FEATURED Marathon SIII: December SuperPi 32M

[speedking]

Very nicely done speedking, that's a great cache clock!

Thanks.

The PC runs through some benchmark with a 4.7GHz cache (SuperPi, 3Dmark series...), buts XTU/Cinebench have to [email protected]. Unless I put through it too much volt at my taste!

 

[ATMINSIDE]

Thanks.

The PC runs through some benchmark with a 4.7GHz cache (SuperPi, 3Dmark series...), buts XTU/Cinebench have to [email protected]. Unless I put through it too much volt at my taste!

Lucky you, my 4770K clocks the core decently, but the cache is garbage. I can barely boot at 44X cache.

 

[Woomack]

Usually cache has to be 200-300MHz below CPU clock and to keep it stable at x48+ ratio it needs higher voltages ( I mean much higher ).

 

[ATMINSIDE]

Usually cache has to be 200-300MHz below CPU clock and to keep it stable at x48+ ratio it needs higher voltages ( I mean much higher ).

That would put me in the 47x-49x range for normal benching, it won't even think about booting at that.

 

[Woomack]

you probably have to bump ring voltage to make it run higher but in most tests it won't make any difference anyway

here I had 5200MHz cache and it was only on ss cooling ( CPU voltage isn't correct ):

 

[ATMINSIDE]

you probably have to bump ring voltage to make it run higher but in most tests it won't make any difference anyway

here I had 5200MHz cache and it was only on ss cooling ( CPU voltage isn't correct ):

Spoiler

View attachment 157212

I have to bump voltage for 43x-44x to boot/run.

 

[Mandrake4565]

CPU voltage isn't correct

LOL I would think if it was Woomack even under SS 2.2 would kill it rather quickly?!?!

 

[Woomack]

Depends from CPU, some haswells don't like anything above 1.75V and some can randomly die even on cold. My best 4790K was running up to 1.85V on dice but died at 1.35V when I was testing it later on water and lower clocks.

 

[Johan45]

I have to bump voltage for 43x-44x to boot/run.

For me I run 1.25v for 45-4600 and 1.32ish volts for 4800 depending how hard the bench is sometimes up to 1.35

 

[ATMINSIDE]

For me I run 1.25v for 45-4600 and 1.32ish volts for 4800 depending how hard the bench is sometimes up to 1.35

Yeah, I'm looking harder and harder at a 4790K after Christmas time passes now...
I can't even get close to that.

 

[EarthDog]

Usually cache has to be 200-300MHz below CPU clock and to keep it stable at x48+ ratio it needs higher voltages ( I mean much higher ).

Seeing how cache makes little to no difference in most benchies, I rarely touch it. That said, even at 5Ghz+ I left it at stock 3300Mhz with no problems. I believe that is a myth. Or if it isn't, I have never run into that issue for some reason. Not in Haswell, not in HW-e either.

 

[devlos]

@devlos -> thats some sick voltages for 5.0 ghz man! Thought mine was good

Thanks, I thought it was decent. I'm new to amd so I don't really know whats good. I also don't know if disabling cores will let you have lower volts. With all cores active it only needs 1.41-1.42v

On the subject of cache voltage I ran my 4690k cache at [email protected] on accident, I meant to hit 1.19v but hit 1.9 I ran xtu a few time with it like that anyway and ended up with my best score.

 

[Johan45]

Thanks, I thought it was decent. I'm new to amd so I don't really know whats good. I also don't know if disabling cores will let you have lower volts. With all cores active it only needs 1.41-1.42v

On the subject of cache voltage I ran my 4690k cache at [email protected] on accident, I meant to hit 1.19v but hit 1.9 I ran xtu a few time with it like that anyway and ended up with my best score.

Just curious what you tested the 5.0 OC with? That's quite low indeed for V_Core. Not that I think it's impossible since a have a CPU that'll do about the same.

 

[devlos]

Cinebenchr15 and 11.5 and a few other benchmarks plus general use and games. I didn't run prime95 on it because it I wasn't testing for 24/7.




and it takes around 1.44v to do anything @5.1ghz all cores.

I'm really not sure about the voltage, I just tried to keep it kind of lowish.

 

[Johan45]

Ahh, OK now we're on the same page. I did a run on cinebench with 1.45v at 5.4

Are you going to stress test it I'm curious as to what it will do. Here's my latest FX http://www.overclockers.com/forums/...Putzen-Round?p=7749764&viewfull=1#post7749764.
If your looking for tips and tricks for that CPU the thread I directed you to is full of them. That's where us "regulars" in the AMD section "hang out"

 

[devlos]

I've been meaning to check that thread out, I could use some tips on getting these AMDs clocked up.

I prob won't stress it for stability other than with benches because it's not going to be used for anything other than benches and overclocking and Rosetta@H

 

[Woomack]

Seeing how cache makes little to no difference in most benchies, I rarely touch it. That said, even at 5Ghz+ I left it at stock 3300Mhz with no problems. I believe that is a myth. Or if it isn't, I have never run into that issue for some reason. Not in Haswell, not in HW-e either.

HW acts totally different than HW-E. For example you can't run CPU=cache clock on HW while on HW-E cache clock can be even higher than CPU clock and it will still run stable. Also cache clock on HW is barely helping while on HW-E it makes huge difference in memory bandwidth ( still CPU clock counts more as memory bandwidth is high even on stock ).

 

[EarthDog]

The point was that I rarely touch cache and on both platforms I left it at stock and was well past the 200-300 mhz 'rule' (and not sure why because so many people state its a rule).

 

[Johan45]

I've been meaning to check that thread out, I could use some tips on getting these AMDs clocked up.

I prob won't stress it for stability other than with benches because it's not going to be used for anything other than benches and overclocking and Rosetta@H

Come on over Devlos, we're a pretty happy bunch of guys and in that thread we have a lot of fun compare ideas and laugh at each other. One thing with the FX after core speed, NB speed it the next important thing when benching.

 

[Woomack]

Haswells on 1150 are losing stability when cache clock is close to CPU clock to the point it won't boot at all but there is no problem to run it at higher clock difference like 800-1000MHz. How close can it be also depends from CPU so it's maybe not a rule but in most better guides all are saying it's about 300MHz. Once someone say it in the larger article and all others think it's a rule
On HW-E there are no limits like that but on most boards you won't set more than 3500MHz cache anyway.