Memory Speed vs Timings

[Carson Dyle]

It's not clear to me how faster timings compare to faster clock speed. Which of the following would be faster on a Z77 motherboard?

16 GB (2x8GB) DDR3 @1866 MHz, 10-11-10-27

16 GB (2x8GB) DDR3 @1600 MHz, 9-9-9-24

 

[jwilliams4200]

Faster for what? The "speed" of any program you run depends on throughput and latency, and those DIMMs have different throughputs and latencies. The only way to see which is faster is to design a benchmark and then run it for both.

But any difference would be small, probably unnoticeable without running a benchmark.

 

[Smoblikat]

I have always been told that intel favors high speed, why AMD favors low latency. Probobly somthing to do with the IMC arch.

 

[aaksheytalwar]

Get Samsung 30 nm. Cheaper and will beat both of them

 

[Smoblikat]

Get Samsung 30 nm. Cheaper and will beat both of them

http://www.newegg.com/Product/Produc...82E16820147096
Are these the aformentioned samsung 30nm? How do they compare to a dtick that does 2133 @ 1.5v CL9 by default?

 

[Carson Dyle]

Another question ... may as well ask here instead of starting another thread.

Are there any advantages to 16GB in 2 x 8GB modules vs 4 x 4GB modules? Enough, that is, to justify the cost difference for the same speed & timing - about $20-25 for DDR3 1600 CL9.

This isn't for a gaming rig, so I'm not concerned about the value of 16GB vs 8GB question. And I don't expect that I'll have a need to upgrade to 32GB during the lifetime of this computer.

 

[cmdrdredd]

Having two slots occupied as opposed to 4 lowers the strain on the memory controller. Generally if you are doing any overclocking, 4 sticks of memory will require a little more voltage or work.

 

[Carson Dyle]

lowers the strain on the memory controller

Pardon me for being thick, but what exactly does that mean? CPU overhead? Wear and tear? Memory throughput?

 

[Makaveli]

I have always been told that intel favors high speed, why AMD favors low latency. Probobly somthing to do with the IMC arch.

Do you have anymore info about this would like to read about it?

 

[TemjinGold]

OP: Let's put it this way, if I put DDR1333 in one machine with crap timings and DDR2133 in another machine with amazing timings and you tried a blind test, you would be hard pressed to tell which is which. I really wouldn't worry too much about the timings unless you are primarily a benchmarker.

 

[Zxian]

OP: Let's put it this way, if I put DDR1333 in one machine with crap timings and DDR2133 in another machine with amazing timings and you tried a blind test, you would be hard pressed to tell which is which. I really wouldn't worry too much about the timings unless you are primarily a benchmarker.

Generally true, but it depends on what you're doing.

To the OP - If you are not overclocking at all, and aren't planning on needing anything beyond 16GB, just pick up 4x4GB DDR3-1600. It will work fine and it's unlikely that you'll notice the difference between that and anything more expensive.

 

[cmdrdredd]

Pardon me for being thick, but what exactly does that mean? CPU overhead? Wear and tear? Memory throughput?

It's ok. Basically the memory controller is what handles the memory speed and voltage. The more sticks it's working with the more it has to do. It's like towing a trailer, the more weight the more work for the engine.

 

[ShintaiDK]

I have always been told that intel favors high speed, why AMD favors low latency. Probobly somthing to do with the IMC arch.

Speed and latency doesnt matter much on Intel due to cachedesign.

http://www.anandtech.com/show/4503/sandy-bridge-memory-scaling-choosing-the-best-ddr3

 

[_Rick_]

Speed and latency doesnt matter much on Intel due to cache design.

This depends entirely on your application.
In most cases there is little difference, but specialist applications can be much more memory bandwidth or latency constrained.
Most benchmarks used in such tests are often compute limited anyway, so it comes as little surprise, that memory has almost no impact, but there are loads that may behave differently, such as simple Photoshop filters, physics simulations with simple arithmetic on a large data set and similarly shaped problems.

Of course, being memory constrained is so rare, that most people that end up in that situation know about it beforehand.

 

[Charles Kozierok]

Speed and latency doesnt matter much on Intel due to cachedesign.

The dirty little truth is that memory speed doesn't matter much period, and hasn't for years, because of caching in general. Especially for enthusiasts who use discrete video.

The difference in real-world performance between the two kits mentioned would be negligible.

 

[Makaveli]

Speed and latency doesnt matter much on Intel due to cachedesign.

http://www.anandtech.com/show/4503/sandy-bridge-memory-scaling-choosing-the-best-ddr3

thank you

 

[lamedude]

If using iGPU 1866MHZ.

 

[ShintaiDK]

True, integrated graphics in the CPUs love more bandwidth. But people buying that memory tends to use discrete GPUs.

And to clearify, iGPU or discrete GPUs dont care about latency, only raw speed.