DDR2 v DDR3

[Cruachan 705]

Hi,Thought some of you might be interested enough to read this interesting and illuminating article:http://www.extremetech.com/article2/0,2845,2328804,00.aspMike

[dadmod 0]

Thanks!Very interesting...Roy

[Reset XPDR 1,271]

The conclusion that DDR2 vs DDR3 makes neglible difference to games performance has been demonstrated for both FS9 and FSX many times in this forum. Whilst the review briefly touches on subjective performance with Vista, it does not do the same for games. The work that Nick_N has done with FSX subjective performance with DDR2 vs DDR3 in this very forum has yet to be repeated by any other review that I know of.Gary

[tfm 169]

Yes, this is quite reassuring. One question: the article points out that performance in a game called "Crysis" - which I know nothing about personally - goes up from c.50 to c.60fps with the faster RAM. The reviewer dismisses this as if it were a flat line: but it's about 20%. Is that really the right conclusion? Or is something more subtle going on here?Tim

[Guest D17S]

It's really not about the speed of the ram. That's all just an internal dynamic that is occurring on the stick. The real result is how fast the system can get a data bit from the memory stick to the FSB (and back again). This is called Latency. It is the "torque" that results from all that FSB and memory buss gear-spinning.For instance, we saw a best case Latency of about 45ns. That means that about 22 million data bits per second were able to xfer between the memory buss and the FSB. In our normal rigs, we generally run latencies of about 70ns. That transfers about 14 million bits per second. You'd think that would make a difference, but it doesn't. It seems even 14 million BPS is more than the system needs. What's really going on here is still Way beyond me, but the evidence is clearly indicating that it does not matter. Increasing ram speed will decrease this latency number by about 1ns per 100Mhz. So the difference between DDR3-1600 and DDR2-800 will decrease latency about 8ns . . . or ~ 4 million BPS. In our forum conversations, we could not objectively observe the difference between 22M/BPS and 14M/BPS. A 4M/BPS change will be indistinguishable (other than to the spender of those big bucks!)If you want to play, the biggest bang-for-the-latency-buck is tRD. Intel only uses every 12th FSB cycle to even try to transfer a data bit between the memory and the FSB. The ol' P35 northbridge can do better . .. a lot better. I have mine set to transfer at every 6th cycle. By this method with my old DDR2-800 ram, I get the same performance increase (none) as using DDR3-1600.Remember too, Nehalem will run a 133Mhz FSB. It's trick will be to use every ~ 3rd FSB cycle. That's the onboard memory controllers claim-to-fame. It can use more closely spaced FSB cycles . . . but the latency will still only be ~ 45ns (cuz the FSB is running slower). We can do 45ns now. So don't expect miracles from Nehalem. We've already seen its onboard memory controller performance enhancement with a P35 @ 45ns, i.e., none.

[Guest Nick_N]

I would not use FSX on anything less than correctly configured DDR3And I have seen your screenshots Samthey were nothing compared to mine in real time for volume of scenery, speed of aircraft and clarity of textures in the same test. And as I recall you were using a prop aircraft with the flaps down, indicating a reduced airspeed where I was taking images at 250kts in a jet.Yes, its subjective to the users tech ability to spec the right parts and set them up right, and it requires cash. I moved to DDR3 in March of 2007 and would never go back to DDR2. I have compared both side by side and there is difference.Most wont make that move until they are required. Nellie will want 3 sticks of DDR3 for correct tri-channel operation. The application is the main problem, not the platform. Its over the top CPU and poorly designed memory calls make for havoc on a system and DDR3 opens to door to allowing a better experience as long as the user does not do something stupid like fall for the marketing part of DDR3 and buy CAS8-9 DDR3 1333-1600 modules to skimp by. At that point there is no difference to DDR2.Games are not FSX. These reports you read about games which use local maps and are optimized a certain way have absolutely NO BEARING on how FSX, a world simulator, runs on such a platform. The application itself is a poorly designed lead weight for a system so anything one does to correctly increase CPU and memory communication, increases their FSX experience.I agree the P35 was a great platform... the DDR3 version that isAll I see is goodness for FSX in Nellie although the primary problem is the application is not designed to take advantage of the new optimizations becoming available. FS11 is. The raw horsepower and removal of the true bottleneck is what will help FSX with that platform.

[c152flyboy 0]

Nick, considering that i, like most people, have used your advice numerous times to optimize FSX, i was wondering if you would consider something for the forum. Alot of us write about upgrading to newer systems but there is so many differnt aspects to look at- a core 2 duo or a quad, DDR2-vs DDR3, 8800GT vs something else.would you consider writing a list of the two or three systems you would personally use for a better FSX experience? there seems to be a lot of confusion especially on processor questions, and lets just consider parts available today. i guess i should also say that we should keep the price tag to $1500-$2000, that would seem to be the average budget from what i have seen here.BTW, if your busy because of the Europe GEXn stuff, i understand.Kyle