Article: Tight Timings vs High Clock Frequencies

[Zim]

Just in case you missed this article, THG has released a study on the impact of memory speed and timings on computer performance:

Tight Timings vs High Clock Frequencies

If you just want the bottom line, then here it is:

"we would not advise anyone to spend a lot of money on highest end memory in the hope of improving computer performance by increasing memory speed. ... as long as you have enough memory - preferably 2 GB - the extra money you pay for more memory speed would be better invested in a faster graphics card"

 

[nealh]

Zebo came up wth this long ago..used to have stickied thread on this in fact

 

[RussianSensation]

Originally posted by: nealh
Zebo came up wth this long ago..used to have stickied thread on this in fact

THG explored effects on timings and processor speeds for P4 and A64 when they first came out. Just because they released this article to refresh the topic on people's minds doesn't mean they just came to this conclusion in 2006. In fact, way before Zebo did his benches, most informed computer builders knew that memory dividers, latency and memory speed are not worth the price premiums. Not to take away from Zebo's excellent post, but the knowledge that most money can be saved on memory out of all computer components without incurring large performance drops was common sense 3-4 years ago.

What Tom's hasn't done well is choose appropriate set of benches. Xbitlabs does a far better job:

Article 1 - 2x1gb vs. 4x512mb vs. 2x512mb
Article 2 - Timings, ratios and speed explored for A64

 

[Madwand1]

I think that Tom's is a very good article. While the X-Bit does include a good number of game benchmarks, nobody's going to really absorb all that information, and it generalizes to to something roughly equivalent. That said, it's good that they did so, so that we can see the commonality. If you are picky about a specific game, I still submit that those benchmarks wouldn't really help you because they're also specific about lower-end settings, etc.

In addition, Tom's shows scaling according to CPU speed, which may or may not have been done elsewhere, but is very clear and interesting in their article.