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

New AMD rig to overclock an unlocked 960T w/ too many questions!!?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
i use the computer all day and i really can't afford to use memtest to test for ram staiblity... lately i've been using prime95 blend but .is there anotehr program to reliably test for ram stability..?

also, i'm hestitant on using memtest mainly because it never reprots any ram instability errors while prime95 blend is reporting ram instability errors. so it looks like prime95 is either more reliable in testing ram instability or faster in detecting ram isntability, or am i wrong here? is memtest not accurately gauging ram (in)stability or is slower?

How do you know the cause of the instability when running Prime95 is the ram?
 
I thought you said in an earlier post you had found several stable ram configurations and you successfully ran Prime95 for 36 hrs. Has something changed? When I go back and reread the thread I find it really confusing and hard to follow. You ask a ton of questions all at one time. I would suggest focusing on the essentials. Ram frequency and timings fine tuning are not among them in my book. Ram speed and latency has a very minor impact on performance. Don't try to squeeze every last inch out of the ram. Find something that's stable and move on.

It would probably be helpful to us if you would post new pics of CPU-z tabs: CPU, Memory and SPD to see where you now are with your settings. What is your ram voltage set to and your CPUNB voltage set to? CPU-z won't show those things.
 
better/most important overall = Latency

thanks!

I thought you said in an earlier post you had found several stable ram configurations and you successfully ran Prime95 for 36 hrs. Has something changed? When I go back and reread the thread I find it really confusing and hard to follow. You ask a ton of questions all at one time. I would suggest focusing on the essentials. Ram frequency and timings fine tuning are not among them in my book. Ram speed and latency has a very minor impact on performance. Don't try to squeeze every last inch out of the ram. Find something that's stable and move on.

It would probably be helpful to us if you would post new pics of CPU-z tabs: CPU, Memory and SPD to see where you now are with your settings. What is your ram voltage set to and your CPUNB voltage set to? CPU-z won't show those things.
sorry for the confusion trents....my reply to your previous post was a response to your question about ram instability in repsect to which program i was using to test ram stability....prime95 blend seemed to be working better for me because memtest was always testing at the stock speed no matter what ram setting i changed so therefore memtest was showing no errors obviously....

as for the myriad questions i wasn't expecting one person to answer them all...just for anyone to pitch in whereever they feel best is their area of expertise...

the ram thing i want to get that outta way now before i oc anything else beacue if later i decide to oc the ram then i know that any instability isn't the ram but something else but only if i find stable max settings for the ram now...and sicne i've alrady ahead and gone and found the maximum stable 24/7 ram configurations i want to pick the best perofrming settings I got outta the 2-3 i've just shown...soooo if latency is the first order of improtance/perofrmacne then the 1760+ settings is the way to go for me (i've compared those several different configs and 1760+ settings seems to net faster latencies)

as if right now..all my settings are stock, except for the following:

NB = 2200mhz / CPU/NB volts = 1.175V
HT = 2200mhz / HT volts = 1.25V
RAM voltage = 1.64V (atm this is unnecessary but it doesn't hurt to leave it at that since i'm going to have to pump 1.64V to the ram anyway)

since i'm pretty much done finding the best performing stable ram settings i will next perform testing on the HT @ 1.25V to see how far it will go until it loses performance...once i reach the 'sweet spot ' then i move on to the FSB, CPU mutliplier or NB ok? ;)
 
Stealth3si said:
if i find stable max settings for the ram now

I don't now and never have bought that thing where you can split the various components out and test them and whatever max is it will always remain that as a max even as other parts of the system are sped up. The theory sounds good and maybe somewhat helpful but I never have put much stock in proceeding one component at a time and believing my maxs hold-up overall.

1. Slow cpu and fast ram means the cpu is not under a strain to contend with the ram frequency. Speed that Cpu up and all bets can be off.

2. Slow cpu and the heat of the cpu is not there on the IMC in the cpu. Add the cpu heat and stability may well go out the door.

There are many reasons that " I " personally do not think I can tune one part of a system and my max individually will be my corporate max when all parts are sped up. Not in this lifetime do I believe it to the point that I can say my cpu is the cause of my instability since I do not have my memory clocked beyond where I tested only it to be overclocked and stable. In all things computing, YMMV.
 
I don't now and never have bought that thing where you can split the various components out and test them and whatever max is it will always remain that as a max even as other parts of the system are sped up. The theory sounds good and maybe somewhat helpful but I never have put much stock in proceeding one component at a time and believing my maxs hold-up overall.

1. Slow cpu and fast ram means the cpu is not under a strain to contend with the ram frequency. Speed that Cpu up and all bets can be off.

2. Slow cpu and the heat of the cpu is not there on the IMC in the cpu. Add the cpu heat and stability may well go out the door.

There are many reasons that " I " personally do not think I can tune one part of a system and my max individually will be my corporate max when all parts are sped up. Not in this lifetime do I believe it to the point that I can say my cpu is the cause of my instability since I do not have my memory clocked beyond where I tested only it to be overclocked and stable. In all things computing, YMMV.
yea idon't disagree with you there buddy...after everything is put together all the max limits for each component will need a final minor adjustment, like a bump on the voltage or bump donw on the speed..whatever..that's why i have that thrid 1760 config i listed above ;) i'm just following what the OC guides are telling me to find the max stable so that wehn i do overclock i can start there and work my way up in voltages or work my way down in speed....
 
When you say you're trying the "HT" at 1.25v I assume you mean the HT Link? I don't know anyone who has found that, unlike the CPUNB, over-volting the HT Link helps with stability. In fact, I have found just the opposite in m own experience. It will overclock a little frequency wise without the need for any extra voltage and giving it more volts makes the system unstable.
 
unlike the CPUNB, over-volting the HT Link helps with stability = I think you meant to write >> unlike the CPUNB, over-volting the HT Link does n0t help with stability.

I say this based on what the next sentence said.

RGone...ster.
 
Sicne my last post i decided to keep the HT around 2000-2200 @ 1.20-1.25V, whichever seems more stable.

next, I did some general testing for the NB frequency combinations.

This is just a general range:

NB: 3000 - 3072 @ 1.287500V - 1.350000V
HT: 2000 - 2048 @ 1.20V-1.25V
RAM: 1667 - 1707 @ ~1.55V

NB: 2982 - 3080 @ 1.287500V - 1.350000V
HT: 2130 - 2190 @ 1.20V-1.25V
RAM: 1704 - 1760 @ ~1.55V - ~1.60V

How many hours should I test it for?

Because IME i found that if the NB is usntable in windows it'll freeze within 4 hours, anywhere from the 5th minute mark, or 20 minute mark, 1 hour or little over 3 hours.

If it's quickly noticeably unstable it will not boot into windows.
 
Sicne my last post i decided to keep the HT around 2000-2200 @ 1.20-1.25V, whichever seems more stable.

next, I did some general testing for the NB frequency combinations.

This is just a general range:

NB: 3000 - 3072 @ 1.287500V - 1.350000V
HT: 2000 - 2048 @ 1.20V-1.25V
RAM: 1667 - 1707 @ ~1.55V

NB: 2982 - 3080 @ 1.287500V - 1.350000V
HT: 2130 - 2190 @ 1.20V-1.25V
RAM: 1704 - 1760 @ ~1.55V - ~1.60V

How many hours should I test it for?

Because IME i found that if the NB is usntable in windows it'll freeze within 4 hours, anywhere from the 5th minute mark, or 20 minute mark, 1 hour or little over 3 hours.

If it's quickly noticeably unstable it will not boot into windows.
You are hitting a brick wall with NB speeds relating to stability. At least benchmarking stability. Im in the same boat, my main machine can only hold prime for about 20 minutes, with the nb at 3000mhz. Once I drop back down to 2800 and then 2600 the system gets further and further along with prime95. But in the real world, none of this matters. If this system is stable for your for all the purposes you need, great, dont worry about stress testing it. Most consumers with average or even above average multitasking and even gaming users rarely tap into anywhere near 100% cpu utilization across all the cores for hour after hour. All Im saying, is, if you cant pass prime its not the end of the world. I've used this as my work machine for 3 years now use it for IT work about 7-8 hours per day. I've rarely if ever had a crash or a BSOD, I could count the times on one hand. Fact is we really need to get a water cooling systems for these North Bridge chips. Id say thats where potential is once you've maxed out your CPU and have decent memory to boot.
 
i just set cpu vcore to 1.33-1.34V but hw monitor shows 1.44V!

ahh ok. i see i have LLC enabled...

im going to disable it to see if it improves my OC
 
Last edited:
wow....i stress tested since last nite and my load temps this afternoon went up to 59C becaus of summer weather and now near sunset the weather is cooling down and the load temp is now 54C

this is seriously going to hurt my final stable OC!

somehow i have to find a way to reduce the load temps...will update later. on...
 
S3si, let me say it this way. If you are at X cpu mhz and "still" stable at 59c under full load of Prime95, then you could keep that frequency day in and day out. Why? Because normally you never fully load the cpu for hour after hour like running 8 hours of P95 would do.

There are scientific applications that can push a cpu into full load for hours at a time, but I doubt you are going to run such applications.

What can happen to you when you take it to the upper temp levels or what is upper temp levels for most is that just out of the blue you have a BSOD or similar then you got to chase down the problem and it likely will have been heat related. But that is never the first thing we think of is how close to mostly recommended 'max stable' temps we have been running.

So if you have 3.8Ghz or so at those temps and voltages and the temps go up with daytime heat and yet go back down with sunset, then you likely can skate by.
 
prime95 small fft crashed shortly (with no errors) prior to the 11th hour. :(

I'm going to give vcore one bump up to ensure 24/7 stability and then proceed on to blend test...
 
Last edited:
i re-enabled LLC and did voltage offset by .1.

idle volts dramatically decreased.

load volts slightly increased.

both idle and load temps noticeably decreased.

i'm 3 hrs 50min into Prime95 blend.
 
Frankly, I think the original questions to this thread has already been answered for the most part. Granted, I've sort of trailed off into overclocking a little while back ago so I'll wrap this up with this:

Thank you all for your patience and answers to my questions. I got what I needed to know about the entire process.

All there is left for me to do is to stabilize my final OC, which I believe deserves its own separate thread for organizational purposes...
 
Back