Not to take away from any of the above post--because it's a great thread--, but isn't this all common knowledge to those of us who have been in this community for a while?As far as hyper threading goes, in practice my i7 960 rig saw no benefit at 4.0GHz with HT on, nor did my 2600k. *regardless of affinity mask*

The groundbreaking part about this thread is the PCIe bottleneck. I've failed misserably at exposing it in benchmarks but now not only do we know there's one because PT told us, Lars has clearly identified it his tests.Whith Ivy Bridge and the new generation of PCIe 3.0 GPU's we'll know for sure if there's a bottleneck at x16 right now or not. It may not translate into higher FPS, maybe it will smooth things out, or help with popping Autogen, I don't know but I know I can't wait to try it, hehe

Zach. I know FSX can be helped by HT with texture loading. Especially good if you use photo scenery. But you will not get better FPS.This testing is done without Reject Threshold or Bufferpools in cfg. I think they will help when we are PCIe limited. Have not had time to do proper testing on it yet. (and won't over the next few weeks either) but innitial check indicates that is the case.

I've never had a blurry in FSX since the 960. Maybe this relates to the type of flying I do? That being said, perhaps I'm being too narrow minded about my experience.

The groundbreaking part about this thread is the PCIe bottleneck. I've failed misserably at exposing it in benchmarks but now not only do we know there's one because PT told us, Lars has clearly identified it his tests.Whith Ivy Bridge and the new generation of PCIe 3.0 GPU's we'll know for sure if there's a bottleneck at x16 right now or not. It may not translate into higher FPS, maybe it will smooth things out, or help with popping Autogen, I don't know but I know I can't wait to try it, hehe

Even then, we have all been aware of that particular shortcoming (the benchmark is markedly better the second, third time around). Or so I thought. I'll let you guys test out IB before I pull the trigger.Either way, I'll be following Lars' post from now on.

Lars, did you use the settings that come with FSMark07 & 11, or your owns? what addons?I'm trying to find test scenario that doesn't scale with CPU or GPU overclock, but I've had no luck so far. Should I try a fresh FSX install?

I have these settings noted down for that testing (if you can work out what I mean, it is basically going through the settings tabs)Graphics:UnlimitedAnisotropicAAVery High (full right)Lens flareNO Light BloomAdvanced AnimationsContinuousAircraft:Medium High3-D cockpitNO cockpit tool tipsHigh resolution 3-D viritual cockpit0% transperacyAircraft casts shadows on the groundNO aircraft cast shadows on itselfAircraft landinglights illuminate groundScenery:Everything MAX/on apart from:Water High 2.xNO Ground Scenery shadowsWether:Cloud Draw distance MAX 110mi/176kmThermal Visualization NoneDetailed cloudsCoverage dencity MaximumNO Download windsNO Disable turbulenceRate at which weather change MediumTrafficHighAirline 15%General aviation 15%Airport vehicle MediumNO aircraft labelsRoad vehicles 15%Ships 35%Leisure boats 35%Sound:everything 50%Realism:HardGeneral:Everything apart fromNO show captioningCompass and pointerPilot voices Pilot 3 MaleUnit of Messurement US systemNorthernWesternJust standard FSX scenery installed over that area.Only wideview and affinitymask (13) in cfg. No other tweaks. And only in game AA at that point.No need to reinstall FSX. Just let it rebuild the cfg and apply these settings and I hope you should be able to get similar results.

I've never had a blurry in FSX since the 960. Maybe this relates to the type of flying I do? That being said, perhaps I'm being too narrow minded about my experience.

Most likely depending on what flying you do. It is very notable if you use photo scenery as the textures have soo much details in them so you will see when they are not fully sharpened and with photo scenery you don't have a lot of autogen hiding the ground textures.If you look closely you will notice that normal textures gets sharper quicker as well. But you'll have to look for it.

I have these settings noted down for that testing (if you can work out what I mean, it is basically going through the settings tabs)Graphics:UnlimitedAnisotropicAAVery High (full right)Lens flareNO Light BloomAdvanced AnimationsContinuousAircraft:Medium High3-D cockpitNO cockpit tool tipsHigh resolution 3-D viritual cockpit0% transperacyAircraft casts shadows on the groundNO aircraft cast shadows on itselfAircraft landinglights illuminate groundScenery:Everything MAX/on apart from:Water High 2.xNO Ground Scenery shadowsWether:Cloud Draw distance MAX 110mi/176kmThermal Visualization NoneDetailed cloudsCoverage dencity MaximumNO Download windsNO Disable turbulenceRate at which weather change MediumTrafficHighAirline 15%General aviation 15%Airport vehicle MediumNO aircraft labelsRoad vehicles 15%Ships 35%Leisure boats 35%Sound:everything 50%Realism:HardGeneral:Everything apart fromNO show captioningCompass and pointerPilot voices Pilot 3 MaleUnit of Messurement US systemNorthernWesternJust standard FSX scenery installed over that area.Only wideview and affinitymask (13) in cfg. No other tweaks. And only in game AA at that point.No need to reinstall FSX. Just let it rebuild the cfg and apply these settings and I hope you should be able to get similar results.

Roger that. Thanks!

...I know FSX can be helped by HT with texture loading. Especially good if you use photo scenery. But you will not get better FPS...

Hi Lars,Are you sure about this? I have been following thyperthreading studies and tests in these forums in regard to FSX since day one. Enormous time and energy with sophisticated comparison techniques have been expended over the years by numerous highly respected individuals that have reported in detail on these AVSIM threads and in other venues. The very near, if not unanimous, informed opinion has been for a long time that FSX is completely hyperthread unaware.Furthermore it is stated and restated in findings over and over that Hyperthreading produces less FSX performance due to, amongst other things, core clash and elevated heat conditions, that must be compensated for by lower clock speeds. There is a lot of weight on their side of this question, which also matches all my personal experience with multiple hyperthreaded CPUs and FSX. I am not saying you are wrong, just that I find it difficult to agree.Thanks for all the interesting information. Quite fascinating with a unique approach!Kind regards,

I have done a bit further quick investigation. (Only one run per test so don't look too detailed at the graphs)I sent my cfg through Bojote's tweaking and tuning tool and re-run some benchmarks (on the conservative setting). Not surprisingly it did help a bit. I also tried and added just RejectThreshold=131072 that was put in the tweaked .cfg to my old .cfg.I attatched another four graphs over the testing.On the first graph displays: the old result, just RejectThreshold=131072 added to .cfg, and a fully tweaked .cfg. As you can see it is mainly the RejectThreshold entry that improves FPS past 260secondsOn the second graph I use the fully tweaked .cfg and run it at different CPU speeds using only in game AA. (I only change the CPU multiple). The speeds are: 1.9GHz (10), 2.9Ghz (15), 3.8Ghz (20) my standard overclock and 4.0GHz (21). Here you can see how well FSX scales with CPU. The 50% inclease from 1.9-2.9 gives you 50% more FPS most of the time. (you can also see that at 1.9GHz FSX tries to keep the PFS from going too low around the 60second mark, on the screen it is shown with reduced autogen and reduced texture loading). But at the second half of the benchmark you can see that CPU speed does not scale at all. We have run in to a bottleneck.On the third graph i show 3.8GHz CPU speed at different GPU clockspeeds. The GPU core speeds are: Downclocked to 455MHz, stock 607MHz and overclocked to 810MHz. Yes there is a small difference but concider that the difference in GPU power is 78% between downclocked and overclocked. The return is 15% more FPS at most. We are clearly bottlenecked by something. But previous graph shows clearly that it is not CPU power during the second half of the benchmark.The forth graph shows the same different GPU speeds but with 8xS AA added. As you can see the FPS is slightly lower with additional AA. But the difference between the different GPU clockspeeds is at the same time bigger. Up to 30% FPS increase is better than what we get without additional AA but it ise still not great for a 78% GPU overclock. We are still bottlenecked.What is this bottleneck? I say it is the PCIe bus.The RejectThreshold entry should limit the traffic on the PCIe bus slightly.I also showed earlier that going to a slower x8 PCIe bus will limit the FPS a lot where that bottleneck is.Could it be the RAM ammount, speeds or timings? Already checked earlier. It's not.The harddrive? Well, I really doubt that. I run on a SSD and programs are always running from the RAM anyway. Too little RAM will cause a few big hold-ups every now and again as RAM is paged.Oh, and if we run in DX10 we get a lot higher FPS during the end of the benchmark...Is there any other thing apart from the PCIe bus that could be the bottleneck?

Hi Lars,Are you sure about this? I have been following thyperthreading studies and tests in these forums in regard to FSX since day one. Enormous time and energy with sophisticated comparison techniques have been expended over the years by numerous highly respected individuals that have reported in detail on these AVSIM threads and in other venues. The very near, if not unanimous, informed opinion has been for a long time that FSX is completely hyperthread unaware.Furthermore it is stated and restated in findings over and over that Hyperthreading produces less FSX performance due to, amongst other things, core clash and elevated heat conditions, that must be compensated for by lower clock speeds. There is a lot of weight on their side of this question, which also matches all my personal experience with multiple hyperthreaded CPUs and FSX. I am not saying you are wrong, just that I find it difficult to agree.Thanks for all the interesting information. Quite fascinating with a unique approach!Kind regards,

Yes, you can overclock slightly more without HT. 4.0GHz instead of 3.8GHz for me. But it works better with more but slightly slower texture loaders. I have attatched a small example of the difference between Affinitymask 15 no HT (top part) and Affinitymask 249 HT on (bottom part). Neither of them are blurry. The one with HT on is at lower FPS. But the one with HT on is as sharp as you will get but the one without HT is at the second sharpest.It's a bit like watching a movie. As long as the story is good you actually don't take notice if it's being shown on a DVD or on Blu-ray.

Oh and regarding the HT bit. If you don't use Hyperthreading the texture will get to its sharpest within a few seconds later compared to if you use more texture loaders with help of HT. But you can fly quite far in a few seconds, so you might not see it...

Very nice comparison there! I think you've proven yourself with HT. Though I'm with Stephen (and you mentioned it). HT means more heat and less overclock. I'm looking for FPS with my flying as TrackIR and the NGX tend to eat away at it.This helps concrete what most of us have been telling a few placebo affected folks. HT won't help the FPS (and in fact will hurt it).

Yes, using more Texture&Terrain loaders will hurt FPS when we are limited by the CPU. But when the GPU or PCIe bus bottlenecks FSX it will make no difference for FPS. You will get the highest FPS with only one T&t loader. (Affinitymask 5 for example) But too few T&t loaders also increases the load time when you launch a flight (and for the light refresh) so it depends on how good your patient is if you won't trade a slight FPS reduction for better load times.And here is where Hyper threaded cores are worse than proper cores. You will get the same FPS reduction (and texture loading) by hyperthreading two T&t loaders on one core as when using two physical cores, but load times will only reduce a little. It's not a proper extra core, remember. And there is the risk of getting fibres or other non FSX processes on the hyperthreaded core sharing resorces with the Main Thread. And that really hurts FPS.But as I have mentioned before. If the crispest textures are what you're looking for. HT really works. Othervise it is just a good way to reduce your FPS and overclock. (unless you only have a Dual core with HT=)

Lars, I tried disabling 2 cores and enabling HT. CPU usage never went 100%. Isn't that a good indication that HT is not doing anything at all?

Lars, I tried disabling 2 cores and enabling HT. CPU usage never went 100%. Isn't that a good indication that HT is not doing anything at all?

No, HT will still work. You should never aim to get 100% CPU utilisation when using HT. The Hyperthreaded core sharing the same physical core as the main thread should be as close to 0% as possible. Therefor reducing the CPU utilasation.If you use only two cores, try it without hyper threading and with affinitymask 3 (or none for that matter). Then try it with HT enabeled and affinitymask 9 (or none, but it should give slightly more fibers on the same physical core as the main thread)Do you see any difference? If you don't, tick the DX10 box (will create more fibres) and see that it does work=)Here are a few more graphs showing what I said about FPS reduction with more T&t loaders. And also backing up what I said about impact of RAM. I did them over a year ago when I did my innitial post. They use slightly different settings and DX10. And that gives the benefit that we are CPU limited pretty much all the time with the GTX470Worst realistic case for the RAM is leaving in dual chanel but setting the slowest speed/latency the motherboard would allow. Absolute worst case is putting them in Single Channel instead of Dual channel as well. Nothing you would do, but you can...Feel free to look back at the first post in this topic and the original topic this one is an update to. I'm away for a week or two now. Enjoy the flying=)

Can you really see the extra T&t loaders doing something with HT on?I ask because most applications where HT helps have all 8 virtual cores working at 100% all of the time.Nick Needham gave an explanation to it, something along the lines of FSX spawing texture threads to every virtual core, but not really loading anythingOne noob question please. I've run some tests with FSMark07 & 11 at different clockspeeds. How do you merge the graphs in Excel using the same time X axis?

Well, I've been trying, but I can't reproduce your results. I'm 100% CPU limited in both FSMark07 & FSMark11 during the entire test. Both with AG set to Extremely Dense

No, HT will still work. You should never aim to get 100% CPU utilisation when using HT.

Respectfully, I don't think so. I just tested at 4.6 GHz, which is as fast as I want to go with HT on. I might be able to squeeze a bit more out at 4.8 GHZ or so and still remain stable, but I am not into it just for this test alone. I ran a number of flights over ORBX PNW with HT on and HT off. Any difference at all would have to fall into the none HT advantage, for subjectively it appeared smoother and textures just as clear or clearer than with it on. Heat was up with HT and the CPU seemed to be working harder with the fans increasing their RPM. Lots of noise and energy, but absolutely no performance or texture loading advantage. If anything both FPS and texture loading was a bit slower with it on.I did not record the numbers, for I don't want to expend more time and energy to verify what has already been known and understood for years. Hyperthreading does not provide any advantage in FSX whatsoever, and indeed the very opposite is the case. In my opinion something must be amiss in your equipment, the monitoring or your analysis if your data shows otherwise. Mistaken findings can lead people into making wrong hardware decisions, something we all want to bear in mind. I would not want to be responsible for someone buying a hyperthreaded CPU just because they thought it would be beneficial in FSX due to having read this thread.That being said, I enjoy your completely independent thinking and view on how FSX works, even if you are not 100% right on everything, including hyperthreading.Kind regards,