@Stephen, I really would appreciate an actual test run of FSMark07 Global High from your system whenever you get a chance - I realize that it is just a single data point and doesn't show the remarkable improvement in smoothness. I wish that maybe there was a "bell curve" variance statistic for the whole run that focused on setting an internal/external fps limit (even though that changes a lot of things in how FSX runs) like 50 fps and then reported on what ratio of frames kept to that. Anyway, thanks for the general report on 30->40% faster fps with respect to a 930 at 4.0 Ghz or similar PC. The company that I am helping has both a 980X @ 4.0Ghz and a 2600K (but only with 1333 Mhz mem as yet) and I measured about 10->20% improvement over the 980X - so they might be nearly equivalent depending on the tuning, but I did think that the 2600K was smoother - no idea why. Again, what intrigues me is the comment in this thread by HLJames on L3 cache (uncore) tuning. I cannot as yet find any discussion on overclocking sites about that issue. He is suggesting a 30% improvement if you understand and tune for that issue.

Hi Dave,Just to clarify I was comparing it to 920s and 930s at 4.2 to 4.3+ over very heavy autogen and traffic.Kind regards,

Hi Friends,Gulftown has a wall at 4400MHz Uncore L3 cache speed; one or two clicks back and you should be about 4377MHz. CPUZ should now report your QPI Link at 4377MHz. This is the maximum performance you can get out of Flight Simulator!Most overclocking mainboards in order to keep you away from this wall will automatically drop the clock multiplier on the uncore L3 cache - DON'T LET THAT HAPPEN!!And now the Good news for Flight Simulator fans: Sandy Bridge has now taken the L3 cache out of the uncore and it now runs at full floating point core speed and voltage.Another way to compare the two microprocessors: gulftown 4377MHz vs Sandy Bridge 5000MHz.HLJAMESCore i7 2600K @ 5.2 GHzAsus P8P67-DCorsair Dominator GT, 4G @ 2133 MHzEVGA GTX 260WD Velociraptor 600GBEnermax Galaxy 1000WANTEC P183Swiftech MCR 220 Drive/Apogee XT/ feser onewindows 7 HP-64bitFlight Simulator x Gold ( no modified files)FSXMark07, Global High Cfg: 90.873 fps avg

Hi Friends,Gulftown has a wall at 4400MHz Uncore L3 cache speed; one or two clicks back and you should be about 4377MHz. CPUZ should now report your QPI Link at 4377MHz. This is the maximum performance you can get out of Flight Simulator!Most overclocking mainboards in order to keep you away from this wall will automatically drop the clock multiplier on the uncore L3 cache - DON'T LET THAT HAPPEN!!And now the Good news for Flight Simulator fans: Sandy Bridge has now taken the L3 cache out of the uncore and it now runs at full floating point core speed and voltage.Another way to compare the two microprocessors: gulftown 4377MHz vs Sandy Bridge 5000MHz.HLJAMESCore i7 2600K @ 5.2 GHzAsus P8P67-DCorsair Dominator GT, 4G @ 2133 MHzEVGA GTX 260WD Velociraptor 600GBEnermax Galaxy 1000WANTEC P183Swiftech MCR 220 Drive/Apogee XT/ feser onewindows 7 HP-64bitFlight Simulator x Gold ( no modified files)FSXMark07, Global High Cfg: 90.873 fps avg

Very interesting info. ThanksDo you have a link to some FSX "review" made with different uncore speeds on a 1366 Nehalem or Westmere please?

@HLJames, Thanks so much for getting back to us on the L3 cache, uncore "tuning" that you spoke about earlier. I found an article on Anandtech : http://www.anandtech.com/Show/Index/3922?cPage=5&all=False&sort=0&page=4&slug=intels-sandy-bridge-architecture-exposed that said:Also unlike Westmere, the L3 cache now runs at the core clock speed - the concept of the un-core still exists but Intel calls it the “system agent” instead and it no longer includes the L3 cache. With the L3 cache running at the core clock you get the benefit of a much faster cache. The downside is the L3 underclocks itself in tandem with the processor cores. If the GPU needs the L3 while the CPUs are downclocked, the L3 cache won’t be running as fast as it could had it been independent. The L3 cache is divided into slices, one associated with each core although each core can address the entire cache. Each slice gets its own stop and each slice has a full cache pipeline. In Westmere there was a single cache pipeline and queue that all cores forwarded requests to, in Sandy Bridge it’s distributed per cache slice.It seems that you are saying that while a little tuning is required for the 980X to get its L3 at just before the wall, the Sandy Bridge doesn't need that L3 tuning. Now for the key question to you! Why are many Sandy Bridge owners like GaryA with a system like the following:HAF932,GA-P67A-UD7, 2600K @4.5,NH-D14, 2x4GB DDR1333 9 9 9 24 1T @ 1.5v, AX750PSU, 600GB Vrap for FSX, W7-64 on 300GB VRaptor, 3Ware 9650SE, GTX580, Asus VE276.Getting the following FSMark07 performance:17975 -- 300000-------- 52---- 62 --- 59.917 But you are getting:FSXMark07, Global High Cfg: 90.873 fps avg I see that you are at 5.2Ghz while he is at 4.5 and you are using 2133 Mhz memory and he is just at 1333 - is it just those factors? In other words, if we can get the early adopters of SB to run FSMark07 Global High, we should see several of them in the 80s ? Also, I would think that your system ought to be able to run complex scenery and PMDG aircraft + severe weather in the 40 fps arena (my fudge factor is half the FSMark07 results). Yes?

Now for the key question to you! Why are many Sandy Bridge owners like GaryA with a system like the following:HAF932,GA-P67A-UD7, 2600K @4.5,NH-D14, 2x4GB DDR1333 9 9 9 24 1T @ 1.5v, AX750PSU, 600GB Vrap for FSX, W7-64 on 300GB VRaptor, 3Ware 9650SE, GTX580, Asus VE276.Getting the following FSMark07 performance:17975 -- 300000-------- 52---- 62 --- 59.917 But you are getting:FSXMark07, Global High Cfg: 90.873 fps avg I see that you are at 5.2Ghz while he is at 4.5 and you are using 2133 Mhz memory and he is just at 1333 - is it just those factors? In other words, if we can get the early adopters of SB to run FSMark07 Global High, we should see several of them in the 80s ? Also, I would think that your system ought to be able to run complex scenery and PMDG aircraft + severe weather in the 40 fps arena (my fudge factor is half the FSMark07 results). Yes?

Looks to me like Gary's results are so close to 60 FPS because of Vsync. He stated he had Vsync on when he run that test

Thanks, that might be it - I have PMd Gary A. to please rerun that test with Vsync off - maybe it might get up into the 70s. If you know of anyone else that has a Sandy Bridge system and would run FSMark07, please encourage them. I haven't seen Doug Horton come out with an article that covers it yet. I do feel that a top-end Sandy Bridge system (5.0 Ghz + 1800 Mhz mem) has a significant improvemnt over the i7 results - and they were pretty good for everyone except the PMDG plus ORBX users. It would be great if we can all achieve "close" to the results that HLJames is getting. I'd like to see the results that JJJallen gets - he is oclocked to 5.3 Ghz I believe. I am using the extra performance for a three monitor system - a single SB PC driving three independent undocked views with wide field of view and well performing scenery like FlyTampa's Hong Kong (no visible 3D aircraft cockpit - a fourth monitor for gauges) - trying for about 50 fps and smoothness! It will be great when the motherboard manufacturers get back to production - glad I wasn't the Intel engineer that screwed up the chipset design.

Thanks, that might be it - I have PMd Gary A. to please rerun that test with Vsync off - maybe it might get up into the 70s. If you know of anyone else that has a Sandy Bridge system and would run FSMark07, please encourage them. I haven't seen Doug Horton come out with an article that covers it yet. I do feel that a top-end Sandy Bridge system (5.0 Ghz + 1800 Mhz mem) has a significant improvemnt over the i7 results - and they were pretty good for everyone except the PMDG plus ORBX users. It would be great if we can all achieve "close" to the results that HLJames is getting. I'd like to see the results that JJJallen gets - he is oclocked to 5.3 Ghz I believe. I am using the extra performance for a three monitor system - a single SB PC driving three independent undocked views with wide field of view and well performing scenery like FlyTampa's Hong Kong (no visible 3D aircraft cockpit - a fourth monitor for gauges) - trying for about 50 fps and smoothness! It will be great when the motherboard manufacturers get back to production - glad I wasn't the Intel engineer that screwed up the chipset design.

Memory speed is really not that important. A few % at most but nothing you would really notice other than in benchmarks.I tested 1333 CL9 vs 1600 CL6 and there was a 4% performance increase at mostI have tons of addons and I really don't want to mess with my FSX installation ATM, otherwise I would run FSXMark07. If I have some time in the weekend I'll run it on a fresh FSX install so we can compare things under the same conditions, but as I already posted, in my testbed flight in UK2000 EGLL + PMDG744 I was getting almost a 40% extra frames. What do you mean by "everyone except the PMDG plus ORBX users"? did those not see a performance boost with SB?

@dazz, "and they were pretty good for everyone except the PMDG plus ORBX users" - I realize that the PMDG+ORBX users are seeing much improvement - not sure whether it is linear vs FSMark07 - but I just meant that many of you complex element users are still at 20->30 fps (but hopefully smoother) - compared to those of us with less complex aircraft and scenery who might be at 60 fps. Obviously, it all depends on what you accentuate (scenery, aircraft, AI, clouds, etc.) for your own interests. Sorry for my choice of phrasing.EDIT:GaryA (4.5 Ghz + 1333 Mhz mem) FSMark07 retest with Vsync off.Same test but vsync off:Frames--Time (ms)--Min--Max--Avg21888----300000----54----92--72.96 This is much more in line with HLJames report of 79 fps with his clock at 4.8 Ghz and 1333Mhz mem (he gets nearly 90 though with 2133 Mhz mem)

http://www.youtube.com/watch?v=kio97WGg1PIFS 2007 benchmark using global Hi config, captured with fraps. Details of config and capture in the movie details.The movie intended to represent "smoothness" with some global reference we can all appreciate. ...I think there's enough headroom in this rig to captured at 60fps full screen.Ken

update:3dMark06 CPU bench @ 5.2 (51x102.6) GHz is 9820 ....geez, I might have the fastest air cooled pc on the planet. :-)Oh, Prime95 and FSX luvs it. 72c max temp (fsx), 67c with Prime95 and 3dMark06. Ambient room temp was 70F.http://3dmark.com/3dm06/15445271

update:3dMark06 CPU bench @ 5.2 (51x102.6) GHz is 9820 ....geez, I might have the fastest air cooled pc on the planet. :-)Oh, Prime95 and FSX luvs it. 72c max temp (fsx), 67c with Prime95 and 3dMark06. Ambient room temp was 70F.http://3dmark.com/3dm06/15445271

Unusual to see higher temps with FSX than Prime.

@HLJames, If you have any details about your BIOS settings (or other tuning parameters) that you could share, it would help us understand how you manage to get such high FSMark07 results - even when you aren't overclocked to 5.2Ghz or so where some are. Also, you upped your results at 2133 Mhz mem a little further, from 87 to 90. Could you please explain what happened there? It seems that your system particularly is well tuned. As far as L3 cache speeds and usage, I wonder if there is anything other than AffinityMask and maybe process priority "higher than normal" settings that we could try to keep the main process of FSX pinned to the same core (and exclude other processes from running on it!) and keep the so-called resident set of instructions as much fully contained in L3 cache as possible - this certainly might help get FSX to run as fast as possible. It was also said that there wasn't much improvement with upping memory from 1333 to 1600 - just 4%. Well, if that continued, from 1600 to 1866, 1866 to 2133, we might just get 12% or so - HLJames provided the data - he went from 79 to 86 or maybe a little more (90?) fps from 1333 to 2133. Personally, that could just be what is needed to get to a better FSX experience.

Unusual to see higher temps with FSX than Prime.

Yeah, with FSX it's really hard to get the cpu past 70% cpu load or about 60 to 70 watts (FS Dreamteam JFK, 100% WOAI traffic (major airlines), GTX 275 1680x1050 with ******* automated tweeks and FPS limiter). At this load FSX temps are in the 50s c. I can get the load up to around 95% (105 watts) with full screen FRAPS capture enabled. The temps at this load are in the hi 60's tickling 70c.Your right about Prime95. I usually only get about 25% cpu load and the temp range is 40's to 50's c.-KenI5 2500k @ 5.2ghz (51x102.6) CPU vid 1.435

I did get a successful boot to windows 5.4ghz. -just reporting that I do have a special processor. :-)I stayed in Windows at 5.4 just long enough to capture CPUid at 5400x100. Did not feel confident that windows was stable enough to run a bench, and I'm not going to force the issue with my Hyper 212 Plus air cooler.Looks like 5.0 will be my default setup.

Anybody interested in creating an "Official" FSXmark07 results thread? I was just about to do it, but figured we should have the details ironed out and agreed upon before doing it so that the "official" rules can be in the opening post. I was thinking brief setup instructions and a screenshot of each of the settings tabs... this was spurred by the fact that the included Global High configuration provides some pretty odd settings here on my computer... and also by the individuals who don't really care about non-FSX benchmarks.Somebody say yes and I'll propose the "official" rules