AMD A II X2 250 : HT3 vs HT1, perf hit ?

[richierich1212]

double check your memory settings

 

[StrangerGuy]

This guy and lots of other people have been able to get 4GHZ on air. And its not a temp problem. I'm getting really annoyed that people always assume that. When I do OCCT, it BSOD's with in a few minutes(sometimes instantly) and temps never get above 53C.

I seem to have screwed up the link above: http://benchmarkreviews.com/index.p...k=view&id=384&Itemid=63&limit=1&limitstart=10

I don't get why people feel they are entitled to an 4GHz+ OC...Did AMD put "guaranteed 4GHz" on their CPU boxes? My 555BE doesnt hit 4GHz at 1.5V and cannot even unlock an extra core, but I don't come here and bitch about it.

 

[DrMrLordX]

Sometimes K10.5 chips are heat-limited and sometimes they are not. My Propus was almost certainly heat-limited, and I'm sure I could have pushed it further with liquid ultra plus the extra fans I bought, though I killed it with a sloppy lap job so I'll never know for sure.

But just slapping on a 100 cfm Delta to my nh-d14 let me drop vcore by about .2v at 3.7 ghz without losing stability.

Then there's my Sargas, which really doesn't like clockspeeds over 3.85 ghz even though it never got above 31-32C at load. Granted, I stopped at 1.48v vcore before bricking the board with a bad flash (meh).

Once my replacement comes back from MSI we'll see how it does at 1.6v. It may be more of an htt limitation than anything else, since the multi is so low.

 

[jvroig]

Well its been shown that ocing the NorthBridge (Which the Memory Controller and L3 Cache run on) can give a solid 10% increase in performance in a couple area's compared to stock 1.8ghz

How much of an OC was observed to give that solid 10% increase in performance from 1.8GHz? What was the final NB speed that was benched that showed 10% performance increase?

 

[jvroig]

Once my replacement comes back from MSI we'll see how it does at 1.6v

Are you not afraid of frying your Sempron 140? 1.6V seems a bit too high for me.

 

[DrMrLordX]

No. It cost me all of $25, and I'd only do that on a suicide run anyway. Besides, cooling a Sempron 140 is so easy that it hurts. Cooling meant for a quadcore CPU is complete overkill on a single-core CPU.

 

[jvroig]

No. It cost me all of $25, and I'd only do that on a suicide run anyway. Besides, cooling a Sempron 140 is so easy that it hurts. Cooling meant for a quadcore CPU is complete overkill on a single-core CPU.

Ah, your giant Noctua NH-D14 is on it. Understood.

 

[veri745]

Ah, your giant Noctua NH-D14 is on it. Understood.

Except no amount of cooling is going to prevent a CPU from degrading quickly if you run it at 1.6V. That's why it's called a suicide run.

 

[jvroig]

Except no amount of cooling is going to prevent a CPU from degrading quickly if you run it at 1.6V. That's why it's called a suicide run.

My comment was about his comment regarding cooling, and not to acknowledge "safetiness" of 1.6V. Personally, anything beyond 1.45V is already beyond my comfort zone.

You do bring up an issue that is important to me. Why would 1.6V "quickly degrade" a CPU? Why 1.6V, and not, say, 1.55V? When does this quick degradation actually start to happen? Does it start at 1.6, 1.55, or 1.5V? Does it start even at 1.45V? Running anything beyond spec will affect the lifetime, but when does it actually shorten it significantly, let's say from 10 years of operation to just 6 months? Who (and by what methodology) ascertained that 1.6V will degrade an existing AMD CPU regardless of operating temps?

I am not arguing that 1.6V is "safe". As previously stated, for my own AMD CPUs (I have an Athlon X2 7750 and a Phenom II X4 965 both currently operating), I do not cross 1.45V regardless of cooling, even for the 7750 that idles at 27C, normally operates at ~33C, and only reaches 39-40C when being stress-tested for stability. But it is more due to a "gut" feel. But in my research, no data has quite clearly said when the voltages clearly cross the line. Visiting various forums will only yield "rules of thumb" from enthusiasts, whose sources are uncited and veracity likewise in question. Some enthusiasts who seem to be respected in their particular forums would say 1.55V is ok for AMD, and they would cite blogs/websites that say likewise, but following up on their sources is impossible, as there is often none.

Unfortunately, I do not see how this can be resolved "officially", since AMD themselves do not seem to offer an official guide. Their Phenom II 965 C3 specs says operating voltage tops at 1.4V, which I interpret to be the "max recommended voltage". Their AMD Dragon Overclocking guide says that 1.3-1.5V is the recommended safe voltage, assuming a high-end air cooler is used, which can then be interpreted to mean 1.5V is safe as long as temps are managed. Following on that, why would 1.4V be the normal safe voltage, 1.5V be safe as long as temps managed, but 1.6V be unsafe regardless of cooling?

I have no wish to run on higher voltages myself, but the question of voltages has hounded me for so long, and the answer to it has so far evaded me, or at least the answers I have found have so far failed to completely satisfy my curiosity.

 

[DrMrLordX]

Ah, your giant Noctua NH-D14 is on it. Understood.

Da. What's funny is that conventional wisdom would dictate that a larger HSF would be wasted on a low-power chip like a Sargas, but it would seem that the newer coolers have improved thermal resistance that makes them (somewhat) useful versus cheaper and smaller HSFs.

HSFs with exposed heatpipes should be the best for applications such as these, but that's just a theory.

Except no amount of cooling is going to prevent a CPU from degrading quickly if you run it at 1.6V. That's why it's called a suicide run.

That's not necessarily true. 1.6v is commonplace for k10/k10.5 chips under water, and they can and will last for upwards of a year under such conditions. Granted, 1+ years is significantly less chip lifespan than the 10+ you can get running a chip at stock, but that still isn't exactly a suicide run.

1.6v on air is riskier, especially if you let temps get too high in the process.

You do bring up an issue that is important to me. Why would 1.6V "quickly degrade" a CPU? Why 1.6V, and not, say, 1.55V? When does this quick degradation actually start to happen?

Some of the more astute engineers could give you a complete and satisfactory answer, but in the meantime, you can always read up on electromigration and Black's equation. Short answer is that electromigration kills chips, and ramping up voltage increases a chip's susceptibility to electromigration in more ways than one.

So, increasing vcore from 1.3v to 1.35v is going to cause less damage to the CPU than increasing vcore from 1.55v to 1.6v.