Hi Folks, I tried the process recommended by McCrash using Catalyst 8.10. On running autotune, I was able to increase the GPU clock from 500 MHz to 830 MHz (40%!), but the memory clock actually dropped a little lower, to 970 rather than 990 MHz. I then applied the recommended settings within Catalyst and FSX.Framerates at Philly, runway 8, with the default Baron, dense autogen and fine weather were between 30 and 60. Not bad, and comparable with my old 8800GT under similar load. I guess I was comparing my old factory overclocked 8800GT with the new card set at stock clock rates. Now that it has been tweaked I'm pretty happy with the card's performance. Thanks McCrash! I owe you one!Cheers,Noel.

>I was able to increase the GPU clock from 500 MHz to 830 MHz 40%!), >but the memory clock actually dropped a little lower, to 970 >rather than 990 MHz. I then applied the recommended settings within Catalyst and FSX.>I guess I was comparing my old factory overclocked 8800GT with>the new card set at stock clock rates. Now that it has been>tweaked I'm pretty happy with the card's performance. So you

I would put it differently. I'm now seeing equivalent performance from the Powercolor HD4870 1 GB card to that I used to obtain from my old 8800GT 512 KO. My first post highlighted a problem with stutters, not with framerates. IMHO, stutters can be worse than low framerates as they detract from immersion more. I was just using framerates as an indicator of the load on the GPU, figuring that if the framerates go up, the stutters may be reduced as well. Cheers,Noel.

CPU speed is obviously hugely important and probably the single most important factor. But opinion is split about whether FSX is meaningfully "CPU-bound". I think, for example, that Nick_N has some useful insights into this topic.For my part - and I don't pretend to be scientific about this - my own experience strongly suggests that RAM speed / latency is another constraint (to a degree that surprised me) and GPU bandwidth a third. And I do not suppose that that is by any means an exhaustive list. So why shouldn't GPU clock also be a factor? In fact, I suspect that the improvement I got when moving last year from an 8800GTS to an 8800 Ultra was the result of a combination of using a GPU with higher-clocked components, bigger vRAM and higher bandwidth.If you measure performance purely by FPS, the extra GPU cycles aren't likely to add much to the experience when you're at cruise altitude, or even during the climb and descent phases of flight. Also, for example, extra cycles in the GPU probably won't help much if you want to run more AI traffic. But in the PMDG 747 (or even MD-11) at ground level on a cloudy day at Stand 414 in the Megascenery Heathrow 2008 with ASX and 15+ cloud layers, or on approach to that airport in bad weather, or with UTX Europe's "night lighting" turned on - or in countless other situations where the system is under highest graphical strain - I wouldn't be surprised to find that the extra GPU cycles come in handy. It might only be half an extra frame per second; it might not even be measurable in the FPS counter at all; it might just be a little less jerkiness in the display (perhaps measurable by the reduced number of "dips" below a particular FPS); it might just be because you find you can add an extra cloud layer in ASX without losing any performance; but one way or another, I wouldn't be surprised to find that it helps, even if only a little bit. Whether you can be bothered to try to measure it objectively (or know how to do so) is of course a separate question. And another question is whether it's worth the risk to the hardware and/or the cost. But I think that to dismiss all possibility of improvement from extra GPU cycles merely because FSX is "CPU bound" is rather sweeping.Tim

>CPU speed is obviously hugely important and probably the>single most important factor. >>But opinion is split about whether FSX is meaningfully>"CPU-bound". I think, for example, that Nick_N has some useful>insights into this topic.>>For my part - and I don't pretend to be scientific about this>- my own experience strongly suggests that RAM speed / latency>is another constraint (to a degree that surprised me) and GPU>bandwidth a third. And I do not suppose that that is by any>means an exhaustive list. So why shouldn't GPU clock also be a>factor? In fact, I suspect that the improvement I got when>moving last year from an 8800GTS to an 8800 Ultra was the>result of a combination of using a GPU with higher-clocked>components, bigger vRAM and higher bandwidth.>>If you measure performance purely by FPS, the extra GPU cycles>aren't likely to add much to the experience when you're at>cruise altitude, or even during the climb and descent phases>of flight. Also, for example, extra cycles in the GPU probably>won't help much if you want to run more AI traffic. But in the>PMDG 747 (or even MD-11) at ground level on a cloudy day at>Stand 414 in the Megascenery Heathrow 2008 with ASX and 15+>cloud layers, or on approach to that airport in bad weather,>or with UTX Europe's "night lighting" turned on - or in>countless other situations where the system is under highest>graphical strain - I wouldn't be surprised to find that the>extra GPU cycles come in handy. It might only be half an extra>frame per second; it might not even be measurable in the FPS>counter at all; it might just be a little less jerkiness in>the display (perhaps measurable by the reduced number of>"dips" below a particular FPS); it might just be because you>find you can add an extra cloud layer in ASX without losing>any performance; but one way or another, I wouldn't be>surprised to find that it helps, even if only a little bit. >>Whether you can be bothered to try to measure it objectively>(or know how to do so) is of course a separate question. And>another question is whether it's worth the risk to the>hardware and/or the cost. But I think that to dismiss all>possibility of improvement from extra GPU cycles merely>because FSX is "CPU bound" is rather sweeping.>>TimHey TimHere are some threads that show what its all about. Read these threads all the way to the end and it should all come into perspective.http://www.simviation.com/cgi-bin/yabb2/Ya...?num=1224240402http://www.simviation.com/cgi-bin/yabb2/Ya...?num=1224509329http://www.simforums.com/forums/forum_posts.asp?TID=28244http://www.simviation.com/cgi-bin/yabb2/Ya...?num=1223693603FSX is not about FRAMES, its about RESOURCES.Its not all about the CPU -or- the video card -or- the memory -or- the drivers -or- FSX setings -or- the addons.. its about ALL OF THEM "working in concert" to produce the end result.The myths such as quads are no better than duals, memory speed and latency do nothing, video cards above 8800GT wont help and hard drives like the Raptor are not worth the purchase that have been perpetuated around here are a royal joke.By the way... VelociRaptor on PCIe 8x -hardware- RAID0http://forums.avsim.net/user_files/193532.jpgI would have to buy 4-5 32MB cache 3GB/s hard drives and run them on the same hardware RAID card to get even CLOSE to the result on 2 VRaps.. and the access time on that hardware RAID array is 4-6ms at 1% CPU use which places it well within 1000-2000 dollar SCSI storage array system perf and is impossible for those 32MB cache drives to achieve on random access.A single VRap will destroy any hard drive out there in random access perf compare and real world FSX use.ANY QUESTIONS? :-hah Going back to the OP... the 4870 is not cutting it in high scenery and weather in FSX the same as the better Nvidia cards. The 4 series is FAR better than the 3 series ATi becasue they corrected TMU issues but unfortunately that card is still not topping the 8800GTX 768 and 280 cards in FSX. Regardless.. you can NOT slam a VC into a tower and expect a result... the result will be based on the CPU and memory subsystem ability working with the VC and that varies from system to system. Selection of a video card is not the only factor one should use when considering a purchase.

Yes, it's partly because of threads like those that I hold the opinions I expressed in my earlier post. My own experience corroborates what the other threads convey: you can't really skimp on anything with FSX; and getting it all in balance is somewhat easier said than done, even with help from you & others.I also agree with you about HDs. For many years I've subscribed to the much-derided view that HD speed has its own modest part to play in improving the FSX experience, although I've vacillated in my own mind as to whether I'm just experiencing a placebo effect (apart from initial load times, obviously). I've now got Hitachi Ultrastar 15k RPM SAS drives running from an 8xPCIe 3ware RAID adapter card. The access time is close to 3ms. It's not a silver bullet - I've given up looking for one of those - but unless it's placebo, it does seem to help things along nicely. I'm not sure it's cost effective, but put it this way: I won't go back to SATA unless / until I have to "sell the farm".Incidentally thanks for a tip you gave in some other post about sacrificing tRFC for stability: that has let me push the FSB back up to 480MHz (to get RAM at 1601MHz on a 200MHz strap and the E8600 running at 4.33GHz) after I'd been having CTD problems. Still, I can't get below about 47ns latency on my DDR3-1600 even after an embarrassingly large number of hours of tweaking. I'm not unhappy about what I've persuaded the PC to do in the few weeks since I started over-clocking, but it sure makes the Everest screenshot showing your <43ns result look pretty impressive.Anyway: I'd better stop dithering on because I have now wandered completely off topic - not that I am entirely sure I was ever on it: I just wanted to speak up for the fellow who said he'd overclocked his GPU and was pleased with the result. Nothing improbable about it, IMHO.Tim

47ns is quite goodThat Everest run was at DDR3 1840 tooSCSI? nice... do note there is such a thing as 'maxed out' for texture loading LOL!Getting those textures from the HDD to the VC is not just about speed.. its also about getting the CPU off that load and focused on whats important.I guess the best way to see the difference between the "placebo' and the truth would be to run FSX on another system which is set up on a typical SATA drive not connected to a hardware card solution and also run the system at increased CPU speed wihout memory tweaking.I think it would all fall into proper perception and perspective at that pointWe all get accustom to seeing a result in FSX over time and even those of us who are running that top end may find we want more or start to wonder if there is more we can achieve. I think a week on 'typical' spec system would act as a good slap in the face for those times. LOL!Thats actually what happens to me when I switch to that old P35 Qx67 system to check perf on GEX changes.. right away I spot where that x48 is doing its job. I moved the 3ware card to the x48 system so I do see scenery related stutters in specific 'high load' scene situations and crossing scenery boundarys that force bach loads of large scenery files as compared to the hardware storage solution which show none.. although they are greatly minimized by using a single 1st generation Raptor on the P35. I tried a single 32mb cache drive some time ago and the 1st gen Raptor produced a significantly better result.I use that P35 tower now to confirm GEXeu design with more typical hardware setups and also vary the settings on it to simulate different hardware ability. For you a 9650 would be a solid upgrade though. Getting off that dual onto the right quad would be an upgrade for you... but I would put the money into the next gen purchase unless I was going to sit on that system for another year.Good luck!

> For you a 9650 would be a solid upgrade though. Getting off that dual> onto the right quad would be an upgrade for you... but I would put > the money into the next gen purchase unless I was going to sit on > that system for another year.With the dual I do sometimes notice a bit of lag in loading the sharpest textures when flying fast at low altitudes, which I didn't get at all (or didn't notice) in the days when I ran FSX on 2x Xeon 5160s at 3.0GHz. This is irritating, but not my priority, because what I like to do is fly online and take the "heavies" to and from complex airports in bad weather. I try especially hard to get the landings right and for these - based on my experience of a "real" BA simulator last year - fluidity is more important than perfection with the textures at low altitude. The E8600 at 4.33GHz with the RAM working nicely is finally giving enough fluidity for a reasonably convincing landing experience in all conditions: but ONLY JUST, so I don't want to give up CPU speed if I can help it. I know that others have been able to get the Q9650 to run at 4GHz, but I have a feeling that not all specimens will achieve that target. Besides, the E8600 gives me 330MHz on top of 4GHz: 8.25% is not to be sniffed at in FSX. It may not be huge but it is a nice little boost in an important component, and one which would almost certainly be beyond a Q9650 with my abilities/patience, even on the Rampage Extreme which generally seems to be tolerant of my mistakes.That said, this PC is really just a bit of an experiment while I'm waiting for Sandy Bridge (or possibly Nehalem if it turns out as well as you hope), so I may yet try dropping in a quad to see what happens. The Q9650, with a 9x multiplier, would indeed be a logical choice given my existing components and settings.Anyway, I'm drifting off course again: I always enjoy a bit of banter with Nick - I've picked up some very good stuff that way - but apologies to the OP for inflicting my trivia on his thread.Tim

I don

> It saves on hair, something I have very little of and losing fast by > the month now LOLDon't get me started on hair. It's a cruel old world.Tim

LOL!!!Almost forgot.. on the subject of going from the dual to the quad my suggestion was based on you maintaining that 4050 GHz range but there is something else you must take into consideration. Right now your dual is dealing with the requests for AG and terrain on cores that are needed for other operations. Between the cache upgrade and the fact that you are moving those requests off the primary core where FSX looks to for its primary support, the 9650 should in fact be a good upgrade for you.But yes, I would wait till you may decide on the timeframe for next gen upgrade4GB+ can be a bit of a bugger to get tuned on a quad too however at the same time you do not need to be quite as tight on pushing FSB. So do keep in that in mind. The type of memory will drive the stability. As long as you are no higher than CAS7 at 1600 on 4GB you should be fine. The rest is in fine tuning tRD.

Thanks: I'll bookmark this page in case I go for a quad later.Tim

Intel is notorious for holding the price high on their better procs however they may do a price cut early next year on P cores to stimulate a push to move them after nel is released so if they drop the price on that 9650 it would be a worth while upgrade should you decide to hold off for another year. Right now that proc has fairly high cost to perf ratio

I had a rather unpleasant experience with the 4870 512MB (Oerclocked ed. from MSI). Although it is a great card for some shooters I occasionally play, I did not see the desired performance boost in FSX. This was true for driver revisions 8.6 (with hotfix) through to 8.10.The bottom line was to swap the Radeon out again and return to my "old" 8800 GTS 640MB.One additional reason for the swap was more than sub-par performance in Armed Assault (including bluescreens) where the 8800 runs circles around the Radeon for some reason.So it seems to be dependent of the system, because I see others not having the problem.It didn't work for me so the card goes out on Ebay.my 2c.

Visiontek 4870 here, absolutely great card imo thus far. Currently playing the new Far Cry 2 with it, looks amazing.