New 8320 Fails IBT at Stock

[Tyerker]

It all depends on the revision.


When? Last screen shot you posted it was still dropping workers at stock clocks?

It still drops one worker after 5-6 tests in Prime95, but the other 7 run with no issue for 10+ runs. Games have been 100% stable, and the system has been running fine, even if one of the cores can't technically take 100% load long-term.

 

[RGone]

It still drops one worker after 5-6 tests in Prime95, but the other 7 run with no issue for 10+ runs. Games have been 100% stable, and the system has been running fine, even if one of the cores can't technically take 100% load long-term.

To me and my rigs that would not be stable at that speed. However the parts are yours. If you deem yours stable...then it is for you. However know that in this AMD Cpu Forum section we work with most thread starters to get 2 hours of full-on P95 Blend, all cores not failing and then call it stable. YMMV.

 

[Tyerker]

Then I guess I should turn mine down to 2.4GHz and see if it's even stable there.

 

[RGone]

Then I guess I should turn mine down to 2.4GHz and see if it's even stable there.

If it is not AMD would be getting an email or phone call if in Usa.
Gone...Again.

 

[Tyerker]

So are you saying it could be the CPU, and not the motherboard?

 

[RGone]

So are you saying it could be the CPU, and not the motherboard?

You said >> Then I guess I should turn mine down to 2.4GHz and see if it's even stable there.

I said at that point that if the dang thing (system) was not stable when running that slow, I would be getting in touch with AMD.

Now what did I mean when I said that? I would not turn my cpu down that far, but I certainly would be testing P95 Blend mode at all 'defaults' as if from a clearing of the CMOS and loading of optimized or Defaults or whatever the Giga bios calls "stock".

Now at those stock settings some of the UD3's had problems with Vishera 8 cores early on. Users adjusted Vcore manually while waiting on slow Giga to change bios. You have updated bios. One would think the issue would be gone of undervolting FX 8 core with default Vcore? But maybe not.

Workers generally fail on a core because of low voltage. Most of the time. Your cpu maybe that crappy. Your CPU_NB could have too low of a voltage. Your ram might need a little extra voltage, like from 1.5V to 1.55V as that has helped when the IMC in the cpu was marginal and a worker died indicating A core problem.

But in the shortest sense of the word, I would test that cpu at stock all around and see what gives.
RGone...

 

[Tyerker]

When I did the initial test, the only options I changed were I turned of Turbo (Performance Boost), C1E, C6, all of that stuff. I guess I can try optimized defaults and see.

But the fact remains outside of Prime95 failing one core, I have seen no issue up to 4GHz. I suppose this is part of the "fun" of overclocking, when something doesn't behave the way that is expected.

 

[Mandrake4565]

But the fact remains outside of Prime95 failing one core, I have seen no issue up to 4GHz. I suppose this is part of the "fun" of overclocking, when something doesn't behave the way that is expected.

It's your rig Tyerker, you do with it what you want. If it were mine, I would be diligently trying to figure out why I cannot run stock clocks and pass prime blend without workers failing. If you don't mind the occasional BSOD, clock that bad boy up and hope for the best, again it's yours. With me, I know if I cannot pass prime blend 2 hours without a worker going to a liquid lunch earlier then the rest, it will crash on me usually at the most in opportune time. Usually it's mid game when I have a good round of BF4/BF3 or BFH going or I'm editing a photo that I haven't saved yet. Your call.

 

[RGone]

I am going to try one more time for a more clear explanation. An explanation that I can write because I WENT thru it myself.

Logged every last Vcore requirement for my FX-8350 that was needed at 4.1Ghz thru 5.4Ghz and using my pretty d*mn good water cooling. Even tested for P95 Blend stable at 5.2Ghz for 2 plus hours.

Then I got a 'hair'. A 'hair' to test the $90.00 odd dollars of air-cooling sitting below my desk for over 7 months. In retrospect it would have been much "easier" just to go to the barber and get a hair cut. By the way you can read about my air-cooling journey in the last 3 or 4 pages of the "does FX-8350 flatline after 4.3Ghz", thread.

Saw all the testing of my air cooling product choice AND knowing how hot FX 8 cores really run, I never bothered to install the Aegir with only a push fan. It got two 98.x CFM fans in push-pull from the day I put it over my FX-8350.

Hey got this Vcore log and nothing to this stuff of overclocking my FX on air. Wrong. Began to move up the Mhz scale. 4.1Ghz. 4.3Ghz and then at 4.5Ghz, I was hollering holy bat shettman, something has gone terribly wrong. Freeken core 8 is running for the side-lines. Retiring from working. Houston we have a problem.

I got these tested Vcore values. WTFO? Time to think not flail. Ah hah. My temps are higher with only an air cooler no matter where it is ranked among the other air coolers. Well just crap in my pants, I know what do about that little problem. Overclocking 101. Add Vcore. Did so and 4.5Ghz had no dropped workers. Good deal? Well sort of. The temps went ^UP^ some more. SpiderMan where is my water cooling rig? Hehehe.

So now I am getting some suspicions because of the clues I am reading in HWMonitor and the reactions of P95 etc. A cpu that most thought of as better than theirs. One that when we get around to posting up are P-State logs for our FX-8350's was not printing out nearly as well as was thought, however it did run pretty good. AND I think I have the answer. I need about 0.03 more Vcore for a given speed to run my stuff on air over what I was using for Vcore running my water rig. Nothing to it now...yeah right!

So I do my video editting at 4.8Ghz. Let me determine if my second ranked air cooler with the ranking done in single fan mode can now with dual fans get me 4.8Ghz stable for two hours of P95 Blend without workers failing?

Now get my thinking so far during all of this? Shett shett my cpu is not any good. On the surface that is accurate. It is not good enough to run X speed without X Vcore added and then that X HEAT monster of FX 8 cores, comes in and bites me square on my left butt cheek. Workers move to sideline. Add Vcore. Add Vcore add heat. Put that in loop mode and now we are into real overclocking as all of us have faced it.

Does FX-8350 flatline after 4.3Ghz shows the results of my 4.8Ghz run for 2 hours of P95 Blend stabillity. Got it too. But not without seeing temps greater than I wanted day in and day out at tht 4.8Ghz. I was going to have to do one of a few possible choices.
1. I could lower my Mhz expectations about stability. (I had to)

2. I could go back to my water rig. (Not yet)

3. I could look for a better cpu. (Good luck with that!)

What did I do? Well sort of #1 above. Over the course of 3 weeks or so, slightly adjusted this. I turned the knob on that. I tested this. I barely bumped that up. How do you spell HOURS? Oh that is it with all capital letters. HOURS of tweaking later; as I gues they call it, I am now stable on air at 4.7Ghz and the dang thing is faster at my video editting at 4.7Ghz than it ever was at 4.8Ghz. Holy bat shettman, that is awesome.

Well okay, maybe it is awesome or maybe just good. Most think that stuff just happens and it all gets lined up by magic. BS!. HOURS of adjusting later and a slightly lesser Mhz desire and I am better off.

What does that mean for you and your FX-8320? The cpu is weak and needs Vcore. More Vcore more cooling. Oh shett right back in that overclocking loop.

Then there is one other thought and it was said by one who built rigs for customers with a number of UD3's and that was this >> In all likelihood the UD3 may clock well but it iwll take 50 "more" tries to get there than with a top of the line and more expensive Asus board. That statement was made in reference to the majority of Rev 1.1 UD3's. Since the move to Rev 3.0 UD3, it has generally been a snorkfookle to get real HiPo results from the UD3 boards. But hey that is life. Win some. Lose some. I know what I don't want. Others may love what I would not have. Strokes for folks.
RGone...ster .

 

[Tyerker]

So I went into the BIOS and restored Optimized Defaults. This includes Turbo, C1E, C6, all of the little power saving things. Here's Prime95 after 20 minutes.



VCore fluctuates down below 1.3V at times, and the CPU will also clock down to 3GHz. So it apparently is nothing with the board or CPU, and nothing with the defaults. For whatever reason, one of those power saving / "performance enhancing" options is actually making the CPU function BETTER than manually configuring everything to the same settings.

I think from here I will turn off Turbo first, just to see if that is the issue.

 

[RGone]

Maybe your best overall resource about that board...

...about that board since few have been seen in this AMD CPU Forum Section.

[Official] Gigabyte GA-990FXA-Series Owners Thread/Club
LINK>>http://www.overclock.net/t/1023100/official-gigabyte-ga-990fxa-series-owners-thread-club

Link above is the #1 page of what seems now some 806 pages of that thread. The first page has some introductory information and links that seemed current. However that thread starts BEFORE the boards are even available in retail. So I am going to look for the page where the fun begins from the onset of users with the boards.

Here you go. Page #24 and the boards are being bought and the users are now beginning to complain.
post #232 of 8055
Getting more Vdrop then I had with my M4N98TD EVO.

M4N only needed 1.45v in bios to maintain 4ghz with a 2.6v VDD

UD5 needs 1.5v in bios to maintain 4ghz, but has no CPU VDD setting to change.

LINK>>http://www.overclock.net/t/1023100/official-gigabyte-ga-990fxa-series-owners-thread-club/230

I expect that bye the last page 806 of that thread that they will have gone thru the revisions of the boards and the bios changes, so they can have CPU_LLC on the boards and DISABLE APM and the like.

A lot to read but back when we were flooded in here with UD3 users, that thread was where most of the 'talk' about the boards were going on. Gradually the Giga users have tapered off and the entry level Asus to the much better Asus high level boards have taken over until just a few weeks ago when some of the cheaper other than Asus brand began to come again.

I think that thread just about covers it all from the lows at release up thru some pretty fair stuff with the Rev 1.1 of the board and onto some more lows with the Rev 3.0 and maybe now the Rev 4.0 boards.

It was far easier for me to find this long-running thread there than to try and piece together the many multiple threads where we tried to help UD3 owners here in the AMD CPU Forum section. Because of the beast the information here would be in many threads not strung together as in that long thread over at OCN.

Of course that does not preclude you using the forum search "here" by board and finding any of however many threads we had here. Just know that most of the users we had coming in here had that really cheap 970A-UD3 and few spent to go with the FXA-UD3.

Best that I can see for you, that as the boards and revisions came out...the users had to do the dance for stability.
RGone...ster.

 

[Mandrake4565]

VCore fluctuates down below 1.3V at times, and the CPU will also clock down to 3GHz.

It's most likely dropping down to 3xxx mhz because something is making it throttle.

 

[Tyerker]

It's most likely dropping down to 3xxx mhz because something is making it throttle.

As I said, I left C1E, C6, Cool and Quiet, and AMD APM Master Mode enabled. I was able to stress test successfully at 3.5GHz with all default voltages.

Turned off all of those features, and I can still stress test successfully with all voltages set to Auto. So it must have been something with manually setting voltages, I guess.

 

[Tyerker]

Welp, now here we are at 4.0GHz 30 minutes Prime95 stable without touching any of the voltages at all.



So, I guess it was either a fluke, or for whatever reason it just really hates manual voltages...

 

[Mr.Scott]

Excuse me in case I missed something but,
HT link speed and NB frequency are both at a different setting when you were having trouble, instead of where they are now.

Voltage may not be the problem.

 

[Tyerker]

Right, I am thinking that may be the case. But I thought they were supposed to be the same clock, I guess. I thought you wanted 2500 on each, so I figured 2400 on each was good, but apparently not.

Let's see where this thing can get with stock volts now. Hoping for 4.4GHz on stock volts if possible. But even if I need to increase voltages a bit for 4.5GHz that is my real goal.

 

[Mandrake4565]

Post #34 looks promising! G/L

 

[trents]

Right, I am thinking that may be the case. But I thought they were supposed to be the same clock, I guess. I thought you wanted 2500 on each, so I figured 2400 on each was good, but apparently not.

Let's see where this thing can get with stock volts now. Hoping for 4.4GHz on stock volts if possible. But even if I need to increase voltages a bit for 4.5GHz that is my real goal.

That was one of the big differences moving up to FX from Thuban; the HT Link no longer must be equal to or lower than the CPUNB and in stock condtion the HT Link is actually higher than the CPUNB.

 

[Tyerker]

So I finally found the limit of stock voltage. I could boot to 4.3GHz, but hard reboot when running Prime95. So, backed it down to 4.2GHz and ran the Heaven benchmark completely maxed out.



I know a lot of the limitations here are the video card.

 

[Mandrake4565]

I know a lot of the limitations here are the video card.

Yeah heaven isn't much of a bench to test out the capabilities of the Cpu. Better off running Cinebench 11.5 or 15 Cpu test.