Hi Guys

 

Being a long term I.T veteran, I can pretty much write down everything I know about FSX on the back of a stamp!

 

So, my question is about memory and FSX specifically...

 

Which one is better:

• 1 lot of matching 4 x 4GB quad channel = 16GB
  
• 1 lot of matching 2 x 4GB dual channel = 8GB
  
• 2 lots of matching 2 x 4GB dual channel = 16GB
  

In my machine currently, I have, mistakenly, 2 lots of matching sets of 2 x 4GB dual channel, thus I have 16GB, but not in matched chips...

 

It works, I can just about hold 30FPS in the 737NGX into heathrow, but, I havent installed any of my Aerosoft scenery yet... Nor FS2Crew...

 

So I'm wondering, would FSX be better if I removed 1 pair of 4GB chips, or replaced all of it with the quad channel 4 x 4GB chips sat on my desk?

 

Its all Corsair Vengance DDR3 1600mHz ram..

If all of the sticks are of the same make and model, they are matching sticks. Even if they're not you can still run them in quad channel, because there's no such thing as "quad channel sets" or "dual channel sets" of RAM sticks.

 

Mixing sticks can create compatibility issues, but most of the times it works just fine.

 

Simply download CPU-Z and take a look at the Memory tab. If it reads 16GB, quad chanel running at 800MHz, you're good.

 

Its not like quad channel makes a difference at all, but still

Indeed it does, all my current 4 sticks are identical, even though its 2 sets of matched pairs meant to be used as 2 x 4GB...

CPU-Z is indeed showing 16GB Quad @ 840MHz

 

Why 840 when its 1600MHz RAM?

even though its 2 sets of matched pairs meant to be used as 2 x 4GB

 

don't sweat it, there's nothing internally in those sticks that makes them any different to a 4x4 "quad channel" set of sticks. The later are just two sets of the former

 

CPU-Z displays the actual memory bus speed, which is 840MHz, but since it's DDR memory (DOUBLE Data Rate) it runs as if it was twice as fast. In your case 840 x 2 = 1680MHz.

 

now, why is it running over it's spec'ed freq is another story!

Which one is better:

 

As you have mobo with Intel X79 chipset and CPU ( i7-3820) with IMC (integrated memory controller) capable to work in four channel mode, the best option allowing you to achieve maximal throughput is using kit of 4 identical memory modules. It can have maximal throughput twice high as set of 2 modules, which would be making your memory subsystem working in dual channel mode. You may see "Intel Core i7 Processor Family for the LGA-2011 Socket Datasheet, Volume 1" for more details.

 

Btw. your CPU has 4 cores each with 2 threads and proper description is 4 cores/8 threads. For your CPU and FSX would be better to disable HT in BIOS. Read impressive post written by SAAB340 - http://forum.avsim.n...ing-ssd-vs-hdd/

If it is too long just jump to "Recap"section at the end and read conclusions.

It can have maximal throughput twice high as set of 2 modules

 

I've never seen a real life application where the channel count makes a difference. Only some synthetic SysSoft Sandra BW benchmark. Actually when Sandy Bridge was released, even on "just" two channels, it beat the crap out of Bloomfield's tri channel chips in most (memory) benchmarks.

 

So I'll take dual channel @ 2400MHz over Quad channel @ 1600MHz any day

So I'll take dual channel @ 2400MHz over Quad channel @ 1600MHz any day

 

There are two ways to improve efficiency of memory subsytem:

- first is raising speed by installing higher frequency (1333, 1600, 1866, 2133, ...) memory modules and as we know it can be expensive way as prices of DDR3 modules are quickly raising with higher frequency and low latency.

- second is raising throughput by using multi-channel memory architecture ( http://en.wikipedia.org/wiki/Multi-channel_memory_architecture )

 

Here ( http://www.legitreviews.com/article/1779/5/ ) you may directly compare benefits from raising frequency vs raising throughput. It is easy to notice that raising throughput brings better results and is more cost effective than raising speed (there is no significant difference in price between 4x4GB kits and 2x8GB kits of the same speed and latency).

 

In theory and in syntetic tests quad channel memory architecture offers highly better performance than dual channel. Real life benefits depend from that how intensively application is using memory. In case when somebody has mobo and CPU capable to work in quad channel mode, downgrading these to dual channel is just wasting money and possible to achieve performance.

 

Quad channel @1600MHz will have better performance than dual channel @2400MHz especially if you notice that internal data processing in Sandy Bridge-E IMC is done with 1600MHz frequency. Btw. what about price comparison between 2x8GB 2400MHz modules with low latency and 4x4GB 1600MHz modules also with low latency?

and in syntetic tests

 

Just one, the one you just posted and I mentioned in my previous post: that SisSoft Sandra benchmark.

For anything else (I've seen) dual / triple or quad channel is a non factor

 

check this out

 

http://www.anandtech.com/bench/Product/287?vs=523

For anything else (I've seen) dual / triple or quad channel is a non factor

 

Real life benefits depend from that how intensively application is using memory.

 

check this out http://www.anandtech...duct/287?vs=523

 

There is no info about memory configuration used for i7 3820 in this case.

Real life benefits depend from that how intensively application is using memory

 

Other memory benchmarks like MaxxMem2 and Aida64 show zero speedup from increasing the channel count, and you won't argue those are definitely memory intensive apps right?

 

At any rate, FSX does not care at all how many memory channels you have. At the end of the day that's all we care about

 

There is no info about memory configuration used for i7 3820 in this case

 

I very much doubt they would test a quad channel system on less than 4 sticks, but ok, here's another one:

 

http://www.techspot....3820/page3.html

First, it is "test vs test" problem:

Test which I presented ( http://www.legitreviews.com/article/1779/5/ ) was done on the same CPU, platform and memory modules with changing only quantity of channels and memory clock speeds. Yours ( http://www.legitreviews.com/article/1779/5/ ) is comparing different CPUs on different platforms.

 

Second, it is problem of trustworthy of testing tools, see these comparision between SiSoft Sandra and AIDA64 - completly incomparable results in the tests done on the same computers:

"SiSoft Sandra is the one test that companies have said show you the best results when it comes to testing the new Quad Channel platform and you can see from the results above that it can't be denied. With our system at stock you can see that performance on the Quad Channel platform skyrockets when compared that of the Dual Channel one."

http://www.tweaktown.com/articles/4416/intel_x79_quad_channel_and_z68_dual_channel_memory_performance_analysis/index5.html

 

"It seems clear that AIDA64 just isn't showing the performance difference between Dual Channel and Quad Channel at the moment."

http://www.tweaktown.com/articles/4416/intel_x79_quad_channel_and_z68_dual_channel_memory_performance_analysis/index3.html

 

Third, we are talking about rigs for FSX, so most important are experiences from flying.

I have built for my work a rig with X79, Xeon E5-1650 (3.2 GHz) and 4x4GB DDR3 Kingston modules. I had it at home for few days on the same desk with my home computer. I installed FSX on it and did some tests. One of these was memory subsystem performance comparision to my home rig (with cpu multiplier set to 32). In Xeon's rig I disabled HT, changed AM to 15 to force FSX using only 4 cores.

I didn't state any noticeble differences in FPS, but noticed faster loading and "shaprening" photoscenery textures, and clearly noticeable improvement in scene "rendering" after camera change. Overall, in my impression, I got better performance on visual aspects but FPS level was practicallly the same. I would like to have such rig as my FSX's computer, especially as I prefere flying over photosceneries, but price of X79 and i7 3930K is a bit too high for home purposes.

First, it is "test vs test" problem:

 

Exactly

 

Test which I presented ( http://www.legitrevi...article/1779/5/ ) was done on the same CPU, platform and memory modules with changing only quantity of channels and memory clock speeds. Yours ( http://www.legitrevi...article/1779/5/ ) is comparing different CPUs on different platforms.

 

I assumed you realised that a I7 3820 & the I7 2600K are essentially the same chip except for the memory channel count (and PCIe lanes, which we'll agree it won't make a difference in single GPU setups)

So it's a perfectly valid comparison. The results are self explanatory

 

Second, it is problem of trustworthy of testing tools, see these comparision between SiSoft Sandra and AIDA64 - completly incomparable results in the tests done on the same computers:

"SiSoft Sandra is the one test that companies have said show you the best results when it comes to testing the new Quad Channel platform and you can see from the results above that it can't be denied. With our system at stock you can see that performance on the Quad Channel platform skyrockets when compared that of the Dual Channel one."

http://www.tweaktown...sis/index5.html

 

So Aida64, just like MaxxMem2 and any other memory benchmarks are wrong, along with just about any real world performance tests out there, while the only synthetic bench that we should consider is Sandra just because it exposes the benefit from that theoretical extra throughput? I'm not sure about that man

 

Third, we are talking about rigs for FSX, so most important are experiences from flying.

I have built for my work a rig with X79, Xeon E5-1650 (3.2 GHz) and 4x4GB DDR3 Kingston modules. I had it at home for few days on the same desk with my home computer. I installed FSX on it and did some tests. One of these was memory subsystem performance comparision to my home rig (with cpu multiplier set to 32). In Xeon's rig I disabled HT, changed AM to 15 to force FSX using only 4 cores.

I didn't state any noticeble differences in FPS, but noticed faster loading and "shaprening" photoscenery textures, and clearly noticeable improvement in scene "rendering" after camera change. Overall, in my impression, I got better performance on visual aspects but FPS level was practicallly the same. I would like to have such rig as my FSX's computer, especially as I prefere flying over photosceneries, but price of X79 and i7 3930K is a bit too high for home purposes.

 

And this is where the "test vs test" thing comes into play, because you tested FSX on that Xeon with a fresh FSX install. That's not apples to apples. Had you cloned your system to the Xeon I would have agreed, but I have tested single vs dual channel here and never saw the increased sharpness you did or any of the other stuff you mention. Please, try it yourself and let me know what you see there. Same install, different channel count.

 

There's a reason why no one (with enough knowledge and common sense) considered Bloomfield anymore for FSX once Sandy Bridge was out.

 

There's a reason why no one (with enough knowledge and common sense) considered Bloomfield anymore for FSX once Sandy Bridge was out.

 

:t0103:

 

Look at these two vid I made with my bloomfield....LOD at 9.5 + a lot of car traffic and a lot of eye candy....see any stutterings in heavy weather....not saying that Sandy ain't better, what I'm saying is that with a well balanced rig you can achieve some pretty good result.

 

 

I said once Sandy Bridge was out Alain. You had your 980X well before that, and of course it's still a very capable chip today

:hi: :friends: