Observations with an FX-8350

[AtenRa]

@AtenRa

I seem to have mistakenly understood that your posted results were HPC on. I see now that you actually said HPC was off. That's rather surprising. What happens when you run those tests and enable HPC mode?

HPC ON,

No difference in power consumption, max Watts at 279W.
Max Temp 62c, ambient at 20c

 

[Stoneburner]

So, Vishera is the perfect CPU for winter.

 

[mrmt]

If those numbers are really correct then we could see whether AMD changed their TDP definition from Thuban to Orochi, or from Orochi to Vishera.

Or maybe they don't consider LinX a commercially useful application...

 

[cytg111]

So, Vishera is the perfect CPU for winter.

- hehe

 

[LurchFrinky]

I don't have any of these powerful cpus, but dissipating 200W with the stock cooler seems suspect to me.
Is the heatsink too hot to touch? Is the fan at max speed and blowing out really hot air?

 

[AtenRa]

I havent been convinced yet that the CPU uses 200W. I will try to investigate this as far as i can.

 

[Puppies04]

I havent been convinced yet that the CPU uses 200W. I will try to investigate this as far as i can.

Same PSU, same killawatt either the mobo is drawing all that extra current or the CPU is. I know where my money is, because asus also make intel mobos and those ones don't seem to magically draw a ton of extra current at load.

But let all all know what your investigation finds.

 

[sefsefsefsef]

Is 100% of the energy consumed by a CPU also going to manifest as thermal energy that needs to be dissipated by a heatsink? Or, are CPUs like lightbulbs with a certain % efficiency?

It seems to me that even if a CPU draws 200W from the wall, the heatsink+fan shouldn't need to dissipate 200W of energy from the CPU socket. My guess is it's more like a 50% efficiency. Does anyone know if this is right?

 

[LOL_Wut_Axel]

OK guys,

FX8350 + ASUS M5A97 R2.0
Default Cooler
2x 4GB Kingston 2133MHz (running at 1866MHz 9-11-9-27)
ASUS HD7950 @ 1GHz core 1500MHz Memory
HDD : 1TB Seagate 7200rpm 64MB cache
PSU : Be-Quite Dark Power Pro 1000W
Win 8 Pro 64bit

HPC disabled

Idle 69W (desktop)

LinX 6890MB 279W

LinX 2048MB 274W

Edit: watch that CPUz is at 4100MHz, Turbo works fine

I will say that IDCs FX8350 is fine.

Sorry if this is off-topic, but I have a question: did you need to update the BIOS of your motherboard for it to recognize your CPU?

I ask because we got the ASUS R2.0 boards at work and I'm kinda hesitant about recommending them to people doing new builds with Vishera because there's a chance they won't POST.

 

[inf64]

Ok to sum up IDC's data:
Load in Linx results in 281W.
Idle is 87W. He has some peripherals attached so they do contribute a bit to both numbers(board contributes the most and ram since they are stressed somewhat during the test).
PSU eff. is 89% according to OP. So we subtract 30W(11%) from 281W and we end up with 251W. Subtract 87W-11%(77W) in idle : 251-77=174W. How much can be the power drawn by motherboard,ram and other stuff connected? According to above,in order for CPU to be in its 125W rating, the other stuff(board,ram,drives,fans,graphics card) have to draw ~50W which is kinda high IMO but possible. I know each stick of memory draws ~4W approx. If OP is running 2 stick then ram constitutes 8W or power draw(for 4 stick it's 16W, approx.) Graphics card can go into low 15W when power saving is engaged. That leaves ~26W for motherboard and drives. I have found out via quick google search that motherboard usually draws from 25 to 40W(I suppose high end boards draw more).

So it's possible that IDC's CPU actually draws 125W or a bit more(130-135W) in this one benchmark such as FMA optimized linpack while the total system power at killawatt is 281W.

 

[Rvenger]

I just ran LinX on my i7 rig below. At it's overclocked state, it pulls 170w from the wall.

At stock speeds it actually has a high default which pulls 161w fully loaded using LinX.

 

[jvroig]

PSU eff. is 89% according to OP. So we subtract 30W(11%) from 281W and we end up with 251W. Subtract 87W-11%(77W) in idle : 251-77=174W. How much can be the power drawn by motherboard,ram and other stuff connected? According to above,in order for CPU to be in its 125W rating, the other stuff(board,ram,drives,fans,graphics card) have to draw ~50W which is kinda high IMO but possible. I know each stick of memory draws ~4W approx. If OP is running 2 stick then ram constitutes 8W or power draw(for 4 stick it's 16W, approx.) Graphics card can go into low 15W when power saving is engaged. That leaves ~26W for motherboard and drives. I have found out via quick google search that motherboard usually draws from 25 to 40W(I suppose high end boards draw more).

I may have misunderstood some of your figures, so let's clarify:
-GPU was also idle during the 89w/77w figure, so that shouldn't count anymore for the 50W figure you are trying to trace. Same goes for harddisks, so they can't be part of it. Both these components are already counted at idle and at load.

-That leaves RAM and mobo only for the 50 watts you are tracing. This is already taking into account the PSU efficiency. I don't see how the RAM+mobo combined can draw an extra 50W while under load ("extra" because while idle they are already using up power, which contributes to the idle power usage) - what mobo components, specifically, will use that kind of power? Again, that is 50w of extra power usage while under load, aside from the "normal" power consumption that happens regardless of loaded state.

 

[inf64]

Yes you are correct,GPU and drives are counted in idle power,my mistake. That leaves 50W in load to be accounted for. For sure the motheboard is drawing somewhat more power when system is loaded Vs when it's idle(electrical components on board are stressed more). RAM is drawing more power since in case of Linx modules are fully stressed. Since RAM draws around ~4W per module it's not a significant contributor. That leaves us more or less with 45-50W for board alone. I have searched and found that older 790FX chipset(alone) was drawing ~10W of power under full load and 2W under idle. Still a lot of difference to make up there.

 

[Idontcare]

I havent seen that you mentioned this before,
The little green circular sticker IS NOT the only thing that seals the box. There mast be a white sticker on the top of the box with the specs of the CPU that seals one side(Back) of the lιd.

Was the white sticker broken on your BOX or not ??

Oh yeah, of course, the white sticker is also present and fully intact. I left it intact to use as a hinge of sorts for the metal lid on the box and only removed the green circle sticker.

I made the remark in another thread but it seems that my thoughts
were not understood.

In short IDC s CPU ID is the one of an OEM SKU not from a regular boxed CPU.

Might well that his CPU was simply exchanged since it s likely
that official Black Editions are binned CPUs that have slightly
better overclock headroom.

I question whether the reported model labeling schema is correct as both AtenRa's (Greece) and my (USA) retail boxed FX-8350's have the same label that your links say means the CPU is a tray/OEM processor.

I will say that IDCs FX8350 is fine.

I appreciate the results and the detail you went into in capturing them. And I concur with your conclusion as well.

A slew of tests today have left me convinced the results are real and basically come entirely down to the CPU's voltage. Tiny changes in voltage when the CPU is at or above 1.35V results in over-sized changes in power consumption and operating temperature. (I'll get back to this)

In the course of testing I found the HPC setting in the BIOS that you mentioned. It was enabled by default for my mobo, but when I disabled HPC the situation remained the same. For me, HPC appears to make no difference on performance or power-consumption.

I also tried other combinations of enabling/disabling. Disabled CnQ for example, no effect on my loaded power usage.

 

[Idontcare]

I'm not sure if this has been mentioned before, but I've heard that Active PFC PSU's make Kill-A-Watt readings inaccurate. Supposedly, connecting the PC to a UPS then the UPS to a Kill-A-Watt is one workaround. If you do this, you'll first want to connect the UPS on it's own to see it's power draw and subtract that from the total.

I'm using the same PSU and the same kill-a-watt for measuring power for all three CPU (2600k, 3770k and fx8350).

To whatever extent your concerns are true, the extent of the error should be present (and nearly the same) in all cases.

That said, I'll do the test as you suggest just to put an absolute number to it.

Lets assume for a second that it's true... That a Kill-A-Watt isn't accurate when an APFC PSU is connected to it. We don't know what that inaccuracy curve looks like. It may not be linear at all. We all know the FX is going to consume more power than SB/IB, that's a given. It is entirely possible that the inaccuracy may increase as load increases, thus compounding the inaccurate readings.

Either way, I'm interested in seeing more data as you harvest it.

Generated the test data today for this as well.

The short of it was that it made absolutely no difference.

==========================
Here are the numbers - first is the arrangement in which the power-train is: Wall Socket <-> Kill-A-Watt <-> UPS <-> PSU <-> Mobo

(the only thing plugged into the UPS is the PSU, all other components are powered by another outlet and power-strip)

With this "UPS before PSU" arrangement here is what I get:

Kill-a-watt (KaW) plugged into wall socket, nothing plugged into KaW = KaW reports 0.0W

UPS powered off, plugged into KaW, nothing plugged into UPS = KaW reports 2.5W

UPS powered on, plugged into KaW, nothing plugged into UPS = KaW reports 3.3W

PSU turned off, plugged into UPS, mobo plugged into PSU = KaW reports 3.3 W (no change in plugging rig into UPS)

PSU turned on, mobo not booted = KaW reports 6.1 W

System booted, fully loaded with LinX = KaW reports 291 W
==========================

When I reverse the order, putting the KaW between the PSU and the UPS the arrangement in the power-train becomes: Wall Socket <-> UPS <-> Kill-A-Watt <-> PSU <-> Mobo

In this case the power consumption of the UPS is transparent to the KaW.

PSU turned off, plugged into KaW, mobo plugged into PSU = KaW reports 0.0 W
PSU turned on, mobo not booted = KaW reports 2.7 W

System booted, fully loaded with LinX = KaW reports 287 W

Basically the only difference in the reported power numbers is the fact that the UPS itself consumes ~3-4W in powering its LED indicators and whatever diagnostic electronics it uses to monitor the battery stats.

 

[notty22]

Seems to be a resistance to OP's observations. Looking at reviews, many were done in ways that are hard to 'apple to apple'. Xbit does a 100% load single threaded and full threaded. They get almost 80+ watts more than the 3770k (8 thread).
A 150 watts above their idle observation for the 8350.

The graphs below (unless specified otherwise) show the full power draw of the computer (without the monitor) measured after the power supply. It is the total power consumption of all the system components. The PSU's efficiency is not taken into account. The CPUs are loaded by running the 64-bit version of LinX 0.6.4-AVX utility. Moreover, we enabled Turbo mode and all power-saving technologies to correctly measure computer's power draw in idle mode: C1E, C6, Enhanced Intel SpeedStep and AMD Cool&#8217;n&#8217;Quiet.

The situation doesn&#8217;t get any better when it comes to AMD FX power consumption under multi-threaded load either. FX-8350 turns out the most power-hungry CPU in our today&#8217;s test session. It consumes 4% more power than FX-8150 under maximum load and is 2/3 more power-demanding than Intel Core i7-3770K.

Had we also tested the old Core i5 processors manufactured with 32 nm process, then the new dual-module quad-core AMD processor would have been comparable against them in power consumption. Although in terms of performance this comparison makes absolutely no sense. In other words, it looks like contemporary AMD processors are hopelessly behind their Intel competitors in terms of performance-per-watt. While in performance tests top eight-core Vishera processors can sometimes compete successfully against Intel&#8217;s quad-core Ivy Bridge CPUs, all illusions vanish into thin air the minute it gets down to power consumption.

The tests by the OP , are not abnormalities, imo.

 

[Idontcare]

I don't have any of these powerful cpus, but dissipating 200W with the stock cooler seems suspect to me.
Is the heatsink too hot to touch? Is the fan at max speed and blowing out really hot air?

It does get hot, both the stock HSF and the air coming out from it.

If those numbers are really correct then we could see whether AMD changed their TDP definition from Thuban to Orochi, or from Orochi to Vishera.

Or maybe they don't consider LinX a commercially useful application...

The fact that "HPC mode" exists for bulldozer/piledriver chips suggests one, or both, of those proposals are correct.

Also consider what DELL has to say about the "HPC mode": (emphasis added)

HPC (High Performance Computing) mode is a new feature introduced in the BIOS which improves the performance of certain workloads on Dell servers based on AMD Interlagos processors. This blog describes how to enable and take advantage of the HPC mode and includes some performance results of the impact of HPC mode on a PowerEdge R815 server.

<snip>

From Table 4, it can be seen that HPC mode provides up to 6% improvement in HPL efficiency. This increased performance is at the expense of higher power consumption and is recommended only for those environments where the power available can support this mode of operation. Another caveat to be noted is that the performance improvement provided by HPC mode for workloads other than HPL is minimal.

Basically "HPC mode" was engineered to improve linpack (HPL) performance, at the expense of elevating power consumption, and it is not expected to improve performane on non-HPL code.

 

[Idontcare]

I havent been convinced yet that the CPU uses 200W. I will try to investigate this as far as i can.

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.

 

[grimpr]

IDC, can you use Maxwell render and the integrated benchmark and not Cinebench? It better utilizes FXs floating point units and integer cores compared to Cinebench that is clearly Intel optimized.

http://www.maxwellrender.com/index.php/benchwell

 

[Idontcare]

IDC, can you use Maxwell render and the integrated benchmark and not Cinebench? It better utilizes FXs floating point units and integer cores compared to Cinebench that is clearly Intel optimized.

http://www.maxwellrender.com/index.php/benchwell

Sure, will do :thumbsup: Thanks for the suggestion and the link!

 

[Abwx]

I question whether the reported model labeling schema is correct as both AtenRa's (Greece) and my (USA) retail boxed FX-8350's have the same label that your links say means the CPU is a tray/OEM processor.

Well , seems that the CPU ID is "enhanced" in the box serial number
with the addition of "box"....

Quite confusing , this say that there s no substancial binning ,
even possible that 8350s are not even cherry picked , the 8320
being the same , just marked according to market demand.

 

[Idontcare]

Well , seems that the CPU ID is "enhanced" in the box serial number
with the addition of "box"....

Quite confusing , this say that there s no substancial binning ,
even possible that 8350s are not even cherry picked , the 8320
being the same , just marked according to market demand.

With any luck my chip will fall under the old adage "a leakier chip is a chip that can OC the best".

 

[2is]

HPC ON,

No difference in power consumption, max Watts at 279W.
Max Temp 62c, ambient at 20c

Just for kicks, I ran Linx on my sig machine using the same settings here. 95watt difference between idle and load values. That's with an OC to 4.2GHz

 

[Idontcare]

I don't have any of these powerful cpus, but dissipating 200W with the stock cooler seems suspect to me.
Is the heatsink too hot to touch? Is the fan at max speed and blowing out really hot air?

For some proof I took a couple of pictures with my IR gun.

Notice the bright red dot straight ahead of the yellow IR gun, that red dot is a laser beam that shows where the temperature reading is taken from. In this case I pointed it at the backside of my GTX460, where the GPU itself is located, showing a nice cool 34.2°C.

Next I turned the IR gun to one of the heatpipes on the FX-8350 HSF when the reported CPU temperature was 58°C:

^ the exterior of the HSF heatpipe was nearly 51°C.

Next I measured the temperature of an area of the heatsink fins:

A nice roasty 51°C.

 

[Idontcare]

IDC, can you use Maxwell render and the integrated benchmark and not Cinebench? It better utilizes FXs floating point units and integer cores compared to Cinebench that is clearly Intel optimized.

http://www.maxwellrender.com/index.php/benchwell

At 4GHz with DDR3-1866 10-10-10-28-T2 I scored 301.45 in 15m26s.

(scroll down in the link to the 15m26s times and you see my score highlighted in the table)

edit: adding some measured values from the run - peak operating temperature was 57°C (19°C ambient), and power usage at the wall was 240W (versus 290W for LinX).