Overclock.net banner

1 - 18 of 18 Posts

·
Registered
Joined
·
1,552 Posts
Discussion Starter #1
hello guys, i need your help to try to reach 4ghz with my cpu, i just upgrade my mobo and with this one i was able to reach 3.8gz (19x200 / 1.55v) stable:


This test last 2 hours..
btw, don´t worry about the temps.. look at my load temps during prime


i could able to boot into windows at 3,9ghz (19,5x200 / 1.55v)but prime fails very afters seconds, and i was even able to boots into windows at 4ghz (20x200 / 1.6v) but bsod after seconds also.

i did not touch the NB V, or any other voltage, only the Ram volatge (from 1.8v to 2v to reach 400mhz x2= 800mhz ddr2 / 5-6-6-18)

in amd overdrive i see so many voltage options i don´t understand, and those options are not present in bios..


my cpu has the unlocking multiplier, so i preffer to let the fsb at 200 and only you the mult, with my old mobo i was able to set my rams at low latence (4-4-4-12) and the NB speed at 2600mhz (13x) but with this one no, i let my NB at 2000mhz (10x) if i increase this value make my 3.8ghz oc unstable


here all the posible options in bios, maybe you could help, i did not try overclock by fsb btw


 

·
Registered
Joined
·
955 Posts
Well, the databases show that the 720 rarely exceeds 3.9Ghz unless there's exotic cooling, though at least one has reached 4.0 at 1.4+ volts, it appears to be the exception, if they're honest about it (might have been boot only, not stable).

Most have required north of 1.5v, to a high of 1.62v, some on phase change.

There's some evidence that voltage bumps on NB or NB/CPU may help a little, which might get you some 50 or 100Mhz, but it's subtle and not always what works.

...there are some that will and some that won't...

Sometimes a different motherboard can squeeze out more from a CPU, and it's not always clear what does it, but it seems the design of the VRM and possibly the quality of the output from the PSU have some effect.

In some cases 2 RAM chips are easier than 4, which hints that possibly some RAM will be smoother than others for an overclock.

Typically, an overclock of 35% is about as good as it gets, 40% is rare on AMD chips. Its difficult to justify a disappointment, though - there are limits to physics for any particular fabrication design, and for a while at the 65 and 45nm nodes, AMD chips have really liked the cold.
 

·
AMD Overclocker
Joined
·
9,178 Posts
Quote:


Originally Posted by Yukss
View Post

well, so far i have achieve 1ghz overclock which is nice, should i stop here ?

It's a really nice oc. The best I can get on my 720 is 3.6ghz @ 1.5v with the 4th core unlocked. But I am on air
 

·
Registered
Joined
·
1,552 Posts
Discussion Starter #7
Quote:


It's a really nice oc. The best I can get on my 720 is 3.6ghz @ 1.5v with the 4th core unlocked. But I am on air

lol you have the 4th core unlocked... great.. mine does not unlocks..

Quote:


Originally Posted by AMDZ
View Post

Increase the CPU/NB Frequency. Your seriously bottlenecking 4GHz with the NB on auto. Also raise its voltage to maybe 1.3. Set it to at least 2.5GHz for starters.

with my old mobo i was able to reach 2600mhz NB (3.6ghz cpu oc) but with this one a little tricky.. please take a look of the amd overdrive screen, and look at the voltage section, can you describe:
CPU VDDC
NB CORE VOLTAGE, NB VID
(are the same ?)
 

·
AMD Overclocker
Joined
·
9,178 Posts
From what i've seen on my system, increasing the nb makes no difference in benchmarks. I got the same cpu score in 3dmark vantage with the nb at 2600 that I did with it on auto. So I just leave mine at auto to save myself the extra heat.

Now for oc'ing higher up, it probably would make a difference. Just giving you a better perspective on the situation.

Also, between those three options, I would say the NB VID is the one you want. Increasing the nb's actual voltage does nothing, you have to increase the cpu/nb voltage, which i'm guessing is the nb vid. The cpu vddc is something else I think. Is the cpu vddc the one that is at like 2.5v?

And I got lucky with my 720 unlocking. It's probably because I got one of the first 720's that was a fully functional deneb that was just branded as a tri-core to meet demand.
 

·
Registered
Joined
·
183 Posts
I had my 720BE Deneb running stable at 4.0GHz for a few weeks on water, booted at 4.1 but would BSOD during Prime95. It took ages to get it there. I was running on the ASUS M4A79T Deluxe with 2x2GB of Mushkin DDR3-PC12800(7-6-7-18). If I recall I had to go through this to get there:

- overclock NB to 2400, Prime95 to stability
- downclock RAM multiplier to 6.66x (1333MHz) and adjust timings accordingly
- bump FSB to 212, adjust RAM timings/voltage and NB voltage until Prime95 stable
- raise CPU multiplier and Prime95 until BSOD, adjust voltage, rinse and repeat

I ended up hitting 4.0GHz at 1.575v, RAM at [email protected] To be fair, this was on a CPU-only watercooling loop in a basement where the ambient temp was around 65'F, and that was an exceptional chip. It's going to be near-impossible to run at 4GHz 24/7 unless you're going sub-ambient on a golden proc, and 1.575 is hardly advisable for long term use. Best of luck to you, though.
 

·
Registered
Joined
·
183 Posts
Don't be disappointed with 3.8GHz, that's still a pretty good overclock for the 720BE. I bought my first 720 within a week of launch, and those early 720's were really solid chips. If you can boot into Windows at 4GHz, I bet you can get it stable with some tweaking. Just keep a BIOS profile saved for your highest stable clock and keep messing with it. Some of them like voltage and some of them like low temperatures, mess around with both until you find a sweet spot :]
 

·
Registered
Joined
·
955 Posts
Quote:


...i was able to boot windows at 4ghz, but few seconds after bsod... 3.8ghz is not bad at all..

That's the spirit!

I reached the same conclusion on my 955 BE
 

·
Registered
Joined
·
955 Posts
In my view, they're all about the same.

You're going to get a lot of differing opinions on that, which I submit is largely based on the false assumption that load somehow verifies stability.

It's indicative, and associative, but it's far from any form of proof.

Many proceed along the assumption that long running tests are the way to establish stability, but then numerous anecdotes appear where such "proof" has been established, only to discover some other context in which the system crashes.

Any test which can cause a crash is suitable, which includes various benchmarks.

However, a lot depends on your approach to the subject of overclocking. I require stability, I'm an engineer, and my goal is to gain value from my investment.

Some, however, try to find the absolute maximum overclock, which by definition means there will be little or no headroom against the slightest origin of instability, and quite possibly creating a situation where the computer passes many tests, but occasionally exhibits instability due to the absence of safety margin.

It is only for this latter case where long running tests apply, and to that extent it is also only logical to run a cocktail of tests, because there are no applications which can test every permutation of usage of every circuit within the CPU. For that, any and every test you can run is best.
 

·
Registered
Joined
·
183 Posts
Nice...I told you it could be done


I see you went for a straight multiplier overclock...that's pretty hardcore. If you went about it with FSB+multi you could probably lower that voltage a bit. You would just have to overvolt the NB as well. 1.6v is fine for quick runs, but wouldn't recommend keeping it there for long. I ran 1.5v 24/7 on water and was still kinda sketched about that.
 
1 - 18 of 18 Posts
Top