and some do not.

[jwenham 1]

yep... that'al do it Still getting them? hmmm .. well I can say there have been a few circumstances where no matter what the user does in a 32bit OS, OOM graphic spikes and errors just will not let up with the heavy hitting aircraft and the 64bit OS is the only solution. We had several of those over at Simforums and all but one were solved with the move to 64bit. One turned out to be a defective video card.If you have the setting 'MEMORY REMAP FEATURE' in your BIOS under the advanced chipset settings... set it to ENABLED for 64bit OS use. Some BIOSes have that setting and some do not.

Sounds good! By changing the switch it changed my crash from a BSOD to a simple error window "Your system has run out of memory ..." but no crash. Lowering the slidders does fix it but hey I got an i7 :(

[jackcnd 0]

Another area I fly ... Toronto, Canada seems to produce a high number of stutters with Scenery Comp at 100%.HT on or HT off.Try taking off east from the Toronto Island Airport and quick fly north low over downtown,then swing back to land on the island.Stutters, stutters, stutters, even with a fresh XP x64 install. I dont recall seeing these with my older quad core setup, maybe I'm just expecting more with i7.Other areas are fine with the i7, but some cities seem to give more pronounced stuttering.They are seen with 64 bit OSes, seems lesser with XP 32, but its hard to prove, might try video taping them and playing back side by side.

[Guest Antarius]

I did testing on my own. The smoothest performance is HT OFF, 1 core OFF, 2-3-4 cores ON.

[Guest Nick_N]

Another area I fly ... Toronto, Canada seems to produce a high number of stutters with Scenery Comp at 100%.HT on or HT off.Try taking off east from the Toronto Island Airport and quick fly north low over downtown,then swing back to land on the island.Stutters, stutters, stutters, even with a fresh XP x64 install. I dont recall seeing these with my older quad core setup, maybe I'm just expecting more with i7.Other areas are fine with the i7, but some cities seem to give more pronounced stuttering.They are seen with 64 bit OSes, seems lesser with XP 32, but its hard to prove, might try video taping them and playing back side by side.

Toronto is a scenery complexity issue however its becuse of the airport more than anything else.. over the city the issue is car traffic especially if you use UTXSet cars to zero, set autogen to 100%, Scenery complexity to 3/4 (very dense), AL/GA traffic to 30, Airport vehicles to either NONE or low and shut off the UTX Canada roads and then try flying over the city of Toronto without the airport in the background view. That airport is a major hit so you can tune settings back from the test settings I listed till you get it tweaked for that specific area. Replacment AI traffic goes a long way for that problem since even 30% traffic during certain times of the day can produce a lot of flights at that airport.

[Guest Nick_N]

My tests show mixed results using the afininty mask tweak. If I drop my CPU speed it seems to have more impact depending on the are than running 4GHzwith the exception of one issue which as far as I am concered is resolved with FSXHyperthreadingIt Does nothing - Turn it off

[jackcnd 0]

Agreed, HT OFF also allows me to run 4GHZ cooler at lower voltage, I only turn it on for video encoding.Thanks for the Toronto comments, seems the areas I've been using for stutter testing need tweaking on even the best systems.Yesterday, I setup FSX Accel on:i7 4GHZ HT off, x58GTX280Win7 Beta181.20 drivers, 16x AFFSX AF, AA ON, TBM 70The base install flys Toronto smooth with 19m terrain, 60cm textures, SC max, AG dense.Made an image and will slowly add extras, and tweak FSX trying to keep things smooth.In the past I was just moving my full FSX installs around.There was too many varibles between hardware, OC settings, drivers and FSX and addon settings.It was all but impossible to find a specific reason for stutters on any given setup.Hoping building slowly on a fresh install will work better.First addon will be GEX to rid those blurry sandlike distant default FSX textures :)

[kttd 12]

I finally got around to this test.I turned HT off and set affinity to 15. Did a few tests with WOAI, did a few without WOAI and My Traffic enabled. I used SeaTac 34, flew north around the city and back, about a half dozen loops per test. Turned off FPS counter and focused on smoothness perception rather than framerate.I achieved the smoothest flying with My Traffic full resolution enabled but less animation (no jetways), WOAI disabled, affinity mask to 15, HT off. 1280x1024. Cars=8, water traffic both at 25, AI planes at 41 - 25 respectively, both scenery at 3/4. Water at 2 low. Clouds medium, weather change low. Airport vehicles, medium. Stock C-172. The only add-on being MT. A pleasant surprise to be so smooth around SeaTac. The only unpleasant factor is artifact spiking the first 1/2 minute or so during each flight, after that none.Vista 64i7 940EVGA 280181.22 drivers, 8x 8xAsus P6T motherboardtons of memoryno overclock

[Guest sillyeagle]

HyperthreadingIt Does nothing - Turn it off

I have to disagree with this. When I run with hyperthreading off it's as though I'm running a Core 2 Quad again, regardless of AffinityMask setting. As with my move from single to dual core, when I moved from C2 Duo to C2 Quad the doubling of threads was apparent in the reduction of micro stutters and in the near doubling of my terrain texture loading power, which halves my propensity for blurries. Moving from C2 Quad to i7 again gives me results I consider consistent with a doubling of threads. Blurries are now a thing of the past, as the highest LOD is snapped in at all times, regardless of speed, autogen density, or frame locks. In fact terrain loads faster now unlocked than it did on C2 Quad locked with 20fps of headroom. Micro stutters are also a thing of the past, as the sim is now as fluid unlocked as it was on C2Q locked with the 20fps of headroom. I attribute both of these improvements to hyperthreading, and when I disable it... hello micro stutters.The only problem I've found with my i7/Win7 setup is the occurance of the herky jerky noted above, which is totally different from the micro stutters experienced with unlocked frames on my last three systems. On my screen the herky jerky looks like the scene is jerking 3/4 inch at a time, and it is something I've not encountered until i7. For me herky jerky only occurs when using an unlimited frame lock. Since i7 seemed to rid all stutters and blurries, I saw little need to lock the frame rate, and I was reluctant to use it, though being the only solution to herky jerky I had no choice. To compensate for the frame rate hit associated with the lock, and to harness the extra texture loading power of i7, I set a very low fiber frame time fraction. Using a low fiber fraction produces a nice framerate gain at the expense of texture loading power, though i7 combined with a locked framerate can spare a good deal of it's texture loading headroom. So what I have found to be the best solution with i7 is to have Hyperthreading ENABLED with a Framerate Lock of 25 or 30, No Affinitymask tweak, and a Fiber Fraction of 10 seasoned to taste, with higher values leaning toward terrain load speed, and lower values leaning toward framerate.Here's my system details and cfg tweaks. At last sitting I was able to fly around FTX using very dense autogen pegged right at my 30fps lock, and the terrain was snapped into the highest LOD at all times, which is very welcome to a low and fast guy like me. A bufferpool over 50MB is nice if you can spare the video memory, and I like the max per cell tweak to change the autogen balance of trees to buildings, as you can thin the trees out quite a bit without it catching your eye due to their scattered appearance, which nets a nice performance boost, while it's good for striking visuals to keep the buildings packed in tight enough to form proper rows. Intel Core i7 920 Quad @ 4.0GHzOCZ Platinum 6GB @ 1600MHz 7-7-6-18 1TEVGA GTX260 Core 216 @ 710/1420/2500MHzNoctua NH-U12P SE1366 w/ 2x 120mmAntec True Power Trio 650w EVGA Intel X58Windows 7

[DISPLAY]TEXTURE_BANDWIDTH_MULT=110 [MAIN]FIBER_FRAME_TIME_FRACTION=0.10[TERRAIN] TERRAIN_MAX_AUTOGEN_TREES_PER_CELL=1700TERRAIN_MAX_AUTOGEN_BUILDINGS_PER_CELL=2000[BUFFERPOOLS]PoolSize=70000000

More pictures of my i7 testing with FTX are at the link.http://orbxsystems.com/forums/index.php?topic=8943.0

[Guest Nick_N]

What a .10 FFTF has done is reduced the attention of the system from the terrain engine and placed it on other things in priority. Can that work? Yes, but it also depends on the scenery being typically flown and other factors. Many can accomplish the same task by using affinity mask in which the application would make the decision as to priority of terrain over other area.The reason I did not add that to the tuning list is because -most- discover eventually that messing with FFTF after SP1/SP2 does not net a 'all scenery-all area' result and as such most will end up removing it from the config ... not to say there are no exceptions to thathoweverLet me make something perfectly clear... I have confirmation that FSX is not hyperthread aware and will not make use of itWhen users get rid of stutters using affinity mask or fiber frame time fraction, which by the way Adam fixed for SP1/2 so the majority of users need not have to deal with the edit, all the user is doing is reducing and cutting out thread collisions which were creating the stutters to begin with due to the poor and 'patched' design of the FSX rendering engine for multithread.And that is the name of that tuneThe scenery installed and preferred by the user will direct what settings may be best for one system over another but be aware HT has no practical performance use for FSX, at all. Its heating up the proc for nothing in the application

[Guest sillyeagle]

Well I tested again, and I get 50% utilization with HT enabled and 100% without, so that appears to prove HT is not used, but when enabled I do get slightly better performance. I'm just amazed i7 can load textures as well as it does on only 4 threads.I also tested the fiber fraction time again using a 30fps lock. With .33 and HT disabled I was seeing a 18fps minimum on my Coffs Harbour pass. With .10 and HT disabled I was seeing a 26 minimum. That's a 45% gain and I was always flying over the highest LOD, so I didn't see any reason not to take the 45%. Other than when passing the scenery elements seen below I was pegged at my 30fps lock with an occasional dip to 28, but when I enabled HT many of those dips did not occur. My minimum would still dip to 26, but was more likely to hover around 28 than without HT. I have not had a chance to test my OC with HT disabled, but maybe I can get a gain that tops the small one I appear to see with HT enabled. Stutters in any form are non existent with a framelock, so to me that's the best the way to go, and with the .10 value there's only a small performance hit when using it. What matters to me is seeing 25-30fps, seeing the highest LOD terrain at all times, and seeing a totally fliud and stutter free presentation. Seeing that makes for the best visuals, and FS is after all a visual simulation to be used as a point of leverage for the imagination. I'll take my visuals as high caliber as they come, and what I've got going achieves my goals with room to improve in any direction, so I'm golden. :)For anyone interested my MDB, CPU, RAM and Air Cooler cost only $820 from newegg.

[Guest Nick_N]

Well I tested again, and I get 50% utilization with HT enabled and 100% without, so that appears to prove HT is not used, but when enabled I do get slightly better performance. I'm just amazed i7 can load textures as well as it does on only 4 threads.I also tested the fiber fraction time again using a 30fps lock. With .33 and HT disabled I was seeing a 18fps minimum on my Coffs Harbour pass. With .10 and HT disabled I was seeing a 26 minimum. That's a 45% gain and I was always flying over the highest LOD, so I didn't see any reason not to take the 45%. Other than when passing the scenery elements seen below I was pegged at my 30fps lock with an occasional dip to 28, but when I enabled HT many of those dips did not occur. My minimum would still dip to 26, but was more likely to hover around 28 than without HT. I have not had a chance to test my OC with HT disabled, but maybe I can get a gain that tops the small one I appear to see with HT enabled. Stutters in any form are non existent with a framelock, so to me that's the best the way to go, and with the .10 value there's only a small performance hit when using it. What matters to me is seeing 25-30fps, seeing the highest LOD terrain at all times, and seeing a totally fliud and stutter free presentation. Seeing that makes for the best visuals, and FS is after all a visual simulation to be used as a point of leverage for the imagination. I'll take my visuals as high caliber as they come, and what I've got going achieves my goals with room to improve in any direction, so I'm golden. :)For anyone interested my MDB, CPU, RAM and Air Cooler cost only $820 from newegg.

It appears to me you are using one area for these tests and possible one type of scenery install.. do the same results occur over default FSX areas such as NYC, LA, Seattle, etc?If so then you have found the settings that work best for your system.. if not, then the scenery you prefer to fly over is directing the settings you came up with which resolve the issues. FFTF lowered for most will blur terrain textures. Not to say lowering it will not work or help however that change and the result is directly related to the system and scenery being flown.In any case it is now confirmed HT is useless to FSX and the original question around it has been answered. Affinity Mask has value but it is defined by the system in use. FFTF is one of those rare changes which may make a difference however every time I have played with that setting since SP1/SP2 I always end up going back to removing the entry because I always ended up with one area looking good and others going to mud @ airspeed

[Guest Nick_N]

===========================================FIBER_FRAME_TIME_FRACTION determines the maximum amount of time per frame that we will run fiber jobs on the primary thread. We measure how long it took to simulate and render and then multiply that time by FIBER_FRAME_TIME_FRACTION to determine how long to run the fibers. For example, if it took 10 milliseconds to simulate and render and FIBER_FRAME_TIME_FRACTION was set to the default value of 0.33, then we would allow the fibers on the primary thread to run for up to 10 * 0.33 = 3.3 milliseconds. For fraction values of 1.0 and 2.0, the time given to fibers would be 10 milliseconds and 20 milliseconds, respectively. The operation of FIBER_FRAME_TIME_FRACTION on single core machines has not changed since RTM. On multi-core machines in SP1, we moved many fiber jobs off of the primary thread and onto secondary threads. Since FIBER_FRAME_TIME_FRACTION only affects scheduling of jobs on the primary thread, it will have less of an impact on the performance of Flight Sim on multi-core machines. In fact, we moved so many jobs off of the primary thread that there probably isn

[Guest Seria17hri11er]

Something I want to bring up in regards to Hyperthreading and FSX. While hyperthreading may not benefit FSX performance directly I have found it benefits MY FSX performance INDIRECTLY. How? Well, I run add-on programs like Active Sky Advance and Radar Contact. It appears that with Hyperthreading enabled XP shifts the processing of these programs to the virtual cores. I have seen better FSX performance when running these programs with Hyperthreading enabled, rather than disabled.