FEATURED Info & Tips >> Using the AMD FX Bulldozer/Piledriver...

[RGone]

reserved

Crap one did reserve but how in hale it allowed such is not known by me.

RGone...

 

[Blaylock]

I think Joe knows this new format pretty well. Maybe shoot him a PM. Change can definitely be frustrating.

 

[RGone]

FYI...

...Something about the motherboards for FX processor has been updated.

RGone...ster.

 

[trents]

I'm a big fan of Frostytech cooler rankings because of the way they test that eliminates a lot of variables. They test the raw ability of a cooler to remove heat and the amount of heat is precisely controlled from one test to another. They don't even use real CPUs. They use a hot plate to produce the heat so it is very controllable and very consistent. I find that their rankings pretty much coincide with my own experience with various cooling solutions.

 

[Mandrake4565]

I was looking at the Cpu Power consumption chart when overclocked. I did a when I saw that .1 v added to the 965 was 40 watts more power consumption.

 

[RGone]

More FYI..

...If you look at the idle power consumption of the full system and 'then' look at what happens when the cpu is called on to do work...well no wonder 8 core FX processors eat weak VRM circuits.


AMD FX-9590 and FX-9370 Review: Socket AM3+ Platform's Swan-song. Page 7

RGone...

 

[bassnut]

300+watts at full load....... crappers man

 

[Mandrake4565]

300+watts at full load....... crappers man

You didn't know that Bassnut? Here is mine at 4.7 with 1.46875 V which is my daily Oc, Running prime.

 

[caddi daddi]

you mean you have'nt noticed we use our fx machines as bathromm heaters in the winter?
by the way, great and very useful thread Rgone.

 

[bassnut]

Yes I knew they drew power and have seen your pic before Manny though I have always thought of that as a complete system / rig power draw. I guess seeing it the way the author put it brought it to a different perspective.

 

[Mandrake4565]

Here another pic at idle with all the power saving features on. I had all the fans at 100% during the test so there isn't any additional wattage drawn from them when the temps climbed. So at the least, when putting a load on the Cpu it's drawing 220 w more for the system. How much of the 380w in total is actually draw from the Cpu I'm not sure. Doing a little calculating, the Bitfenix Spectre Pro fans are .2 A at 100% each so I'm figuring about 30w total for the 12 fans at 100%. I did this test for the Cpu at 4.3, 4.5 and 4.7 the only requirement was to be able to pass 20 minutes of prime blend. Though, at 4.7 it is my known 2 hour stable Oc. Looking back at the test it seems that for every .1 increase in Cpu V the Fx 8xxx will draw about 60 more watts.

 

[RGone]

Yes I knew they drew power and have seen your pic before Manny though I have always thought of that as a complete system / rig power draw. I guess seeing it the way the author put it brought it to a different perspective.

Seeing it as the author put it does and should bring things into much clearer focus. As humans we often want things that look like do A and B and C is the result but that is not easily done in testing, for current to the cpu only, as we imagine it in our minds.

But taken like this it is the reason why after such a long time I finally had it clear enough in mind to SEE what is happening. If I measure load at 4.3GHz be it system load or measured on the 8 pin to cpu as have seen a few do the test...the load is the load. Now if I jump to 4.5GHz and measure a load well what caused the load to increase. Right the load caused by the cpu speeding up because that is what I changed. This based on simply moving the multplier by a number to change the cpu speed. So Ram speed did not go up. Video cards did not suddenly jump into another gear. Drives did not catch another gear upward, so the increased load is the result of the cpu speeding up and can be attributed to that fact alone when we keep the playing field level as such.

Now we can say well what about the VRMs and them having to do more to supply power to the cpu and if that cpu speed increases and more power is drawn, well what about that? No matter. For our purposes of considering load on VRMs and knowing that overclocked FX processors draw a ton of power...the added load of VRM working to keep up with an overclocked FX processor is still a function of speed needing voltage, needing current to maintain power within cpu to do the work. So power to get the job done is still power to do the job and as noted is pretty darn high for overclocked FX 8 core processors and the total power is still the total power needed and must be furnished by the motherboard VRMs. And the total power is inclusive with any extra work the VRMs must do as seen in the increased demand at the watt meter.

I hope my babbling did not muddy the water. Just trying to emphasize that increase in watt meter indicated draw is really showing that the power needed to do X speed with an FX processor is truly great and rises rather quickly with increase in speed and also with voltage.

RGone...

 

[Blaylock]

It didn't muddy the water at all.

Something I'm curious about as well is if a CPU pulls "X" wattage at a set OC on a CH-V will "X" increase on a Sabertooth? What about an even lesser board like a M5A97? I think a test like this could show how much harder the VRM's have to actually work. This is of course assuming that "X" does actually increase on lesser boards. I haven't seen any testing of this sort.

 

[Johan45]

I would think if the clocks were identical and once you've gotten to the point where the VRM are getting hot on the lesser board then yes it would have to pull more power. IIRC when the circuits get hot first the efficiency is compromised and the hotter it get the more that happens to the point where the circuit is useless because of the heat. I don't know specific numbers but as the temp goes up the delivery goes down. That's why they start to throttle. Say at 80c the circuit is only 75% effective so what started as 16A delivery is now only capable of 12A because of the heat. So the circuit needs to pull more power to maintain a certain delivery level.

How's that for muddy.

 

[t1nm4n]

X will increase on a lesser motherboard after a certain point. X being heat, and heat being the factor that increases resistance for energy to pass through adding to the never ending cycle of heat build up to something (I can guess what it is) Lessor motherboards have a few things going against them from the start, No added cooling (via heatsinks on VRMs) and in most cases fewer Voltage regulation Modules to cut the load each one has to work at. The interesting part is where do lessor boards really start to fall off on power delivery resulting in poor performance overclocks and the higher end MBs start to shine, is it the point at which the crazy ones start to dip there boards in water proof poly(something-a-rather) and use LN2 or can it be noticeable on Air level OC's?

My apologies if that was were you were going Blaylock, or if that was the point of the entire discussion from the past few days.

Edit: update.

This is what I get for not reading the next post saying the same thing Johan says and not noticing.

 

[Johan45]

ShrimpBrine has take a n 8 core FX well over 7G on a 4 phase board with LN2. The cold reduces the power requirements enough that the board can still keep up. Just not practical for everyday use.

 

[Mandrake4565]

This is what I get for not reading the next post saying the same thing Johan says and not noticing.

Be very careful of that Johan character t1nm4n he is a sniper

ShrimpBrine has take a n 8 core FX well over 7G on a 4 phase board with LN2. The cold reduces the power requirements enough that the board can still keep up. Just not practical for everyday use.

Shrimpy might debate you on that one.

 

[RGone]

I think it is...

...is whoa RCBoyz. Take an FX-8350 and clock it to 4.5GHz and the power it needs to run is exactly that...how much power it needs to run. That cpu does not know what motherboard socket it is placed into. It will "need/want" that much power no matter the board be it cheap or expensive and capable or incapable.

The onus is on the VRMs to keep up. If the VRMs cannot keep up then we usually see the results as failing P95 Blend first since we all test that way. We fail P95 so we DO something. Others come to the forum and say >> shett shett my stuff is not stable or some tale of woe. Hehehe.

The best and easiest example I can give is to attach a 6V light bulb to a 6V battery and experience great light. Then take a 12V light bulb and hook it to that same 6V battery. Shett not so much light. Well that is how it would be if the VRMs cannot keep up...poor results. You cannot supply a 12V need with only 6Volts, the results are poor if you were in need of good light.

I need to stop there because it only gets stickier and muddier the further you go and sort of pointless. The cpu called NEED is being fed by the FEED and if the FEED struggles to keep up the NEED stays the same but the FEED gets hotter and thus has to work harder and because FEED cannot keep up, poor NEED falls into disarray and quits doing 1s and 0s accurately.

RGone...ster of the 'muddy waters' tribe.

 

[Blaylock]

My apologies if that was were you were going Blaylock...

I was referring to the wattage draw at the wall being "X" and determining if a lesser board would show in increase at the wall due to in-efficiency. Increased wattage would naturally create increased thermals as a side effect though.

 

[bassnut]

Johan and tinman you are both right about how the efficiency of a component or circuit work just as RGONE in his explanation is as well.

From looking at the author chart it looks like he is comparing the 8350 to the new 9xXx chips at stock clocks. Now I am going on memory so feel free to correct me if wrong. What really surprised me was the difference between the 8350 at something like 120 - 140 watts compared to the 9590 which was well over 300 watts. This comparison all one should bring sense to what shrimpy has been trying to say as well as how many issues we have seen lately with said chips even on good boards. Maybe better said then my first statement on this thread this am.