OP
- Joined
- Jan 10, 2012
- Thread Starter
- #21
i'll let woomack answer that.
-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
not what i had expected
- Thread starter caddi daddi
- Start date
- Joined
- Dec 19, 2012
I get it now you're adjusting nb along with your ram frequency those are the numbers you're posting. Sorry I just got a bit confused
I've seen a lot of posts where people suggest keeping the HT in tandem wit the NB but I find that it works a tad bit better if my HT is behind my NB
I've seen a lot of posts where people suggest keeping the HT in tandem wit the NB but I find that it works a tad bit better if my HT is behind my NB
OP
- Joined
- Jan 10, 2012
- Thread Starter
- #23
that's what i am doing.
with a number of posters it's " stutter while i game", with 1600 ram.
raise the cpu/nb voltage above 1.3 and speed to 2500 and the ht link to 2500 and they report that it improves, next step is 1866 ram and it seems to correct the issue.
with a number of posters it's " stutter while i game", with 1600 ram.
raise the cpu/nb voltage above 1.3 and speed to 2500 and the ht link to 2500 and they report that it improves, next step is 1866 ram and it seems to correct the issue.
- Joined
- Dec 27, 2008
Well, this may be a case where the extra bandwidth of the higher frequency ram is not able to be efficiently used by the six core CPU so the quicker timings of the lower frequency ram wins out. Extra bandwidth is not useful unless it is needed any more than extra freeway lanes are needed when traffic is light.
Last edited:
OP
- Joined
- Jan 10, 2012
- Thread Starter
- #26
trents i get that concept and can apply it in the benchmarks.
- Joined
- Apr 20, 2004
- Location
- JAX, Mississauna
Dang internet down way too long...
I wanted to post this yesterday but at 4:00AM beautiful comcast was still down in my 'area'. Wow.
C_D and I normally setup our systems pretty close to alike. He does like to play with ram timings more than I do but when he is using Linux to do real work and not play, then is where he sees results or no results from his tweaking since he has an application with a lot of calculating to do and he has timed the app to completion maybe 500 plus times so he knows when he is getting positive results from his tweaking.
Let us say he is a, "I want to buy it now" type of person and has at least 3 of the pretty major online outlets available to him as 'brick and mortar' stores. However they don't seem to sell G Skill ram which is my suggestion for AMD systems. Anyway long story to short, he had a set of ram to fail at least one stick so he was getting short on ram. Went to brick and mortar to get some. Got DDR1866 in Corsair brand and came home and inserted and the results were not as expected.
This new DDR1866 ram just plugged in and overclocked quite vigorously did not really produce for him any better "benched" performance. Not a lot less peformance but not the increase he expected.
I suggested a couple of things to him on the first page of this thread and he was able to try a couple but still without the results he was "expecting". My second suggestion to him was to set tRCD to 10 from 11 but his "new" ram would not sustain such behaviour yesterday. By phone I suggested that he lower the "newer" ram overclock and adjust tRCD to 10 then and test. Doing so should result in him getting his "expected results", just as I did when I tested his original setup. As you will be able to see by my posted captures, my F3-1866C9D-8GAB ram was able to use tRCD at 10 instead of only 11 *even* when overclocked as he had tried.
So by phone, I made the suggestion to lower his ram clock and then be able to use the tRCD to 10 setting. I said that he might give his "new" ram a little more voltage in order to allow the tRCD of 10 setting while overclocked OR to adjust cpu speed by whatever means to get his 4587Mhz and "have" the "new" ram running at only DDR1900ish or so. Doing this of course will require making minor changes to FSB and the other busses as well, but maintain his cpu clock speed. Of course maintaining high CPU_NB and HT frequencies is what he will try to do and thus make as great an effort as possible to make his subsequent testing as close to apples to apples as possible. Within reason since he does need to change the ram speed and that will necessitate an FSB/HT Ref change.
I only write this in such a long-winded appearing manner since I am fairly cetain more than just one of us has come face to face with this issue. The issue of buying faster and looser timed ram and not seeing "the expected" results.
Had his "ram" been AUTO adjusted by the motherboard bios and/or been able to run with timings suggested as my ram is able to run, then C_D would have seen his "expected results" as outlined below in my uploaded test captures.
RGone...
My G Skill ram setup like his original ram was setup with a DDR1700 ram speed and his original 9, 9, 9 timings. Setup from what I remembered since my Internet was down for over 1/2 day.
How he wished his ram could have run when overclocked but did not and gave him "unexpected results".
I wanted to post this yesterday but at 4:00AM beautiful comcast was still down in my 'area'. Wow.
C_D and I normally setup our systems pretty close to alike. He does like to play with ram timings more than I do but when he is using Linux to do real work and not play, then is where he sees results or no results from his tweaking since he has an application with a lot of calculating to do and he has timed the app to completion maybe 500 plus times so he knows when he is getting positive results from his tweaking.
Let us say he is a, "I want to buy it now" type of person and has at least 3 of the pretty major online outlets available to him as 'brick and mortar' stores. However they don't seem to sell G Skill ram which is my suggestion for AMD systems. Anyway long story to short, he had a set of ram to fail at least one stick so he was getting short on ram. Went to brick and mortar to get some. Got DDR1866 in Corsair brand and came home and inserted and the results were not as expected.
This new DDR1866 ram just plugged in and overclocked quite vigorously did not really produce for him any better "benched" performance. Not a lot less peformance but not the increase he expected.
I suggested a couple of things to him on the first page of this thread and he was able to try a couple but still without the results he was "expecting". My second suggestion to him was to set tRCD to 10 from 11 but his "new" ram would not sustain such behaviour yesterday. By phone I suggested that he lower the "newer" ram overclock and adjust tRCD to 10 then and test. Doing so should result in him getting his "expected results", just as I did when I tested his original setup. As you will be able to see by my posted captures, my F3-1866C9D-8GAB ram was able to use tRCD at 10 instead of only 11 *even* when overclocked as he had tried.
So by phone, I made the suggestion to lower his ram clock and then be able to use the tRCD to 10 setting. I said that he might give his "new" ram a little more voltage in order to allow the tRCD of 10 setting while overclocked OR to adjust cpu speed by whatever means to get his 4587Mhz and "have" the "new" ram running at only DDR1900ish or so. Doing this of course will require making minor changes to FSB and the other busses as well, but maintain his cpu clock speed. Of course maintaining high CPU_NB and HT frequencies is what he will try to do and thus make as great an effort as possible to make his subsequent testing as close to apples to apples as possible. Within reason since he does need to change the ram speed and that will necessitate an FSB/HT Ref change.
I only write this in such a long-winded appearing manner since I am fairly cetain more than just one of us has come face to face with this issue. The issue of buying faster and looser timed ram and not seeing "the expected" results.
Had his "ram" been AUTO adjusted by the motherboard bios and/or been able to run with timings suggested as my ram is able to run, then C_D would have seen his "expected results" as outlined below in my uploaded test captures.
RGone...
My G Skill ram setup like his original ram was setup with a DDR1700 ram speed and his original 9, 9, 9 timings. Setup from what I remembered since my Internet was down for over 1/2 day.
How he wished his ram could have run when overclocked but did not and gave him "unexpected results".
OP
- Joined
- Jan 10, 2012
- Thread Starter
- #28
- Joined
- Apr 20, 2004
- Location
- JAX, Mississauna
Odd and no clue the real why.
Latest efforts with fast ram show CPU 2 Red Valley higher and HDR1 Canyon flight higher than the slow ram, but the rest of the scores seem lower. Completely strange as it is understood. Crap I meant not understood. Maybe working with a bench that is just not what we need to be using for the looksee. FPS done to 9 decimal places are just not showing anything over 1% difference if looking at it correclty. That is within what we call a margin of error.
Then when comparing to the faster ram speeds you posted, the latest bench shows a higher overall score than the first fast ram setup and shows greater performance in almost every single bench area. Now the ram speed is slower but timings are different than first fast ram setup but even then the original slow ram overall score is not bested everywhere. Beaten is some areas but not all areas. Who the freek knows? Not Hehehe. I am ticked.
Could it just be shett ram compared to what mine will do? How to know? Some other bench of memory only performance maybe? One thing is for sure...performance of my ram precludes me from seeing the wonky results you are seeing C_D.
RGone...
Latest efforts with fast ram show CPU 2 Red Valley higher and HDR1 Canyon flight higher than the slow ram, but the rest of the scores seem lower. Completely strange as it is understood. Crap I meant not understood. Maybe working with a bench that is just not what we need to be using for the looksee. FPS done to 9 decimal places are just not showing anything over 1% difference if looking at it correclty. That is within what we call a margin of error.
Then when comparing to the faster ram speeds you posted, the latest bench shows a higher overall score than the first fast ram setup and shows greater performance in almost every single bench area. Now the ram speed is slower but timings are different than first fast ram setup but even then the original slow ram overall score is not bested everywhere. Beaten is some areas but not all areas. Who the freek knows? Not Hehehe. I am ticked.
Could it just be shett ram compared to what mine will do? How to know? Some other bench of memory only performance maybe? One thing is for sure...performance of my ram precludes me from seeing the wonky results you are seeing C_D.
RGone...
OP
- Joined
- Jan 10, 2012
- Thread Starter
- #30
