>Excellent prose as always.>>Signed, >>The Novice>haha, you're no novice Noel.I think with all of our (technically savvy) minds working on the same common goal, we're bound to have super-smooth sims...Someone coined the term "wholistic system building" one time, and that's what it's all about.Wait til I get my Core2 built. All of you have guinea-pigged Core2's and various hardware for me, now I swoop in and benefit from it all.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

>Which reminds me, again I don't get an intelligent answer with a very simple question >posed: can you please define CALWI? Not what the acronym represents, but in simple >terms what relevance does CALWI have? This is where I am getting frustrated.. I don

>This is where I am getting frustrated.. I don

Nick, please pardon me for any frustration you may have incurred from the questions I have been posing. I tell you, this discussion is improving my knowledge base on really what matters with this subject. I truly appreciate your patience, stretched as it has been.Here's my last question on this topic, which will put it all together. I am asking you for an opinion based on theory only, since as you point out, you can't know the details of the components involved. Anyway, here goes . . .Which of these two option *should* run FSX better:Option 1: my current mushkins, overclocked beyond all recognition as I don't know how this is happening with DDR2-800 modlues: FSB 440/5-5-5-12/DRAM 1103Mhz @ 4:5 divider/2T Command Rate/CPU @4.19MHz. Everest scores at 53.8ns/9350Mb/sec read rate. CALSI value of 9.1ns. OR . . . .Option 2: this is OCZ's fastest rated 1GB DIMMS:FSB 400/3-4-4-15/DRAM 800Mhz @ 1:1/1T Command Rate/CPU at 4.2Ghz.This OCZ Titanium DDR2-800 is rated at 3-4-4-15 and users are reporting 1T command rate, and some have even said these timings are flying with 4 modules are installed. CALWI value of 7.5ns.This comparison wants to pit CALWI against DRAM frequency, and FSB frequency for their relative importance. This being said, I'll go out on a limb and guess highest DRAM & FSB Freq , will always beat lower DRAM freq & FSB given the same OPTIMIZED CALWI of say 7.5ns.Tell me your best prediction on this, provided the OCZ performs as rated. This OCZ is not too exorbitantly priced at $78/kit, but I STILL don't have a sense of how important CALWI is versus FSB/DRAM freq & for example, Everest scores. I hope you can appreciate how the answer to this question is not so obvious.And, here is an ultra low latency Kingston that would have the same CALWI if it performs according to its stated timings:Kingston Technology 1GB KHX6400D2UL/1G 800MHz DDR2 Non-ECC Ultra Low-Latency CL3 (3-3-3-10) DIMM Memory ModuleThese are $50, but don't appear to come in 2 x 1GB kits. Is this an issue, buying modules in kits? I don't know if these can do 1T command rate, but the OCZ's aren't of course guaranteed to perform there.Thanks Nick for your insights.

It's a toss up because of the bandwith vs the latency --but-- if the OCZ sticks will indeed run CAS3 and 1T I would put my money on the lower bandwidth, lower latency because it is hitting all the 'notes' over CAS 5 2T @ 1100Noel, its simple.. the lowest CAS at the highest memory speed which places the memory in a 7.0-8.0ns range, the lowest tRD and 1T will net the best result, period. FSB is based on what gets CALWI where it belongs, be it 350FSB or 450FSB however the formulas dictate 400 minimum for DDR2 and 450 minimum for DDR3 and that is because of how the divider (STRAP) works. This is all done with NO BIOS memory optimizers enabled. All that remains OFF until it is found that they may be needed because formula CALWI/CAS is not possible. Where CPU sits is based on where it is stable and not running hot with a multiplier increase after FSB/STRAP/tRD/1T/CALWI is achieved. It need not be 4.XXMHz to accomplish the best result. It

>Noel, its simple.. the lowest CAS at the highest memory speed>which places the memory in a 7.0-8.0ns range, the lowest tRD>and 1T will net the best result, period. FSB is based on what>gets CALWI where it belongs, be it 350FSB or 450FSB however>the formulas dictate 400 minimum for DDR2 and 450 minimum for>DDR3 and that is because of how the divider (STRAP) works.>This is all done with NO BIOS memory optimizers enabled. All>that remains OFF until it is found that they may be needed>because formula CALWI/CAS is not possible.Well, if the OCZ 800's perform as it sounds most are verifying, then they should generate the CALWI value of 7.5ns with a 1:1 ratio, FSB at 400 and 1T command rate. So that sounds like it is in the Window of Interest as it were, or so the chart suggests. Anything over 400 at 1:1 is bonus it would appear. >Where CPU sits is based on where it is stable and not running>hot with a multiplier increase after FSB/STRAP/tRD/1T/CALWI is>achieved. It need not be 4.XXMHz to accomplish the best>result. My CPU is ultra cool, with my cooling solution in place it stays in the range of 52 to 57C max at absolute full load at 4.3Ghz, and that is with the max specified vCore at 1.45v. So, I'm guessing this is better than 3.8Ghz, provided overheating, heat-induced down throttling are not coming in to play. And from what I understand, they wouldn't be at those temps. Correct me if I'm wrong please . . .>However since you see no differences in your tests with FSX I>really see no reason to bother. You have to remember though, and I'm taking this from hearing your presentation of CALWI, all other modules and configurations I've tested have not been in the CALWI HG zone of 7.5 to 8.5ns as you kindly pointed out. So, this is why I ask, how much of a role does CALWI 7.5 to 8.5ns play versus better read rate for example. I've not had a chance to make this comparison so it's intriguing. Fortunately, my experimenting will only end up costing $50 or $60 to test and learn, so that is in my budget with no remore :()Here's food for thought: since ultimately latency and bandwidth matter, how about a score based on CALWI value multiplied by some other variable or two, such as bandwidth, read rate, to generate a potential configuration score? Now we're talking! Use the reciprocal of CALWI since lower is better, and perhaps another variable to take into account the relative values of staying in CALWI zone versus going for more bandwidth in one can make this deterimination. It's just a thought, and maybe someone has done it.Also, as it turns out, Kingston has some 3-3-3-10 DDR2-800 also, but I don't know command rate it will do. They sell for about $50 per 1Gb module. Do you happen to know how critical it is to get a "kit" of two modules, versus 2 same model modules from different packing? Nick I'm tempted to keep going with this and see what role hitting CALWI plays, in itself, with bandwidth that is similar to what I have now. Your arguments obviously peequed my interest and so, as you say, it's fun to play . . .What I don't know (waiting for a comment on the mushkin forum) is whether or not setting my DRAM voltage setting at AUTO overpowers my lowly modules at 1103Mhz. This config produces the best Everest scores, which I understand now are only so important. I get a pretty hefty improvement in latency & read rate with this, but it's troublesome not knowing if my memory is being slow cooked. It's certainly not hot though, or the heat sinks aren't--they are COLD to the touch, which has to help . . .

>Not sure if there is any way to improve this, but I am using>trilinear filtering in game and 16xAF in the video driver. >Water reflection just looks so marginal compared to games like>Far Cry & Crysis. Any way to improve the shimmering in dense>forest areas in the NW and also with water reflection? Pretty>marginal for this day and age. I don't have this issue in>FS9.>>Noel>>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 installedTry the 'clamp' option in nvidia control panel for 'texturing filter - negative load bias'.

> Your arguments obviously peequed my interestThey are not arguments. They are facts, electronic engineering is not a theory.As for the OCZ's... you would be better off raising CAS to 4, maintaining 1T and setting their speed to DDR2 1000 with a ratio of 5:4 if they will do it. If not, you're better off at DDR2 800 as specified. You will however need to increase Vdd to stabilize.http://forums.avsim.net/user_files/188112.jpgI have not worked 'regularly' with DDR2800-1000 in over 2 years so I have no basis to make any assessment of your results. It was before FSX was delivered to the market. The last DDR2 system I ran was DDR2 1200+.

"Speed costs. How fast do you want to go?" 10% performance increases will always be available at an increasingly extravagant cost. This is the nature of this hobby and will always be. (Let's not get started on cascade cooling!) That CPU clock increase from a Q6600 at 3.6 to a QX at 4.0 cost an additional $750. Given that, the cost for this 10% (under discussion) seems entirely reasonable. The X48s are what, ~ $350, and bleeding edge ram is probably about the same. Once again, that's ~ $750 for a 10% bump. What Is the big deal? Intel gave us 50% for free with the Core2's headroom, however this industry's generosity does have its limits. Once a builder burns up all the freebees, $1000 for first set of ten-percent increases is the way it goes. The next set of increases will cost 3X that. For instance my Q6600 just won't go to 4.0Ghz: If I want that performance I will have no choice but to pony-up and buy capable hardware. My advise (to me) would be, "Just go buy the dang stuff," 'cus, If not now, when?

> My advise (to me) would be, "Just go buy the>dang stuff," 'cus, >>If not now, when? LOL!!!!!!!!And thats why I just go buy the dang stuff.. cuz I want the fastest + the 20-40%of course I do get comp'ed from time to time when I am testing which is fun too.

>> Your arguments obviously peequed my interest>>>>They are not arguments. They are facts, electronic engineering>is not a theory.Ahh, but there are opinions and arguments mixed in. Obviously the engineering speaks for itself, and while I appreciate the technical elements much better thanks to you, Nick, I still see there are other factors that go into the question of what it takes to run FSX to a very high level, even relatively speaking. I have already proven to myself the very lowly and cheap DDR2-800 with the QX delivers quite acceptable performance as defined by what matters, and also that while the highest end memory, properly tuned, will always get you better or best, the difference is affected by the complexity of the situation such that even the highest end will whimper when load is increased beyond a certain level. Sam's "10%" is probably quite relevant. And I'm not thinking about frame rates. I'm thinking about all the dynamic things you have mentioned previously would make a useful benchmarking tool were one to create one that did in fact tested all that FSX *could* demand in a given situation, minus the background throttling it apparently employs. Put another way, 85-90% of your optimized system's "performance" may in fact be quite workable, if smoothness and image clarity matter are what we are looking for. So, is it worth bothering with any further for me. I think for the other $160 it will cost for 4 x 1GB modules of 3-4-4 OCZ DDR2-800, it's worth it just to satisfy the curiosity alone. Plus, since my cheapie mushmemory is willing to play at 1103 and turns in a remarkably good performance, I'm a little fearful it will fry. It's not hot, but it must be well above the rated vDIMM when left on AUTO vDIMM. Apparently a diff BIOS may report vDIMM, but unfortunately all other BIOS' for my mainboard report CPU temp incorrectly. If the OCZ CAS 3 performs essentially the same, I'll send it back, pay a few bucks for the fun, and leave it be until it's time to pawn off the mainboard. Or perhaps I'll keep it so I can o'c my mush and not worry about frying the DIMM without having a backup. I took 2 modules out of my system just to have as back up, in case these fail. It will be awhile until I can use 8Gb (in GigaStudio 4.0)By the way, how fragile are DIMMS in terms of overvolting? Mine are lowly rated at 1.80v. But they must be running much higher than that right now, when I leave it on AUTO. One can afford to be a little less careful with $23 1GB modules over $100 modules. How are DIMMS binned? How are the rated vDIMMs determined? Is it possible my mushins can handle 2.3v no problem? I'm guessing part of the higher cost of high performance memory is so they can be run at higher vDIMM, thereby adding cost to the manufacturer when the warranty claims come. They stay cool, this i've checked. BTW, do DIMMs store their max voltage every exposed to?

Cas 3 is will get you that 7.5ns access, but it is Still OnLy running at 800Mhz And a T2 command rate. That Cas 3 setting won't last long if you try to speed up that DDR2 squirrel cage. Once you have to drop to Cas 4, the game's over. If you are trying to make a Discernible, real world difference you will still need this 7.5ns latency, but: 1) The memory will need to run Significantly faster than 800Mhz (1200Mhz?) AND 2) Be able to achieve a T1 command rate AND 3) Still Maintain that 7.5ns latency. This really IS bleeding edge stuff and it costs Plenty. If one is a computer hobbiest, give up with this DDR2 stuff! Any DDR2. DDR2 at Any speed or Any timing will NoT provide ANY discernible increase in performance. This has been my point all along to our flying-airplanes oriented audience. "You will never see any discernible difference with Any DDR2 ram at Any speed or Any timing." Memory based performance increases are Extremely expensive and Technically oriented. Speed costs. In this case, it'll be about $750. Will that be Cash or Credit?

Sam, as you so correctly pointed out, I did fork over some serious cash for the QX. So, to eak out another 15% of *performance* for $160 doesn't sound to out of line, no?!I'm pretty sure me bro will need this mainboard, GPU & ram, but it won't be for a year or more. At which point I can keep the QX as it's decent plus I've become somewhat attached to it . . . emotionally . . . , pick up a high end DDR3 and board, and take some advantage of what I've picked up here. In the interim, it gives me some stuff to play with and run as fast as able, even if some smoke comes out the back on occasion!Since even Nick's benchmark machine did not reach absolute perfection, one can argue amping up one's particular performance closer towards theoretical perfection has some value. It could be best timed and performing DDR2 can arguably have some value . . .Also, CAS 4 at 1000 should go with these modules. That gets CALWI in order plus adds some more bandwidth as Nick pointed out. 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

That (undefinable) 10% is what might be occurring with that platinum priced 1200Mhz/7.5ns/T1 stuff. I think you're dreaming if you think that DDR2-800 is going to hang at Cas4 up to 1000Mhz. Even if it does, it'll still only be a 1000Mhz/8ns/T2 set . . . . of two, 1 gig sticks (Remember also, that Command Rate is a lead 'bone in this modern Jazz band. T2 is game looser). Also, my flights always load above 3 gigs. My system is starved at 2.A 15% performance boost for $160 worth of DDR2 ram? Show us that magic memory trick. We're all hopeful.

Well, I'm not in the magic CALWI zone right now. According to testimony from other users of that ram, 1T is happening at least with some, though probably less likely with 4gb i'm guessing. So, who knows. The % improvement isn't going to be in frame rates, it should hopefully be in other measures of performance, else the discussion on CALWI is mostly academic. And so it may be. I can always return it, and it's been a fun ride in any case. I'm having decent *performance* with DDR2 @ 800, 4-4-4-10, 2T. If I can do 800 @ 3 & 1T, it may be worth it. Pretty difficult to say when *performance* is, essentially subjectively determined, since we have no benchmark that measures image clarity, texture updating rate, or any of those dynamic performance measures. 15% improvement towards "perfection" is what I mean by this.Command Rate being a lead bone sounds like a matter of contention, at least it wasn't ranked very high in Nick's interpretation. And users are saying they are getting 1T, at least some are.I'm going to try 4 sticks Sam. If it flies with 4 sticks, at 3-4-4 at 800+, I may keep it, provided there is some subjective sense of improvement. I bought the pairs for $78 apiece. 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