SP's AMD build!

So... The higher the NB the better regardless of RAM speed I have found.

Here's where I have stopped:


And here is the comparison between the NB frequencies:

I'd love to run 2420 NB, but I can't seem to get it fully stable.
any tips? hammer the NB volts? CPU NB volts?

Silver, I just had quite the discussion with RGone.
Would you be willing to try a new PSU? We both believe that it could be the source of your high vCore problems.

If so, see if you can snag an EVGA SuperNova G2 (750W or 850W would be plenty) and see if you can drop the core voltage.

Doesn't he have one on that machine? See sig...?

ATMINSIDE said:

Silver, I just had quite the discussion with RGone.
Would you be willing to try a new PSU? We both believe that it could be the source of your high vCore problems.

If so, see if you can snag an EVGA SuperNova G2 (750W or 850W would be plenty) and see if you can drop the core voltage.

Click to expand...

EarthDog said:

Doesn't he have one on that machine? See sig...?

Click to expand...

I do have the one in the sig running this rig.
EVGA Supernova 850 GS I have the review model from Oaklahoma Wolf


Curious though, what is the PSU you ran this setup with Austin?

SuperNova G2 850W is what I ran the day before I shipped it.
I had forgotten about you picking up that PSU from OW though.

ATMINSIDE said:

SuperNova G2 850W is what I ran the day before I shipped it.
I had forgotten about you picking up that PSU from OW though.

Click to expand...

Cool

The Supernova I have now ought to be enough for this chip and board right?
Quite a step up from that Coolermaster PSU I had.

Looking back at your subs with this chip, I agree my voltages look crazy....
I dunno. Perhaps the chip just needs more Vcore to be prime stable...


I'm sitting at 4817Mhz @ 1.584Vcore Small FFT stable, 1 WHEA error....
Large FFT was a mess of WHEA errors...
Blend was fine until all the cores failed with rounding errors after 4 hours.

I'm running the RAM at stock frequency and I lowered the NB to 2115 Mhz to eliminate those as culprits.

Personally, I see nothing wrong. It's an average chip, that's all. Pounding it with voltage isn't gonna help. It's gonna be happiest in the 4-6-4.7 range, just like millions of others. You're chasing ghosts.

Mr.Scott said:

Personally, I see nothing wrong. It's an average chip, that's all. Pounding it with voltage isn't gonna help. It's gonna be happiest in the 4-6-4.7 range, just like millions of others. You're chasing ghosts.

Click to expand...

Seems like it.
Came back to WHEA errors while running large fft with stock RAM and lower than stock NB

Damn. I was hoping for 4.8 daily at least.
If only I knew what voltages Austin was running this chip at for 4.8 daily..

Kyle you should keep in mind those subs have nothing in common with prime stability. Those are just benchmark scores and only stable for that benchmark. For benching clock it as high as you can, volt it as much as you need and cool it as much as you're able. For 24/7 find a sane stable OC that keeps the volts and temps low for a longer lasting chip.

Blaylock said:

Kyle you should keep in mind those subs have nothing in common with prime stability. Those are just benchmark scores and only stable for that benchmark. For benching clock it as high as you can, volt it as much as you need and cool it as much as you're able. For 24/7 find a sane stable OC that keeps the volts and temps low for a longer lasting chip.

Click to expand...

I ran it for over a year at 4.8GHz. It was nowhere near 1.6V.
But, unfortunately, I couldn't save those settings and I long-since don't have them written down any more.

I'm perplexed that he can't run higher, honestly.
There's no way Prime knocks you down 300MHz from Cinebench with a slight increase in vCore.

Can you post some bios pics for us please? I know these chips aren't 83x0s bjt it seems like it's not right somehow

any tips? hammer the NB volts? CPU NB volts?

Click to expand...

While overclocking the actual NB or chipset, the CPU/NB or memory controller setting does nothing for higher clocks.

You need just the NB or chipset voltage increase. Default is 1.1v Was the same on AM2+ Asus boards was called chipset voltage. Later changed to another NB name instead of being left as it was to avoid confusion.

The CPU/NB does nothing for the chipset directly speaking in terms of overclocking the actual NB chipset. You should test this overclock as a stand alone while implementing and keeping a lower stable Cpu clock.

However, it is rare to see over 2500mhz for a daily stable clock as perhaps the 9 series the chipset works HT differently than the 7 series. The 7 series chipsets for example, wouldn't allow Cpu HT link to be above the actual NB speed. The board simply would not post. Never the less, 3ghz NB speeds where obtainable in most cases with the chipset over-volt.

I use a simple guideline with chipsets 7-9 series. 1.2v max not actively cooled (without fan) and up to 1.3v (+) (with fan +waterblock+ monitor temps closely.) Then push the NB/Chipset a little.

Also, Prime95 isn't a tell all stability test. Try a bunch of others. I mean heat is only what you create of it. F@H may run a full load with a lower temperature for example.

Blaylock said:

Kyle you should keep in mind those subs have nothing in common with prime stability. Those are just benchmark scores and only stable for that benchmark. For benching clock it as high as you can, volt it as much as you need and cool it as much as you're able. For 24/7 find a sane stable OC that keeps the volts and temps low for a longer lasting chip.

Click to expand...

Indeed there is nothing in common, but like Austin and bob said, it just doesn't seem right. There has to be something wrong that I need much more Vcore than to run a bench, especially cinebench. Perhaps the power from the wall is dirty? Can't be though, I'm running this, my lamp and my monitor off a surge suppressor... Unless the suppressor itself is causing my grief?

ATMINSIDE said:

I ran it for over a year at 4.8GHz. It was nowhere near 1.6V.
But, unfortunately, I couldn't save those settings and I long-since don't have them written down any more.

I'm perplexed that he can't run higher, honestly.
There's no way Prime knocks you down 300MHz from Cinebench with a slight increase in vCore.

Click to expand...

Damn wtf.

Either I'm feeding the board dirty power or I screwed something up really bad. I'm at 1.58Vcore to get 4.8Ghz Prime small fft stable.

bob4933 said:

Can you post some bios pics for us please? I know these chips aren't 83x0s bjt it seems like it's not right somehow

Click to expand...

I agree here, I'll get some pics asap of what I'm currently running. Which is stable and WHEA free btw.

@Shrimpy
Okay up to 1.3v.... I set the NB to 1.2 and it has been there ever since IIRC, you said that you leave it at 1.2 anyway. I have a nice Corsair AF120 cooling the NB and VRM's. Both are barely warm to the touch.

And here are my BIOS settings.
I dropped the multiplier by .5 last night and upped CPU NB v to 1.4. I also upped the VDDA back to 2.8. Stable without WHEA errors, though worker #2 dropped out after 2 hours of Large FFT. The remaining cores #1 and 3 continued on until I stopped them 6 hours after.



I can try dropping the Vcore. I'll run IBT instead this time. I think it's better than prime for CPU only stressing.

How about this, let's try from scratch again.

All cores, stock speeds and voltages with the exception of the CPU itself.
Try 24x200 and see how much voltage it takes to stay Prime stable

ATMINSIDE said:

How about this, let's try from scratch again.

All cores, stock speeds and voltages with the exception of the CPU itself.
Try 24x200 and see how much voltage it takes to stay Prime stable

Click to expand...

Sure. I'm stumped on this as well.

Thats a hefty fsb dude. Definitely agree on trying to roll with the multiplier first. You might be pushing the ram and NB too hard requiring a LOT of voltage to get it stable.

bob4933 said:

Thats a hefty fsb dude. Definitely agree on trying to roll with the multiplier first. You might be pushing the ram and NB too hard requiring a LOT of voltage to get it stable.

Click to expand...

Really?

Here I was thinking it's wayyy to low.
That's a real possibility.

We really just need to hit this from the basics. Go step by step like we would do for anyone else that's having trouble.

ATMINSIDE said:

We really just need to hit this from the basics. Go step by step like we would do for anyone else that's having trouble.

Click to expand...

Yes.

Initial testing is... disappointing. I'm back up to 1.584Vcore and I'm finally not dropping workers within the first 2 minutes of prime blend but I'm getting lots of WHEA errors in the event log....
I found my notes of this chip too. 1.584 Vcore is in line with what I found when I first started OC'ing this chip.