Overclock.net banner

1 - 20 of 88 Posts

·
Registered
Joined
·
73 Posts
Discussion Starter #1
I have Mushkin DDR2 HP2-6400 (2x1GB) in dual channel.
Stock timings: 4-5-4-11
Stock voltage: 2.1v

I was originally running this at 800 MHz, but then did some overclocking:

From 266x10 -> 333x10

I don't understand why people say always use 1:1 fsb:ram ratio/divider? I understand that it will let me overclock my processor higher as the ram won't be the bottleneck...but If I have my CPU at it's max overclock, can't I increase the divider to get more out of the ram? (2:3 or 4:5?)

I don't know how this is possible but after changing to 333x10 and relaxing ram timings to 5-5-5-12, I was able run the ram at 1000 MHz...Am I breaking my ram? should this even be possible?...I barely changed the timings from stock and didn't change the voltage at all.

Can someone shed some light on if what I'm doing is okay or if I am damaging my ram?
 

·
Registered
Joined
·
73 Posts
Discussion Starter #4
Okay I read the guide but still am unsure about how I was able to overclock the memory that well. After reading the guide, I can use better terminology:

This is DDR2-800 memory that I am running at DDR2-1000 by relaxing timings from 4-5-4-11 to 5-5-5-12 and leaving the voltage at 2.1v. Also, the divider is obviously not 1:1, everyone says 1:1 is best..but all that would do would be lower my memory bandwith....why would I want to do that...
 

·
Premium Member
Joined
·
13,422 Posts
Quote:


Originally Posted by Nerogk
View Post

Okay I read the guide but still am unsure about how I was able to overclock the memory that well. After reading the guide, I can use better terminology:

This is DDR2-800 memory that I am running at DDR2-1000 by relaxing timings from 4-5-4-11 to 5-5-5-12 and leaving the voltage at 2.1v. Also, the divider is obviously not 1:1, everyone says 1:1 is best..but all that would do would be lower my memory bandwith....why would I want to do that...

You don't. The faster memory speed is better. You are giving yourself more bandwidth.
 

·
Premium Member
Joined
·
5,395 Posts
I second Pauldovi here.

Speaking of which, paul, do you want any help in creating some real world evidence of greater ram speed rather than 1:1 being better?
 

·
Registered
Joined
·
1,966 Posts
Quote:

Originally Posted by Nerogk View Post

Can someone shed some light on if what I'm doing is okay or if I am damaging my ram?
I have 2Gb Geil DDR2 800, stock timings at 4-4-4-12 and can easily overclock them to 1000 4-3-3-4 @ 2.1v, rock stable. I think your Mushkin sticks also use Micron D9GMH chips, so you should be able to do the same.
 

·
Registered
Joined
·
3,499 Posts
dukes think of this, have any proof that 1:1 would be better?

now think of this when has making something run at a faster speed bad? the higher frequency is going to be able to transmit more information up to the limit of the FSB. even then though it still has the advantage of getting lower latency for having the same settings.

sorry I know not quite an answer to your question but eh...
 

·
Registered
Joined
·
1,966 Posts
I believe (someone correct me if I'm wrong) that keeping the RAM and the FSB at a 1:1 ratio allows for smoother computing, as the CPU won't have to wait for the RAM and vice versa. If both are at the same speed, there is no waiting for the other.
 

·
Premium Member
Joined
·
13,422 Posts
Quote:

Originally Posted by Ihatethedukes View Post
I second Pauldovi here.

Speaking of which, paul, do you want any help in creating some real world evidence of greater ram speed rather than 1:1 being better?
Bandwidth increases even with increased latencies.

@ 1:1 DDR2-400 Memory bandwidth = 3224MB/s
@ 2:3 DDR2-600 Memory bandwidth = 3774MB/s
@ 1:2 DDR2-800 Memory bandwidth = 4047MB/s





Quote:

Originally Posted by rx7speed View Post
dukes think of this, have any proof that 1:1 would be better?

now think of this when has making something run at a faster speed bad? the higher frequency is going to be able to transmit more information up to the limit of the FSB. even then though it still has the advantage of getting lower latency for having the same settings.

sorry I know not quite an answer to your question but eh...
A little of topic, but your avatar is disturbing! Real men know Honda's are the best cars around. (This is coming from a guy who's family owns 2 Honda Civics, 2 Honda Accords, 1 Honda Odessy, 1 Honda generator, and 2 Honda lawn mowers.



Quote:

Originally Posted by sgtdisturbed47 View Post
I believe (someone correct me if I'm wrong) that keeping the RAM and the FSB at a 1:1 ratio allows for smoother computing, as the CPU won't have to wait for the RAM and vice versa. If both are at the same speed, there is no waiting for the other.
This is just nonsense.
 

·
Registered
Joined
·
3,499 Posts
paul I'm also the proud owner of a honda as well. 1976 accord
. it's just more for the "ricer" type of honda owners.

and forgot about one important thing with ram. even if the FSB gets saturated you still get the advantage of faster responce but also faster copy bandwidth
 

·
Premium Member
Joined
·
13,422 Posts
Quote:


Originally Posted by rx7speed
View Post

paul I'm also the proud owner of a honda as well. 1976 accord
. it's just more for the "ricer" type of honda owners.

and forgot about one important thing with ram. even if the FSB gets saturated you still get the advantage of faster responce but also faster copy bandwidth

Again OT, but I am no ricer fan. I am a fan of Honda because they are build well, they last forever, they are safe and efficient. Best cars ever made. I have gotten 35.4 MPG on my unmodified 3.0 V6 Accord...
 

·
Premium Member
Joined
·
5,395 Posts
Quote:


Originally Posted by pauldovi
View Post

Bandwidth increases even with increased latencies.

@ 1:1 DDR2-400 Memory bandwidth = 3224MB/s
@ 2:3 DDR2-600 Memory bandwidth = 3774MB/s
@ 1:2 DDR2-800 Memory bandwidth = 4047MB/s

As I said I agree with your position. However, synthetic bandwidth benches =/= real world performance. Which I why I was offereing.
 

·
Nobody
Joined
·
6,466 Posts
well 1:1 makes the system stable during overclocking of the fsb, take for instance a 400 fsb with a 1:2 ratio, your ram would try and run at 1600Mhz instead of 800Mhz if it was on a 1:1 ratio. just an example. i run mine at a 4:5 ratio,which is fsb of 325Mhz so my ram is 406 before DDR, it all depends on how much you want to oc your ram, or how much it can handle.

thats my 2cents anyway.
 

·
Registered
Joined
·
1,574 Posts
I would like to add something to this.....I recently purchased and installed Super Talent ddr2-800. According to the website the ram runs at 800 mhz with 4-4-3-8 timings at 2.2v. If i run it at these settings(745 mhz actually with 1/1.33 divider in bios) my sandra results are 7800 mb/sec. If i up the divider to the 1/1.66(902 mhz 4-4-4-12 timings), my sandra says 6800 mb/sec. So i am thinking that added bandwidth isnt always better?
 

·
Premium Member
Joined
·
13,422 Posts
Quote:

Originally Posted by jcharlesr75 View Post
I would like to add something to this.....I recently purchased and installed Super Talent ddr2-800. According to the website the ram runs at 800 mhz with 4-4-3-8 timings at 2.2v. If i run it at these settings(745 mhz actually with 1/1.33 divider in bios) my sandra results are 7800 mb/sec. If i up the divider to the 1/1.66(902 mhz 4-4-4-12 timings), my sandra says 6800 mb/sec. So i am thinking that added bandwidth isnt always better?
Do you have a screen shot?
 

·
Registered
Joined
·
1,574 Posts
nah, but i could make one for you real quick...
 

·
Registered
Joined
·
1,574 Posts
here are my screenies......

Its making a liar out of me right now....i dont know what happened exactly but i did have 7800 mb/sec at the slower speed. But there is no change at all now....i dont know what to make of that.
LL
LL
 

·
Banned
Joined
·
16,364 Posts
Quote:
Its making a liar out of me right now....i dont know what happened exactly but i did have 7800 mb/sec at the slower speed. But there is no change at all now....i dont know what to make of that.
DDR2 @ 900mhz is not capable of 7800mbs. Do the math.
 
1 - 20 of 88 Posts
Top