DDR3 VS DDR4 and overclocking ram

[fudgecakes99]

What is the difference i mean in terms of just general application use. Or more so these two sticks specifically.

http://www.newegg.com/Product/Product.aspx?Item=N82E16820231677&cm_re=ddr3-_-20-231-677-_-Product

http://www.newegg.com/Product/Product.aspx?Item=N82E16820233752&cm_re=ddr4-_-20-233-752-_-Product

So what i do know is that generally ddr4 is double the bandwidth of ddr3 and ddr4 uses less power to run at the same or higher speeds then ddr3. I understand that much outside of a server environment i don't see any benefits. So what i'm asking is a couple things.

At the speeds that will be useable for ddr4 on skylake 170 mobo's is it worth upgrading? Will i have a noticeable difference in loading times general application use between those two types of ram? Is it overall worth switching from ddr3 to ddr4 i'm assuming it's not but i just want to reaffirm my beliefs.

If i oced ddr4 ram would i get noticeable improvements or is it just like ddr3 where it's not actually accountable unless you use copius amounts of ram 64gb+.

 

[Tradesman1]

Corsair did some testing between DDR3 and DDR4 and the differences aren't that great, when you say 'double' the bandwidth it is much higher between dual channel and quad channel which would be obvious, but when testing quad DDR3 vs quad DD4 there isn't much

http://www.corsair.com/en-us/blog/2014/september/ddr3_vs_ddr4_synthetic

 

[fudgecakes99]

Corsair did some testing between DDR3 and DDR4 and the differences aren't that great, when you say 'double' the bandwidth it is much higher between dual channel and quad channel which would be obvious, but when testing quad DDR3 vs quad DD4 there isn't much

http://www.corsair.com/en-us/blog/2014/september/ddr3_vs_ddr4_synthetic

I would imagine that would be the same for overclocking having 1600 mhz would be the sweet spot on ddr3 ram right?

 

[Tradesman1]

For Haswell the sweet spot is 18666/2133 price to performance (and can often fined 2400 sets at the near same price), I wrote an article looking at a number of 2400 sets here:

http://www.tomshardware.co.uk/dram-benchmark-fluctuations,review-33154.html

 

[fudgecakes99]

For Haswell the sweet spot is 18666/2133 price to performance (and can often fined 2400 sets at the near same price), I wrote an article looking at a number of 2400 sets here:

http://www.tomshardware.co.uk/dram-benchmark-fluctuations,review-33154.html

So the same would apply for skylake then?

 

[Tradesman1]

😉 Have to consider if going DDR4, that the entry level point is 2133 - Skylake has a stronger MC than Haswell

 

[fudgecakes99]

yeah but is their really that much of a massive difference between two sticks of 1600-1866 mhz of ddr3 vs 2133 of ddr4? At least in terms of the applications i plan on using it seems like a waste. Plus why spend the extra cash on ddr3 memory when i know for a fact higher speeds will be released later in the life cycle. I figure save some cash on ddr3 now buy ddr4 ram later when its cheaper and better?

 

[fudgecakes99]

😉 Have to consider if going DDR4, that the entry level point is 2133 - Skylake has a stronger MC than Haswell

if i had to choose between a stck of 2600 and 2800 both had ddr4 and both had 16 cas latency would the 2800 one be the better choice? or would it not matter?

 

[Tradesman1]

Yes the 2800 will provide more bandwidth

 

[fudgecakes99]

Yes the 2800 will provide more bandwidth

but then the question is would it warrant it. I mean 2600 mhz ddr4 ram same cl timings at 144 vs 2800 mhz at 229 doesn't seem to justify that when i could almost buy twice the amount of 2600 for the same price.

 

[Tradesman1]

Any time you look into DRAM one needs to look at price to performance, a lot of it depends on what you do and what benefits you or the rig might gain from upping data rates - i.e. most people don't need really high data rates, they just don't do enough with the rig to justify the expense, then there are others that will really work a rig, say doing video work for a living or extremely often, those that heavily multi-task work with very large data sets or memory centric apps, running multiple VMs, etc