Archived

This topic is now archived and is closed to further replies.

GSalden

Memory bandwith more important then processorspeed

Recommended Posts

Hi all,Since a couple of days I have been trying to tune my pc to it's maximum potential.Finally it came down to 2 configs.System :E8500 --> stock 3.15GhzAsus Maximus Extreme moboOCZ 2x 2Gb Platinum memoryXFX 260 Black Edition ( oc ) Thermaltake 700W Power SupplyThermaltake Kandalf Liquid Cooling SystemHaving pushed my system to it's limits I had to slow down the memoryspeed to have the pc running.Here the results :3DMarkVantage ( without PhysX ) :Processor 4.05 GhzMemory 1700 Mhz ( 8/7/7/7/28/82 )Cpu 1084,45 plans/sGpu 32.76 fpsProcessor 4.5 GhzMemory 1600 mhz ( 7/6/6/6/26/72 )Cpu 1218,93 plans/sGpu 31.8 fpsThe same with FSX.FSXBenchmark07 and FrapsProcessor 4.05 GhzMemory 1700 Mhz ( 8/7/7/7/28/82 )Min fps 28Max fps 41Avg fps 31.24Processor 4.5 GhzMemory 1600 mhz ( 7/6/6/6/26/72 )Min fps 27Max fps 44Avg fps 31.49Several times I did the tests to be sure.I have had contact with OCZ about these results.It seems that when scenery gets very detailed more memory bandwith has a more positive influence than more processorspeed.The differences are small in my tests , but keep in mind that with the second config the processorspeed was 12,5% higher , while the memoryspeed was only 6% lower.So when buying a new mobo try buying the fastest memory it can handle and then buy the processor.Many people buy the fastest processor they can get and then buy slower memory than could be used on the mobo.

Share this post


Link to post
Share on other sites
Help AVSIM continue to serve you!
Please donate today!

Been saying it for a long time.. but it does require locating the correct settings in conjuction with the speed such as timing and in the case of Q/P/C2 the northbridge latency through STRAPThe main problem is some people read hardware websites and see information about how memory speed and latency has little result in games... well, there is only one problem with that assessment and that is MSFS is not a game application in the technical definition of a 'game'. As such it DOES respond to higher memory speed/lower timing as any application that takes advantage of that part of the memory subsystemPhil Taylor posted very early on about MSFS and how much it relies on memory speed and timing. The main issue with the application is it produces in laymans terms a scrambled mess for the system to deal with. Therefore the lower the latency and the higher the memory speed, in correct balance with the CPU delivers the best result to the screen.That is not to say people should drop their CPU speed down very low to try and clock memory very high... that is to say people should BUY the motherboard, CPU and the MEMORY to run the highest memspeed/lowest latency stable. Its those who skimp and try to push hardware not really designed to do the job, and, understand how the system works that end up with false results which perpetuate forum myths.The results can not be judged ny any benchmark program simply because the better result is quantified to the visual expereince, something the FPS readout will never display as a linear result with what is experienced on the screen.

Share this post


Link to post
Share on other sites
So when buying a new mobo try buying the fastest memory it can handle and then buy the processor.Many people buy the fastest processor they can get and then buy slower memory than could be used on the mobo.
I think that is a wonderful, concise summation of a smart approach to component selection sequencing. It seems to me the reason there is any discussion whatsoever about what components or conditions FSX needs to run the best is that FSX is so close to the edge in ALL three areas, heck make that FOUR, of processor speed, memory subsystem speed, GPU and even HDD/file system, that the specific flying situation itself puts demands on the entire system such that at any one moment, ANY ONE of those variables can become the weak link for that moment. That was the Mother of All Run On Sentences :( Noel

Share this post


Link to post
Share on other sites

Once a set of modern components are assembled, FS performance is going to be 95% CPU influenced. Once a modern 7200RPM harddrive/88-9800GT Vcard/4G-ram/64bit op system is established, any remaining 5% available improvement remains (only arguably) ram speed based. And intractably, any 5% difference will always remain only subjectively arguable. (Runnon? - Dang, it's contagious! ~ )

Share this post


Link to post
Share on other sites
Once a set of modern components are assembled, FS performance is going to be 95% CPU influenced. Once a modern 7200RPM harddrive/88-9800GT Vcard/4G-ram/64bit op system is established, any remaining 5% available improvement remains (only arguably) ram speed based. And intractably, any 5% difference will always remain only subjectively arguable. (Runnon? - Dang, it's contagious! ~ )
Hi Sam,I am running Vista 64 and so far I kinda like it. FSX seems to run best ever, and I don't have even one "tweak" set up in Vista. Just Vista 64 SP1. I am dual booting using the BIOS' F8 BBS popup. I also picked up an EVGA 280GTX since it was down to $289 on newegg. Seems to run everything well so far.Noel

Share this post


Link to post
Share on other sites
Once a set of modern components are assembled, FS performance is going to be 95% CPU influenced. Once a modern 7200RPM harddrive/88-9800GT Vcard/4G-ram/64bit op system is established, any remaining 5% available improvement remains (only arguably) ram speed based. And intractably, any 5% difference will always remain only subjectively arguable. (Runnon? - Dang, it's contagious! ~ )
Sam, I think when it comes to raw frame rate, your analysis applies well. When it comes to smoothness, I think my comment may apply well. A fast CPU can make it all happen faster, but the timing issues, readiness states, delays, etc that can come out of any of the other basic components can affect smoothness. Or so it seems, intuitively, to me.As with discussions with NickN, I like to think "performance" in FSX is a product of IQ + smoothness at an "acceptable" frame rate. When frames get too slow, they become the show stopper. But as we all know, have herky-jerky motion, and fuzzy IQ can be a show stopper even at high frames. Having a high performance experience using this definition arguably is one that most simmers would find the most enjoyable/immersive/whatever.

Share this post


Link to post
Share on other sites
. . . but the timing issues, readiness states, delays, etc that can come out of any of the other basic components can affect smoothness. Or so it seems, intuitively, to me.
Those are all ram/buss related, which in turn relate to ram/buss performance's singular contribution, latency . . . and this contribution remains entirely subjective (or < 5%). We all recognize the frame counter is an irrelevant metric.

Share this post


Link to post
Share on other sites
Those are all ram/buss related, which in turn relate to ram/buss performance's singular contribution, latency . . . and this contribution remains entirely subjective (or < 5%). We all recognize the frame counter is an irrelevant metric.
http://forums1.avsim.net/index.php?s=&...t&p=1529654When you answer the simple questions placed directly to you, then you have a basis to post this nonsense... otherwise the only thing that is subjective here is the suggestion you may actually know what you are talking about.Catch 22If you do answer those questions correctly, you wont have any point to make, which is why you did not answer them when asked and instead continued to market an opinion.

Share this post


Link to post
Share on other sites
Those are all ram/buss related, which in turn relate to ram/buss performance's singular contribution, latency . . . and this contribution remains entirely subjective (or < 5%). We all recognize the frame counter is an irrelevant metric.
That sounds like a cop out Sam. Stuttering/jerkiness from grossly apparent to subtle can be huge show stoppers to user experience and framing them as subjective in no way detracts from their impact on "performance."

Share this post


Link to post
Share on other sites

". . . Stuttering/jerkiness from grossly apparent to subtle can be huge show stoppers to user experience and framing them as subjective in no way detracts from their impact on "performance."", and once a set of modern components are assembled, FS performance is going to be 95% CPU influenced. Once a modern 7200RPM harddrive/88-9800GT Vcard/4G-ram/64bit op system is established, any remaining 5% available improvement remains (only arguably) ram speed based. And intractably, any 5% difference will always remain only subjectively arguable.

Share this post


Link to post
Share on other sites
Once a set of modern components are assembled, FS performance is going to be 95% CPU influenced. Once a modern 7200RPM harddrive/88-9800GT Vcard/4G-ram/64bit op system is established, any remaining 5% available improvement remains (only arguably) ram speed based. And intractably, any 5% difference will always remain only subjectively arguable. (Runnon? - Dang, it's contagious! ~ )
I disagree :( Read my post over here: http://www.simforums.com/forums/forum_posts.asp?TID=29276Scroll down, it's the last post (for now at least).The performance boost I get from replacing my old 8800GTX card with a GTX285 OC:ed is obvious. It's up to you to calculate the performance increase in percent :( With the good advice from Nick, I've now completed the build of my new system :( If I should have followed your advice, I'd be stuck with a C2D Quad and a 8800GT card :o

Share this post


Link to post
Share on other sites

Finally it comes down to hardware that will work with eachother, without a piece that holds performance back ( the wellknown bottleneck ) Also having tuned the software in a proper way will add to better performance and smoothness.If a pc has a q6600 processor than the diffrence between a GTX800 and GTX280 will not make much difference in framerates , like they tested at TomsHardware.However , if you have a faster setup than replacing the "bottleneck" card with a faster one will give a gain in performance , like UlfB found out.It's a pity that there are lots of tweakersforums where people try to achieve the highest processorspeed possible , where on the other hand they have to cut way back on their memory speed to get it running.Helpfull guys like Nick try to make clear what will help and what doesn't.

Share this post


Link to post
Share on other sites
I disagree :( Read my post over here: http://www.simforums.com/forums/forum_posts.asp?TID=29276Scroll down, it's the last post (for now at least).The performance boost I get from replacing my old 8800GTX card with a GTX285 OC:ed is obvious. It's up to you to calculate the performance increase in percent :( With the good advice from Nick, I've now completed the build of my new system :( If I should have followed your advice, I'd be stuck with a C2D Quad and a 8800GT card :o
Can you quantify the difference between the 8800GT and the GTX285?Jim Harnes

Share this post


Link to post
Share on other sites
Once a set of modern components are assembled, FS performance is going to be 95% CPU influenced. Once a modern 7200RPM harddrive/88-9800GT Vcard/4G-ram/64bit op system is established, any remaining 5% available improvement remains (only arguably) ram speed based. And intractably, any 5% difference will always remain only subjectively arguable. (Runnon? - Dang, it's contagious! ~ )
Sam,Check this thread out: http://forums1.avsim.net/index.php?showtop...42613&st=15How about that?@GSalden: I agree :(

Share this post


Link to post
Share on other sites
Can you quantify the difference between the 8800GT and the GTX285?Jim Harnes
I read the forum and it looked like more than 25% (on an 8800GTX rather than GT. That is not due to the GTX285 on its own, but rather in a system with best cpu and best memory.RegardsSimon

Share this post


Link to post
Share on other sites
I read the forum and it looked like more than 25% (on an 8800GTX rather than GT. That is not due to the GTX285 on its own, but rather in a system with best cpu and best memory.RegardsSimon
Simon,I've made several flights and I might add that the performance and smoothness is very much better with this OC:ed 285 card than with the 8800GTX OC:ed card. I just flew through a valley in the Vancouver+ scenery and it was so smooth :( With my previous card the amount of trees on the mountains would cause stutters. I guess you need all your components to be top notch to get this kind of performance boost by replacing a 8800GTX with a GTX285. With my old pc (X6800 at 3.4GHz and Asus P5W DH Deluxe) I guess the GTX285 wouldn't have made that much difference. But you have to remember that all this come with a big price tag :( Is it worth the money? I'll guess it's not the "best performance for the buck", but it's a lot of fun :( I really wish that everybody could afford this kind of pc or even better!

Share this post


Link to post
Share on other sites

So does this mean that i7 users should be persuing higher Bclocks/lower multiplier vs. higher multiplier/lower Bclocks? It seems that higher Bclocks provide significantly lower Everest (ver4.6 beta, haven't experimented with ver 5.00 yet) latency numbers at the same or higher bandwidth. Or is it six-in-one, half-dozen-in-other kind of thing?-jk

Share this post


Link to post
Share on other sites
Simon,I've made several flights and I might add that the performance and smoothness is very much better with this OC:ed 285 card than with the 8800GTX OC:ed card. I just flew through a valley in the Vancouver+ scenery and it was so smooth :( With my previous card the amount of trees on the mountains would cause stutters. I guess you need all your components to be top notch to get this kind of performance boost by replacing a 8800GTX with a GTX285. With my old pc (X6800 at 3.4GHz and Asus P5W DH Deluxe) I guess the GTX285 wouldn't have made that much difference. But you have to remember that all this come with a big price tag :( Is it worth the money? I'll guess it's not the "best performance for the buck", but it's a lot of fun :( I really wish that everybody could afford this kind of pc or even better!
Hi UlfThe smoothness itself is a compelling reason to upgrade. I love the forests of FTX, but with my current set up (Q6600@3.5Ghz, 8800GT) it can be a a struggle to push the autogen slider to the right. I'll ride the year and see what hardware comes along, 32nm cpu and 40nm gpu offer lots to look forward to.RegardsSimon

Share this post


Link to post
Share on other sites
". . . Stuttering/jerkiness from grossly apparent to subtle can be huge show stoppers to user experience and framing them as subjective in no way detracts from their impact on "performance."", and once a set of modern components are assembled, FS performance is going to be 95% CPU influenced. Once a modern 7200RPM harddrive/88-9800GT Vcard/4G-ram/64bit op system is established, any remaining 5% available improvement remains (only arguably) ram speed based. And intractably, any 5% difference will always remain only subjectively arguable.
So really, the argument becomes how much does ram speed, or some other mem sub sys element, factor into the smoothness AND IQ component of performance. In my interpretation of user experience, CPU power best translates to frame rate, NOT smoothness or IQ per se, though it does factor in.

Share this post


Link to post
Share on other sites

My understanding of frame rate has evolved to only count unique frames per second. Perhaps a new acronym might be helpful. How about "UFPS" or Unique Frames Per Second. For instance, when stutters are obvious but the frame counter is at 30+ FPS, clearly the same frame is repeating over and over. That's a stutter. The Vcard just repeats the last dataset it received from datapath over and over again, while it patiently waits for the next Unique frame's dataset. This data set can only come from one of two sources. Stored data from a ramstick, or directly from the CPU. Given this dynamic, it's hard to imagine how ram transfer speed (latency) could affect the - Amount - of data that is available to the Vcard, at least for very long. Ram is a short term storage buffer. If the CPU could not keep the ramstick "refilled" fast enough to keep up with that (speedier) ramstick's ability ability to transfer data onto the datapath, that ram-buffer would soon be depleted (and we're talking 'in a matter of nanoseconds'). Once this 'burst capability' was depleted, the system would be back to a 1:1 transfer capacity between the CPU and its users (ram, Vcard, et al). If ram subsystem "flow-through" was an - actual- pacing factor (ie, ram latency), faster ram transfers would directly translate into increased dataflow to the Vcard, ie, fewer repeated frames and therefore more UFPS. Users claim this is occurring, but it is taking 100% increases in ram subsystem performace to provide only subjectively arguable results. If this was really happening, a decrease in ram latency from 80ns to 40ns would produce obvious results. It doesn't.

Share this post


Link to post
Share on other sites

repeating frame? LOL!!!!!!!!!!!!!!!!OK... I'm not going there... I have seen some whoppers but that is some grade A prime cut BSThis is simple folks.. If anyone on a Q/C2 wishes to check this on their own go into your BIOS, leave CPU speed alone and reduce the the memory speed. Set it as low as you possibly can and dont worry about the timing just leave it as it is, remember the professor here said memory speed and timing doesn't matter or equate to much of anything. Reboot and fly your sim. On the same settings in a flight you know is smooth in a dense scenery area running your normal memory speed see if your typical frame rate remains the same.. much less the image and flight qualityand remember... a the professor says what you are seeing is VERY subjective. Just the suggestion that you have reduced the memory speed is enough to make you think the sim is not running anywhere near as well, including frames,.. but its all in your mind. :(

Share this post


Link to post
Share on other sites

reduce the the memory speed. Set it as low as you possibly can and dont worry about the timing just leave it as it is, remember the professor here said memory speed and timing doesn't matter or equate to much of anythingImho you don't have to be a rocket scientist to know that slower memory will result in worse performance. Otherwise I would like to have my SDRam 133 mhz back..... :(

Share this post


Link to post
Share on other sites
reduce the the memory speed. Set it as low as you possibly can and dont worry about the timing just leave it as it is, remember the professor here said memory speed and timing doesn't matter or equate to much of anythingImho you don't have to be a rocket scientist to know that slower memory will result in worse performance. Otherwise I would like to have my SDRam 133 mhz back..... :(
Yes, you might think it would be obvious. But some people say that RAM speed is either irrelevant, or of such marginal relevance as to be insignificant. Nick is making the sensible suggestion that if you have a configurable BIOS, you can find out easily enough for yourself.This topic has been done to death on many, many occasions. Nick and others persuaded me to give faster RAM a try. And sure enough, in my experience, faster / well-tuned RAM gives appreciably better results. It's not like the difference between a 2GHz and a 4GHz CPU (all other things being equal): but nobody said it would be, and every little bit counts with FSX. Some people, though, will never be convinced: but that's their privilege and provided everyone visiting the forums gets to see that there is a contrary (and, I think, better) opinion then, personally, I couldn't give a hoot.Tim

Share this post


Link to post
Share on other sites
So does this mean that i7 users should be persuing higher Bclocks/lower multiplier vs. higher multiplier/lower Bclocks? It seems that higher Bclocks provide significantly lower Everest (ver4.6 beta, haven't experimented with ver 5.00 yet) latency numbers at the same or higher bandwidth. Or is it six-in-one, half-dozen-in-other kind of thing?-jk
Bump on this Q

Share this post


Link to post
Share on other sites

But some people say that RAM speed is either irrelevant, or of such marginal relevance as to be insignificant. Nick is making the sensible suggestion that if you have a configurable BIOS, you can find out easily enough for yourselfOfcoarse if you are having a "slower"processor and buy the fastest memory than this memory will not make much difference compare to slower memory.Bandwith OCZ DDR1600

Share this post


Link to post
Share on other sites