DDR3 RAM performance question

[michaelg]

I'm a newcomer to this forum, and an overclocking noob, so hi all!

I've just been looking at the sections on memory and CPU o/clocking, but can't quite get my head around ram with faster clock speed but higher CAS latency, versus slower RAM with lower latency. (I get it that CAS is measured in clock cycles, so it is also relative to clock speed, but I don't know how much weight to give latency, as it's only one aspect of overall RAM performance).

So I have a specific question based on my intended purchase as follows:
Asus ROG Maximus VI HERO Intel Z87 Socket 1150 motherboard
Intel Core i5 4670K.

Assuming I don't overclock, will there be any difference between the following 3 memory options - in terms of performance or stability?:

16GB (2x8GB) Corsair DDR3 Vengeance Pro Series Blue PC3-12800 (1600), Non-ECC Unbuffered, CAS 9-9-9-24, XMP, 1.5V
16GB (2x8GB) Corsair DDR3 Vengeance Pro Series Blue, PC3-14900 (1866), Non-ECC Unbuffered, CAS 9-10-9-27, XMP, 1.5V
16GB (2x8GB) Corsair DDR3 Vengeance Pro Series Silver, PC3-17066 (2133), Non-ECC Unbuffered, CAS 11-11-11-27, XMP, 1.5V

And if I did overclock, which would be the best of those options - firstly in terms of how suitable they are for overclocking in my configuration, and secondly in terms of performance and stability?

I appreciate that the real-world performance difference is unlikely to be dramatic, but on the other hand, there's only a small price difference between all 3 options.

TIA,

Michael...

[kalniel]

Assuming I don't overclock, will there be any difference between the following 3 memory options - in terms of performance or stability?:

16GB (2x8GB) Corsair DDR3 Vengeance Pro Series Blue PC3-12800 (1600), Non-ECC Unbuffered, CAS 9-9-9-24, XMP, 1.5V
16GB (2x8GB) Corsair DDR3 Vengeance Pro Series Blue, PC3-14900 (1866), Non-ECC Unbuffered, CAS 9-10-9-27, XMP, 1.5V
16GB (2x8GB) Corsair DDR3 Vengeance Pro Series Silver, PC3-17066 (2133), Non-ECC Unbuffered, CAS 11-11-11-27, XMP, 1.5V

Realistically, no.

And if I did overclock, which would be the best of those options - firstly in terms of how suitable they are for overclocking in my configuration, and secondly in terms of performance and stability?

All the same. RAM has no effect on overclocking.

[MrJim]

Some types of software may be particularly sensitive to memory latency & bandwidth: one commonly cited example is compression software like 7-Zip and alike. But in real-world usage, it's unlikely you'll notice much difference, it's true.

[wolfsSpirit]

i just got an MSI Z87-GD65 Gaming board and when i clock my processor , the system/board clocks down the memory lolz
right now i have vengeance 1866 memory in it , but has little use as the system clocks my memory down , i could as well have bought the slowest available haha

[MrJim]

i just got an MSI Z87-GD65 Gaming board and when i clock my processor , the system/board clocks down the memory lolz
right now i have vengeance 1866 memory in it , but has little use as the system clocks my memory down , i could as well have bought the slowest available haha

You should be able to adjust your memory parameters manually, or if your memory supports it, just enable the XMP profile (which will adjust the parameters automatically).

[Phoenixx]

I would always aim to get the cheapest but well known brand at the correct size and speed and only worry about the timings if two otherwise similar modules were at a similar price. I've never noticed timings make a difference in real life.

[michaelg]

Many thanks for the replies!

[Lee H]

For anyone else trying to decide what memory speed to opt for - the following is a great read

http://www.xbitlabs.com/articles/mem...well-ddr3.html

[michaelg]

Lee, that is by far the most comprehensive advice on the matter that I have seen.

The gist of the article as I understand it, is that high frequency RAM can make quite a difference to the overall performance of a high-spec Haswell build. They reckon 2400 is the current price/performance sweet spot.

Also, higher frequency RAM is far more important than low latency RAM. For several reasons:
1) Latency is measured in clock cycles, so faster RAM will have less latency than slower RAM, even with the same CAS rating.
2) Latency values are higher than they used to be, so shaving off 2 cycles from CAS 11 is less than 20%, whereas in the old days of CAS 4 RAM, shaving off 2 cycles was a 50% improvement.
3) Latency accounts for a small proportion of total RAM reading/writing. A 20% drop in latency will make far less than a 20% performance improvement. A 20% increase in frequency will be far more significant overall.
4) Modern CPU architecture has caching and data pre-fetch algorithms that are designed to mitigate latency to a large extent anyway.

Many thanks for the excellent link.

Rgds,

Michael

[kalniel]

The gist of the article as I understand it, is that high frequency RAM can make quite a difference to the overall performance of a high-spec Haswell build.

But not realistically - they had to really push some odd things to get it to show any difference in gaming for example. In realistic game settings there was no difference.

Memory speed doesn’t affect the frame rate much now. The difference between the slow and fast memory configurations amounts to just a few percent.

And for the application tests it depends very much on the application, don't forget for the test they were using as slow as 1333mhz and comparing up to 2933, so check the absolute figures for the specs you would be considering.

[michaelg]

Fair comment, kalniel, in that for some - maybe most - applications, the difference is less than a percentage point or two. But bear in mind that the price difference between 1600 and 2400 RAM is often just a percentage point or two also. And as a percentage of the total system price, it's really a tiny fraction .

And for some common applications like Photoshop and WinRar, the performance difference is significant. Photoshop was 8% faster using 2400 CAS 11 RAM compared to 1600 CAS 9, and WinRar was 15% faster!

I'd say that's a pretty good boost for just a few pounds extra spent on RAM. Unless there is a downside specifically?

[kalniel]

If you're using those applications then go for it - there's no downside until you get to higher overclocks of the CPU.

It's odd that other reviews don't find any difference. Xbit state they are overclocking the processor but it's not clear whether they're overclocking the processor at the same time as the ram or whether it's overclocked to the same level throughout.

[MrJim]

The problem with 2400 RAM is that it is all rated at 1.65v - at least, I've not been able to find any rated lower than that. I wouldn't want to pair it with a Haswell CPU myself.