Hello I was looking at these, as the cas and MHz matches my present RAM this much ram is likely over kill and I might just end up replacing the ram I have instead. Am I better off looking at a lower cas or a higher MHz? http://www.scan.co.uk/products/16gb...00)-non-ecc-unbuffered-cas-9-?ProductId=66989
I maybe totally wrong but i thought higher timings are offset by the higher speed. Theoretically 333Mhz Cas 2, 666Mhz Cas 4, 1333Mhz Cas 8, 2666Mhz Cas 16, etc, etc, would all have the same response time.
Basically im looking for an experts opinion. 1600Mz Latancy Window is up to 9-9-9 before it starts getting out preformed by lower MHz moduals with lower Latency. Should I look at 1800MHz (CAS 10-10-10 window) or 1600 MHz as my price range is around the £110-120 mark. Is the difference worth it.
When I was looking into something similar I found the following article very useful http://www.anandtech.com/show/7364/memory-scaling-on-haswell Particularly this diagram on the last page. You can see that, for example, 2400MHz CAS 9 performed the same as 2666MHz CAS 10
It used to be with Sandy and Ivy Bridge, but Haswell really can utilise faster memory. The difference in framerate between 1600MHz CL9 and 2400MHz CL 10 can exceed 20% in some scenarios. Minimal and average fps: Far Cry 3, Watch Dogs
The difference wouldn't be down to using Haswell, it would be down to the 1600MHz CL9 taking longer between the moment a memory controller tells the memory module to access a particular memory column on a RAM module, and the moment the data from the given array location is available on the module's output pins. As i alluded to before you can work out the differences in latency between different speeds and different CL ratings, this post on superuser.com goes into details of the calculations...
Sandy Bridge Memory Scaling - the differences between different memory speed except Physics are next to none. Well... in theory. In practice Sandy Bridge with 1600MHz CL9 can sometimes be faster than 1866MHz CL8:
That's because the "difference" isn't different in the anandtech article, within a few nanoseconds all the memory anandtech tested had the same latency. The purepc.pl article you originally linked to shows a difference because the memory is not running at the same latency. An old article on tomshardware.co.uk details the maths.
Now I have to admit I am quite lost. The graph above with Crysis: Warhead results is from Anandtech. They used an i7-2600K, all common memory from 1333MHz up to 2133MHz and compared with all common different latencies: CL 7, CL 8 and CL 9. As we can clearly see the results are the same within an error margin. I understand your point with maths, speed and latency. We can see it in the Aida memory read, write, copy and latency tests, but there is no real gain in practical use with older cpus. However new z97 chipset and Haswell architecture (especially Devil's Canyon) brought some changes. The tests on PurePC with i5-4690K and GTX 780 Ti shows (imho) significant difference in minimal framerate in Far Cry 1333MHz CL9 : 1600MHz CL9 : 2133MHz CL9 : 2400MHz CL10 = 44fps : 48fps : 56fps : 58fps or Watchdogs = 32fps : 33fps : 38fps : 42fps Anyway, if OP wants to change his current RAM and beat his highscores in synthetic tests then the best way is to choose the calculation provided by Corky42. In everyday use he won't see any real gain from the memory over 1600MHz and CL9 with his Lynnfield cpu (Sandy or Ivy too). However if OP has a plan to upgrade into Haswell then a faster memory clock is almost always better choice than a tighter CL time (within the common sense ofc).
No offence to pure pc but I'd trust anandtech a lot more than them. The cost of cl7 memory modules is crazy and anandtechs own article shows 0 benefit over cl9. Infact 1600mhz cl9 is faster in certain games than 2400 cl9. There's also other issues of high MHz memory of your trying to hit high overclocks it can hold them back as the memory stuff is all cpu based. On a single direct gpu there's no benefits on SLI or 2 gpus there's no benefit now if you own 3-4 gpus there's scaling at the super high end 198 fps vs 212 for example.
I do also have a lot respect to anandtech, but there are few things worth to mention: - anandtech test is over a year old on z87 mobo and "...All benchmarks are also run at 1360x768 due to monitor limitations..."
@bartiszon, ahh i think i get what your saying with Haswell, if you are depending on the IGPU then yes moving to faster ram would net you a nice performance gain. If on the other hand we are just comparing like for like: If a lower CAS is better than faster MHz (or visa versa) in terms of just the performance of the ram, then neither is strictly true as the latency of ram, and thusly the performance depends on a combination of the two. Once you know that CL is how many ticks, how many of those Mhz it takes to do its thing, you can work out the latency. I.e 1000Mhz = 1000 times a second, so CL 10 would take 10 nanoseconds. (that's if my maths is right) But as you say this is pretty much useless outside of benchmarks as we are talking about tiny improvements in real world scenarios, unless you are heavily dependent on the rams performance.
Id still say to anyone theres better ways to gain noticeable performance. More so if gaming is the priority as 16gb of 2400mhz cl9 ram is the cost of a Nvidia 970 gpu for most brands.
