Observations with an FX-8350

[Abwx]

And you didn't mention this at all during the course of the original thread?



Thanks for your permission in allowing me to defend a poster I find credible. It means a lot.

There really isn't anything for me to post. The issue you presented was discussed by the posters. They even came to a conclusion. It's there for you to read.

Say what you want but to stay on topic adress the quote below, or just aknowledge that you re clueless of what is discussed, all other posts would be what you already posted, that is thread crapping, but certainly not the slightest valuable argument :

The FX was new, there was no precise knowledge of its power management, what pointed me in the good direction is Hardware.fr 2012 review since they published both VIDs and probed voltage, at stock settings and for their ocking needs.

Both the Asus boards they used at a 2 year distance to test the same 2012 FX8350 show the same behaviour, either you put the voltage management LLCC on auto and the chip will work at optimal voltage, 1.24-1.27V at 4GHz for a FX8350 circa october 2012.

Or you can tweak the thing by forcing out of specs voltages for overclocking stability needs, on setting "LLCC high" the board will force a voltage that is 0.04V below the VID, on "LLCC extreme" the board will force a voltage 0.05V above the VID, the forced increased voltage will compensate for the increased voltages losses when ocking as the currents running from the VRMs to the socket will be often out of specs in this case.

In the case that interest us the VID was set as 1.3375 and the probed voltage was 1.377, this say that LLCC was forced to a setting between "high" and "extreme" resulting in the CPU being fed a forced voltage equal to the VID.

 

[railven]

Say what you want but to stay on topic adress the quote below, or just aknowledge that you re clueless of what is discussed, all other posts would be what you already posted, that is thread crapping, but certainly not the slightest valuable argument :

Again, what else can I add? I can't time travel. The thread covered your questions, like I said - it's there for you to read:

I'm convinced it really it the CPU that is burning through the power, and not the mobo. And I am convinced it is completely due to the stock voltage value.

The VID for my FX-8350 is 1.3875V.

When I set the CPU LLC to "High" for my mobo, the measured voltage under LinX load is 1.374V (a droop of 1.3875-1.374 = 0.0135V). Peak temperature is 61°C with a 19°C ambient. And power-usage peaks at 290W.

When I set the CPU LLC to "Medium", one notch lower than the "High" setting, the measured voltage under LinX load droops to 1.348V (a droop of 1.3875-1.348 = 0.0395V). Peak temperature falls to 55°C with a 19°C ambient. And power-usage peaks at 262W.

The I set CPU LLC back to "High" to reduce the Vdroop and I manually lowered the voltage to 1.300V, which under LinX load drooped to 1.291V (a droop of 1.300-1.291 = 0.009V). Peak temperature further fell to 53°C with a 19°C ambient. And power-usage peaks at 252W.

For one final test I manually lowered the voltage to 1.250V which under LinX load drooped to 1.249V. Peak temperature did not get above 45°C with a 19°C ambient. And power-usage was a mere 225W. (but the system wasn't stable at that temperature, voltage, and clockspeed, it eventually rebooted)

So that is a relatively huge range in power-consumption and peak operating temperature for a rather minimal range in operating voltage.

The reason my FX-8350 is sucking down the juice is because the spec'ed VID is so high for stock clockspeed. Luck of the draw I guess. At least I know I have lots of room to under-volt to reduce power usage.

But that pretty much rules out the mobo as the suspect, if the mobo was the source of the excessive power consumption then I would not expect the CPU's operating temperature to vary so much in correlation with the variation in the power consumption.

Oh, I did run enough tests to pull together the following graph:

This shows us nothing we didn't already know, but it does put numbers to it for this specific LinX app. Shows the CMT-tax of loading 2 threads per module versus 1 thread per module for the first four threads with LinX.

I am now running some more benches with traditional benching programs like cinebench to get a feel for the power-numbers when not using a power virus like LinX. Will report back.

You claim he did it out of malice. You can't validate that because of something you know now that you didn't then. You can argue ignorance but if this info was buried for over 2 years - WTH do you expect from someone who was doing this out of interest not financial responsibility?

You're wheels are spinning with no traction. At this point there is nothing that can be said. You'll attempt to branch out of this thread (which is what you are contesting) using hindsight. Well d'uh! That doesn't make the errors of a person intentional malice. But you'll continue to imply it.

Have at it!

 

[Abwx]

Again, what else can I add? I can't time travel. The thread covered your questions, like I said - it's there for you to read:

Dont know, the answer is somewhat in the quote you posted :

The VID for my FX-8350 is 1.3875V.

When I set the CPU LLC to "High"......

When I set the CPU LLC to "Medium"....

Then I set CPU LLC back to "High"...

So only user settings but no automatic LLCC, but also :

For one final test I manually lowered the voltage to 1.250V which under LinX load drooped to 1.249V.

Of course it cant be stable since the automated regulation loop is not activated, the regulation will not compensate micro variations, hence the instability at what is actualy the normal voltage at 4.0, notice that not a single time the regulation is set on default mode, that is auto, it s user settings from start to finish...

Well d'uh! That doesn't make the errors of a person intentional malice. But you'll continue to imply it.

Have at it!

I m not implying that, i m saying that under the guise of technicality this review was actualy very unprofessional to say the least, no research has been done by the OP to check what was going on, actualy i believe that he didnt knew how the CPU voltage regulation was working, otherwise he wouldnt had reduced the voltage to 1.249V without enabling the automatic compensation, i can explain here how it works exactly and how it manage to reduce TDP, AMD has recently implemented another scheme that will be used in Carrizo, rather than compensating using voltage like in the FX they will compensate using frequency as variable.

 

[railven]

Perhaps you missed it, but here is what I addressed was my issue:

Perhaps it s time to revive timely what appear retrospectively as one of the most fuddy thread ever wich has done great damage to AMD, not because those latter did actualy underspec their TDP ratings but because the OP was unaware of how a CPU voltage regulation loop work.

What caught my attention is that the probed voltage is exactly equal to the VID, wich is rarely the case, unless one remove the LLCC automatic setting and goes fully manual, in wich case the CPU can no more set the voltage to the optimal number since in this case the user is the one that decide of what is the optimal level.

By removing the automated regulation the OP, deliberatly or not, forced the CPU to eventualy work out of specs, in this case the CPU was overvolted from 1.24V up to a forced voltage of 1.377 wich is the value of the VID wich is forced as default value since the automation is removed.

With this trick the CPU was overvolted by a ratio 1.377/1.24 = 1.1105 leading to a ratio 1.233 more power then necessary, using the 125W official figure it s not difficult to compute that this inflated the power up to 154W, now pass thoses 154W through the MB VRMs efficency and the PSU efficency wich are about 90% each and you ll get 192W power variation at the main from idle to load, this is correlated by the OP measurement that showed a delta of 194W.

Hardware.fr review from 2012 with overclocking results, VID voltages and probed voltages of the FX8350 on an ASUS MB, like the OP, the first line is the defaukt setting :

http://www.hardware.fr/articles/880-5/overclocking-undervolting.html

Even if truth is now known i guess that the bad has already been done by what i consider as an infamous review, and i back it with numbers that say so.

You're straight out insinuated that IDC lied or misrepresented his findings. "Fuddy" implies FUD which is malice by intent. You go to claim this is an infamous review when it's a guy posting on a forum, stating his findings with a huge disclaimer right at the start:

Figured I'd make a thread regarding the trials and tribulations with my FX-8350. This is a new area for me, so expect some noobness to my efforts, feedback and pointers will be much appreciated as I attempt to scale the new learning curve

At no point during the thread did you spring this information, nor any other user, and it remained unmentioned for over 2 years. But this is an "infamous" review because it smears AMD, yet it doesn't mislead, or lie, or imply anything but show the user's findings.

Hell, there are countless back and forth (peppered by your snark trying to discredit) discussion by users about the results. Even they seemed satisfied with the conclusions reached.

Again, why didn't you bring this up? You only acted with a "GOTCHA" after you learned of it, yet even in the thread IDC discloses his settings/results. Where is the FUD? Where is the incompetence?

You have problems with posts debunking what is ultimately viral marketing, be it deliberate or not.?.

Before being a necro it has served some purpose, that is, to spread falsehood and misinformation, how many people where influenced by this review thinking that the OP was competent in this matter.?

How many people used this caricature of review as a "prove" of AMD CPUs being allegedly power hungry.?

Besides, this necro was quoted recently by the very OP that opened it....

Where is the viral marketing? You've made a bold claim - time to back it up.

Dont know, the answer is somewhat in the quote you posted :

So only user settings but no automatic LLCC, but also :

Of course it cant be stable since the automated regulation loop is not activated, the regulation will not compensate micro variations, hence the instability at what is actualy the normal voltage at 4.0, notice that not a single time the regulation is set on default mode, that is auto, it s user settings from start to finish...

D'uh, because IDC disclosed his findings. And did you raise your hand when that post went live two years ago? If no, why?

I m not implying that, i m saying that under the guise of technicality this review was actualy very unprofessional to say the least, no research has been done by the OP to check what was going on, actualy i believe that he didnt knew how the CPU voltage regulation was working, otherwise he wouldnt had reduced the voltage to 1.249V without enabling the automatic compensation, i can explain here how it works exactly and how it manage to reduce TDP, AMD has recently implemented another scheme that will be used in Carrizo, rather than compensating using voltage like in the FX they will compensate using frequency as variable.

Read what I put in bold. You've done more than implied - you've written it out. At this point you are labeling IDC a shill because he posted his personal review (not even a professional/paid for one) and you, two years later, finally got a "GOTCHA!" AMD did their own damage. If you think these observations by IDC damaged AMD's reputation more than JFAMD (which IDC defended) you're out of your mind.

And at that, I've seen three AMD Reps come through here, and yet not a single one stopped and took issue with this thread. I guess even AMD is as competent, viral marketing, and spreading FUD as you claim IDC of being. They just let it go uncontested (unless that's why you're here ).

I'll let IDC addressed your allegations, nothing left to say (again the info is there in the thread).

 

[SlowSpyder]

Nevermind.

 

[SlowSpyder]

Hardly scientific, but some quick numbers.

Set the CPU to 4.0 / 4.2 turbo at a few different voltages and ran LinX. LLC was set to 'Auto', but I think I'm going to try some different settings and report back.

At 1.35v the CPU ran at 4.0GHz (never saw turbo kick in). The system jumped up 248 watts from idle.

At 1.30v the CPU and the same 4.0GHz during the stress test, the system jumped up 214 watts from idle.

At 1.25 volts and again staying at 4.0GHz during the test the system used an additional 180 watts over idle.

A few things I could have done different. My memory is rated at PC2400 speeds and 1.65v. I ran it at 1600Mhz and still 1.65 volts. Since this bench fills the ram, I'm sure I was wasting power there. I also think my LLC settings could use some fine tuning. Max temp during the 1.35v water according to coretemp was 38C (but that's not very reliable).

 

[Arachnotronic]

Lol, Idontcare is as much "viral marketing" as the world is flat.

 

[monstercameron]

Lol, Idontcare is as much "viral marketing" as the world is flat.

I knew a guy that fell off the edge of the world once.

 

[SlowSpyder]

Some more interesting results. The LLC settings on my board are a bit wonky. I had it set to auto for the three runs above, changed it to 50% for the new test. That's what I had the best luck with whether I was running 5.3GHz+ beating on it, or running 4.4GHz 1.25v for normal use. I saved quite a bit of power too, just ran the same three tests again.

1.35v 4.0GHz/4.2GHz power went up from idle 189 watts running LinX.

1.30v 4.0GHz/4.2GHz power jumped from idle 166 watts.

1.25 v 4.0GHz/4.2GHz power use went up a more reasonable 147 watts. Also worth noting, Coretemp gave me a max temp of 38C when I ran 1.35v and auto LLC, this test gave me just 21C max. I think there is a lot of room to tinker with the FX chips.

I'm still running my ram at 1.65v, going to drop that down a bit and see if there is a measurable savings.

*edit - Memory at 1.45v vs. 165v didn't do much. Saw the Kill-o-Watt read no more than three watts lower.

*edit 2 - I ran this CPU at 4.4GHz @ 1.225v before, so figured I'd go lower. 1.20v, 4.0GHz, power climbed 130 watts over idle. 2C cooler than 1.25v too.

 

[AtenRa]

Some more interesting results. The LLC settings on my board are a bit wonky. I had it set to auto for the three runs above, changed it to 50% for the new test. That's what I had the best luck with whether I was running 5.3GHz+ beating on it, or running 4.4GHz 1.25v for normal use. I saved quite a bit of power too, just ran the same three tests again.

1.35v 4.0GHz/4.2GHz power went up from idle 189 watts running LinX.

1.30v 4.0GHz/4.2GHz power jumped from idle 166 watts.

1.25 v 4.0GHz/4.2GHz power use went up a more reasonable 147 watts. Also worth noting, Coretemp gave me a max temp of 38C when I ran 1.35v and auto LLC, this test gave me just 21C max. I think there is a lot of room to tinker with the FX chips.

I'm still running my ram at 1.65v, going to drop that down a bit and see if there is a measurable savings.

*edit - Memory at 1.45v vs. 165v didn't do much. Saw the Kill-o-Watt read no more than three watts lower.

*edit 2 - I ran this CPU at 4.4GHz @ 1.225v before, so figured I'd go lower. 1.20v, 4.0GHz, power climbed 130 watts over idle. 2C cooler than 1.25v too.

Nice thanks,

Could you leave Vcore on auto and change the LLC and report the differences ??

 

[FX2000]

Read in this thread : Everyone thinking TDP = Powerdraw.
TDP = Cooling, if a chip has, let's say, a 125tdp, and draws 300watt from the wall, that's fine, because it's not ALL converted into heat.

 

[ShintaiDK]

Read in this thread : Everyone thinking TDP = Powerdraw.
TDP = Cooling, if a chip has, let's say, a 125tdp, and draws 300watt from the wall, that's fine, because it's not ALL converted into heat.

Everything it draws from the wall is converted into heat.

 

[Erenhardt]

Everything it draws from the wall is converted into heat.

That is not true.

 

[ShintaiDK]

That is not true.

How much is not made into heat? The kinetic energy from fans and HDs not displaced in the computer/room plus the radiation from components doesnt exactly make up much to say it mildly.

 

[Phynaz]

That is not true.

Of course it is. Where else does it go?

 

[Torn Mind]

Ever wonder why a PSU needs a fan? Because of some of the electrical energy that goes through the PSU is not converted from AC current to DC current. Or in other words, not all of it remain as electrical energy. What form of energy is it converted to? Heat energy of course.

The DC current that comes out of the PSU goes to the motherboard and eventually the CPU. I don't have intimate knowledge of what inside of the PSU converts electrical energy to heat and what does not, but a lot it is and that is why CPU needs a heatsink and fan.

 

[AtenRa]

Of course it is. Where else does it go?

Wall ---> PSU

The Higher the efficiency of the PSU the less you loose,

PSU -----> Motherboard VRM

The higher the efficiency of the VRM/electrical circuit the less you loose,

So not everything from the wall goes to the CPU

 

[III-V]

So not everything from the wall goes to the CPU

The statement was "everything from the wall gets converted to heat."

And this is the truth. It's very basic physics.

The kinetic energy from fans and HDs not displaced in the computer/room plus the radiation from components doesnt exactly make up much to say it mildly.

Well, the kinetic energy just gets converted to heat in the end anyway.

 

[SlowSpyder]

Nice thanks,

Could you leave Vcore on auto and change the LLC and report the differences ??

Sure, but I have an FX 9370, not an FX 8350. Not sure where it'll put the voltage, but I'll give it a go.


I believe technically, some of the power is turned into some type of radiation, but only a very small percentage. The other 99%+ is converted to heat. Pretty sure it works like that.

Also worth pointing out that I'm using a big power supply that has different efficiencies at the loads I'm testing: http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story3&reid=258

 

[frozentundra123456]

well, i suppose you could argue some is converted to light energy for case lights, etc, and some into kinetic energy for fans, an small amounts of other types of electromagnetic radiation, but i would agree, for all practical purposes, it is basically converted to heat. And even some of the other types of emissions would eventually be converted to heat.

 

[SlowSpyder]

Nice thanks,

Could you leave Vcore on auto and change the LLC and report the differences ??

Sorry, doesn't look like I have an 'Auto' option for power. Just different number values to choose from. When I set everything to auto for my CPU and go into the voltage section, it is set to 1.5v.

What is factory voltage for an FX 8350, I thought it was 1.35v, do you know for sure?

 

[TurtleCrusher]

Sorry, doesn't look like I have an 'Auto' option for power. Just different number values to choose from. When I set everything to auto for my CPU and go into the voltage section, it is set to 1.5v.

What is factory voltage for an FX 8350, I thought it was 1.35v, do you know for sure?

My FX-8350 factory voltage for 4GHZ is 1.4V. I can do 1.35 but any lower I had stability issues. That was with my failing PSU that I replaced, so I may have to run it again a couple times and play with the voltage.

 

[SlowSpyder]

Just for fun, decided to beat on it a bit.

4.8GHz / 5.0GHz turbo @ 1.45v- 286 watt power use. Max temp 42C.

5.0 / 5.1GHz turbo @ 1.5v - 323 watts power use, max temp 52C.

I could run lower voltages possibly, but just guesstimated on past experiences about what it'd need. I only ran for five minutes and watched max values, actual normal power use was probably a bit less as I looked for the absolute lowest and highest value on the Kill-o-Watt. Never saw a turbo frequency. Going to try for higher, looks like temps are ramping up quickly now.

 

[Phynaz]

Holy crap, you have a big set. There's no way I'd crank the voltage up to 1.5!

 

[TurtleCrusher]

Just for fun, decided to beat on it a bit.

4.8GHz / 5.0GHz turbo @ 1.45v- 286 watt power use. Max temp 42C.

5.0 / 5.1GHz turbo @ 1.5v - 323 watts power use, max temp 52C.

I could run lower voltages possibly, but just guesstimated on past experiences about what it'd need. I only ran for five minutes and watched max values, actual normal power use was probably a bit less as I looked for the absolute lowest and highest value on the Kill-o-Watt. Never saw a turbo frequency. Going to try for higher, looks like temps are ramping up quickly now.

It just occurred to me as odd, but your turbo clocks work with an overclock and are possibly editable? When I make a change, turbo gets disabled and I get the clock I put in. So when I put in 4.2Ghz it is my max clock at load, nothing more. Not that I'd really want any higher due to the power draw. Maybe this is a quirk for getting a 970 based board? I've got the Biostar TA-970.