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

Just Putzen Round

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Wish I had some of Johan's Dice
DICE is so 3 months ago Blaylock, he's gone colder. :D

That's a pretty hefty OC for that voltage Blaylock nice! Whatcha coolin it with?


I'm a little jealous of these guys dice and ln2 setups. Not gunna lie
There's plenty of people that will help you if you want to go down that road Ray
 
That's a pretty hefty OC for that voltage Blaylock nice! Whatcha coolin it with?

I'm still just ambient, using my Zalman CNPS95000LED. I need a few things still before I can go liquid.

I think this Sempron is a pretty nice chip. Everytime I expect it to BS it just keeps goin. Eventually I'll need to bump the volts, but for now it seems to be doing fine. Probably the most surprising is the CPU-NB volts. I think I'm close to 3000Mhz and it's still at the stock 1.175v.

Oh, and I still need to try unlocking the second core still. I'm going to see what I can get on the solo core first though. :)

20141015_164643.jpg
 
There's plenty of people that will help you if you want to go down that road Ray

I'm definitely interested but no clue where to start. Haven't done a whole lot of research yet though. Just watched some YouTube videos just to see ln2 and dice in action. Since it's starting to get colder I'm about to start doing some more to see what all I can do.
 
Last night I gave it another go with the LN2. I had thought/hoped that I just had a bad mount on my pot the last time but that doesn't seem to be the case at all. I don't know exactly what it is maybe a weak core but once I was over 6.0 I had a hell of a time getting it to work on all 8 cores. Light benches were OK but even WPrime I could run 32M at 6.4 but 1024 would crash. SuperPi though I ran 1M and 32M at 6.9 without issue. It seemed as long as I kept the voltage at 1.7v or less It would keep running.As soon as I went over 1.75v it would idle for maybe 20-30 seconds then just lock up even at lower speeds. So the only thing now is to try a different CPU next time I guess. I wish I still had my piggy. Here's a couple of shots from last night.

api 11m 06.043.JPG

5.349.JPG

I also managed to run HWBot Prime at 6651 but that was the limit for eight core work and it's a very light bench http://hwbot.org/submission/2655098_
 
Thanks Ray, I'm just stumped on speed and work. Couldn't make it do anything difficult at higher speed. It was fun but a bit disappointing at the same time.
 
No worries Ray, I'm sure some of it will come with experience, I'd just like to be able to figure out if it's the CPU or me. I'm out of LN2 for now so I can't test anything else ATM. But I will in a few weeks.
 
basically any bus speeds like 2400ish ht and 2000ish nb, even memory around 1700ish, cannot clock over around 5.05ghz...... temps do not appear to be the issue, could be though...... tried many different timing configurations.... best was around 12 minutes of prime 95 around 5.1 ghz.... I did get the cpu/nb down to 1.26 ghz using the offset and lowering "load line calibration" to high for cpu and regular for cpu/nb...

musta had 12 hours in it attempting.....

:screwy:
 
Where do you guys get the ln2 for this testing? Doesn't seem like something that would be easy to get
Praxair in London, the thing is you have to have a dewar to contain it. I just got one two weekends ago so had to try it out. Got 30L for about $50. There's only some stipulations around transportation. Like it has to be outside the vehicle. Not in the cabin or it'll kill you if you have an accident and it gets out of it's container.

basically any bus speeds like 2400ish ht and 2000ish nb, even memory around 1700ish, cannot clock over around 5.05ghz...... temps do not appear to be the issue, could be though...... tried many different timing configurations.... best was around 12 minutes of prime 95 around 5.1 ghz.... I did get the cpu/nb down to 1.26 ghz using the offset and lowering "load line calibration" to high for cpu and regular for cpu/nb...

musta had 12 hours in it attempting.....

:screwy:

When you have faster NB and Memory the CPU naturally works harder so as the speed goes up you'll just have to pump voltage to it.
 
I'm still just ambient, using my Zalman CNPS95000LED. I need a few things still before I can go liquid.

I think this Sempron is a pretty nice chip. Everytime I expect it to BS it just keeps goin. Eventually I'll need to bump the volts, but for now it seems to be doing fine. Probably the most surprising is the CPU-NB volts. I think I'm close to 3000Mhz and it's still at the stock 1.175v.

Oh, and I still need to try unlocking the second core still. I'm going to see what I can get on the solo core first though.
Sure seems like a goodun Blaylock :thup: Fingers crossed it will unlock!

I wish I still had my piggy.
I was thinking the same, that 8350 was a little btch on ambient cooling but really shined when it went sub ambient. Shame it had an early termination, may the lil piggy R.I.P.!

There's only some stipulations around transportation. Like it has to be outside the vehicle. Not in the cabin or it'll kill you if you have an accident and it gets out of it's container.
Never thought of the implications of transporting it Johan, I can see that being a problem if it's inside and you crash. At least your body will be preserved! :D


basically any bus speeds like 2400ish ht and 2000ish nb, even memory around 1700ish, cannot clock over around 5.05ghz...... temps do not appear to be the issue, could be though...... tried many different timing configurations.... best was around 12 minutes of prime 95 around 5.1 ghz.... I did get the cpu/nb down to 1.26 ghz using the offset and lowering "load line calibration" to high for cpu and regular for cpu/nb...
As Johan said the harder you push the more voltage you will need for the next XXX Mhz . Show us a Hwmonitor SS under load and the Spd Memory and Cpu tabs again, maybe well see something your missing. I can tell you on my chip when I started pushing for 5.1 "stable" I had to back the voltage dump truck up and just start pouring it in. I do not remember exactly how much I needed because I do not try to run it above 4.7 for anything other then benching and I'm only concerned with being stable for that bench when doing so. It was was a lot though.
 
I don't like to run more than 1.45v for my daily driver. For benching.......whatever it takes. (P.S. I don't bench my daily rig either, well almost never.)
 
I don't like to run more than 1.45v for my daily driver. For benching.......whatever it takes. (P.S. I don't bench my daily rig either, well almost never.)

It is hard to resist that urge sometimes. I feel that one Blaylock.
 
@Johan: Wow 7G's. Nice going.

You PM'd for input, so I wanted to reply here... Maybe not any help to others, but cant hurt. I'm a bit rusty, but last I checked I still hold the world record for highest frequency Superpi32M run on any CPU ever on my FX-8120... 7.74Ghz. That submission isn't worth anything, but still one of my favorites.

I guess I should start with facts, so I've loved a number of AMD chips to death. Keep this in mind with any advice I give. Maybe I petted them too hard. I don't think it was the voltage.

One thing I often did on Intel was compare other submissions, and see if anything of mine looked out of whack. This is harder on AMD, as there are fewer good examples of people really running it well, but still it can be helpful.

Here are all my 8120 submissions for reference:
http://url.hwbot.org/102PREf

And my 6120 subs:
http://url.hwbot.org/10LLL3K

There isn't anything immediately jumping out in the screens I looked at. Except CPUNB freq. I don't recall what I always ran, but sub 3Ghz looks low for those core freqs. For example, I dropped from 6.5s to 5.9s in wprime32m on the FX-6100 with more vcore and ~500mhz higher cpunb. I forget if anything else was different, but I attribute that improvement to better stability from those two changes. Mostly voltage I think made it run better, as typically more cpunb only helps a lot with superpi32M. You'll notice my 32M sub's were pushing around 3.7cpunb I think. Other could get up above 4, but this depends on various factors like ram, chips, etc.

One trick, that never made sense to me, but worked consistently enough that I have to say its a real thing... More CPUNB volts. I hit hard walls a number of times, where it seemed that's all the chip could offer. Then add CPUNB volts, even if I wasn't increasing or pushing CPUNB freq, and out comes more core frequency. Go figure. I blame Dolk.

How much CPUNB volts you giving?

What are your digi VRM settings? All safeties/over volt should be off. I used extreme for all the rest. I think.

Did I mention I've killed my share of AMD chips?

One thing I would do, assuming you aren't too LN2 limited, is test each core individually at load. I would do this by enabling 2 cores at a time, keeping one core low - 5Gs. Then taking the other core as high as you can go and run spi. See if you have any cores that can't keep up with the rest. Usually all mine were within 100mhz of the rest. Takes a little time to do this, but you might learn about the chip in the process also. If you have one weak core, just keep it a little lower and let the rest go high.

Wprime clocks should be around 300-500mhz of spi, for anything phenom or after. That held true for all chips I pushed to their limits. So if you can get spi higher by focusing on single core, wprime should follow.

Turn down for what? Superpi... Only one core should be clocked up, the rest turned off or around 5Gs.

If none of that helps, try 1.9vcore and 1.7vcpunb. This is approaching insta-death territory. Ive gone there and higher, but some chips stopped talking to me for good afterwards. It wasn't anything personal but I guess they felt differently.

Most chips I killed just gave up at some point, nothing that aggressive really, perhaps just their time. 2 chips turned up dead after running voltages they had no business running at as long as they did. Mainly I say this because you shouldn't kill anything with any of my advice, I just wouldn't be that surprised if you did. :)
 
Last edited:
Back