8320FX

[RGone]

We d0n't need Open Hardware Monitor capture. We d0n't need Core Temp capture. We d0n't have to have P95 capture..just that it failed in X mins with what error.

But we need HWMonitor open far enough to see Voltages and include Package/Core temps. And we need CpuZ > CPU TAB > Memory Tab > SPD Tab and those four captures will get 'er done.

Without the needed stuff we are just peeing north into a north wind. Messy and little good.

Stable how long with P95 BLEND Mode? 2 hours? That is what we consider stable in just about every situation.

Looking at HWMonitor and seeing Value of 1.344V for cpu voltage...I sort of doubt that you can run 2 hours of P95 BLEND mode error free. That is normally just not enough cpu voltage to support ~4.6Ghz. I have passed the IBT before for 20 runs and failed P95 BLEND mode test and had to add Vcore to the cpu. Others have had it the other way around as you are having and the issue was not enough Vcore for the cpu speed.

In addition, I notice your HT Link Speed is now at 2716.66Mhz and can be an issue.
RGone...

 

[MyNameIsWill]

Ok you said "it seems" you are stable running prime blend, how long did you run it. 2 hours is usually what you'll need to see if it's truly stable. What I would do is this, if it really isn't stable. Drop the Ht link speed to 25xx then set your ram timings to the JEDEC#6 profile. Set the Cpu Nb voltage to 1.2 not the "Nb Voltage", leave that one alone and run prime blend 2 hours to see if you pass. Another option is to the above but instead of changing your ram timings drop the ram freq to 666 which will be 693 with the Bus Speed at 208.

Prime can run for 2hours+, It just cannot run the intel burn test because that uses a lot of ram. and im sure it's because of my unstable ram. You said to lower HT to 25xx, does that effect performance? i thought the highest settings is best, or does lowering the HT increase performance? Thanks

 

[MyNameIsWill]

We d0n't need Open Hardware Monitor capture. We d0n't need Core Temp capture. We d0n't have to have P95 capture..just that it failed in X mins with what error.

But we need HWMonitor open far enough to see Voltages and include Package/Core temps. And we need CpuZ > CPU TAB > Memory Tab > SPD Tab and those four captures will get 'er done.

Without the needed stuff we are just peeing north into a north wind. Messy and little good.

Stable how long with P95 BLEND Mode? 2 hours? That is what we consider stable in just about every situation.

Looking at HWMonitor and seeing Value of 1.344V for cpu voltage...I sort of doubt that you can run 2 hours of P95 BLEND mode error free. That is normally just not enough cpu voltage to support ~4.6Ghz. I have passed the IBT before for 20 runs and failed P95 BLEND mode test and had to add Vcore to the cpu. Others have had it the other way around as you are having and the issue was not enough Vcore for the cpu speed.

In addition, I notice your HT Link Speed is now at 2716.66Mhz and can be an issue.
RGone...

Yeah, im 100% sure it's not the CPU Vcore, i went all the way up to 1.5V just to see if it will stabilize it, but it didn't. I'm almost 100% sure it's the ram. And that is exactly what i am looking for, so you are saying HT over 2600 = worse? i thought decreasing the HT would effect speed and lower the speed overall on the computer. is 2400>2600 for HT? And can NB Frequency be higher than HT, is that good or bad? or do they have to balance. Sorry for so much questions. And my Prime95 doesn't fail in tests, even when left for 3+ hours. Only fails in blend test and im sure the reason is the ram like i said.

 

[RGone]

IF we can get on the same page...

...maybe we can adjust to sustain IBT.

Well not sure why you are running Bus Speed of 208 when the stock is 200. That little bit is
also raising the Memory speed and the HT Link Speed and the CPU_NB speed by some amount.

I also see that the ram you have is that new AMD stuff we have zero experience with...good or bad.

It shows to be DDR3-1866 at 1.65V and when only running about DDR3-1600 it calls for 1.5V to ram.
I would raise the ram voltage to about 1.6V for DDR3-1600 just to make sure there is enough voltage to the ram since we know nothing of that ram and it specs are up to 1.65V.

HT Link speed of more than the "stock" 2600Mhz will do nothing for you unless you are running 3 video cards in SLi or CFx. That is what is said. I never ran that many cards so no idea if it will help even then.

As little as we know of that ram, I would set 10, 10, 10, 27, 43 and the increased ram voltage I mentioned and see if IBT does any better.

It might do you well to raise the CPU_NB not NB voltage from a default of ~1.175 to 1.25-ish. Just so the buss over which the ram talks to cpu has enough Voltage to be most stable.

CPU_NB increased Mhz can help performance but as you approach >2500Mhz most certainly you should increase CPU_NB voltage as said in item above.

I would correct FSB/HT Ref Freq (200) and ram voltage and timings to those suggested above first and test.

I would also make a signature so any that come in Late to the thread, have an idea of what is in the case. Make sure since it seems a Giga board that you put model AND the Revision from the lower left corner as the board is mounted in a normal case location. It is toward the front edge at lower left.

Put up Sig at OCF.

New Shortcut method for putting a Signature with your system information following your every post so people can know what is in the case that they are trying to assist with. You can use something like what is shown in my signature as a good template of needed information Thank you.

RGone...

 

[MyNameIsWill]

...maybe we can adjust to sustain IBT.

Well not sure why you are running Bus Speed of 208 when the stock is 200. That little bit is
also raising the Memory speed and the HT Link Speed and the CPU_NB speed by some amount.

I also see that the ram you have is that new AMD stuff we have zero experience with...good or bad.

It shows to be DDR3-1866 at 1.65V and when only running about DDR3-1600 it calls for 1.5V to ram.
I would raise the ram voltage to about 1.6V for DDR3-1600 just to make sure there is enough voltage to the ram since we know nothing of that ram and it specs are up to 1.65V.

HT Link speed of more than the "stock" 2600Mhz will do nothing for you unless you are running 3 video cards in SLi or CFx. That is what is said. I never ran that many cards so no idea if it will help even then.

As little as we know of that ram, I would set 10, 10, 10, 27, 43 and the increased ram voltage I mentioned and see if IBT does any better.

It might do you well to raise the CPU_NB not NB voltage from a default of ~1.175 to 1.25-ish. Just so the buss over which the ram talks to cpu has enough Voltage to be most stable.

CPU_NB increased Mhz can help performance but as you approach >2500Mhz most certainly you should increase CPU_NB voltage as said in item above.

I would correct FSB/HT Ref Freq (200) and ram voltage and timings to those suggested above first and test.

I would also make a signature so any that come in Late to the thread, have an idea of what is in the case. Make sure since it seems a Giga board that you put model AND the Revision from the lower left corner as the board is mounted in a normal case location. It is toward the front edge at lower left.

Put up Sig at OCF.

New Shortcut method for putting a Signature with your system information following your every post so people can know what is in the case that they are trying to assist with. You can use something like what is shown in my signature as a good template of needed information Thank you.

RGone...

You're the man! thanks it helped a lottttt, my ram is running stable @ 1886 now. Still fails at IBT but im passing Prime95 like usual. Programs launch quicker and such. Thanks a lot man =)

I'm using the 11-12-12-32-42 timings, do you think i could get the timings tighter without causing a failure?

 

[Johan45]

I have to ask , is that the P95 blend torture test you're running or one of the FFT tests. Just seems odd that you pass P95 but not the Intel burn test is all. As for the ram that can be tightened up but you need to be stable first.
I ask about the P95 since the blend test is usually harder to pass than most other tests, uses quite a bit of ram and I noticed in some of your earlier posts that the ram was at 10% usage while P95 was running which doesn't seem to be enough.

 

[Tyerker]

Using Extreme LLC is not going to help your OC. AMD has put all their eggs in one basket so to speak when they calibrated their sensors, what I mean is they have the sensors calibrated very well when under load but not so well for when the CPU is sitting idle ...... pay no attention to the temp while idle as they can be very off so to speak. The only temp you need to worry about is while under load.

I have found Extreme LLC has improved the stability of my OCs, particularly if you aren't trying to brute-force the voltage. Extreme LLC at 1.3V can actually be more stable than Normal LLC at 1.35V, because too large of a fluctuation in the voltage can cause memory errors, rounding errors, etc that will give you blue screens, or at the very least instable Prime95.

It may not be universal, but I have seen the difference, personally. I'm currently running at 4GHz below stock voltage with Extreme LLC. Even with increased voltages, I'm still hitting a wall at 4.3GHz I can't overcome with safe temperatures, so I think I'd rather run 4GHz undervolted than 4.4GHz at 75C load.

 

[bassnut]

Every system is different based on the equipment used in it. My personal experience as others how really push their gear have found going beyond ultra high not necessary and sometimes a hindrance. A while back there was a pretty good discussion on this topic.

 

[Mandrake4565]

Every system is different based on the equipment used in it. My personal experience as others how really push their gear have found going beyond ultra high not necessary and sometimes a hindrance. A while back there was a pretty good discussion on this topic.

I agree, I find setting to "extreme" overshoots the Cpu V by more then my liking, I'd rather just raise the Cpu V manually and and have the LLC on High.

P95 blend torture test

I have been wondering the same Johan.

 

[RGone]

Every system is different based on the equipment used in it. My personal experience as others how really push their gear have found going beyond ultra high not necessary and sometimes a hindrance. A while back there was a pretty good discussion on this topic.

Yeah, I hear you "bassnut". I got to thinking about it last night as I was making that last post and the reason I asked him to make a signature and put the model and Revision of Giga board he has. If he is using a UD3 of Revision 1.1 OR 4.0 then we might be able to finish helping him along. But man oh man if he has that Rev 3.0 and Giga has not made it easy to find that BETA bios "tong" was using...then all bets are sort of off.

I tell him to set 10, 10, 10, 27, 34 and he sets 11, 12, 12, X, X and then wants to know if he can tighten the timings. Oh man. He is not following along with what we specifically are asking and I suspect he is listening to more than one forum.

"johan45" is wondering if instead of P95 Blend; if he is using small FFTs and that uses n0 ram really and he might pass small FFTs and fail IBT. Who really knows since I do know he (OP) is not really reading what we post and doing just that. Other things are being added into the mix. So there is no way to really see the monitor screen at his desk from our locations.

Anyway you are correct in saying that using high LLC can hinder some clocks and even in the long-run damage a cpu. I mean the whole LLC is new for AMD and the FX series of processors. A first for AMD with FX where as with Intel it has been a spec for a number of years now.
RGone...

 

[Johan45]

Yeah, I hear you "bassnut". I got to thinking about it last night as I was making that last post and the reason I asked him to make a signature and put the model and Revision of Giga board he has. If he is using a UD3 of Revision 1.1 OR 4.0 then we might be able to finish helping him along. But man oh man if he has that Rev 3.0 and Giga has not made it easy to find that BETA bios "tong" was using...then all bets are sort of off.

RGone...

I went back and reviewed this stuff earlier and he said a rev 1.1 in one of the posts.

 

[RGone]

I went back and reviewed this stuff earlier and he said a rev 1.1 in one of the posts.

I "thought" I had read that same thing, but then I went back to call myself verifying the Rev 1.1 but I have not found it again. Maybe I am just freeken blind. Hehehe.
RGone...

 

[Tyerker]

I "thought" I had read that same thing, but then I went back to call myself verifying the Rev 1.1 but I have not found it again. Maybe I am just freeken blind. Hehehe.
RGone...

Post #6

 

[RGone]

Post #6

There it was, just waiting to be read again. Hehehe.

Thanks.
RGone...

 

[MyNameIsWill]

I have found Extreme LLC has improved the stability of my OCs, particularly if you aren't trying to brute-force the voltage. Extreme LLC at 1.3V can actually be more stable than Normal LLC at 1.35V, because too large of a fluctuation in the voltage can cause memory errors, rounding errors, etc that will give you blue screens, or at the very least instable Prime95.

It may not be universal, but I have seen the difference, personally. I'm currently running at 4GHz below stock voltage with Extreme LLC. Even with increased voltages, I'm still hitting a wall at 4.3GHz I can't overcome with safe temperatures, so I think I'd rather run 4GHz undervolted than 4.4GHz at 75C load.

This right here saved my life -_-. Oh my god, ive spent hours trying to stabilize my CPU in intel burn test and it finally happened. I had my LLC on Medium and when i saw that you said Extreme i tried it, VIOLA! I set it to Ultra High and it's stabilized aswell! i don't dare to go lower because i am happy with what i have now and i read Ultra High doesn't do nearly as much damage as Extreme. The crazy thing is i dropped down .2ghz but i also dropped my Vcore a TON, i can run at 1.23Vcore and run Prime95 and IBT AVX on High 10 passes! i raised it to 1.26V just incase it ever needs a little extra juice. Man i have learned a ton the past couple of days, thanks to everyone who helped me i truly appreciate it. I will post a 25minute pass of Prime95 once i run the test after i type this. I am so relieved.

PS: I was never ignoring or going against the suggestions you guys gave me, i tried them and if they failed i just never posted that it went wrong. Sorry on my part because it was out of frustration. Right now im at 1600 on 4x4 16GB Ram 9-9-9-27-35 Timings. THANK YOU ALL!

 

[RGone]

IF MNIW is happy, then we are slap tickled nearly to death. Luck man.
RGone...

 

[MyNameIsWill]

IF MNIW is happy, then we are slap tickled nearly to death. Luck man.
RGone...

Thank you sir
Here are Prime95 with HWmonitor since all of you love HWmonitor xD

and sorry im posting from a different website but the attachment window won't load to upload the picture =/... Just right click the image and select "open in a new window" and it will fullsize.

 

[MyNameIsWill]

IF MNIW is happy, then we are slap tickled nearly to death. Luck man.
RGone...

Thank you sir
Here are Prime95 with HWmonitor since all of you love HWmonitor xD

and sorry im posting from a different website but the attachment window won't load to upload the picture =/... Just right click the image and select "open in a new window" and it will fullsize.

 

[Johan45]

Nice work MNIW , I would just like to point out one reason we prefer HWmonitor and a bit about LLC. If you notice your first post using HWinfo it reads your V_Core as 1.264 which is what you had set in bios , correct? BUT with LLC set to Ultra it added .05v to your setting under load or the 1.312v you'll see when using HWMonitor to log volts. So your CPU is actually using 1.312v for stability not 1.264v. This is what we were discussing earlier about LLC and how it will overshoot you set voltage when used in the Ultra and extreme modes.

 

[RGone]

Nice work MNIW , I would just like to point out one reason we prefer HWmonitor and a bit about LLC. If you notice your first post using HWinfo it reads your V_Core as 1.264 which is what you had set in bios , correct? BUT with LLC set to Ultra it added .05v to your setting under load or the 1.312v you'll see when using HWMonitor to log volts. So your CPU is actually using 1.312v for stability not 1.264v. This is what we were discussing earlier about LLC and how it will overshoot you set voltage when used in the Ultra and extreme modes.

"johan45", I think one of the big differences is that Rev 1.1 of the Giga board. From what we know, the VRM/LLC ciruit is a little up and down from the reports back when that board was big time and before the newer Revs came out.

What you and I do now; as do some others on Asus upper tier boards, is use the Offset Vcore and limit the LLC and thus keep the overshoot to a minimum. Keeping the overshooting spikes from hitting the cpu is what AMD and Intel both wish to eliminate.

But he seems to have a system for his Giga board and maybe there is nothing for it but to run as it is.
RGone...