Conclusion
We looked at different memory speeds for the LGA 1156-based Core i7-870 and chose to run DDR3-800, -1066, -1333, and -1600 at fast, as well as relaxed, timings. Although the differences were typically very small, there were a few applications that obviously benefited from faster memory. This wasn’t surprising, as we already did similar comparisons on most of the other popular platforms:
DDR3 Memory Scaling: Intel’s Core 2 Quad Examined
Core i7 Memory Scaling: From DDR3-800 to DDR3-1600
DDR3 Memory Scaling on AMD’s Phenom II X4
In all cases, we’ve seen significant performance differences when looking at the synthetic or low-level benchmarks. Memory bandwidth does increase considerably if you speed up the memory transfer rate, and tightening timings also improves performance by cutting latencies. However, only a marginal fraction of these benefits actually arrive at the application level. Even going for the fastest memory available will give you a performance boost that is probably smaller than the effect a faster processor speed bin would deliver.
Nevertheless, there are some applications that are more sensitive to memory performance differences than others. Some 3D games (Left 4 Dead in our case) show a noticeable performance boost, likely because it isn't being bottlenecked by graphics performance. Memory-intensive applications, such as Adobe Premiere Pro CS4 and WinRAR, ran quicker, as well. However, the majority of our benchmarks saw little to scarce performance improvement when going for faster memory, so we tend to stay with our original recommendation: go with brand-name memory at mainstream speeds, which still are in the DDR3-1333 space.
However, memory prices have dropped quite a bit, making even DDR3-1600 products more attractive and bringing even some DDR3-2000 products within range. We believe that it’s acceptable to spend a little more on faster memory today if you’re about to invest in other valuable components. Here is our recommendation list (in this order):
- Make sure you have 4GB of RAM. Two 2GB DIMMs are favorable to ensure you can use the tightest timings.
- Make sure you pick a branded product of at least DDR3-1333 speed and timings of CL8 or faster.
- Go for a faster product if you find DDR3-1600 memory that provides the same timings as your preferred DDR3-1333 RAM kit at only a little price premium. Don’t do it if you could get a faster processor for less or the same extra money.
Related Stories :
- Core i7-870 Overclocking And Fixing Blown P55-Based Boards
- Deals May 20: Alienware X51 Core i7 mini Gaming PC $999
- Habey Launches Pro-6820 Fanless Core i7 Quad-Core System
- Next Intel NUC Will Carry Core i5 & i7 CPUs
- Deals April 11: HP ENVY Core i7 Laptop $599.99
- Deals March 13: Dell 15.6'' Core i7 Ultrabook $749.99
- Deals March 9: Dell 15R SE Core i7 1080p Laptop $699
- Deals Feb 26: Dell 15R Core i7 Laptop $649
- Deals February 14: Toshiba Core i7 Ivy Bridge Laptop $549
- FX Vs. Core i7: Exploring CPU Bottlenecks And AMD CrossFire
Great read and it answered some questions for me.
Considering jumping to an i7 soon since the prices are getting pretty reasonable.
might want to check those cpu-z screenshots, mixed up a few...
For me, the conclusion is a bit different:
Buy 8 gigs of some cheap DDR3-1066.
Patrick and Achim , dont show relative performance , show the actual fps or time benchmarks . even 0.6 seconds is 20 percent than than 0.5 seconds , so the benchmarks can not be relevant actually but in relative graphs , they seem (falsely) phenomenal .
u have only mentioned seconds once in the CS4 benchmarks , not very helpful .
as an example , in left for dead , it shows a difference of 10 percent from ddr3 800 to ddr3 1600 , but how many actual fps does that mean ?
Patrick and Achim , dont show relative performance , show the actual fps or time benchmarks . even 0.6 seconds is 20 percent than than 0.5 seconds , so the benchmarks can not be relevant actually but in relative graphs , they seem (falsely) phenomenal .u have only mentioned seconds once in the CS4 benchmarks , not very helpful .
Actually, all they have to do is to mention the FPS of only one of the memory modules for each chart in for example the text, or say how the FPS ranged from min to max between the modules.
Lets do some math... If the normal frame rate is 60, then 5% difference would be 3 frame. So actually nothing...
Yeah, it'd be great to see the raw scores for comparison purposes. If I ever run these benchmarks at home, I'd like to be able to compare my system to this one (same socket, i5 750).
Lets do some math... If the normal frame rate is 60, then 5% difference would be 3 frame. So actually nothing...
Isn't that the ultimate conclusion of the article? I don't see why you are so paranoid.
If the framerate of any game increase 3 frames from memory alone it's huge. We see many graphic cards being called "superior" with those slim margins.
Isn't that the ultimate conclusion of the article? I don't see why you are so paranoid.
If the framerate of any game increase 3 frames from memory alone it's huge. We see many graphic cards being called "superior" with those slim margins.
Yes it is the ultimate conclusion. This was to those who said that they needed real frame rates, not relative speeds...
I was glad to see this article - hoping it would put another nail in the coffin of the seemingly endless debate between faster RAM or tighter timings. Unfortunately, I don't think this article adds much to the current debate. I agree with comments that actual results would be helpful. I would include them in addition to the percentage comparison. Maybe you could overlay the two in a line and bar graph to keep from having pages of graphs to scroll through. Also, the last table in the article is helpful only in comparing all RAM to the 800 CL6 but is misleading otherwise. Take the 1600 CL8's vs 1333 CL7's. The graph makes the 1600s look 1.4% better than the 1333s. But the relative difference is much smaller.
One note for the editors: right up front you say the testing was done on an i7 860. But then in the testing section it shows you used an i7 870.
I want to see someone test the higher end DDR3.
Again another reason why I will never spend top dollar on RAM...
Again another reason why I will never spend top dollar on RAM...
Agreed. I've never really understood why some people spend 2-3x as much on their RAM when it gives a negligable increase in performance. I usually just assume they don't know any better. I seem to be doing that a lot lately...
What about buying faster RAM in order to facilitate overclocking?
But will it speed up Crysis?
But will it speed up Crysis?
honestly speaking , i love it when someone jokes about this for the Nth time , brings all the memories of 2007 back . almost 3 years and still unconquered in graphics .
What about buying faster RAM in order to facilitate overclocking?
Or save money and drops the speeds, and then they'll be bumped back up by the FSB. I still think the money saved on cheaper ram would get you more performance in another component, cpu or gpu.
Even with o/c the 1156 or 1366 platforms you can almost always go down a multiplier so as to NOT o/c your ram. Spending extra money on 2000 ram IMO is not worth it. Worse seeing so many rma working ram because of user misunderstanding on setting timings and their effect on performance.