Help me get my Deneb above 4Ghz.

[Plimogz]

Good Day.

I popped a new 965 C3 into my UD5P (790fx, ddr3) and went about trying to get this to work @ 4 Ghz with Vista 64. And, so far I've encountered a whole lot of fail in my attemps.

I'm not really going to outline all of my settings, and all that other good stuff here, suffice it to say that 3.9Ghz is pretty easy to hit with stock volts (was stable when I fell asleep with it linpacking, but it rebooted while I wasn't looking) and I'm thinking that 1.425 would be good for 3.9 stable on this chip.

I've tried up to 1.475v in my vain attempts to get 4ghz stable, but the chip doesn't seem to want to budge. Anybody have any good ideas which they wouldn't mind sharing?

CPU : AMD 965 BE C3
Mobo: Gigabyte GA-MA790FXT-UD5P (BIOS version F7)
RAM : OCZ PC3 12800 7-7-7 @ 1.9v Platinum Edition OCZ3P16004GK 4x2GB sticks
PSU : PC&P Silencer 750 Watt Quad Red
VGA : HD4890
HSF : CoolerMaster Hyper212 with Push-Pull High CFM Delta fans
HDD : WD Raptor 74GB (OS) with a OCZ Vertex thrown in there

 

[formulav8]

Some reviewers has been getting C3 to almost 4.2ghz in 64bit os. But, your chip just may be a 4ghz dud.

You can run up to 1.55v(At least I have before) on that cpu if you have real good cooling.

Otherwise just make sure your ram, htt links, nb and such are at very low settings to make sure its not something else besides the cpu keeping you from getting 4ghz.


Jason

 

[shaolin95]

First, no chip is guaranteed an overclock.
Second:

I'm not really going to outline all of my settings, and all that other good stuff here, suffice it to say that 3.9Ghz is pretty easy to hit with stock volts (was stable when I fell asleep with it linpacking, but it rebooted while I wasn't looking) and I'm thinking that 1.425 would be good for 3.9 stable on this chip.

I don't mean to be rude but if you are not willing to take time to post your settings, load temps, heatsink, etc...do you expect a lot of people to take time to help you?
We want to help but we need details ;-)

 

[Plimogz]

First off, thanks for the replies, gentlemen. Just had another BSOD, but you all know how it goes...

Nah, I won't post my settings, because I'm kinda going at this pretty half-assed for the time being, didn't even update my BIOS yet, which is what I was thinking I'd do while I rebooted, just now.

I don't know if it's a dud or not, hard to say, but I doubt it. 3.9 was just too easy for 4.0 to be this hard, it's got to be something I'm missing. I won't try to brute-force the thing with 1.5+v just yet.

I think i'll start up a pot of coffee and take this slow and methodical. Thanks again for the replies, I'll post back in a couple. oh, and the heatsink I'm using is a CM 212 with a couple of high CFM fans push-pull. Ambient is like 45F.

Does anyone have any clue as to what the actual TDP of a Deneb @ 4GHZ / 1.5V might be? From some of IDC's posts I seem to remember that Power scales exponentially with both clock and volts, but I wonder how much of a pig a 4ghz phenomII really is...

 

[Accord99]

Does anyone have any clue as to what the actual TDP of a Deneb @ 4GHZ / 1.5V might be? From some of IDC's posts I seem to remember that Power scales exponentially with both clock and volts, but I wonder how much of a pig a 4ghz phenomII really is...

Xbitlabs measured close to 250W for theirs but they needed almost 1.6v.

http://www.xbitlabs.com/articles/cpu/display/phenom-ii-x4-965-c3_10.html#sect0

 

[Spikesoldier]

i havent tried amd since the K8 days.

but have you tried loosening memory timings? i have read that the PhII's have problems with 64bit OS while heavily OC'ed.

maybe you can try a 32bit OS temporarily just to see how high you can crank it up?

 

[Hey Zeus]

Good Day.

I popped a new 965 C3 into my UD5P (790fx, ddr3) and went about trying to get this to work @ 4 Ghz with Vista 64. And, so far I've encountered a whole lot of fail in my attemps.

I'm not really going to outline all of my settings, and all that other good stuff here, suffice it to say that 3.9Ghz is pretty easy to hit with stock volts (was stable when I fell asleep with it linpacking, but it rebooted while I wasn't looking) and I'm thinking that 1.425 would be good for 3.9 stable on this chip.

I've tried up to 1.475v in my vain attempts to get 4ghz stable, but the chip doesn't seem to want to budge. Anybody have any good ideas which they wouldn't mind sharing?

This is all you need

 

[Plimogz]

@Accord99: Many thanks for that link! extremely relevant information there and exactly what I was wondering about -- kinda begs the question: how much excess power circuitry capacity did Gigabyte engineer into the GA-MA790FXT-UD5P...

@Hey Zeus: LOL. Yeah, I'm already on the phone with suppliers, my 60 gallon tank of LN2 should be here next Tuesday. I'm planning on recoup-ing the cost of it going around town stealing bikes by pouring the stuff in up-turned U-Locks and smashing them to bits... (J/K, of course. i do not even know if that would actually work)

seriously though. This 965 is currently churning out LINPACK @ 3997Mhz/1.504V (reported under load in CPU-Z), so I guess it must be eating up somewhere around 220-240W... and it has been stable for 10 loops, using 4096 MB, so it is looking good.

@SpikeSoldier: I'm running pretty loose timings at the moment... I think 1333 7-7-7 2t. (the ram is DDR3-1600 7-7-7-24 @ 1.9v, but I run it with 1.75v and it'll do 6-5-5 with that voltage up to about 1400mhz). I have thought about trying out a 32-bit O.S., just to see how much more Mhz I can get out of the PII, but I'm not in a hurry to do so because I have 8GB of RAM and wasting 4GB RAM for an extra 200Mhz is a definite deal-breaker.

FWIW, BIOS v.F7 helped somewhat. I was running v.F5M before, and I am pretty sure there was a bit of an improvement with the new version (which did mention something about improved support for C3 steppings, so...) but it may all be in my head...

 

[TJCS]

Nah, I won't post my settings, because I'm kinda going at this pretty half-assed for the time being, didn't even update my BIOS yet, which is what I was thinking I'd do while I rebooted, just now.

lmao, this is ridiculous. It's like asking the doctor to give you a cure, and when he asks you what's wrong you tell him: "sorry can't say, because I got a half-ass disease."
​

 

[Plimogz]

@TJCS: yeah, it is pretty ridiculous, but whatever... If i did post my settings, what then? I mean, take this for example: I was stable @ 19.5x205=3997mhz w/1.504v just now. I increased the reference clock from 205 to 207, kept the same volts and now IBT isn't returning accurate results... Can you see how an extremely slight variation in settings is currently the difference between "works" and "doesn't work" in my overclock? am i really going to post up every last BIOS setting and ask people to try to "remote tweak" for me? Not only would it not work (probably) but it would be time consuming and frustrating...

Nah, I'm just hoping that maybe somebody from the forum has recently spent some time getting a PhenomII to 4+Ghz and happens by this thread and cares to contribute any particular insight he/she might have gained on their own.

Thanks for replying though.

Back on Topic:

Would anyone care to confirm that I can get higher clocks with less volts if I overclock using the Base Clock & the Multi instead of just going for a straight Multiplier overclock? I've read that it might be the case but all my previous experiments have yielded less than conclusive results.

I ask because i would really rather not increase vCore anymore than I already have (currently 1.5v) but would also like to actually break the 4Ghz barrier. I'm so close! (3997Mhz lol)

Update Edit: Yeah... It is as I had feared: even with the vCore turned up another notch (so 1.525v) I can't get 20x200=4027Mhz anywhere near Linpack stable. this sucks... this is too much voltage anyhow, (1.5v is going to be my self-imposed limit, because I don't have the will-power to respect AMD's 1.425v upper-limit spec) but... sigh.

 

[Spikesoldier]

@Accord99: Many thanks for that link! extremely relevant information there and exactly what I was wondering about -- kinda begs the question: how much excess power circuitry capacity did Gigabyte engineer into the GA-MA790FXT-UD5P...

@Hey Zeus: LOL. Yeah, I'm already on the phone with suppliers, my 60 gallon tank of LN2 should be here next Tuesday. I'm planning on recoup-ing the cost of it going around town stealing bikes by pouring the stuff in up-turned U-Locks and smashing them to bits... (J/K, of course. i do not even know if that would actually work)

seriously though. This 965 is currently churning out LINPACK @ 3997Mhz/1.504V (reported under load in CPU-Z), so I guess it must be eating up somewhere around 220-240W... and it has been stable for 10 loops, using 4096 MB, so it is looking good.

@SpikeSoldier: I'm running pretty loose timings at the moment... I think 1333 7-7-7 2t. (the ram is DDR3-1600 7-7-7-24 @ 1.9v, but I run it with 1.75v and it'll do 6-5-5 with that voltage up to about 1400mhz). I have thought about trying out a 32-bit O.S., just to see how much more Mhz I can get out of the PII, but I'm not in a hurry to do so because I have 8GB of RAM and wasting 4GB RAM for an extra 200Mhz is a definite deal-breaker.

FWIW, BIOS v.F7 helped somewhat. I was running v.F5M before, and I am pretty sure there was a bit of an improvement with the new version (which did mention something about improved support for C3 steppings, so...) but it may all be in my head...

i actually think that running 8gb is a burden to the CPU, try yanking two sticks out and see what happens.

additionally why 2T? i know that in the k8 that going from 1T to 2T was a big performance hit. is it because of the 4DIMMs?

 

[Plimogz]

i actually think that running 8gb is a burden to the CPU, try yanking two sticks out and see what happens.

additionally why 2T? i know that in the k8 that going from 1T to 2T was a big performance hit. is it because of the 4DIMMs?

ok. good advice. I'll try that once my current run of IBT fails (sigh) I'm going for 19x211=4009Mhz w/1.504v. I've also turned pci-e speed back to auto as well as HT link-width back to auto. I think my Mobo is quirky like that. not sure though... ANYWAY.

--->I agree with you that 8GB is a burden on the CPU, but I'm not interested in benching this maching, rather I'm looking for a nice 24/7 overclock (i.e. reasonable voltages, all 8GB of RAM installed, 50+ loops Linpack w/6000Mb ram assigned stable, etc.) I guess i might try pulling two sticks to see what effect that'll have, just to try it out, for kicks, as it were.

As for the 2T: well, I was just thinking that I would try to take a bit of a load off of the IMC, y'know? In fact, my previous 720BE unlocked to X4 never had any serious issues with 1T timing, even with all four DIMM slots populated AND with 1333 6-5-5 timings, but I was just going for 4Ghz on the core for the time being, UNcore and RAM will come after.<---

Hmmm... I've been writing this response to you for the last, what, 10 minutes, and IBT hasn't failed yet... I swear... there's something quirky about the PCI-E frequency setting on this Mobo. Or maybe it's the HT Link Width setting... Any body out there have any thoughts on the matter?

Ah, it just failed 5 loops in...

 

[TJCS]

1.

First, no chip is guaranteed an overclock.

2. The more relevant info you share, the more people have to work with to help you. Besides voltage and clock speeds there are plenty of other settings that can destabilize an overclock.

3. Wouldn't it be great if other people who came to this forum looking for the same knowledge was able to benefit for all the time you & others spent to get it to work?

This is just my 2 cents :hmm: Good luck with your rig.

 

[Plimogz]

Alright Man.
I'll do whatever. post whatever. But you first:

What do you think I should put in here?
BIOS Version/Settings? Hardware Configuration? OS? applications I'm using to stress test?

The reason I said I was being half-assed and not putting up my specs is that rounding all the above info up is going to take a while. during which time I'll to stop trying to get this OC to work...
Mind you, at this point I am getting somewhat turned off by my resounding failure

(IBT is returning
Time (s) Speed (GFlops) Result
54.749 48.2669 1.#QNAN0e+000
54.950 48.0903 1.#QNAN0e+000
54.213 48.7438 1.#QNAN0e+000
right now lol...)

So I'll take a break, gather up all the pertinent info i can think of (would be helpful if you post back to tell me of anything I might be forgetting about and didn't mention higher up. You know, besides hardware configuration and BIOS settings) and then I'll post back here. Hopefully I can get this little piece of $#!7 to break 4Ghz with 8GB and 64-bit later on with a little help from some of you out there, of course

Cheers

Edit: For the record, normal IBT results always look something like
Time (s) Speed (GFlops) Result
55.802 47.3557 3.573501e-002
55.819 47.3415 3.573501e-002
55.781 47.3731 3.573501e-002

Oh, and the immediately preceeding fail was running only 4GB (2 dimms) @ 4Ghz.

 

[brandonwh64]

I have some simular problems on mine too! i can hit 3.9ghz on my C3 by just changing the multi 19x and a HT of 205 and a Vcore of 1.45 and it linpack stable over night but it doesnt matter what voltage or setting i throw at it, It WILL NOT get stable at 4ghz maybe we have the same batch?

 

[Plimogz]

Off-hand I can remember that my chip was a week 45. But I didn't pay attention to the batch code. (that's the one that goes a little like CAZWM, right?) Anyway. I don't believe in batches...

I mean, sure, some weeks will be better than others. Some wafers will be better than average, that's not really what I don't believe in. My point is more often than not, success or failure, being able to wring out that last little bit of performance from the chips, will come down to the skill and patience of the individual overclocker/tweaker and not so much some fabled golden batch or whatnot. And also what tests a person uses determine stability...

I notice that you too are using linpack to test for stability. And I think that there is a pretty good consensus among clockers that linpack is pretty much the most stressful test out there right now. So could it be that by using that as our metric for stability we are perceving that even these C3 deneb have trouble around 4Ghz, when others who are more easily satisfied with "lighter" stress tests are touting 4+Ghz results at lower volts, are in fact deluding themselves as to the real stability of their overclocks, to an extent in any case...

just a thought

 

[shaolin95]

@TJCS: yeah, it is pretty ridiculous, but whatever... If i did post my settings, what then? I mean, take this for example: I was stable @ 19.5x205=3997mhz w/1.504v just now. I increased the reference clock from 205 to 207, kept the same volts and now IBT isn't returning accurate results... Can you see how an extremely slight variation in settings is currently the difference between "works" and "doesn't work" in my overclock? am i really going to post up every last BIOS setting and ask people to try to "remote tweak" for me? Not only would it not work (probably) but it would be time consuming and frustrating...

Nah, I'm just hoping that maybe somebody from the forum has recently spent some time getting a PhenomII to 4+Ghz and happens by this thread and cares to contribute any particular insight he/she might have gained on their own.

Well, the reason I asked about settings is because I AM that person that you were looking for.
I took a 955 C2 to 4Ghz on air (cold air though) on windows 7 64 bits and it was a huge battle to do so.
I then sold it and got a 965 c3 which I took to 4.1 Ghz prime stable on W7 64 bits as well. It was a LOT easier than the c2 revision. Heck I could bench at 3.4 on that thing.
So you see , I was the person you wanted but no way I am going to try to guess things if I do not know all your details so good luck on your quest.

 

[Plimogz]

@shaolin95: Ok. I'm in a bit of a lull because of my repeated failures. so here goes.

CPU : AMD 965 BE C3
Mobo: Gigabyte GA-MA790FXT-UD5P (BIOS version F7)
RAM : OCZ PC3 12800 7-7-7 @ 1.9v Platinum Edition OCZ3P16004GK 4x2GB sticks
PSU : PC&P Silencer 750 Watt Quad Red
VGA : HD4890
HSF : CoolerMaster Hyper212 with Push-Pull High CFM Delta fans
HDD : WD Raptor 74GB (OS) with a OCZ Vertex thrown in there

that is pretty much it as far the hardware is concerned, unless i missed something.

BIOS settings are at defaults, except that power saving features (C1E, C&Q are disabled)
ram timings are 1333 SPD. Voltages are default except 20x multi, vCore@1.5v vNB-CPU@1.15v (though i could have left it at default 1.1 for the time being, i guess) vDimm is 1.75v (I have tested these sticks with this mix of Mhz/timings/volts, it works well)

and, am I missing something here... anyway, I can edit this post to add any important info which you think is important as we go along

Thank you for taking the time to read this.

oh, and temperatures top out around 50C when I push 1.525v @ 4Ghz. I have also probed the area around the Power delivery components using a IR thermometer, and nothing on the board gets hotter than 50-ish C either. Ambient temperature is 45 degrees F.

 

[Spikesoldier]

@shaolin95: Ok. I'm in a bit of a lull because of my repeated failures. so here goes.

CPU : AMD 965 BE C3
Mobo: Gigabyte GA-MA790FXT-UD5P (BIOS version F7)
RAM : OCZ PC3 12800 7-7-7 @ 1.9v Platinum Edition OCZ3P16004GK 4x2GB sticks
PSU : PC&P Silencer 750 Watt Quad Red
VGA : HD4890
HSF : CoolerMaster Hyper212 with Push-Pull High CFM Delta fans
HDD : WD Raptor 74GB (OS) with a OCZ Vertex thrown in there

that is pretty much it as far the hardware is concerned, unless i missed something.

BIOS settings are at defaults, except that power saving features (C1E, C&Q are disabled)
ram timings are 1333 SPD. Voltages are default except 20x multi, vCore@1.5v vNB-CPU@1.15v (though i could have left it at default 1.1 for the time being, i guess) vDimm is 1.75v (I have tested these sticks with this mix of Mhz/timings/volts, it works well)

and, am I missing something here... anyway, I can edit this post to add any important info which you think is important as we go along

Thank you for taking the time to read this.

oh, and temperatures top out around 50C when I push 1.525v @ 4Ghz. I have also probed the area around the Power delivery components using a IR thermometer, and nothing on the board gets hotter than 50-ish C either. Ambient temperature is 45 degrees F.

1.9v on ddr3, that is a pretty extreme vdimm you are pushing there. i assume you are actively cooled?

also im not sure about this next statement, so correct if im wrong: high vdimm on IMC style CPU's is dangerous compared to pushing the limit on a traditional CPU/NB/DRAM setup?

 

[Plimogz]

hey, what can I tell you, these sticks are speced by OCZ as 1600 7-7-7-24 @ 1.9v
I could actually push 5% more volts through them (per their extended voltage protection) and still technically be within my warranty safe zone lol. but yeah, 1.9v for ddr3 is kinda asking for trouble.

but, as I've specified here and there throughout the thread, I'm running them at 1.75v 1333Mhz. With PhenomII speeds below, say, 3.8Ghz Core, 2.6-2.8Ghz UnCore, the extra bandwidth resulting from running 1600Mhz RAM instead of 1333Mhz is largely wasted and equal or better real-world performance is more easily achieved through the use of tighter timings/lower speeds. OCZtony tested and reported this months ago when the PhenomII architecture was debuting.

Anyway, with regards to High vDimm / IMC style architectures... I remember there being some debate with regards to this back in the day on the defunct DFI-Street forums (A64 was just picking up steam at the time) and iirc the idea back then was to always keep the IMC voltage below that of the Core. Or else, as some would put it, bad things would happen. Nowadays, with Intel having gotten on-board the IMC bandwagon, people with i5-i7's apparently do have to keep an eye on the delta between IMC voltage and vDimm -- I think the idea is that high vDimm would damage the IMC of these intel chips. I suppose that this is what has led to the rise of low-voltage DDR3.

I'm functioning under the impression that all this is kinda moot on the AMD side of the fence however, given that 1:AMD SOI has a higher tolerance to voltage and 2:with default voltages for AMD parts being somewhat higher than nahalem-type intel chips, the need for really low voltage DDR3 is, for the time being at least (until Bulldozer) somewhat mitigated.

 

[Rhoxed]

if i read right your running 1.15V NB volts?

for both of my chips (720BE unlocked and 940BE) it took a heafty push to 1.3 on the NB to get to that magical 4ghz running correctly

try 1.5V on the CPU and 1.3V on the CPU-NB (assuming you aren't overclocking the NB much past 2.6, then i would suggest upping it further)

 

[Plimogz]

First off, yeah you read that correctly.

Next off, I took a break from trying to get my 4.0GHz Linpack stable. I've been messing around with superPI 32M and Cinebench10 (I'll get around to posting some results to the Cinebench10 thread later on) and haven't had any crashes or incorrect results in those test FWIW... this is at 20x200=4000MHz, 1.5v vCore.

also, I'm up to 2.8NB-CPU with 1.25v so...

so I'll try out your advice: lower my NB-CPU to 2.6GHz, raise the NB-CPU voltage to 1.3v and try another round of Linpack @ 4.0GHz...

Honestly, and I do thank you for your input, and pray do keep it coming, but I'm not too hopeful about this... anyway, it sure is worth a try.

wish me luck!

Edit: Nope. No 4.0GHz linpack-stable love w/ NB-CPU voltage @ 1.3v

 

[Nemesis 1]

Why not just go for highest stable O/C and leave it alone . There is No point to push those AMDs to higher unstable clocks . None what so ever . 2 core Intels can bitch slap them . Set it at Stable and enjoy your very nice cpu. I can tell you from experiance that taking a 4 core nehalem to highest stable O/C and running 24/7 is an expensive thing to do .. The AMDs cost even more for 24/7 stable O/C . I here the guys here talking about Value . It really makes me and the wife smile and laugh. We know what it cost and its not cheap. Value my ass.

 

[Nemesis 1]

Why not just go for highest stable O/C and leave it alone . There is No point to push those AMDs to higher unstable clocks . None what so ever . 2 core Intels can bitch slap them . Set it at Stable and enjoy your very nice cpu. I can tell you from experiance that taking a 4 core nehalem to highest stable O/C and running 24/7 is an expensive thing to do .. The AMDs cost even more for 24/7 stable O/C . I here the guys here talking about Value . It really makes me and the wife smile and laugh. We know what it cost and its not cheap. Value my ass.

 

[DrMrLordX]

Since most C2 chips had problems getting past 3.8 ghz on 64-bit OSes, 3.9 ghz on a C3 chip should come as no surprise (nor should it be seen as a disappointment).

A few other miscellaneous notes:

2T command rate isn't as big a performance hit on DDR2 K8 and DDR2/DDR3 K10 as it was on DDR K8.

The whole reason to be concerned about high vdimm on any cpu with an imc has less to do with the RAM's rated voltage tolerances and more to do with the delta between imc voltage and RAM voltage. While AMD memory controllers don't seem to burn out as easily as Intel IMCs, if you're going to raise vdimm, you're still risking an imc burnout unless you raise voltage on the imc. On a K8, this was accomplished by raising vcore, but since the imc is on a separate voltage plane on K10 processors, you'd probably need to raise NB voltage to compensate.

If you can get away with 1.9v vdimm, however, then more power to you.

Nem: Intel dual-core CPUs do not "bitch slap" The Phenom II X4 965BE when overclocked. Certainly not the i3-530 or i3-540. Making such statements is hardly constructive given the OP's dilemma.

Also, OP, I get the impression that you may need better cooling. Is your CPU lapped? Can you get a better HSF? The Thermalright Venomous X may provide better cooling.