PART 1 starts here if you missed it: http://forums.avsim.net/dcboard.php?az=sho..._id=37203&page=There is no mistaking the visual difference between the lower speed memory and higher speed/lower latency memory settings. Here are compares of the better Test 2 results with the test 3 results:DDR2 900 / DDR3 1440http://forums.avsim.net/user_files/187085.jpghttp://forums.avsim.net/user_files/187086.jpghttp://forums.avsim.net/user_files/187087.jpghttp://forums.avsim.net/user_files/187088.jpghttp://forums.avsim.net/user_files/187089.jpgThose do not show a placebo :)lets talk about the flights now.... --SYNOPSIS OF RESULTS-- Avg Test 1 DDR3 900 (tRD 12) + FRAPS = 17.13Test 1 DDR3 900 (tRD 12) - FRAPS = 17.65 Difference = +.52 Combined Avg = 17.39Flight Synopsis: Overall, stutters and blurs throughout the flight (cockpit and exterior), poor terrain quality up close and in the distance althought it started to clear up (somewhat) after the city and airport area were left behind. Autogen was slow to come into the scene forcing me to really lose one aspect in the sense of reality autogen provides during a sim flight. Some textures would snap in sharp as approached however a large percentage remained in a low LOD state and never reached their 1M detail even passing over them. None of the textures in the images presented obtained much of any clarity. Avg Test 2 DDR3 900 (tRD 8) + FRAPS = 18.63 Test 2 DDR3 900 (tRD 8) - FRAPS = 20.80 Difference = + 2.17 Combined Avg = 19.72Flight Synopsis: Overall, reduced stutters and blurs as compare to Test 1, cockpit stutters were greatly reduced especially in turns however exterior flight still displayed detrimental stammer. The speed and volume of autogen appearing increased somewhat and terrain texture details were better defined but still unacceptable for what would be considered an enjoyable flight experience. This flight was much better than the first since cockpit stutters decreased and if a user remained in the cockpit at all times their experience would be much better. None the less, high impact areas such as the airport and around downtown car traffic still produced unacceptable performance issues. Some of the textures in the images presented obtained a better clarity. (shown above in the compare) The resulting visual and performance increases that occurred were due to one BIOS/memory timing change, tRD, nothing else. Avg Test 3 DDR3 1440 (tRD 7) + FRAPS = 20.11 Test 3 DDR3 1440 (tRD 7) - FRAPS = 21.45 Difference = + 1.34 Combined Avg = 20.78Flight Synopsis: Smooth, sharp detailed flight, little or no blurs or stutters. Autogen speed/volume was quick to appear in the distance, terrain textures maintained their 1m LOD within the defined boundary and distance visuals were clear. A few very minor stutters and distance blurs were noted (recovered quickly) when approaching/crossing large scenery changes (two areas) and high disk access was invoked, otherwise no stutters were observed anywhere else. The aircraft responded as if it were floating gracefully on air as it should. This would be the type of flight users would wish to strive for given the current hardware limitations and FSX. All of the textures in the images presented maintained defined 1m clarity. (shown above in the compare)No significant difference between FRAPS and the FSX frame counter notedThe frame rate increase between DDR3 900 (tRD12) and 900 (tRD8) = +13.4% The frame rate increase between DDR3 900 (tRD8) and DDR3 1440 (tRD7) = + 5.38% CONSIDERATION OF RESULTS:I would probably pay more attention to those +3% to +5% TRUE average increases in benchmark frame rates in conjunction with the lower timing, lower latency memory products

Almost forgot.. You will find information about how to deal with the FSB/STRAP(divider, mem-speed and CAS in relation to tRD here for most systems:http://www.anandtech.com/mb/showdoc.aspx?i=3208&p=9The graph as the bottom (green areas) can help most DDR2 users find that sweet-spot. DDR3 users should look at the 5:8 RATO CAS 6 450FSB chart (2nd chart) at DDR3 1333 and abovePlease be careful if you memory is not of good low latency and requires a Vdimm increase. You can review CAS in relation to memory speed and CPU here: http://www.thetechrepository.com/showthread.php?t=195The first step is to find the lowest value the memory woo run @ CAS within same memory voltage. Usually the manufacture of the sticks will have good information about the limits in their forums.You can get to tool MEMSET and review a list of definitions for memory timings here:http://www.tweakers.fr/timings.htmlbe VERTY careful with that tool!

That took some effort Nick. Thanks!Now, is it hopeless with my system to get the sort of performance you describe? I am running low end DDR2 at 1600 (400x4) and have increased the timings from the default 5-5-5-18 to 4-4-4-12. I haven't tested to see if timings can be improved. I don't really understand any of these variables. And I know from reading other forums that this DDR2 cannot be clocked signficantly higher than 400.Given what you see in this image, what should I try adjusting to optimize my configuration for the best I could get? I definitely don't want to start over with a new mainboard and pricier memory, but would like to optimize my system. I am using a P5E and there is good control over memory timings on this board, I think . . .I haven't installed FSX yet. http://forums.avsim.net/user_files/187093.jpgA couple of questions though . . .It's too early in the day for me to fully critique the validity of your report. If you look at the difference between the fastest timings in the DDR3 setup, versus the lowest timings and FSB that you see in my setup, what sorts of percentage differences are we seeing, maybe in terms of efficiency? 15%? 30%? I always wonder about validity & objectivity when ANY subjective tests are done with FS9. You can set up a scenario (at least with my older slower P4) and find two "identical" test runs perform very differently in terms of smoothness, sharpness, or other subjective measures, despite your best effort at setting up identical scenarios. How much of your analysis is subject to this sort of subjectivity, or just variability of actual outcome due to unknown factors?Anyway Nick, if you could advise me on setting up my snail-driven memory to get something that will be at least optimized for my hardware that would be much appreciated! Maybe a short treatise on the basic principles of optimizing the various timings would help. I plan to run my QX at 4.3Ghz with FSX.NoelQX9650 w/ Retail HSF|8GB Muskin PC-6400|ASUS P5E|EVGA 8800GT @700|Seagate SATA 2 x 4|Seagate Cheetah 15K.x|XP Pro|Vista 64--soon to be installed

Whoa. This is heavy stuff. I am going to have to take some time and really get into this. I do think, in 2 years, DDR2 will be a memory. Pun intended.RhettAMD 3700+ (@2585 mhz), eVGA 7800GT 256 (Guru3D 93.71), ASUS A8N-E, PC Power 510 SLI, 2gb Corsair XMS 3-3-3-8 (1T), WD 150 gig 10000rpm Raptor, WD 250gig 7200rpm SATA2, Seagate 120gb 5400 rpm external HD, CoolerMaster Praetorian

Excellent posts Nick, Thank you. My memory is the one area that I want to improve but, is also the area in which I have the least experience. This is a big help.Best Regards,JeffWhoa! Where's the tylenol:-lol

Nick,Thanks for putting this together for our benefit. I am knee deep in uni assignment work at the moment, so I won't be able to read the whole thing in detail until the weekend and provide some comments. Once again, I am sorry that our recent discussion in this area went south and, at face value, it looks as though the work you have done here will be useful for everyone.Gary

Is there much of a difference between DDR and DDR2 in FS9?RH

GaryPlease think of what was shown in an every day situation, a traffic flow/jam situation with everyone trying to get to work with a twist.. The twist is they are all going to a business that has an unlimited number of employees and the profitability of that business is defined by how many employees get to work by leaving at 7am and arriving at 7:30am and there are no legal speed limits between point A and point B.The distance the workers need to travel is exactly the same for all of them to get to their destination and start work for the day. That distance is hindered by one traffic light and the width of the road and number of lanes, and, the engine abilty if the cars. In this twist, the business can only be as efficient and profitable as the number of workers who arrive and start their day, on time (7:30am) and they all leave home at 7AM. The road = is the bandwidth. Its size and number of lanes represents how much traffic can use that roadThe cars = are the data requests. Those requests increase as the sliders and the demand for more is increased by the user.The traffic light = the latency or the amount of cars (data requests) that can successfully reach their destinations in X time based onthe width of the raod and is defined as:The RED/GREEN light = MCH READ DELAY (tRD) and FAB4 The CPU/FSB = the engines the cars all have. The same throttle position of the accelerator pedal therefore defines a different speed in which the cars travel down the road, and, how may get through the light when it is green because their acceleration ability has been significantly increased.LOGIC:If the number of cars (data requests) increases and their engine ability (CPU/FSB), the bandwidth (road/lanes) and the traffic light timing (latency) remains the same, the resulting profitability of the business will remain the same because the road can only handle so much traffic. The cars can only go so fast, and, the light will only allow so much through in a

Noel I don

FPS analysis can only have value if the user is getting the total flight they expect from FSX, and, they know what to look for when testing.To apply a true and defined FR number to hardware perf is impossible in that respect. Even if the user is getting the flight they expect with a FR value, that value is subjective to the user.As mentioned in the past, if we had access to a report that displayed what was actually being successfully rendered in the flight, and more important, the time the system was taking to render the different elements of a flight, we would have data to pit against FPS for analysis. In that, a performance value can be assigned. We must have the frame rate pitted against and the other factors of the flight which have been tabulated to produce a correct

Nick N,You are da

Nick, thanks for simplifying the interpretation of all that. I think my low-budget ($39 for 2GB modules) mushkin DDR2 is really doing about as good as it can at stock settings. Windows will not boot with only a small increase to 410, upping volts from 1.8 to 1.86v. I've read this before, about this memory that is, from other users. Works great at default settings, and that's about it. Not good for overclocking. Sam convinced me there is really no need to consider FSB bandwidth, so this is why I chose to go with an X38 w/ DDR2 memory. It was also unable to run with a command rate of 1, however again this is at 1.86v. So, I was able to boot with tRAS of 10, but I haven't done any stability testing. Pretty soon I will run FSX on Vista 64, and hopefully I'll see decent performance at 4.3GHz with these memory settings. I think what truly matters, is smoothness, with a frame rate adequate to prevent a slide show. When you drop down below a certain threshold in FPS, smoothness becomes secondary. But beyond maybe 18 or 20, smoothness per se is probably the bigger concern. I can tell you FS9, fully tricked out with max everything, PMDG heavy, runs truly as smoothly as you could detect, and very fast, in the 60-70 fps rate, or higher. And that's only with a CPU clock speed of 3.8GHz. I guess it's time to wait n see.Thanks again, truly . . . QX9650 w/ Retail HSF|8GB Muskin PC-6400|ASUS P5E|EVGA 8800GT @700|Seagate SATA 2 x 4|Seagate Cheetah 15K.x|XP Pro|Vista 64--soon to be installed

Here's a couple of plain-jane DD2-800 shots. [email protected] / 8800GT@700 / 4g-O-DDR2-800 5-5-5-16/ P5K-e, 400mhz FSB, 1.4Vcore and All other bios settings in auto/default. GEX ( . . . that saved this sim. Sincere thanks). FSGenesis. UT and UT Urban LandClass - on top. SceneryTech LC - on deck. All left side scenery sliders full right ('cept water).http://forums.avsim.net/user_files/187150.jpghttp://forums.avsim.net/user_files/187151.jpghttp://forums.avsim.net/user_files/187152.jpghttp://forums.avsim.net/user_files/187153.jpgBoy, I don't know. I can see it, but it's like my gazillion $$$ stereo. The difference is simply stunning . . . to me. In any case, the visual fidelity discrepancies seem to not be entirely dependent on ram characteristics. There is clearly something else reducing the fidelity of the "before" shots. I think it more depends on what airplane you fly. That P51 is soooo sweet!

Sam what the heck are you saying with regard to those screenshots? I still don't know what to believe. I asked Nick this question: what is the percentage of mismatch between the max ouput from the CPU and the capacity of the FSB bandwidth? Will my processor's max output potentially exceed the carrying capacity of my FSB, thereby amplifying the effects of memory bandwith on overall performance or smoothness? I think this is the valid question, and would love get a sense of the numbers involved, the percentage of mismatch if you will. QX9650 w/ Retail HSF|8GB Muskin PC-6400|ASUS P5E|EVGA 8800GT @700|Seagate SATA 2 x 4|Seagate Cheetah 15K.x|XP Pro|Vista 64--soon to be installed

NoelI am not going to get into an argument with anyone. The information posted in this thread can be repeated on any systemMy system was set up using the config edits and the settings listed in part 1 and I did not vary from the FSMark07 default flight aircraft. I booted that flight directly and allowed it to select the aircraft. There were no instructions about selecting an aircraft after the FSMark07 flight started so I left the aircraft which booted by default run the test since the FLT file calls Sim=Bombardier CRJ 700 Paint1 for that test.My config file and FSX settings are set up to deliver the goods with the DDR3 system @ 1440The bottom line is, the DDR2 type memory speed and timing could not handle my settings and configuration file edits.You can believe what ever you wish. This test and the parameters of it can be repeated on any system however in order to duplicate the result, a user must use ALL the FSX settings AND edit config file exactly as I did and they must be using UTX and the default AI traffic, also set as I specified, and run the FSMark07 benchmark correctly as Gary specified in his instructions. the only difference would be the WIDESCREEN entry if the user is not working with a wide screen display. I am also running 1920x1200 which may not be possible for another system to do and may also play into a compared result.There will naturally be differences in how one system may respond over another. The motherboard, BIOS and attached products will all define a different result. Some motherboards have a good BIOS programmer behind them and as such the BIOS on some systems will have well considered timing math behind them. This presentation was not about proving anyone wrong. It was about showing the benefit of high speed, low latency memory use in conjuntion with the correct STRAP and CPU speed, and, to look at the frame rate to see if there may be a linear response to the changes, that is all.Even if I remove the flights that were not FSMark07 specified (exterior view and non-fraps frame recorded) the results as specified by Gary's benchmark and FRAPS recorded are the same in nature and display the same result..Flight 1 Average FPS = 16.70 - DDR 900 tRD12Flight 2 Average FPS = 18.65 - DDR 900 tRD 8Flight 3 Average FPS = 19.79 - DDR 1440 tRD 7Increase of FPS from Flight 1 to Flight 2 = 11.68% = Small change in performance and visual qualityIncrease in FSP from Flight 2 to Flight 3 = 6.1% = Large change in performance and visual qualityThis tends to suggest the frame rate increase is not linear with the performance displayed. We do not have any readout information that will tell us what is going on with FSX behind the scenes so there is no way to compare the frame rate to the sucessfully completed programming requests and how fast those code requests were completed, therefore visual became the tool for this examination.