trying to get to 4.2 on a 955be

[Mandrake4565]

So I've been trying to get a stable 4.2 overclock and have been getting random bsod's. I have it 3 hours prime stable at 4.1, with 1.43 cpu v, 1.225 nb v and am able to play any game i throw at it no problem. When I get to 4.2 I ran prime for 3 hours again with the same volts and lost a worker so I upped them to 1.48v "goes up to 1.5 under load" and NB to 1.25. Prime ran fine no problem temps were exactly the same. Went to benchmark "3d mark vantage" and it bsod at the 1st cpu test. The screen shots are all from the 4.0 overclock, i only changed the multi so everything remains the same other then the v core and nb v. Looking at my timings do i need to relax the tras and trc to 25 and 40?

 

[Just a nickname]

What test you use for prime?

 

[Mandrake4565]

Blend torture test

 

[Just a nickname]

I don`t know much about AMD cpu but I see no one replied. Common sense tells me you should first try with ''stable'' settings and see when you begin to crash in the benchmark (at which frequency). It`s way easier to find the problem when you start from a stable point and you gradually increase the speed so you can isolate easily what is the source of the problem.

 

[Mandrake4565]

I have been running it at 4.1 for a while now, It has been totally stable at that speed. Passed Prime, any benchmark and all the games I play no problem. I am just trying to see how far I can push it and also learn what adjustments are needed the higher you push. So far 4.2 has been un obtainable for a stable clock.

 

[keny]

Looking at you're screens there is not a great deal that I can see wrong as you're only using the multi up to now so everything else is still on stock, maybe you could drop the ram to jedec#4 to see if the cpu's IMC is stressed a bit as they are only rated for 1333mhz and you're running 1600 as it is now, It also may be that you could start creeping up the reference in small increments (2 at a time) and see if you can get a bit more out if it that way, but remember changing the reference clock, aka FSB, will also change the ram and ht speed with it, so they will also need adjusting but only if you get instability as they can normaly cope with small amounts of clock increase, and I would say that you won't get more than 210 anyway so you should be ok

 

[trents]

The Denebs most often would not run stable with the ram at 1600 mhz. Thubans will usually but not Denebs, at least not typically.

 

[Mandrake4565]

ok I'll try dropping the ram to 1333 and test it again tomorrow. Getting late so I'm going to call it a night. Thanks for the replies. I'll report back in the morn.

 

[trents]

If you drop the memory frequency you may be able to get away with tighter timings and mitigate some of the bandwidth loss.

 

[Mandrake4565]

Dropped the ram to 1333, I'm going to run prime for 3 hoirs just to make sure that it's still stable at 4.1. Now the question is do I try and tighten the timings before I try 4.2 via multiplier or do I use the fsb to get to it. Am I going to see any real world difference in the pc with the ram at 1333 9-9-9-24-33 vs 1600 9-9-9-23-39?

 

[trents]

I would not tighten the timing first. Make one change at a time. If you are going for a higher CPU clock that is more important. With a little extra ram voltage you may be able to get away with CL 8 but wait and check for stability at higher latencies before you try that.

 

[Mandrake4565]

Ok thanks Trents, 1 1/2 hours into blend. Temps at 45 max settled at 41 for now, cpu v 1.45 cpu/nb 1.25. Letting it run 3 hours before I take the next step. I'm going to try a lil fsb first and see how that works if not I'll try the multi.

On a side note I was reading your post on the 960T about TDP. When the rig is at idle with the fans maxed, I'm using 195 w according to my watt o meter. Right now I'm at 297 running the blend, therefore my processor is generating roughly 100w of heat, correct?

 

[Mandrake4565]

3 hours blend max temps 45 stayed at 41-42 most of the time, no errors. Just upped the fsb to 205 and cpu v to 1.475 reading 1.488 on cpu-z 4216.6mhz on core. I'll report back after this runs to see if it's stable.

 

[skpwn]

Don't forget pictures. We (I) love pictures .

 

[Mandrake4565]

here they are

 

[Mandrake4565]

4 1/2 hours into prime #4 worker quit....I'm going to put the cpu v up to 1.5 even and run it again tomorrow. I figure I'll give the cpu a break for the day

 

[skpwn]

4 1/2 hours into prime #4 worker quit....I'm going to put the cpu v up to 1.5 even and run it again tomorrow. I figure I'll give the cpu a break for the day

Considering that you made it passed the 2 hours threshold, I'd say you're good to go. You're not going to be crunching the CPU for 4 hours at 100% max again while running games or applications!

 

[Mandrake4565]

Sorry my post wasn't totally clear about the 4 1/2 hours. The first 3 hours were at the 4.1 overclock, then I upped the volts on the cpu and brought the FSB to 205 which gave me 4216mhz so the worker quit 1 1/2 hours into the 4.2 clock. So not stable yet. I also decided to run 3d mark vantage again and it BSOD again right when it started the 1 cpu test, which was giving me the problem before. I then upped the cpu v to 1.5 and ran it and it passed so tomorrow I'll run blend again at 1.5 and see if it passes 3 hours. I report back. Thanks again all.

 

[trents]

Ok thanks Trents, 1 1/2 hours into blend. Temps at 45 max settled at 41 for now, cpu v 1.45 cpu/nb 1.25. Letting it run 3 hours before I take the next step. I'm going to try a lil fsb first and see how that works if not I'll try the multi.

On a side note I was reading your post on the 960T about TDP. When the rig is at idle with the fans maxed, I'm using 195 w according to my watt o meter. Right now I'm at 297 running the blend, therefore my processor is generating roughly 100w of heat, correct?

Yep, roughly anyway as my calculations were based the assumption that at "idle" the CPU is producing 0 watts of heat, which is in turn based on the assumption that the gadget CPU activity meter shows little or no activity at idle. Those assumptions are probably not strictly correct as obviously, some juice is flowing through the CPU, even at idle.

 

[Mandrake4565]

I'll get back to you tomorrow and let you know how many more watts are being used with the cpu v at 1.5 i forgot to check earlier. Till tomorrow. Thanks again.