Command Rate 1T vs 2T

[UNhooked]

So been out of the loop on memory overclocking but back from the DDR1 days, memory that ran at a lower frequency but with a command rate of 1T was either on par or faster than ram that ran at a higher frequency but with a command rate of 2T.

Has this changed with DDR3 or is it still the same.

 

[ShintaiDK]

You want higher frequency today. Compared to DDR1, CL ratings and command rate means nothing today.

 

[Insert_Nickname]

You want higher frequency today. Compared to DDR1, CL ratings and command rate means nothing today.

Really...?

But you're right in the sense that its mostly an academic discussion. You will not be able to feel the difference outside benchmarks...

 

[UNhooked]

I know typically Command Rate of 1T would normally net you a few more FPS in games but I suppose it doesn't really impact too much.

 

[razel]

Yes, I've been out too for about 10 years. I remember how important those were back in the AMD Athlon days. AMD essentially removed their memory bottleneck by moving the memory controller on the CPU with the Athlon 64s. Intel eventually did that starting around the Core i series.

Because of that, the memory controller in CPUs is designed for all the bandwidth required. So these days memory timings and frequency is the least of your worries. Memory timings can still help but it won't mean much as insert_nickname mentioned outside of benchmarks. Higher frequency timing will only matter if your CPU needs data that is beyond the bandwidth the frequency is limiting it at. And that doesn't happen very often.

Of course that won't stop you or me from trying and testing. I enjoyed tweaking and overclocking computers when it was a hobby.

 

[ShintaiDK]

Really...?

But you're right in the sense that its mostly an academic discussion. You will not be able to feel the difference outside benchmarks...

Not sure why you link it? Its only a benchmark in speed, not testing commandrate or CL. Not to mention only IGPs tested for gaming.

 

[Insert_Nickname]

Not sure why you link it? Its only a benchmark in speed, not testing commandrate or CL. Not to mention only IGPs tested for gaming.

Lets try again then...

http://www.xbitlabs.com/articles/memory/display/ivy-bridge-ddr3_3.html#sect0

And as I wrote in the first post timings matter, but its not by the huge margin it once was. You said it yourself, clockrate is more important then tight timings...

 

[Soulkeeper]

Likely the most definitive memory option comparison i've seen. It is in korean, but the pictures tell the story.
tRFC had one of the biggest performance effects in his testing.

Out of all the memory i've owned over the years (dozens of sets) i've never had one that wouldn't do 1T (2 sticks 4 sticks, ddr1/2/3). Maybe I just got lucky...

Lately, after doing some reading i've taken to using forumlas for calculating my timings.

 

[ShintaiDK]

Likely the most definitive memory option comparison i've seen. It is in korean, but the pictures tell the story.
tRFC had one of the biggest performance effects in his testing.

Out of all the memory i've owned over the years (dozens of sets) i've never had one that wouldn't do 1T (2 sticks 4 sticks, ddr1/2/3). Maybe I just got lucky...

Lately, after doing some reading i've taken to using forumlas for calculating my timings.

Still, less than 1% from the best to the worst.

 

[Vectronic]

Google Translate does an ok job there.

The non-typical tests were interesting. It's a shame that most benchmarks/reviews only test the same 5 timings, rather than the other 20 or so. That's why I don't use XMP profiles, they underclock the non-typical timings.

 

[Stu @ MSD]

Timings are as valuable as always, it's just that the improvements they make pale into insignificance when compared with the improvements another 200 or so MHz made. If you have come from the celeron 133 era as I have, it can be a tricky one to get your head round.

CAS timings are still king, but secondary and tertiary timings are all still worth getting right. I have my PC1600 ram running at 2520mhz.

 

[Soulkeeper]

Still, less than 1% from the best to the worst.

yep
If you're lucky maybe 5% _total_ on a specific app after dialing in every setting from default.

Also worth noting is many of the timings should be changed in lockstep with others, not sure if that korean reviewer did that (doubtfull). Some are actually frequency dependant as well (considering tburst).

 

[Soulkeeper]

Here are some formulas I use:

tburst = 4*2000/ddr3freq
tRAS = CL + tRCD + tburst
rct = tRP + tRAS

rouding up where needed.
consider these as min that make sense

EDIT: sorry if I over-nerded things
default bios settings are good for 99% of people also

 

[Vectronic]

Here are some formulas I use

Well, if you're going to nerd-out... go full nerd. What are ALL your formulas?

 

[Soulkeeper]

This 2010 anandtech article is still relevant today. It has some nice pictures in it. You can glean some other interesting info if you search for the datasheet on your particular chips as well.
I won't list any other formulas because I don't fully understand them myself and can't verify if they are right.

I just like messing with bios settings and dialing things in. Not because I need to, but because I enjoy it