FX its tougher than you think.

I like this thread.. pushing 32nm silicon to its limits.
Quick question: If one buys an FX chip today, I take it that the manufacturing process is at its most matured? and thus will more than likely be a better OC performer with regards to voltage required for certain GHz values and acceptable thermals?

Sologuy said:

I like this thread.. pushing 32nm silicon to its limits.
Quick question: If one buys an FX chip today, I take it that the manufacturing process is at its most matured? and thus will more than likely be a better OC performer with regards to voltage required for certain GHz values and acceptable thermals?

Click to expand...

It's still a silicon lottery, but it seems as if a lot of the newer 8xxx chips especially the 8370's will Oc to X Ghz on less voltage then the earlier chips that came out. Note that this doesn't necessarily apply to pushing to extremes on subambient cooling. When pushing on DICE/Ln2 higher leakage chips seem to be able to push further then low leakage ones.

EarthDog said:

thats.. well, tough at best. It would take forever to gather the data.. to the point where, imo, we use the same rules of thumb from our already gathered experiences. I don't even think you mentioned what temp we are talking about and what application to read it from. You talking hwmonitor and the cores or package? I don't know amd well, but do know that you need to specify what you want and it needs to be relevant (not sure what that is specifically..Intel guy here)

@ shrimp - good refresher on the info. But that isn't the data we need/that kenrou was trying to gather. In the end though, did you notice their thermal shutdown was all the same @ 85c?

Click to expand...

Yea, that's pretty much the core threshold at 85c. CoreTemps reports a TJmax of 90c with my 9590. I've seen the Cpu shut down at 87c, 85c... but never has reached 90c.

The info is already there actually. It would be listed under TCase temps. In the examples of 8370E, 8370 and 9590, the TCase differs for each cpu indicating a higher Max P-state for turbo and the recommended temperature to sustain stability according to AMD white sheets.

It's not a rule of thumb so much as it would be simple mathematics.

Notice the above tCTLMax = 70 is the same for the 3 listed processors above? Well I'm not sure where people get 72c or 75c from, but I'm guessing that's what they see in AMD OverDrive. However, AMD overdrive is displaying a core max recommended temp based on how the software was written. I base my clocks on TCase max recommended temperatures being pretty accurate based on the algorithm being used. And since the TCase changes from cpu to cpu based on clock speed and p-state voltage so does the algorithm.

So the most basic rule of thumb if you will would be..... Raise clocks and voltage = must lower temps.

Running sustained temps of 80c at the socket I do not recommend at any clock speed while this is far beyond tCTLMax of all the FX processors.

Will heat kill your Cpu? Yes indeed. When? Just a matter of when not IF.

At 300w, we are using 25 amps. That's a lot of current. TCase based on these numbers would likely need to be below 50c sustained at the Cpu/Socket temp to obtain long term stability. Not to mention factoring the NB, HT and RAM overclocks which all have a temperature threshold as well including but not limited to VRM TDP.

As VRM adds heat, it travels to Cpu via copper inlay in the motherboard PCB and the CPU also returns heat to the VRMs in the same fashion.

At 4.7ghz - 5.0ghz, TCase max should stay below 57.0c and that is quite difficult.... to sustain stability.

tCaseThermalLimit - Asus sets this in bios to 65c to prevent stability problems and then throttles cpu through Cool and Quiet and C1E ect. (auto settings) / (Manual settings shuts this off)

Let me add the tCTLMax is Cpu socket temp, while TCase is a recommended Socket temp and THERMTRIP is always 85c at the Core temp.
Above is why I de-lid processors.

I like the direction this thread is going. I too am curious to see the REAL, but practical, limits of these FX chips.

Synn21588 said:

I like the direction this thread is going. I too am curious to see the REAL, but practical, limits of these FX chips.

Click to expand...

Real but practical would be the FX-9590. Limited to cooling and voltage out put of VRM section, only a couple of boards handle these 4.7ghz and above speeds even with overclocking lower end 8 series chips.

So would be safe to say most overclockers can achieve 4.7 - 5.0ghz for daily use and still be within the thermal limits.

Running 4.7ghz ~55c 1.425v/1.440v fine for gaming/stressing or 4.8ghz 1.462v/1.488v ~62c fine for gaming but becomes slightly unstable for stressing (passes IBT AVX max 10 runs, reboot fails, reboot passes, always passes Prime95 blend 8h). We can definitely see something there.

Kenrou said:

Running 4.7ghz ~55c 1.425v/1.440v fine for gaming/stressing or 4.8ghz 1.462v/1.488v ~62c fine for gaming but becomes slightly unstable for stressing (passes IBT AVX max 10 runs, reboot fails, reboot passes, always passes Prime95 blend 8h). We can definitely see something there.

Click to expand...

Been busy lately but back to this thread..

Going by those numbers, my system has some 'OC headroom' left in it imo.. gaming for me atm with the system in my sig, HWmonitor indicates 45C socket temp and 42-43C for cpu... glad I've got a BIG case..lol.. however, this is a 2 yr old build & I"ll be upgrading the cpu fans and gpu next week (if it doesn't get held up by Xmas post rush..lol..)

Let me borrow a spread from RGone, might be a bit out of date :

Kenrou said:

Let me borrow a spread from RGone, might be a bit out of date :

*image*

Click to expand...

I take it this is with all cores enabled, bit out of date? joking? not having a go at you Kenrou.. just say'in..a quick search around enthusiast's forums does not indicate FX-6300 with those Vcore values for a given clock speed imo unless that chip was cherry picked.. how was this tested? looks like the values for booting into windows... info in that chart is not serious indication of what is actually needed in 'real world' stress testing.

you can bet those are stable, those were my chips at one time.

Yes that was RGones own testing not some chart pulled off the webs.

as a matter of fact I pulled both of those chips out of my rigs and sent them to him for his testing.
I still have that 8350 in a rig here and I think manny or bassnut got the 6300.
so they were "cherry picked", right out of my rigs.

Yes I have the 6300 and am typing on it as we speak.

I take it this is with all cores enabled, bit out of date? joking? not having a go at you Kenrou.. just say'in..a quick search around enthusiast's forums does not indicate FX-6300 with those Vcore values for a given clock speed imo unless that chip was cherry picked.. how was this tested? looks like the values for booting into windows... info in that chart is not serious indication of what is actually needed in 'real world' stress testing.

Click to expand...

You guys have the need and rights to know that RGone used a chilled loop to achieve those clock speeds and voltages and are not relevant to actual ambient cooling.

And because I have to state that because the thread that came from does not, the above voltages are not correct to the clock speed at hand.

It is super very ultra unlikely that RGone and or Mandrake or Caddi Daddi hit 5.4ghz with 8 cores with any kind of stability on ambient temps.

The chart may be accurate up to 5ghz, but not likely past that and certainly not with a "high end air cooler" as another user in this forum has recently stated with A Noctua could only had 4.8ghz and probably on a good day.

I'm sorry. I do not mean to offend, but that chart would be accurate for chilled liquid cooling and I personally do not follow it.

It must be taken with a grain of rice that the figures depicted are strictly tested to his like and set up. results will heavily vary.

And I'd be willing to bet Mandrake would not run 1.7100v with that 6300 at 5400mhz on ambient temps as we speak. (sorry to call this out, but hey.... you know)

Let me edit:

If so... Lets see that 2 hours Prime95 at 5.4ghz at 1.7100v (this I gotta see)

ALL charts must be taken with a grain of salt, i thought that was a given since all setup differ slightly. And yes the NH-D15 tops out at 4.8ghz with good case ventilation, i can attest to that.

Granted this is a sample of one, but i tend to butt heads with my father who loves he's 6300 (has gone through 3 of them) and he regularly reports he's voltages to be slightly higher then mine at the same clock speed

ShrimpBrime said:

You guys have the need and rights to know that RGone used a chilled loop to achieve those clock speeds and voltages and are not relevant to actual ambient cooling.

And because I have to state that because the thread that came from does not, the above voltages are not correct to the clock speed at hand.

It is super very ultra unlikely that RGone and or Mandrake or Caddi Daddi hit 5.4ghz with 8 cores with any kind of stability on ambient temps.

The chart may be accurate up to 5ghz, but not likely past that and certainly not with a "high end air cooler" as another user in this forum has recently stated with A Noctua could only had 4.8ghz and probably on a good day.

I'm sorry. I do not mean to offend, but that chart would be accurate for chilled liquid cooling and I personally do not follow it.

It must be taken with a grain of rice that the figures depicted are strictly tested to his like and set up. results will heavily vary.

And I'd be willing to bet Mandrake would not run 1.7100v with that 6300 at 5400mhz on ambient temps as we speak. (sorry to call this out, but hey.... you know)

Let me edit:

If so... Lets see that 2 hours Prime95 at 5.4ghz at 1.7100v (this I gotta see)

Click to expand...

Actually Shrimpy that chart was done on his big loop, ie 700 Gph, pond pump gallons of water and a 360 rad but was at ambient temps, not chilled.

In addition all speeds up to 5.2 Ghz were run for stability using Occt or P95 I can't remember which one.

5.3 and 5.4 were done because CD lent the 6300 to Rgone and wanted him to push the 6300 as far as it could go and be able to bench. So no it's not a daily stable Oc.

No I have never tried or ran my Fx 8350, 8120 or 6300 which was the chip in the article I purchased from CD, above 4.7 daily on ambient cooling. My Fx 8350 I have tested up to 5.1 using Prime 95 blend and had passed 2 hours at that speed but the voltage necessary to do vs performance gained, in my eyes wasn't necessary. Therefore I run my 8350 @ 4.7 with 1.46 v.

I doubt anyone would be able to pass 2 hours P95 at 5.4 with 1.7v and be able to keep it under 70c with ambient cooling.

sorry shrimps, prime stable @5.4 ambient water on the 8350, benched @5.6 with the 8350, massive cooling system in the garage on a cold day of course.
I can't remember what the temps were but it doesn't get below 20F here in Georgia.

caddi daddi said:

you can bet those are stable, those were my chips at one time.

Click to expand...

That depends on how the user defines "stable". Plenty of arguments about that on any enthusiast's forum.

yep.

Methinks everyone has their own definition of what stable means, you can start with IBT AVX very high 20m, but to be sure, better do that 2h Prime 95 blend, oh wait 1st 20m Prime95 small fft to check max temps... but to be really sure you have to of course do the overnight 8h Prime95 Blend, and to be REEEEEEEEEEEEEEEEEEally sure you MUST do 24h Prime95 Blend... But wait... even that might not be enough, i have seen guys state they run 72h to be absolutely 100% sure its stable... so... take your pick

For me atm is 20m blend between voltage/clock switches and overnight when you find sweet spot, do you think its enough