Overclock.net banner

1 - 20 of 48 Posts

·
Registered
Joined
·
102 Posts
Discussion Starter #2
wow I've been able to get it to 6-7-5-18-42-1
Not sure if I should tighten it more
There seems to be no difference between 6-8-6-22-1, 6-7-5-10-1 and 6-7-5-18-1
LL
LL
LL
LL
 

·
Premium Member
Joined
·
20,621 Posts
Looks pretty good; I'd suggest bumping that NB clock up to get the full potential, I have RAM running at 6-8-6 with a 2700Mhz NB and I can get 10GB/s + memory score and less than 50ms latency scores
 

·
Registered
Joined
·
102 Posts
Discussion Starter #4
yep that makes sense. I tightened even more (didn't test the limit yet though) and still very little difference. In the last screen, I upped NB by only 250mhz and a huge difference, compared to lowering timings
LL
LL
LL
LL
 

·
Premium Member
Joined
·
20,621 Posts
6-7-5-15, oh my... seems really impressive
 

·
Registered
Joined
·
102 Posts
Discussion Starter #6
haha yeah lol it seems that it only seems impressive. I saw so many people regurgitating that lower timings are better than higher clocks for amd, even saying that 1 timing was equal to 150mhz or something. Take a look at this screen shot.
So basically going from 6-7-5-15-1-25 at 803mhz (NB 2400, CPU 3374mhz) to 11-11-11-29-40-1 at 1000mhz (NB 2500, CPU 3500mhz) gives better results. Also, for 2000 ram I can keep 1.5v, however if I want to tighten at 1600 even by the smallest increment I have to bump the voltage
LL
LL
 

·
Premium Member
Joined
·
9,823 Posts
Quote:

Originally Posted by olegplanets View Post
haha yeah lol it seems that it only seems impressive. I saw so many people regurgitating that lower timings are better than higher clocks for amd, even saying that 1 timing was equal to 150mhz or something. Take a look at this screen shot.
So basically going from 6-7-5-15-1-25 at 803mhz (NB 2400, CPU 3374mhz) to 11-11-11-29-40-1 at 1000mhz (NB 2500, CPU 3500mhz) gives better results. Also, for 2000 ram I can keep 1.5v, however if I want to tighten at 1600 even by the smallest increment I have to bump the voltage
The increase you're seeing is from an increased CPU/NB as well - 2400 vs 2100. I get the results of your 1000Mhz speeds @ 1600 @ 7-6-6 @ 2700 CPU/NB.
 

·
Registered
Joined
·
102 Posts
Discussion Starter #8
I think you missed some of the screens. It's not 2400 vs 2100, it's 6-7-5-15-1-25 at 803mhz (NB 2400, CPU 3374mhz) vs 11-11-11-29-40-1 at 1000mhz (NB 2500, CPU 3500mhz)
 

·
Premium Member
Joined
·
9,823 Posts
Admittedly I didn't read the whole thread or all of the posts - but...

Quote:

Originally Posted by olegplanets View Post
yep that makes sense. I tightened even more (didn't test the limit yet though) and still very little difference. In the last screen, I upped NB by only 250mhz and a huge difference, compared to lowering timings
This post has a SS's that shows a CPU/NB speed of 2100~ for 1600Mhz RAM at tight timings.

EDIT: One SS (third) shows 2400 CPU/NB.

And...

Quote:

Originally Posted by olegplanets View Post
haha yeah lol it seems that it only seems impressive. I saw so many people regurgitating that lower timings are better than higher clocks for amd, even saying that 1 timing was equal to 150mhz or something. Take a look at this screen shot.
So basically going from 6-7-5-15-1-25 at 803mhz (NB 2400, CPU 3374mhz) to 11-11-11-29-40-1 at 1000mhz (NB 2500, CPU 3500mhz) gives better results. Also, for 2000 ram I can keep 1.5v, however if I want to tighten at 1600 even by the smallest increment I have to bump the voltage
This post shows a SS with RAM @ 2000Mhz with normal/loose timings and mentions the CPU/NB speed at 2500.

Did I miss something?
 

·
Registered
Joined
·
102 Posts
Discussion Starter #10
well exactly, post #4 screen 4, I have 6-7-5-15-1-25 at 803mhz (NB 2400, CPU 3374mhz)
post #6, two screens of the same configuration, which is 11-11-11-29-40-1 at 1000mhz (NB 2500, CPU 3500mhz)
NB and CPU MHz are almost the same in both instances. However, a slightly higher ram clock 1000(2000) vs 800(1600) produces the same result with timings that are almost half more loose. In essence what I found is that CL6 at 1800 is equivalent to CL11 in 2000, which doesn't seem to follow what I've read so far...
 

·
Premium Member
Joined
·
9,823 Posts
Well - I personally don't use MaxxMem because I find it doesn't correctly identify my speeds. But - if you have Everest it might be more beneficial in reporting with the cache/memory benchmarks.

As to CL6 @ 1800 versus CL11 @ 2000, seems to me the point is proven. The delay time in CL6 is greatly reduced in contrast to CL11, at any speed, and this is where AMD is happier. If it were Intel, the delay would not contrast so differently (due to different IMC architecture).

EDIT: Latency ~ is a better word than delay.
 

·
Premium Member
Joined
·
20,621 Posts
I'm quite sure CAS6 at 1800 will perform much better than CAS11 at the small speed difference of 2000.
 

·
Registered
Joined
·
2,770 Posts
Well, from reading this thread you are comparing two things that are close to equal, but not equal. A NB frequency of 2400 is not the same as a NB frequency of 2500. Also having the CPU at 3.37 is not the same as having it at 3.5. Yes, they are close but not equal. In order to compare you need to make things equal.

Why not try stepping those timings down one by one at 2000 speed and see what happens. I would probably run the ram at lower speed and tighter timings if performance is on par.
 

·
Registered
Joined
·
4,237 Posts
Quote:

Originally Posted by olegplanets View Post
haha yeah lol it seems that it only seems impressive. I saw so many people regurgitating that lower timings are better than higher clocks for amd, even saying that 1 timing was equal to 150mhz or something. Take a look at this screen shot.
So basically going from 6-7-5-15-1-25 at 803mhz (NB 2400, CPU 3374mhz) to 11-11-11-29-40-1 at 1000mhz (NB 2500, CPU 3500mhz) gives better results. Also, for 2000 ram I can keep 1.5v, however if I want to tighten at 1600 even by the smallest increment I have to bump the voltage
Shouldn't you compare apples to apples?

So you would want to run multiple tests to see the effects:
Test 1:
6-7-5-15-1-25 @ 803mhz
NB 2400
CPU 3374mhz

Test 2:
6-7-5-15-1-25 @ 803mhz
NB 2400
CPU 3374

Test 3:
11-11-11-29-40-1 @ 1000mhz
NB 2500
CPU 3500mhz

Test 4:
11-11-11-29-40-1 @ 1000mhz
NB 2500
CPU 3500mhz

The goal is to figure out the gain for each performance enhancement, then from there you can figure out what yields the best results.
 

·
Registered
Joined
·
102 Posts
Discussion Starter #16
Quote:

Originally Posted by xd_1771 View Post
I'm quite sure CAS6 at 1800 will perform much better than CAS11 at the small speed difference of 2000.

How do you know? How can I check?
 

·
Registered
Joined
·
102 Posts
Discussion Starter #17
Quote:

Originally Posted by PeaceMaker View Post
Shouldn't you compare apples to apples?

So you would want to run multiple tests to see the effects:
Test 1:
6-7-5-15-1-25 @ 803mhz
NB 2400
CPU 3374mhz

Test 2:
6-7-5-15-1-25 @ 803mhz
NB 2400
CPU 3374

Test 3:
11-11-11-29-40-1 @ 1000mhz
NB 2500
CPU 3500mhz

Test 4:
11-11-11-29-40-1 @ 1000mhz
NB 2500
CPU 3500mhz

The goal is to figure out the gain for each performance enhancement, then from there you can figure out what yields the best results.
tests 1 and 2 are the same as well as tests 3 and 4. And that's what I did, no?
 

·
Premium Member
Joined
·
9,823 Posts
Quote:

Originally Posted by olegplanets View Post
How do you know? How can I check?
You need equalized settings with both to separate the results down to just the RAM speed/timing.
 

·
Registered
Joined
·
102 Posts
Discussion Starter #19
Quote:

Originally Posted by Asmola View Post
tRP (5) does almost nothing on AMD-system..

Try if those can do CL7 2000MHz!

Here's the best I got at 2000.
So yeah, 1000mhz at 8-9-9-28 outperforms 6-7-5-15-1-25 at 803mhz.
Where are all the people who said timings are better? What did I do wrong?
LL
LL
LL
 

·
Premium Member
Joined
·
9,823 Posts
Quote:

Originally Posted by olegplanets View Post
Here's the best I got at 2000.
So yeah, 1000mhz at 8-9-9-28 outperforms 6-7-5-15-1-25 at 803mhz.
Where are all the people who said timings are better? What did I do wrong?
I think the stock timing on most 2000 kits, usually rated for Intel and otherwise, are at 10/11 and that's where vs CL6/CL7 1600 RAM will show better results. When you bring 2000 down to CL8, you're just confirming that 'lowering timing' does improve your RAM speed.

You're just doing it at a higher RAM clock - 1000Mhz. I'll give you an example:

1333 @ CL6 performs on and/or about the same as 1600 @ CL8 - but 1600 @ CL7 or CL6 outperforms both.
 
1 - 20 of 48 Posts
Top