Sign in to follow this  
Followers 0
sathenzar

Is DDR3 3000oc worth it?

14 posts in this topic

Just as the title describes, I'm asking if DDR3 3000oc is worth it. It seems to be around $600. I think I'm using DDR3 12800 or something like that atm. I'm going to build a new computer at the beginning of next year so I'm doing my research. Wanted to know how much of a difference it would make. I am a programmer so I actually use my computer to do things other then just play games.

Share this post


Link to post
Share on other sites

Honestly, higher RAM speeds will give you very very little, if any, gains in noticeable performance.

 

A faster SSD, GPU, or even CPU will be much much more worth your money.

Share this post


Link to post
Share on other sites

One word: No. (Unless you are a benchmark ######, it will give you a lot of trolling power on 3dmark forums)

Share this post


Link to post
Share on other sites

In benchmark scores, yes. In reality, not very much. 

Share this post


Link to post
Share on other sites

Curious, when has faster RAM been better than MORE RAM???......

 

.......

 

 

Yeah, never :p

Share this post


Link to post
Share on other sites

No, I run 8GB of DDR3 1333 RAM, I'm happy...

Share this post


Link to post
Share on other sites

no it isn't worth it.

 

CPUs care more about latency than bandwidth. just disassemble some programs,and all you'll see is a ton of load and store instructions. if the size of these memory operations is smaller than the cache lines, then for example writing 32 bytes will make the cpu read the whole cache line,modify it,then write it back. such a waste. even if you use sse, avx or non temporal instructions that bypass the caches, there is still this overhead. the only way to get the true potential out of this bandwidth is to use rep stosq and rep movsq string instructions, but this is only beneficial if youre transferring large amounts of data ,most of the time the cpu is working on smaller data sets,and in the real world, this is what the cpu does the majority of the time.

1 person likes this

Share this post


Link to post
Share on other sites

CPUs care more about latency than bandwidth. just disassemble some programs,and all you'll see is a ton of load and store instructions. if the size of these memory operations is smaller than the cache lines, then for example writing 32 bytes will make the cpu read the whole cache line,modify it,then write it back. such a waste. even if you use sse, avx or non temporal instructions that bypass the caches, there is still this overhead. the only way to get the true potential out of this bandwidth is to use rep stosq and rep movsq string instructions, but this is only beneficial if youre transferring large amounts of data ,most of the time the cpu is working on smaller data sets,and in the real world, this is what the cpu does the majority of the time.

 

See, the driver hooks a function by patching system call table...

 

(if URL is borked, 1:33 to 1:44 :p )

 

 

Anyway, access time is a function of frequency and an inverse of latency. So by increasing either access time is decreased.

Is that the gain from Ivy Bridge defaults 1600/9-9-9-24 to 3000/CL12-14-14-36 is very little - about 2 ns, because latency has been raised to maintain the frequency.

1 person likes this

Share this post


Link to post
Share on other sites

Fast RAM matters a lot for integrated GPUs like the current AMD Trinity/Richland chips. I personally witnessed my framerates skyrocketing simply by enabling the memory profile in the BIOS, and many benchmarks online confirm this. I don't know what the situation is for Intel's iGPUs but I would assume it is similar. GPUs are very simple processors that deal with large amounts of data without complex caching mechanisms; accessing memory quickly is of the essence.

 

For pure CPU performance it doesn't matter much; performance is all about keeping things in cache. Accessing system memory is dreadfully slow from a CPU's perspective - a C++ compiler expert at BUILD 2013 compared it to taking a coffee break -, and shaving a few nanoseconds here and there is not what makes a program drastically faster. Even overclocking RAM by large amounts only show marginal improvements in CPU benchmarks.

 

So in short, if you're going with a traditional CPU + discrete GPU with its own VRAM, then RAM speed doesn't matter much; but it does matter on integrated GPUs and it will matter more and more as iGPUs get better. The next gen consoles are basically integrated GPUs similar to AMD Richland/Kaveri, btw, so the trend might end up on PCs sooner than later. Keep an eye on Kaveri when it's released early next year.

1 person likes this

Share this post


Link to post
Share on other sites

if your going to be overclocking, the RAM speed is important, if your going to be overclocking a huge amount, you will also need very good cooling, 

 

if you do not overclock, RAM that is over 1333mhz is pointless. 

 

when you put RAM that is higher than the default FSB speed of the CPU and what is set in the motherboard, say you put in 1600mhz RAM and are not overclocking, your ram will run at the default speed set by the FSB in the bios, which will be 1333mhz or 1066Mhz, you will only get the 1600Mhz out of your ram if you overclock or change the settings in the bios.

 

higher speed RAM is not really worth the price, even if your overclocking, 1600Mhz should be more than enough to get a good overclock.

 

if your a programmer, the higher speed ram is of no use, you need more RAM not higher speed.

Share this post


Link to post
Share on other sites

if your going to be overclocking, the RAM speed is important, if your going to be overclocking a huge amount, you will also need very good cooling, 

 

if you do not overclock, RAM that is over 1333mhz is pointless. 

 

when you put RAM that is higher than the default FSB speed of the CPU and what is set in the motherboard, say you put in 1600mhz RAM and are not overclocking, your ram will run at the default speed set by the FSB in the bios, which will be 1333mhz or 1066Mhz, you will only get the 1600Mhz out of your ram if you overclock or change the settings in the bios.

 

higher speed RAM is not really worth the price, even if your overclocking, 1600Mhz should be more than enough to get a good overclock.

 

if your a programmer, the higher speed ram is of no use, you need more RAM not higher speed.

 

Hey hey, I'm a programmer in both c++ and opencl and ram speed does matter, not much, but does matter nonetheless. Games and other task normally don't see the benefit of higher speed ram though.

Share this post


Link to post
Share on other sites

if you plan to OC then yes

Share this post


Link to post
Share on other sites

I run 16gb of ram and I think I might have overdone it XD

Share this post


Link to post
Share on other sites

Better use 1866 CL9, still cheap, but better than most >2000MHz because of low CL.

 

Of course you CPU/MB pair can clock >1600MHz ?

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!


Register a new account

Sign in

Already have an account? Sign in here.


Sign In Now
Sign in to follow this  
Followers 0

  • Recently Browsing   0 members

    No registered users viewing this page.