Final OC X2 3800 @ 2.6

[DeFJaE]

2.7 is the maximum I can take my processor. I tried lowering my multiplier to 9X but the same amount of voltage (when I had it at 10X) is required to keep it stable with P95. Either way, I can go with high HTT w/divider or straight up 1:1 @ 270 X 10. My biggest hurdle is the CPU temp. While stressing my CPU, I noticed my temps climbed as high as 55C and that is with my tornador cranked to max on a SLK-948 heaksink. My temps hovers at 29C on idle with 1.51 vcore which is great. I really think that the drastic spread on temps is being caused by

 

1: Inadequate cooling solution or

2. The IHS and lack of thermal compound is hampering thermal transfer to heatsink or

3. This one I leaning on......I think this processor gets VERY hot with high vcore. I believe this may be the case because I actually touched my heatsink when MBM5 read 55C and it was hot.

 

 

I think I need a better cooling soulution to get 2.8. I am thinking about removing the IHS and see if this improves my clocks, but many people who tried saw no improvements.

 

Yeah seems like you do need better cooling. 55c that is very hot, also try not to go over 1.50 on your vcore, I read that some X2's doesnt like higher vcore. I found my sweet spot at 1.40 with a special of 102%, dfi still manages to under volt it. I dont advise removing your IHS just because you are still on AIR. You will see little to no difference. Save that if you are in Water as it will benefit you more. My X2 doesnt like lower mutlipliers, I tried using x9 with 312 seeing that it yeild an extra .8 I am not sure if that cause any stability. X10 @ 280 is the most stable I was able to get this cpu - 10hrs of x2 prime large fft.

[RLM]

Defjae, the thing is personally I use my kit mostly for 3D artwork. These chips are awesome for that, as both cores are used to the max. Rendertimes close to half that of single cores.

 

Gameswise, you can only hope that the developers incorporate multi-threading in the future. Be daft not to wouldn't it?

 

One thing though, and it relates to games. The newer drivers by nvidia for their graphics cards, have dual core optimizations as part of their feature list. I don't know about ATI. These dual core optimizations have been a real thorn in the side for 3d artist, as there have been all sorts of conflicts with opengl. Displays freezing etc. As a solution we have to use a registry tweak to switch it off.

 

Would of thought for games, however, you would notice some sort of improvement with these features, otherwise the whole dual core optimization thing is a total waste of time.

 

Actually, thinking about all this, I'm tempted to ask over on the 3d forums, whether they know if their colleagues in game developing are actually incorporating multi threading or not. Anyone else know?

 

RLM

 

ps. Not pretending to know my arse from my elbow here, as I'm just copying what I've read, but apparently a clue to whether your x2 likes the voltage or not is indictaed by the letter in the bottom corner ie. "t" or "n". 't' apparently likes more voltage. Here's the thread

 

http://www.xtremesystems.org/forums/showthread.php?t=92978

 

Edit. Just had a look at my x2 4200 and it has an 'n' in the bottom corner, so the volt theory, albeit interesting, is still debatable for me.

[rhiridflaidd]

I've got a new x2 3800 daa5bv

Ldbfe 0607upmw

 

at 2700

It booted into windows and passed memtest at stock volts, but was dual prime unstable.

1.475volts allowed it to run Prime for 10 mins - but I was a chicken and backed off.

 

Now I'm supposedly on 1.40v and 2600, and I've been playing ut2004 without crash for a few hours ; which picks up CPU instability very well and far more amusing than Prime - of which I got 30 mins without error.

 

These chips seem really easy to overclock.

 

I'd imagine that each chip is stable at 1.35v up to a certain speed; then needs exponentially more voltage to go from there.

 

I'm lucky in that I know my memory is stable up to 285mhz; so I can run 1:1 accross the CPU's whole speed range.

 

So the real tricky question is.... I'm happy to marginally overvolt. But what is my volts.

 

BIOS says 1.4, smartguardian says 1.36.

 

The next question is what's safe.

 

Ok I know it's all relative; but on water, what do people reckon?

 

With Winchesters there was a consensus that 1.6v ish was fine .1v over spec.

 

But with these x2's there is loads of diferent opinions; many happy with 1.55v; others keeping it to stock?

 

So I'm a bit puzzled and baffled as to what to aim for.

 

Either.

a. Find the lowest voltage that will keep me stable at 2600 or play more at 1.475v?

[twivel]

No one can answer for sure the safe voltage question, though many have asked. Please search for safe voltage in these forums and you can get more detailed answers.

 

Some have theorized for dual core chips that < 1.5 is safe on air, others have said 1.55. or less. My comfort point for water is < 1.6.

 

I trust multimeter first, BIOS setting next, BIOS Reading third, then software in an OS last. What do you trust? It's only faith and guessing until you pull out a digital multimeter to make sure.

[rhiridflaidd]

I'll dig it out!

 

I know there aint hard and fast answers; and I know that we can never guess what will be fact 9 months from now. It's just the whole range of oppinions that I find baffling - I've done the searches. Many stay just over stock; others crank it to 1.6. So I'll compromise on 1.45 and see how far I'll get. That puts me .1 over stock - as I did with my winnie.

 

But I will dig out a multimeter for good measure.

[MiLK_MD]

It's only faith and guessing until you pull out a digital multimeter to make sure.

 

Agreed.

 

Others have pointed to the thread at Anandtech about safe voltages. The link is somewhere around here (sorry, too lazy to look it up at the moment.) Twivel has summed it up well, but the answer you're probably not looking for is safe voltage is only as risky as you want to be. Voltage kills the CPU fastest, then frequency, then heat. Do you measure risk in terms of your chip dying within a few years? Months? Weeks? What is acceptable to you? From what I have been able to glean, and mind you this is IMHO, with a V of 1.55 (for an X2) you can measure your chip life to an order of magnitude of a few to five years. I remember a designer saying that spec life for stock V is around ten years. A 10-15% increase reduces it to 5 years. Blast the chip with a V of 150% and you can expect you chip to last minutes to days.

 

I suppose that when people talk about save V and they quote something like 1.55V for the X2, the chip life of 3-5 years is acceptable since most overclockers will upgrade sooner than this interval. But it is up to you to determine risk,

 

The above numbers are big time guesses on my part so you may want to ask more experienced/knowledgable people, but the above seems to be consensus.

[DeFJaE]

I suppose that when people talk about save V and they quote something like 1.55V for the X2, the chip life of 3-5 years is acceptable since most overclockers will upgrade sooner than this interval. But it is up to you to determine risk,

 

I agree, more often we all upgrade to the next best thing. I am already looking forward to what conroe has to offer. If all the benchmark is as accurate or better, when it is release I can only imagine the performance increase when we over clock that bad boy. Then again we have to see how overclocking friendly conroe can be. For now most games is not dual core or CPU bound, it's all about MHZ. Until games take full advantage of Dual Core, we wont see any dramatic increase in performance switching from single to dual.

[DeFJaE]

Each CPU has a voltage limit, there's a thread in HardOCP I think I linked it in my first post in this thread. They pretty much explained what each letter and #'s means in your CPU ID (ADA3500DAA4BP). That's just an example, ok since it will be alot easier for us to actually look at it here. I copied and paste the ID Codes.

 

Identification

to break things down, the first line on your heatspreader or cpu should look something like this:

ADA3500DAA4BP

AAABBBBCDE1FF

 

AAA = which line of cpu's

ADA = Desktop

AMA = DTR

AMN = Mobile

 

BBBB = PR number

 

C = which socket and lid type

A = s754, lidded

B = s754, unlidded

C = s940, lidded

D = s939, lidded

 

D = Voltage rating

A = variable (1.3-1.4v)

Q = 1.2v

K = 1.35v

I = 1.4v

E = 1.5v

C = 1.55v

 

E = Temperature rating. basically, your cpu is rated stable at stock speeds + voltages up to this temperature

A = variable (51-71c)

X = 95c

P = 70c

K = 65c

I = 63c

 

1 = L2 cache size

6 = 2048kb

5 = 1024kb

4 = 512kb

3 = 256kb

2 = 128kb

 

FF = Core type

AP = C0 Clawhammer, s754

AR = CG Clawhammer, s754

AS = CG Clawhammer, s939

AX = CG Newcastle, s754

AW = CG Newcastle, s939

BI = D0 Winchester, s939

BP = E3 Venice, s939

BN = E4 San Diego, s939

BU = E5 Newark, s754

BX = E6 Sempron, s754

LD = E0 (?) Lancaster, s754

BV = E6, Manchester, s939

CD = E6, Toledo, s939

[rhiridflaidd]

I'm superficially stable at 2700mhz and 1.45v. That'll do me fine.

 

The PC is slicker and smoother than a single core -and that's a subjective, not an obvective thing (but an extra 150 mhz over the winnie can't co amiss either- assuming I can stabilise it at this.)

 

I'm setting up for UT 2007. If I was setting up for today's games; i'd be better off getting a new graphics card.

[MiLK_MD]

2700 is fine and dandy! Nice oc. Let us know how it goes with the games and stability.

 

@Defjae

Thanks for the [H] post. I go there too, but I didn't find that thread before.

[rhiridflaidd]

Just been playing Ut2004 for 2h solid. Cpu temp 26 deg idle - 38's the highest it's gone; but it seems be be coming down as the thing's bedding in.

 

So the codes make mine a Manchester.

 

So according to the above, my chip is rated 1.3 to 1.4v - well an extra .05 is trivial.

 

So I'll prime it now till the morning and see if it gives errors.

[RLM]

Did you read the thread/link I posted rhiridflaidd?

 

The voltage equation ain't so simple, it seems.

 

edit: http://www.xtremesystems.org/forums/showthread.php?t=92978

 

The guy in the above thread did a bit of research of a 1000 or so processors. According to his findings, 't' chips have been found to go up in smoke(so to speak), if overclocked on lower voltages. I'm not sure if that means sticking more through them is a good idea or not, but it's a thread worth reading IMHO.

 

RLM