Is CAS Latency performance noticable?

December 5, 2012 8:49 AM

What's the real world difference between CAS 9 and 10 for modern dual channel memory? For the last six months DDR3 1600 has been roughly double the price for CAS 9 v. CAS 10.

All the articles I've found are at least a year old and aren't helping with the current product landscape.

Is CAS 9 really going to be snappier?

Please do not link to http://www.anandtech.com/show/4503/sandy-bridge-memory-scaling-choosing-the-best-ddr3 - I have read that and (I think - though I may be wrong) it doesn't addresses my current decision.

I have 2x4GB in my machine and I was going to buy another 2x8GB to add in - my existing RAM is DDR3 1600 CAS 9. I suppose I could purposefully slow it down to CAS 10 and see if I notice any difference, and if it is noticeable, spring for the CAS 9 in the additional chips.

In terms of where it would be used, I do a lot of RAW photo processing and that's about it. I actually have a sneaking feeling my budget video card may be a better target for performance upgrade dollars (I already run the system on an SSD - buying a huge one to to store photos seems like it is going to be a huge hassle in terms of clearing them off after processing is done), but I don't game at all and the card is already good enough for HD video playback, in addition to being lowish voltage/temp/noise.

posted by mzurer to Computers & Internet (7 answers total)

Best answer: Have you seen these charts? You're looking at about a 10% increase in memory cycles for 100% price increase. Since any operation you do is going to be using multiple cycles, it's really doubtful you'd be able to notice any difference in your day-to-day workflow.
posted by blue_beetle at 9:18 AM on December 5, 2012

Best answer: You won't notice a difference with delicate RAM timings like this. You're talking about a 1 clock cycle latency difference at 1600MHz.
posted by zsazsa at 9:20 AM on December 5, 2012

Where are you buying your ram? On newegg, the G.Skill DDR3 1600 2x8GB are going for $67 for CL10, and $74 for CL9 under the "Ares" series-- which is no where near 100% price premium.

I see crucial, corsair, and patriot pretty close to $74 as well.
posted by Maxwell_Smart at 10:05 AM on December 5, 2012

Response by poster: You're right - I missed quite a few offerings that are much cheaper. What accounts for that price difference from the many chips that do cost $120? I would think that prices would be driven down on all comparably specced chips.
posted by mzurer at 10:19 AM on December 5, 2012

What accounts for that price difference from the many chips that do cost $120?

People miss quite a few offerings that are much cheaper, and thus some of them will pay $120?
posted by RustyBrooks at 11:09 AM on December 5, 2012 [1 favorite]

Best answer: You're looking at about a 10% increase in memory cycles

This is way over-optimistic. Most memory accesses involve bursts of 2, 4, or 8 transfers with most being the longer because of code and data cache fetches. The CAS latency only applies to the first transfer in a burst. All subsequent transfers in the burst have zero latency. So by eliminating 10% of CAS latency, your overall memory performance increase may be as little as one-eighth of that or barely a 1% performance increase.
posted by JackFlash at 11:35 AM on December 5, 2012

I looked at some of the other chips on Newegg- in the $120 range, I see:

registered, ECC ram (server ram)
CAS 7 RAM
32GB of RAM that mistakenly came up in the search for 16GB
"Gamer" ram with built-in LED lights

for a couple of things that push the price toward $120 on Newegg.
posted by Maxwell_Smart at 2:00 PM on December 5, 2012