which RAM setting is better?

[Bananasplit_00]

basically i spend "some time"(6h or so....) doing some RAM overclocking and i ended up with this:

 

 

 

basically is this any better then stock which is this:

 

clocks are looser on the overclock and i dont have the 1T comand rate either but its running 300MHz faster. which one is the better one? all timings were set to auto and i got this, i can get higher speeds but it loosens the timinngs even more. i cant run stock timings with the 2201MHz speed so i just stuck them all on auto for now. anyway which one is the better one?

[Skiiwee29]

honestly, best way to test is to run some benchmark tests like SuperPI or something that will put a stress on your RAM to see how it does. Judging from the limited knowledge I have, I would say the first would be better. 

[Bananasplit_00]

Just now, legacy99 said:

honestly, best way to test is to run some benchmark tests like SuperPI or something that will put a stress on your RAM to see how it does. Judging from the limited knowledge I have, I would say the first would be better. 

yah iv just been makeing sure they work fine with memtest86(which takes about an hour per run ) i havent had time to run any benchmakrs, il do that when i can but i wanted to know if anyone could tell by just that because i have no clue

[Skiiwee29]

They likely will be close, but I know tasks that use ram like them MHz. CAS latency is only 1 difference so I think the extra MHz will be the best. 

[airdeano]

MHz times the CAS (1100.5 x .89 (100 minus CAS 11) = 893 data transfers per second) times two (DDR) 1786MHz

MHz times the CAS (933.8 x .90 (100 minus CAS 10) = 840 data transfers per second) times two (DDR) 1680MHz

CAS is how slow the performing frequency is affected per second. more CAS induced, the slower the processing data is slowed. visually, it is fractional.

 

[Bananasplit_00]

3 minutes ago, airdeano said:

MHz times the CAS (1100.5 x .89 (100 minus CAS 11) = 893 data transfers per second) times two (DDR) 1786MHz

MHz times the CAS (933.8 x .90 (100 minus CAS 10) = 840 data transfers per second) times two (DDR) 1680MHz

CAS is how slow the performing frequency is affected per second. more CAS induced, the slower the processing data is slowed. visually, it is fractional.

 

so the overclock is basically a bit better?

[airdeano]

moving to 2200/CL11 (stock 1.5v?) yes. make you voltage max 1.65/1.675v and work the speed to 2666MHz. then working the timings tighter.

prolly end up on 2600/CL14

1600/CL9 - 1456

1866CL10 - 1680

2133/CL11 - 1898

2400/CL13 - 2088

2666/CL14 - 2293

note the increase in CAS tends to yield the same as last base/CL so essentially, they are all the same just a bit faster in data transfer speed.

using p95 custom is a RAM stability test (http://www.overclock.net/t/838244/prime95-a-quick-dirty-guide-to-the-custom-settings/0_30)

[Bananasplit_00]

6 minutes ago, airdeano said:

moving to 2200/CL11 (stock 1.5v?) yes. make you voltage max 1.65/1.675v and work the speed to 2666MHz. then working the timings tighter.

prolly end up on 2600/CL14

1600/CL9 - 1456

1866CL10 - 1680

2133/CL11 - 1898

2400/CL13 - 2088

2666/CL14 - 2293

note the increase in CAS tends to yield the same as last base/CL so essentially, they are all the same just a bit faster in data transfer speed.

using p95 custom is a RAM stability test (http://www.overclock.net/t/838244/prime95-a-quick-dirty-guide-to-the-custom-settings/0_30)

ok il be sure to try more setting then, im running the stock 1.5V right now

[airdeano]

great! see what you can buy on 1.5 for grins, but eventually the faster speed/lower CAS the voltage will need to be increased.

[Bananasplit_00]

Just now, airdeano said:

great! see what you can buy on 1.5 for grins, but eventually the faster speed/lower CAS the voltage will need to be increased.

should i keep the other setting on auto and just manually change the CAS or keep everything on auto or what? also does the 1t vs 2t comand rate do much?

[airdeano]

in most cases AUTO over applies.

best to hit the max speed with loose timings, then start squeezing the timings.

command rate is whether the data fetch from a DIMM bank is 1 clock cycle or 2 clock cycles. 1T is sensitive vs 2T which is more forgiving, but adds some mushiness in latency.

[Bananasplit_00]

Just now, airdeano said:

in most cases AUTO over applies.

best to hit the max speed with loose timings, then start squeezing the timings.

command rate is whether the data fetch from a DIMM bank is 1 clock cycle or 2 clock cycles. 1T is sensitive vs 2T which is more forgiving, but adds some mushiness in latency.

ok, thanks for the info. il get to my max speed on auto on stock volts, then see what i can do on 1.65V with auto and then il start trimming of the timings then, should i do a memtest86 pass every time i try new settings or should i use something else?

[airdeano]

memtest, superPi, p95, cinebench can ll torture memory for stability

[Bananasplit_00]

1 minute ago, airdeano said:

memtest, superPi, p95, cinebench can ll torture memory for stability

ok il run Cinnebench then seeing as thats what iv got installed and then when i find how hard i can go and pass Cinnebench and iv tweeked my timings down as far as i can il do a few passes of memtest and superPI  thanks for all the help  

[Bananasplit_00]

5 hours ago, airdeano said:

in most cases AUTO over applies.

best to hit the max speed with loose timings, then start squeezing the timings.

command rate is whether the data fetch from a DIMM bank is 1 clock cycle or 2 clock cycles. 1T is sensitive vs 2T which is more forgiving, but adds some mushiness in latency.

well this is what i ended up with, picked up about 30 CB points in just getting these timings lol

 

 

wont boot at any speed higher then this, even at 1.7V and any tighter timigs wont pass memtest so this is what iv ended up with. somehow ended up gaining about 30 CB points from decresing the timigs just a little bit and about 5fps in OpenGL....  very strange.