• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Could use some help with my RAM

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

knoober

Member
Joined
Mar 18, 2015
My sig is current ... for today :)

Im trying to eek a bit of extra out of my ram, not for performance sake, but more for experience. I purchased the ram without any consideration for the cpu (and its max supported speed of 1600 apparently) so I figure why not try to get a little extra out of what Ive got right? Worst case scenario I might learn something about reading first :)

I understand you dont really have to play with the FSB to overclock a BE chip, but this is how I intended to reach beyond the limits set by the cpu -- ramp up the FSB so I can speed up the ram and then downclock everything else to ranges that I know will work (which as of right now is 3.8 stable for the cpu , but I think I can do better with some tinkering)

Before we get into what exact settings Ive used or anything I just want to know if what Ive done will work (theoretically). Is it good theory ? Have I misunderstood something? Is this something that only sounds good on paper but is truly impossible in the real world, or an I OC'ing incorrectly?

By doing things this way I can bring the cpu up to 3.8 no problem (Using the FSB instead of the multi) , but it wont boot unless I set the RAM close to stock (which defeats the purpose of OC'ing the long way right?)

Edit: I see I forgot to mention that the RAM is 1866 ! whoops
 
Knoober a few comments I have. For the Phenom chips I believe max supported is 1333 Mhz, that said most will run 1600 Mhz without issue. Additionally, with the correct timings and additional voltage some will run higher then 1600 Mhz. I have had my Ram slightly over 1800 Mhz on my 955be though for benching purposes only, so I do not know if it would run there as a stable daily Oc.

No you don't "have to" use the FSB but, you may find a stable OC at X mhz with a little less voltage using the FSB as opposed to using only the multiplier. Trying to get the Cpu higher or Memory Oc using the FSB is a good idea just keep other changes to a minimum. So try and keep the Ht link and Nb freq as close to where they were when stable. The less frequencies you change at one time the easier it will be to try and find stability. So if you're looking to see how far you can push the ram. Try to keep all other Frequencies static or as close as you can. It will be a lot of trial and error but you will learn a great deal about your set up, doing this.

Notes notes notes, I cannot stress enough, take good notes of every change you make. When you really start pushing especially when using the FSB, it is easy to get lost. Also have a "fall back Oc" so if you do get lost you have a known stable Oc you can revert to.

If you're planning on doing this on your daily OS make sure you have a back up/image. I have borked more OS installs trying to OC ram then anything else.
 
Knoober a few comments I have. For the Phenom chips I believe max supported is 1333 Mhz, that said most will run 1600 Mhz without issue. Additionally, with the correct timings and additional voltage some will run higher then 1600 Mhz. I have had my Ram slightly over 1800 Mhz on my 955be though for benching purposes only, so I do not know if it would run there as a stable daily Oc.
You're right. I went to double check after I read that (both because I wanted to know and because i wanted to be able to get you that info for future reference) and max supported is 1333. I swore that I had read 1600 when I last checked. Seems I was already getting better than I should. Atleast the BIOS has been set for 1600 for quite some time. Ive never paid a whole lot of attention to what CPUz reports in the OS

Notes notes notes, I cannot stress enough, take good notes of every change you make. When you really start pushing especially when using the FSB, it is easy to get lost. Also have a "fall back Oc" so if you do get lost you have a known stable Oc you can revert to.

Thank you for the reminder. I had a goal once to document every change in writing for my PC. Ive changed motherboards and all other components 3 times since then, and have gotten out of the habit.

ANy other tips are appreciated, but the main purpose of this post has already been taken care of. Ill post back when further help is neccessary. I was just getting frustrated with the "no boot" scenarios I guess. It is nice to have the 1333/1600 knowledge corrected as well.
 
If you're Ocing the ram, raise the Dram voltage .05-.1 over stock, if you have an extra fan lying around put it on top of the Ram. Raise the Nb Freq to 1.25-1.3, you may also want to actively cool the VRM/NB area. Drop the cpu, ht link and Nb freq then raise the FSB 5 mhz or less. If you're looking for 24/7 stability test 20 min prime blend, if just looking to push, run Sp32 or Hyper Pi 32m to test. If you can't raise more then 5 mhz without an issue raise the timings 1 at a time, rinse and repeat tests.
 
You posted just as I blue screened! excellent timing

First try ended at about 19min with a BSOD and I went in and tightend (lowered right?) the timings -stock was 10-10-10-11 and lowered to 9-9-9-10- and got almost to half an hour before the BSOD. From reading ive done this should be alright for what I do, but I'm still going to try for a little more stability.

I'm showing different settings in CPUz and the bios (for voltage timings etc- just the ram is off though, cpu settings match), and for some reason I've been kicked out of dual channel mode - even though im running matched sticks in the slots designated by the manual.

Drop the cpu, ht link and Nb freq then raise the FSB 5 mhz or less.

are you saying that I might have to give up some clocks on the cpu to reach higher with the RAM? just checking because that is how Im reading that,
 
You posted just as I blue screened! excellent timing

First try ended at about 19min with a BSOD and I went in and tightend (lowered right?) the timings -stock was 10-10-10-11 and lowered to 9-9-9-10- and got almost to half an hour before the BSOD. From reading ive done this should be alright for what I do, but I'm still going to try for a little more stability.

I'm showing different settings in CPUz and the bios (for voltage timings etc- just the ram is off though, cpu settings match), and for some reason I've been kicked out of dual channel mode - even though im running matched sticks in the slots designated by the manual.



are you saying that I might have to give up some clocks on the cpu to reach higher with the RAM? just checking because that is how Im reading that,
A couple of things, when pushing the memory you may need to loosen the timings ie raise them not tighten them As far as giving up cpu speed, it will depend on the chip. If your goal is getting the memory speed higher then drop everything else until you find what speed you can run the memory at, to your liking. Then try raising the cpu speed back up. Trying to raise the cpu speed and the other frequencies at the same time will lead to a lot of frustration because you will not know what is causing instability. Only change one thing at a time. If your goal is trying to push the cpu above 3.8 using the FSB then drop the other frequencies. If you raise both the Cpu and Memory at the same time you aren't going to know if the Cpu is causing the issue or the Ram.
 
Final verdict: Cpu 3.8+ (3850 or close to it)
RAM: 1704MHZ stable (well stable enough for everyday use, passes 30 min of blend on prime)

I'm going to call it a day here because Ive got a new case and a new mobo to switch too soon. New mobo has better power phasing (8+2 as opposed to the current 4+2) so Im hoping to see some gains there. Ive havent done any long term stability tests but I can run the small fft's for a couple hours at the current settigs. Good enough until I get the new equipment put together. Thanks for the crash course, it gave me what I needed to know for now. I'll be back if it doesnt stay stable
 
Final verdict: Cpu 3.8+ (3850 or close to it)
RAM: 1704MHZ stable (well stable enough for everyday use, passes 30 min of blend on prime)

I'm going to call it a day here because Ive got a new case and a new mobo to switch too soon. New mobo has better power phasing (8+2 as opposed to the current 4+2) so Im hoping to see some gains there. Ive havent done any long term stability tests but I can run the small fft's for a couple hours at the current settigs. Good enough until I get the new equipment put together. Thanks for the crash course, it gave me what I needed to know for now. I'll be back if it doesnt stay stable
Just an FYI small fft doesn't stress the Ram Blend does. With the addition of a much better power section you might find you can achieve X Oc with less voltage then on the 970A D3p.
 
Does it make sense that I cant pass more than a half an hour of prime95 blend but can get through memtest for almost 12 hours? It just seems to me that if Im not stable enough to get past the 30 minute point in prime, then I would for sure throw some errors in memtest. But anyway six passes of memtest and no errors. do I dare to tighten the timings down any further or should I be gratefull for what Ive got? 1704mhz and 9-9-9-24 is what Im running right now. On a different day I might try to see what I can get anyway, but I dont want to play roulette like that today :)
 
Does it make sense that I cant pass more than a half an hour of prime95 blend but can get through memtest for almost 12 hours?

It makes perfect sense. Memtest only tests the ram, under virtually no load.
P95 is stressing the CPU, IMC, and the ram all at the same time.
IMO, your stability problem is CPU/IMC related, not memory.
 
It makes perfect sense. Memtest only tests the ram, under virtually no load.
P95 is stressing the CPU, IMC, and the ram all at the same time.
IMO, your stability problem is CPU/IMC related, not memory.


I guess I was misunderstanding what each test did.

I've already pushed the voltage on the ram up to 1.7v and I really don't know that I care to go any higher, or if it would give any gains worth having. The CPU voltage is definitely higher than I would like to be as well (a bit below 1.475v IIRC) so I guess this truly is the end of the road for this OC on this board. Unless there are suggestions that don't involve raising the voltages - and my already barely acceptable temps? Thanks are appropriate either way soo.... Thanks :)
 
You can try raising the Cpu Nb Vid and see if it helps. This may raise temps though.
 
You can try raising the Cpu Nb Vid and see if it helps. This may raise temps though.
It will. ;)

OP- keep in mind that CPU MHz is king. Don't sacrifice CPU speed for what you think is better ram speed. It makes little difference in everything other than benching.
 
I was just re-reading Dolks guide.... again. Every time I look at that thing I learn more about what I need. The guide agrees with both of you guys. I might try to give it a nudge, but I already bumped it up a little (CPU NB vid) Im a NervousNellie when it comes to adding volts. If anoth +.1 or so on the voltage (CPU NB Vid that is) doesnt give a little more stability then I will either downclock back to stock or accept it as it is. Ive already got 20min of prime blend to say that Im reasonably stable (according to post #4 ^^ ) Ive just heard folks talk about getting hours of stability on prime blend and Im shooting for max stability (well max stability above stock).

As of right now I dont think Im giving up any CPU speed for the ram. I couldnt push the cpu much higher before I started messing with the ram in the first place. I was only using the multiplier then, but I still topped out around 3.8 and some extra Mhz. if I try to push it to 4+ I can boot but i get BSOD as soon as i start prime - with every config of NB and voltage I tried. I hoping this changes with the new board, but even if it doesnt, 3.8 is a respectable OC for my daily driver :)
 
knoober you will "hear" a lot of things on the internet, what is reality is what counts. A few things off the top of my head, from my experience Ocing and spent on these forums. The PII 965 chips seem to need more voltage for X Oc then a 955 Be. I may be wrong in this statement but I'm just going by what I've seen, as I do not own a 965be and do own a 955Be.

As far as stability goes, my gauge for a daily run rig is 2 hours prime blend. With what I do on my Pc, I know if I pass 2 hours blend I will not crash. I will also not post E-Peen Oc's and say it's stable there. Yes my 955 has been as high as 4.7 on Air and 5.0 on chilled H2O, but it was no where near stable. It's just stable enough to run what ever bench I'm attempting. Additionally, as you have found out the components surrounding the chip will play a big part in the overclocking process. Every setup is different so even the exact same components will Oc differently. Will your chip go higher on the new motherboard, hopefully but only testing will tell.

As far as NervousNellie read this guide from AMD it gives a lot of useful information. Page 18 also give recommended tuning targets as far as Mhz and Voltages goes. Does this mean if you follow them to the T. you will not hurt your chip, no. We always run that risk when overclocking but these AMD chips are pretty tough. My 955 has been running since some time in 2011 at 4.0Ghz on a Cooler Master 212 approximately 10 hours a day 5 days a week.
View attachment AMD_Dragon_AM3_AM2_Performance_Tuning_Guide.pdf
 
The PII 965 chips seem to need more voltage for X Oc then a 955 Be. I may be wrong in this statement but I'm just going by what I've seen, as I do not own a 965be and do own a 955Be.
I own both, and you are correct. :)
 
I own both, and you are correct. :)
Thanks Scotty was hoping you had input on that. I was going by the ones I've seen come through here and other forums. They sure do see to need a lot more voltage for X mhz compared to 955's.
 
Ive been doing some reading that says that a lower clock speed on the ram with tighter timings is better than a higher speed with looser timings because the higher speed with loose timings will not provide the enough bandwidth to make up for the difference in timings. Im not really sure what this means. I had thought that the bandwidth of any given stick was static - there's only a set number of pins on each stick after all, isnt that what the bandwidth is determined by- so I am confused with "more bandwidth" at higher clock speeds and even more confused by more bandwidth being available with a given OC. Can you guys shed some light on this for me? Thanks in advance.
 
I am no memory expert knoober but what I can say is this. There is a trade off between timings and speed, when pc is doing memory intensive tasks. For example If you have a set of say DDR3 1600 sticks with 7-7-7 timings they may be able to perform tasks faster then say a set of 2133 11-11-11 sticks. If you compared those same DDR3 1600 Mhz sticks to a really good set of 2133 with tight timings the 2133's would most likely be faster. That all said it also depends on the processor one is using, for example the PII 965. Most will not run ram speeds much higher then 1700-1800 Mhz under normal conditions. So having a set of ram that will run tight timings around 1600 would be beneficial compared to a set that has looser timings.
 
I am no memory expert knoober but what I can say is this. There is a trade off between timings and speed, when pc is doing memory intensive tasks. For example If you have a set of say DDR3 1600 sticks with 7-7-7 timings they may be able to perform tasks faster then say a set of 2133 11-11-11 sticks. If you compared those same DDR3 1600 Mhz sticks to a really good set of 2133 with tight timings the 2133's would most likely be faster. That all said it also depends on the processor one is using, for example the PII 965. Most will not run ram speeds much higher then 1700-1800 Mhz under normal conditions. So having a set of ram that will run tight timings around 1600 would be beneficial compared to a set that has looser timings.
What Mike said here, plus I'd like to add that AMD's prefer timings over speed and Intel's prefer speed over timings, as a general rule of thumb.
 
Back