Question about memory latencies

Hi guys,

Today I did a test of my Corsair Dominator Platinum 2x8GB memory kit (part number in my profile) with AIDA64 (Extreme v 4.00.2700) and it turned out that my latency is quite high on XMP profile (1866 MHz 9-10-9-27 C2) - it was around 117 ns. As far as I've seen that usually todays DDR3s standart (without custom overclock) kits have around 30 - 50 ns in the same tests. Could anybody explain me - is it ok in my case, if not, what could be the reason of this high latency and how to solve it?

P.S. basically system is working ok, without any issues - just those test results doesn't make sense :confused:

What latency is a version 3.xx of Aida64 reporting in that test?

Maybe they have changed the algorithm again...

What latency is a version 3.xx of Aida64 reporting in that test?

 

Maybe they have changed the algorithm again...

Actually, I didn't think about that, haven't tried yet, but will definitely do this later in the evening and will give a feedback.

P.S. :idea:

I don't know, if that make sense, but I asked one of my friends to test his RAM (KHX1600C9D3K2/8GX)x2 with the same version of AIDA64 on his system (GA-Z68X-UD3H, i7 2600k, 16GB RAM) and he had something like 74 ns, which is around 60% lower than I had. So, while I am not a pro in RAMs and all this latency stuff, but I hardly believe, that in my case there really is a problem in algorithm, if quite similar RAM, with close latencies (friends Kingston has 9-9-9-27) is doing much better latency wise than my Dominator.

Anybody else have some opinion on this :question:

/--------------------------/

P.S.S.

Done the test on AIDA64 v3.200 - results are even worse 125 ns. As I assumed, there is nothing wrong in AIDA64. Some other ideas?

P.S.S.S.

Did the SiSoftSandra 2014 (Lite) Cache & Memory Latency test and the Aggregated score is: 30.2 ns. Is Aggregated score the same as AIDA64 Latency test result?

At 1866 mhz the timings need to be loosened to compensate for the overclock. They are all perfectly normal.

If you look at the specs for your modules you hjave two sets of timings . Tested and SPD timings.

Tested timings are what is used at default settings(1333mhz) Thn you have the SPD timings which are the overclock or XMP timings which are ggenerally a little looser , like i said to compensate for the overclock.

Nothing says you cant experiment and try to lower them some. Just try one value at a time and retest witth memtest to be sure the system is stable. Rinse and repeat until you are unstable, then back it off one notch.

At 1866 mhz the timings need to be loosened to compensate for the overclock. They are all perfectly normal.

 

If you look at the specs for your modules you hjave two sets of timings . Tested and SPD timings.

 

Tested timings are what is used at default settings(1333mhz) Thn you have the SPD timings which are the overclock or XMP timings which are ggenerally a little looser , like i said to compensate for the overclock.

 

Nothing says you cant experiment and try to lower them some. Just try one value at a time and retest witth memtest to be sure the system is stable. Rinse and repeat until you are unstable, then back it off one notch.

I hope you are right and there is really nothing wrong with the RAM or MB.

So basically, what you are saying, if I would run a test for default settings (1333 MHz), then results should be much better latency wise?

AT 1333mhz they will be 9-9-9-27. Your memory is set just the way it should be.

http://www.corsair.com/en/memory-by-product-family/vengeance-pro-series-memory/vengeance-pro-series-16gb-2-x-8gb-ddr3-dram-1866mhz-c9-memory-kit-cmy16gx3m2a1866c9r.html

You cant compare your memory benchmarks with your friends. They are two totally different systems even though they are using similar memory. Other than the benchmark numbers is your system giving you any other reason to believe you have something wrong. Crashes? Unstable?

AT 1333mhz they will be 9-9-9-27. Your memory is set just the way it should be.

http://www.corsair.com/en/memory-by-product-family/vengeance-pro-series-memory/vengeance-pro-series-16gb-2-x-8gb-ddr3-dram-1866mhz-c9-memory-kit-cmy16gx3m2a1866c9r.html

 

You cant compare your memory benchmarks with your friends. They are two totally different systems even though they are using similar memory. Other than the benchmark numbers is your system giving you any other reason to believe you have something wrong. Crashes? Unstable?

To be honest, no nothing is wrong with my system - everything is working very stable, I even overclocked it once to 2400Mhz - (10-12-12-31 1.65v) and it was still stable, also latency droped a bit (I'll attach a print scr.).

I also did a test of 1333 MHz (default) and latency got worse - 143 ns, while timings actually were lower - 9-9-9-24 CR1. Don't understand anything.

Actually - only thing why I am interested is the difference in benchmark numbers of latency. I would understand, that there is a slight difference (lets say some 2-5%), but if you see more than 2 - 3 times the difference of similar systems, this makes me feel incomprehensive! I understand that nothing happens without a reason and I just want to know the reason.

P.S.

This is my memory kit: http://www.corsair.com/en/memory-by-product-family/dominator-platinum-ddr3-memory/dominator-platinum-with-corsair-link-connector-1-5v-16-gb-dual-channel-ddr3-memory-kit-cmd16gx3m2a1866c9.html

To be honest, no nothing is wrong with my system - everything is working very stable, I even overclocked it once to 2400Mhz - (10-12-12-31 1.65v) and it was still stable, also latency droped a bit (I'll attach a print scr.).

Thats impressive!Out of an 1866mhz kit , WOW! It's nice to see some good overclocks with these modules. Believe it or not we really dont get many people that give feedback like that.

But that should give you a good indication that your system is working just fine. Ignore the benchmarks. They are an irrelevant number really. It's not just memory that is going to impact those scores, but just about every other component in your system. Your hard drives are going to be another huge factor too, MB, individual CPU's and such. The drop could be caused by any number of them.

Run a different benchmark , like P95 and see what you guys get. What I'm getting at is don't sweat one benchmark number. It's just that, a number that in no reflects real world ,everyday usage, performance.

Thats impressive!Out of an 1866mhz kit , WOW! It's nice to see some good overclocks with these modules. Believe it or not we really dont get many people that give feedback like that.

 

But that should give you a good indication that your system is working just fine. Ignore the benchmarks. They are an irrelevant number really. It's not just memory that is going to impact those scores, but just about every other component in your system. Your hard drives are going to be another huge factor too, MB, individual CPU's and such. The drop could be caused by any number of them.

 

Run a different benchmark , like P95 and see what you guys get. What I'm getting at is don't sweat one benchmark number. It's just that, a number that in no reflects real world ,everyday usage, performance.

:) Ok, thanks for your help, suggestions and support. Will check the P95 bench.

Thats impressive!Out of an 1866mhz kit , WOW! It's nice to see some good overclocks with these modules. Believe it or not we really dont get many people that give feedback like that.

 

But that should give you a good indication that your system is working just fine. Ignore the benchmarks. They are an irrelevant number really. It's not just memory that is going to impact those scores, but just about every other component in your system. Your hard drives are going to be another huge factor too, MB, individual CPU's and such. The drop could be caused by any number of them.

 

Run a different benchmark , like P95 and see what you guys get. What I'm getting at is don't sweat one benchmark number. It's just that, a number that in no reflects real world ,everyday usage, performance.

Hi,

Just to let you know - I've solved or to be more precise, found an actual reason of low latency indicators. In my case problem was in usage of CPU/GPU, as on the moment of doing tests, I was still awaiting for my discrete video card - so I was kind of forced to use integrated video. :idea:

As soon as video card was dispatched - I did a test once more. Now everything seems fine :roll: Test results attached.

I believe that this thread can be closed as solved.