Oh So Efficient: FSP Announces 400 W 80 Plus Titanium PSU

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
i personally feel like the need to lose the fan, i mean at worst this PSU will create 40w of heat it should be easy to went passively. and for the price point you could really appreciate the minimum noise level of a fan less psu.
 
i personally feel like the need to lose the fan, i mean at worst this PSU will create 40w of heat it should be easy to went passively. and for the price point you could really appreciate the minimum noise level of a fan less psu.
Click to expand...
I have a feeling it shuts the fan off at low loads. I hope so
 
Going from not having any efficiency requirements below 20% load with Platinum to having to hold 90+% all the way down to 10% load with Titanium is a fairly big deal.
Click to expand...
I don't feel it is as big of a deal as you're making it out to be at very low loads - especially if you spec your PSU properly. 10% load on a good quality 80 Plus Gold/Plat unit getting 80-85% efficiency vs Titanium at 90%. For a low-power system drawing 40W or less? You're talking about a handful of watts difference.

Firefoxx says they saved 13-18W at idle. That's a LOT at such low draw. Much more than the "leap" you'd get going from Gold/Plat to Titanium. So either their old Raidmax was VERY inefficient or was higher wattage than the replacement unit (or both).
The big catch is that a ~40W server is well below the 20% mark for most PSUs and nothing below 80+ Titanium (the highest standard) is required to meet any particular efficiency goal below 20% load. At 10% even for a 400W PSU, you are already well into la-la-land on anything short of a Titanium-class PSU.

While some non-Titanium PSUs fare better than others at 10%, the results are rarely flattering. The reason for that is that internal circuit operation ends up accounting for a large chunk of total power at very light load unless the designer goes through considerable extra trouble to specifically improve low-load efficiency which most people pay little to no attention to.
Click to expand...
La-la land? Not necessarily. As you yourself pointed out, many high-efficiency Gold+ units already have decent efficiency at those ranges. They're not REQUIRED to, but many do. Proper wattage selection is at least as important as what 80 Plus cert the unit meets.

Of course, doing both wouldn't hurt, if you can find such a thing. I agree that there aren't a lot of good quality 80 Plus Bronze and above units in the sub-300W market, and that can be tough for low-power systems.
 
I'd like to see a laptop AC adapter that is 80 Plus Titanium rated so they could run cooler.
Laptop AC adapters use the 'efficiency level' rating. The latest is V or 5. It's around 85% efficiency I think.
 


You clearly misunderstood. I meant going from Platinum to Titanium must not have been much difference in cost for FSP, hence why they went ahead and made a Titanium unit.
 
I like to see higher efficiency power supplies. Maybe I'm delusionnal, but in my mind higher efficiency means higher quality and longer lasting.

Fanless designs are also great, no moving parts = better. Espescially in a high dust environment. But if not possible then a large fan is better than a small one.

400W is more than enough for most modern gaming systems. Unless you have more than one graphic card. Systems with the 980 GTX or even the 290X are bellow 400W on peak load power consumption.
 

With Platinum, there is still plenty of slack at loads below 20% to afford the losses of simple and inexpensive RCD snubbers and simple gate drive circuitry. With Titanium, you need to use more complex and expensive "loss-less" snubbers (can't afford wasting 3-5W on snubber resistors and diodes) and have much more precise proportional gate drive electronics. Getting any of those parameters slightly wrong can quickly add 5-10W of extra losses.

You probably cannot tweak a Platinum design into a Titanium one unless your former Platinum designs were either intended to go beyond Platinum in the first place or were failed attempts at Titanium because achieving Titanium requires throwing everything including the kitchen sink at the design.
 
"PSUs need to manage 90 percent efficiency at 10 percent load, 94 percent at 50 percent, and 90 again at 10 percent."

"PSUs need to manage 90 percent efficiency at 10 percent load, 94 percent at 50 percent, and 90 again at **100** percent."

Who cares what efficiency is at 10% load ~ 40 watts .... Let's say 5% difference between Titanium and Gold.

40 watts / 90% = 44.3 watts
40 watts / 85% = 47.0 watts

(47.0 - 44.3) = (2.7 / 1000 ) kw x 24 hours / day x 365 days / yr x $0.10 per kw-hr = $2.37 per year

A 4 bay NAS generally pulls 40 - 45 watts.

I have little concern about efficiency or whether a PSU is Gold Bronze or Titanium .... not at $2 a year. What I am most concerned about is can it maintain voltage stability and provide clean ripple free power throughout it's service range

This 400 watt Seasonic unit for example maintains a worse case 1.67% voltage variance at 5% of load, miniscule ripple and near perfect power factor. These criteria, not efficiency, are always my main concerns. Without this information, the article is of little value outside of a press release.


 


There is more to it than power cost. A more efficient supply will also create less heat,
with a variable fan controller, less noise. And in theory, last longer.
 
Yes.... it creates 2.7 more watts of heat..... that motor for that 120 fan on the FSP, (the referenced Seasonic unit doesn't have one) will create about 1.7 watts of heat.

As for the noise, note that the Seasonic PSU has no fan and will therefore undoubtedly be quieter than the Titanium unit.

http://www.silentpcreview.com/article1062-page5.html

With less parts and no moving parts (no fan or fan controller) the Seasonic will, in theory and practice, last longer.
 


I disagree that 400w is enough for a 290X system. My Core i5 and 280X require more than that. While it may run at 400w, if you utilize enough of your CPU and GPU simultaneously you will crash your system using a 400w PSU. The card alone can draw about 300w if taxed and a processor can do over 100w on it's own. If you count in memory, motherboard, USB devices, a hard drive or two, system fans and lights, etc you will need more in the range of 600+ watts. I had to check and double check to see if my 650w PSU would run my 280X. It does, but not with a whole lot of margin.

 


There's more to a system then the GPU. Here's nVidia's recommendations:

http://www.tomshardware.com/answers/id-2311121/power-supply-requirements-nvidia-gpus.html#14243229

http://www.guru3d.com/articles_pages/msi_geforce_gtx_970_gaming_review,7.html

Here is Guru3D's power supply recommendation:

GeForce GTX 970 or 980 - On your average system the card requires you to have a 500 Watt power supply unit.
GeForce GTX 970 or 980 in 2-way SLI - On your average system the cards require you to have an 800 Watt power supply unit as minimum.

If you are going to overclock your GPU or processor, then we do recommend you purchase something with some more stamina.
Click to expand...

This is Guru3D's generic power supply recommendation for the R7 and R9 series:

AMD R7 260X - On your average system the card requires you to have a 450 Watt power supply unit.
AMD R7 260X Crossfire - On your average system the cards require you to have a 650 Watt power supply unit as minimum.

AMD R9 270X - On your average system the card requires you to have a 500 Watt power supply unit.
AMD R9 270X Crossfire - On your average system the cards require you to have a 700 Watt power supply unit as minimum.

AMD R9 280X - On your average system the card requires you to have a 550 Watt power supply unit.
AMD R9 280X Crossfire - On your average system the cards require you to have a 750 Watt power supply unit as minimum.

If you are going to overclock GPU or processor, then we do recommend you purchase something with some more stamina.
Click to expand...

While it's true in a general sense that better quality electronics is more efficient, this is not an absolute. Same with fan size... all things being equal, larger will mean lower rpm and therefore quieter but blade design, rpm and other factors are rarely equal so again, while true in a general sense, not an absolute.
 

The only way to push a 290X over 300W is FurMark.

Here are Anandtech's numbers for a 290X + i7-4960X overclocked to 4.2GHz with 4x8GB RAM:
http://www.anandtech.com/show/7457/the-radeon-r9-290x-review/19

375W total system power measured at the wall while running FurMark. That would include absolutely every component inside the case including PSU losses and port-powered external devices. With the AX1200 being ~93% efficient at 300-400W, this means actual output is around 350W. So, in principle, there should be about 50W to spare.
 
Furmark is designed to test your GPU .... it puts a very minimal load on the CPU....I usually observe about 40 watts of CPU power draw when my system is pulling 750+ watts from the wall running Furmark...... in an X79/X99 system, the CPU can draw 230 watts on its own.

http://www.guru3d.com/articles_pages/msi_radeon_r9_290x_lightning_review,11.html

Subjective obtained GPU power consumption = ~ 287 Watts

Here is Guru3D's power supply recommendation:

AMD R9-290X - On your average system the card requires you to have a 550~600 Watt power supply unit.
AMD R9-290X Crossfire - On your average system the cards require you to have a 800 Watt power supply unit as minimum.

If you are going to overclock GPU or processor, then we do recommend you purchase something with some more stamina.
Click to expand...
''

287 watts stock will have to well over 300 overclocked....add in 230 watts for the 3960x / 4960x .... 10 watts for HS, 10 for RAM ... you can easily pull close to 600 watts from the wall

Also, you must discount capacitor aging ... so take off 15% for even the best PSUs. And finally, you never want to be operating near the rated limit of a PSU. For one, it's at its lowest efficiency point and at it's loudest. And let's not forget voltage stability and ripple are at their peak the closer you get to the rated capacity.

 
Exactly right. You need to look at the max power on each rail. Once you max out any of the rails, that's it. Doesn't matter how much capacity is remaining on the other rails. And if you ever see the full rated capacity of a power supply being drawn at the wall, you're almost certainly overloading at least one of the rails.

I've seen several web apps to calculate power supply wattage requirements, based on system spec. For a while, the one on newegg had a bug for newer Radeon cards. The recommendations were way overkill.

Using better quality power supplies usually translates into better system stability & longer component life. Cheap PSUs are a bad place to cut corners.
 

I think you have the wrong CPU in mind since the i7-4960X and i7-5960X are only rated for 130W and 140W respectively with actual power draw closer to 90W at stock clock and voltage. The AMD FX9xxx (or massively OC'd FX83xx) are the ones hitting 230+W.

If you see numbers over 200W associated with Intel's chips, you are most likely looking at total system power measured at the wall; not CPU-only numbers.
 
My post was very clearly not about running at stock settings.... it's about "If you are going to overclock GPU or processor". So ya gonna need to do some research and learn a little about the effect over overclocking on power usage.....TDP rating at stock speed is irrelevant with an overclocked box. The power draw to a CPU when overclocked can be double that of stock settings.

http://extreme.outervision.com/psucalculatorlite.jsp

Select:

CPU Brand = Intel
Socket = LGA 2011
CPU = 396x
Overclock MY CPU = check
Overclock CPU speed = 3300
Overclock CPU voltage = 1.2
Press the Overclock Button and we see that we get the exact Intel stated TDP of 130 watts at 0% overclock, thereby proving its accuracy.

now lets redo the calculation using the settings here:

http://www.guru3d.com/articles_pages/core_i7_3960x_processor_amp_msi_x79a_gd65_review,11.html

Overclock CPU speed = 5000
Overclock CPU voltage = 1.5
Press the Overclock Button and we see that we get 308 watts

I used a more modest overclock of

Overclock CPU speed = 4600
Overclock CPU voltage = 1.35
Press the Overclock Button and we see that we get 229 watts which I rounded to 230.

Want more proof ?

http://www.guru3d.com/articles_pages/core_i7_3960x_processor_amp_msi_x79a_gd65_review,10.html

Measured Power Consumption at wall for 3960x @ stock settings under full load = 201 watts w/ 130 watt TDP
Measured Power Consumption at wall for 3960x @ 4.8 Ghz overclock under full load = 401 watts

Added wattage from overclock = 401 - 201 = 200 watts x 90% PSU efficiency = 180 watts

180 added overclocked watts + 130 TDP = 310 watts

lets go back and put those numbers into the calculator again

Overclock CPU speed = 5000
Overclock CPU voltage = 1.50
Press the Overclock Button and we see that we get 317 watts pretty darn close to that 310

Wanna try the 4960k ? Let's start here:

http://www.guru3d.com/articles-pages/core-i7-4960x-processor-review,9.html

Leave baseclock for what it is right now
If optional in the BIOS, increase your TDP limits of the processor to 250 Watts (by that you are allowing a higher power draw)
Leave your base multiplier at default e.g. 34
Set the per core Turbo multiplier at a maximum of your liking, we applied an MP of 50 on all six cores
Increase CPU voltage, though setting AUTO might work fine, we applied 1.5V on the processor cores
Click to expand...

The fact that the BIOS allows you to even set the TDP to 250 watts refutes your claim all by itself. Let's read further .....

index.php


356 overclocked - 212 at stock = 144 watts x 90% PSU efficiency = 130 watts of PSU output

So the OC alone **adds** 130 watts and that is on top of the stock load ... 130 for OC + 130 TDP stock = 260 watts

Let's go plug those numbers into the calculator

Overclock CPU speed = 5000
Overclock CPU voltage = 1.50
Press the Overclock Button and we see that we get 260 a dead match for the measured results.


Lets look at 5960x

http://www.guru3d.com/articles_pages/core_i7_5960x_5930k_and_5820k_processor_review,19.html

If optional in the BIOS, increase the TDP limits of the processor to 250 Watts (by that you are allowing a higher power draw), most motherboards have this pre-configured pretty okay these days though.
Increase CPU voltage; though setting AUTO might work fine, we applied 1.3V on the processor to reach 4100 MHz and needed 1.4 Volts to reach ~ 4.5 GHz
The 'Load Overclocked' was at ~4500 MHz which requires 1.425 Volts on the eight CPU cores.
Click to expand...

index.php



Again we see that the system load increased from 189 watts @ stock settings to 378 watts overclocked.

378 - 189 = 189 .... overclocking DOUBLED the power draw

189 x 90% = 170 watts

170 from OC + 140 TDP = 310 watts

Again,lets go to the calculator resetting for LGA 2011-3 and 5960k

Overclock CPU speed = 4800
Overclock CPU voltage = 1.425
Press the Overclock Button and we see that we get 298 watts pretty darn close to that 310.

So now I definitely do not don't have the wrong CPUs in mind, I have measured this behavior myself as my desktop is plugged in to a wall power meter 24/7. The data provided by the calculator is confirmed by actual measured test results in the published reviews. The suggestion that CPUs run below their published "stock TDP" is erroneous. In summary:

-My stated 230 watt load for the X79 / X99 CPUs if right on target with a "modest OC" of 4.6 Ghz at 1.35 volts

-Overclocked 3960x (5.0 Ghz) results ran 317 calculated and 310 based upon physical measurements, 238% of the 130 watt stock TDP

-Overclocked 4960x (5.0 Ghz) results ran 260 calculated and 260 based upon physical measurements, 200% of the 130 watt stock TDP.

-Overclocked 5960x (4.5 Ghz) results ran 310 calculated and 298 based upon physical measurements, 213% of the 140 watt stock TDP.



 

Notice how practically every article you quoted mentions TOTAL system power? So, on top of PSU losses, you also have motherboard losses, DIMMs, the CPU's VRM, the GPU, HDD/SSDs and whatever else is in the test system. Those can throw out your "CPU power measurements" off by a wide margin.

I would not give so much credit to psucalculator since the "accuracy" stems simply from their model being calibrated to match Intel's spec at stock settings. This does not prove anything about real-world results, stock or OC'd, where each chip is different and few of them will be anywhere near as bad as the spec sheet allows them to be. Your "near perfect match" is purely coincidental since it fails to take into consideration all of the factors above.

While CMOS devices' power draw is fairly predictable, you cannot apply it to total system power measured at the wall like you attempted to do with your numbers. Here, you are comparing the hypothetical power draw of overclocked CPUs to power of whole overclocked systems, which is completely silly.
 
Please read everything before responding:

1. The calculated numbers are CPU ONLY..... so that in and of itself refutes your assumption.

2. The measured numbers are AFTER deducting system power and are confirmed as they almost EXACTLY match the calculated results. So please explain that "coincidence".

3. The measured results already accounted for the things your list:

Let me take you thru it slowly.

a) We have 189 watts system load recorded for the stock settings
b) We have 378 watts system load recorded for the overclocked system.

The difference between that is 189 watts. Accounting for PSU efficiency means that the PSU is outputting an extra 170 watts over and above the TDP at stock settings which we apparently agree is 140. Do the math. 140 + 170 = 310

So please tell me how they measured 170 watts EXTRA from the CPU alone over and above the full load draw at stock settings w/o breaking the 140 watt TDP ?

Did they add RAM or increase RAM load when overclocking ?
Did they add VRM to the MoBo or significantly increase VRM load when overclocking ?
Did they add another MoBo when overclocking ?
Did they add a SSD or HD when overclocking ?

No, minimal, no and no. So no, their presence throws out nothing...their presence is immaterial.

Ever see what happens at a weigh station at a scrap yard ? The measure the weight of truck going in.... they measure the weight of the truck going out and subrract the two. Your argument is that we don't know what the load weighed because both measurements included the truck.

We know the load of the system overclocked when subject to a CPU only load
We know the load of the system stock when subject to no load (idle)

The difference between the two tells is how many more watts the CPU consumes over idle.... very plain, very easy, very simple.

Again, doing this in simple steps.....

c) Adjusting for PSU efficiency (90%) , we have 170 watts at stock settings under load, we have 69 watts at idle and 378 overclocked. The system uses a 1200AXi a platinum PSU

Above, power consumptions in respect to other processors and platforms. In an IDLE state the PC (X99 / 5960X / 16GB DDR4 memory / GeForce GTX 780 Ti / SSD / LCS) consumes roughly 70 Watts. Mind you, we measure the ENTIRE PC, not just the processor's power consumption.

When we place load on the CPU and we see the power draw rise, the system now consumes roughly 190 Watts.
Click to expand...

So it's clear ......

Entire Stock System Load Stock = 69 watts
Entire Stock System Load Loaded = 189 watts
Entire Overclocked System Loaded = 378 watts

And let's assume CPU at 10% of TDP at idle = 13 watts

So, let's do the math:

The increase in SYSTEM wattage caused simply by loading the CPU @ stock settings = 189 watts - 69 watts = 120 watts.

-As the CPU benchmark places no additional load on the RAM, so 0 watts can be attributed to RAM.
-As the CPU benchmark places no additional load on the GPU, so 0 watts can be attributed to GPU.
-As the CPU benchmark places no additional load on the HS/ SSDs/Opticals (which aren't installed but would be irrelevant of they were) , so 0 watts can be attributed to storage.

Since the Max Wattage consumed by the entire MoBo when Overclocked is 40 watts .... lets assume 5 of that comes from an increase under load and another 5 watts VRM load from overclocking.

So again ... taking the system from idle to load @ stock settings

The increase in SYSTEM wattage caused simply by loading the CPU @ stock settings = 189 watts - 69 watts = 120 watts. From that we have to subtract

0 watts for RAM, same RAM and same RAM load both before and after
0 watts for GPU, same GPU and same GPU load both before and after
0 watts for storage, same storage and same storage load both before and after
5 watts for MoBo, minimal load increase to VRM from before and after

And we have to add 13 watts for the idle power consumption of the CPU.

So 120 - 0 - 0 - 0 -5 + 13 = 128 watts attributable to the CPU and the CPU alone.

==========================================================

So again ... taking the system from idle to load @ overclocked settings

The increase in SYSTEM wattage caused overclocking the CPU to 4.5 Ghz = 378 watts - 69 watts = 309 watts. From that we have to subtract

0 watts for RAM, same RAM and same RAM load both before and after
0 watts for GPU, same GPU and same GPU load both before and after
0 watts for storage, same storage and same storage load both before and after
10 watts for MoBo, minimal load increase to VRM from before and after

And we have to add 13 watts for the idle power consumption of the CPU.

So 309 - 0 - 0 - 0 - 10 + 13 = 312 watts attributable to the CPU and CPU alone.

or to put it another way, what we have done is measure the total system wattage overclocked minus the total system wattage at idle which is equal to (o = overclocked / i = idle)

The measured difference in System wattage =
(CPUo - CPUi) + (RAMo - RAMi) + (STORo - STORi) + (GPUo - GPUi) + (MOBOo - MOBOi) = 309 watts

Since:

(RAMo - RAMi) = 0
(STORo - STORi) = 0
(GPUo - GPUi) = 0
(MOBOo - MOBOi) = 10 watts
CPUi = 13 watts

Plugging in the numbers
(CPUo -13) + 0 + 0 + 0 + 10 = 309 watts

CPUo = 309 + 13 - 10 = 312 watts at the wall

312 watts * 90% = 281 watts

Now lets see what the calculator says is the actual CPU ONLY power consumption @ 4500 and 1.425 colts
http://extreme.outervision.com/psucalculatorlite.jsp

298 watts .....

Read the calculator page ... that 298 watts is CPU only no PSU adjustments no MoBo, no Storage no nothing JUST the CPU.

You also didn't address the fact that in order to overclock ALL those CPus, you have to go into the BIOS and raise the power to the CPU socket to at least 250 TDP ? Why is that needed if the CPU maxes out at 140 watts ?


Want more ?

http://rog.asus.com/365052014/overclocking/rog-overclocking-guide-core-for-5960x-5930k-5820k/

PSU Requirements

For overclocking 5960X processors, we recommend PSUs that can supply a minimum of 30 amps to EPS 12V. At 4.6GHz a 5960X can draw close to 25amps from the EPS12V connector under software load. Minimum recommended PSUs for Haswell-E are upwards of 1,000W if using more than one high performance GPU.
Click to expand...

The EPS Power connector feeds the CPU and ONLY the CPU .... again let's do the math....

12v x 25 amps = 300 watts ....

Looks like Asus agrees pretty much with the 298 from the calculator
Looks like Asus agrees pretty much with the 281 from Guru3D's measurements

Another source ... from 60 watts system power at idle .... to 349 watts system watts overclocked under load. That's 289 watts

power_psu_idle.png

power_psu_load.png


Once you're pushing 4.7 GHz in lightly-threaded apps (and 4.5 GHz in workloads that tax the whole CPU), system power exceeds 300 W. That's not even taking into account graphics power, since only our processor is subjected to a load.

Since "only the processor is subjected to a load" the only thing the difference in power usage can be attributed to is the processor.
Click to expand...

To go back to the weigh station analogy .... I measure the weight of your pickup truck with you in it and 5 wooden boxes.... one of which is marked CPU and just has the packing foam and stuff in it and the car weighs 3200 pounds with everything in it.... I know the empty box weight 10 pounds. Then I put the items in that box and now the car weighs 3480 pounds. You're saying I can't know how much the stuff in the CPU box weighs with all the stuff in it .... I can ... it's 3480 for truck after - 3200 for truck before + 10 pounds for the empty box or 290 pounds. I know that cause nobody put anything in the other boxes.

To suggest that you can add measure 289 watts difference at the wall while staying below a 130 - 140 watt TDP by loading ONLY the processor is beyond ridiculous.

-The wall results are consistently well above 230 even at moderate overclocks for CPU only
-The calculator is consistently above 230 watts for CPU only
-The reviews are telling us that you MUST raise the CPU TDP to 250 watts which is above 230 watts
-Asus is telling us that the cable feeding ONLY the 5960x socket can draw 300 watts... that's bigger than 230 watts

-You can't argue with measured at the wall results....their actual measured results.
-You can't make believe that overclocking a CPU somehow increased consumption of other components that are not placed under any additional load.
-You can't ague that you can in any way attempt overclocking one of those CPUs to those levels w/o changing the TDP supplied to the CPU socket to 250 watts.
-You can't argue with the world's biggest MoBo manufacturer when they tell you that a 5960x is capable of pulling 300 watts thru a cable the provides power to the CPU ONLY.


 

140W is only what the spec sheets says. Real-world measurements would be lower since almost no chip would be at the absolute worst limit of the spec but nearly all reviews use whole system figures because they lack the necessary measurement tools and knowledge to do it properly.

If you double the VRM's output current, you more than quadruple I2R losses through the PSU leads, VRM, motherboard and socket pins. Depending on how efficient the VRM is, this can account for more than 50W extra power loss. By overclocking, over-volting and putting a heavy load on a quad-channel DIMM setup, you can have another ~15W more due to DRAM and DRAM-related off-CPU circuitry. The fans will work harder and depending on the fan controller setup, that may mean another 10-30W extra, a few of which may be the PSU's own fan, etc.

Your assumption that all extra power gets dissipated by the CPU is fundamentally flawed; there are several other components whose power draw may be both directly and indirectly affected by CPU overclocking.
 
1. Yes, I now what the spec sheet says, and again you clearly didn't read the previous post.

Please read:

The increase in SYSTEM wattage caused simply by loading the CPU @ stock settings = 189 watts - 69 watts = 120 watts.
Click to expand...

Now, with the difference in wattage being 120 watts we add the 13 watts idle watts and we are at 133 "at the wall". The PSU efficiency is low at this point ... figure 88% at low (6%) load.... 133 x 85% = 117 watts. Now the big questions .... is the 117 watts less than 140 watts ? Thank you for confirming the math.

2. Who said anything about overclocked RAM ?.... that is not in any of the system calculations so your adding in a load that doesn't exist in the measurements. Lets go back to the CPU calculator...with nothing filled in, hit calculate and you get 38 watts. Now let's add 4 x 4Gb of DDR4 and recalculate .... and we get a whopping 43 watts ... 5 watts for 4 sticks. So your argument is overclocking RAM will add 15 watts .... so OC'ing RAM takes us from 5 to 15 watts with 0% increase in voltage or TRIPLE the stock load but by some magic a 42% increase in CPU speed and 54% increase in CPU voltage has no effect.

3. You said double the VRM's output current, you more than quadruple the impact .... so that's a square function. So what happens when you increase system voltage from 0.92 (stock - 140 watt TDP) to 1.50 ??? A linear increase alone would talk us to 230 watts. But waita sec... the primary function of the VRM on Haswell -E is what ?

An extreme platform demands extreme control over the purest power, and nothing delivers more control or purer power than Extreme Engine Digi+ IV — Rampage V Extreme's exclusive voltage-regulator module (VRM). MicroFine alloy chokes lower overall core loss for lower temperatures, while the VRM has been upgraded with the brand-new PowIRstage® IR3555. This chip integrates the driver, high-side and low-side MOSFETs and features the latest RDS On current-sensing technology. Extreme Engine Digi+ IV's high-frequency pulse-width modulation (PWM) implementation scales to 1MHz allowing the DRAM VRM to achieve a 40% increase in stability!
Click to expand...

Again, RAM isn't overclocked in those tests and therefore is not in the "at the wall measurements.

4. Now you are throwing fans in, that's a bit desperate. Let's go to actual real measurements. Just tested my box....

Fan rpm @ idle = 430 rpm
Fan rpm at 4.6 Ghz 1.51 volts = 630 rpm.

Fan = 1.68 watts at 1250 rpm

The power load is roughly equivalent to the speed ratio cubed....

430 / 1250 = 34% speed ~ 4% of full load = 0.07 watts
630 / 1250 = 50% speed ~ 13 % of full load = 0.21 watts

Each fan adds 0.14 watts x 5 radiator fans = a whopping 0.70 watts. 30 watts of **extra** power for fans ? Really, the exaggeration is a bit outta hand. I have 10 fans installed 5 case fans and 5 radiator fans.... I turned off the case fans for the above test because they don't even reach their starting threshold when CPU testing..... so even if my fans ran at full tilt at 1250 rpm.... that's a whopping 8 watts. The PSU fan ? Really ? Did you take a look at that PSU used in the test ... the PSU fan never turned on.... It don't turn on until PSU output exceeds 360 watts .... with 378 watts at the wall PSU output was only 340 watts.... off at idle, off at overclocked load.... difference in watts = 0 as 0 - 0 = 0.

5. First ya said that HDs/SSD's, and all these other things were in the mix, now you are adding fans and all these other things after that argument was knocked down. What's the point anyway.... ? We are talking about a 400 watts PSU. If we are adding 50 watts there and 30 watts here as you claim, then how the heck is this 400 watt PSU going to power all these things ? Please tell us how this PSU is going to handle a 287 watt PSU your magic 140 watt overclocked CPU, this 50 watts of extra power on top of the original 40 for the MoBo, this extra 15 watts for the RAM on top of the original 15, and these super special fans with an extra 30 watts in top of their original 30 watts ...By my math thatz > 600 watts. Now add 15% to account for capacitor aging... you need a 700 watt PSU !

Again, read what was stated and look at the links. The PSU calculator calculates CPU Draw and MoBo Draw separate.... the maximum total wattage draw for a high end desktop is reported at 55 watts. Look at any water cooling guide and show me one suggesting an additional 50 watts for MoBo overclocking ....

But for the sake of simplicity, forget all the math, forget the PSU calculator, forget the actual measured results at the wall, forget the water block ratings and their manufacturer recommendations, forget all the water cooling guides.....and forget that they all say the same thing within single digit % points. Answer me this:

A. If the CPU doesn't exceed 140 watts then why is it that you must

http://www.guru3d.com/articles_pages/core_i7_5960x_5930k_and_5820k_processor_review,19.html
increase the TDP limits of the processor to 250 Watts (by that you are allowing a higher power draw)
Click to expand...

Note the bolded type...not the fans, not the PSU, not the RAM, not the HD/SSD, not the GPU, not the MoBo not the VRMs

B. Why is the largest MoBo manufacturer on the planet saying that

For overclocking 5960X processors, we recommend PSUs that can supply a minimum of 30 amps to EPS 12V. At 4.6GHz a 5960X can draw close to 25amps from the EPS12V connector
Click to expand...

Again "a 5950x can draw 25 amps" .... not a 5960x plus fans, plus VRMs, plus PSU fans, plus HDs, plus SSD's, plus GPUs .


http://www.hardwaresecrets.com/article/Everything-You-Need-to-Know-About-Power-Supplies/181/3

EPS 12V Connector: This 8-pin connector has the same function as ATX 12v i.e. to provide electrical current to the system CPU.
Click to expand...

Again, not the CPU plus PSU fans, case fans, radiator fans, SSDs, HD's, GPUs, RAM or anything else.... just the CPU
 
Again "a 5950x can draw 25 amps" .... not a 5960x plus fans, plus VRMs, plus PSU fans, plus HDs, plus SSD's, plus GPUs .
Click to expand...
Small correction:
Haswell VRMs are integrated I believe. And measuring at the CPU power connector would count the VRM power loss anyway.
PSU fans are probably already factored in for power capacity
 


The IMC is included on the die.... there are still VRMs on the MoBo but they need to do a lot less and consume less power than they did before.

http://www.overclockers.com/asus-maximus-vi-formula-motherboard-review/

There is a Digi+ VRM assigned to the two memory power phases that allow the user finite control over the memory power delivery. Even though the Digi+/EPU chip lost a bit of its luster when Intel moved the CPU’s VRM on die, it still provides a seamless interface and extensive control over its functions.
Click to expand...

But again, if you are not overclocking the RAM....there is no additional memory VRM load between the CPU stock and CPU Overclocked system wattages so it is irrelevant.


The argument about the PSU fan was as far as I can tell that when measuring power at the wall.... this maximum 2 watt fan was somehow going to effect the measured system loading negatively.... the argument being that the the at the wall measurements from the test would include the additional load from the fan ramping up. However, this is bogus cause the fan in the test PSU (AX1200i) doesn't even turn on until system wattage breaks 360 watts (30% load)

http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story3&reid=300

Due to the high efficiency and lower heat output, the AX1200i is capable of running in fanless mode with up to roughly a 30% load (360W).
Click to expand...

Corsair's graph on the 1st page of the review shows 32% but 32 certainly fits in "roughly 30%" so I used the 30% to calculate. So as we see in the previous post with the system load graph, 378 watts was measured at the wall at the full overclock which means that about 340 watts (378 x 90%) was being put out by the PSU. So the fan consumption was 0 watts at idle and the fan consumption was 0 watts at full load. As a result:

- the effect of the fan at idle was 0.00 watts
- the effect of the fan at OC loaded was 0.00 watts

The PSU fan therefore has 0 impact
 

What is the input voltage of Haswell's FIVR? 1.8V stock.

Here are some more accurate numbers for you:
http://www.tomshardware.com/reviews/intel-core-i7-5960x-haswell-e-cpu,3918-11.html

Out of the 280W peak going into the EPS12V connector measured with current probes, 85W is lost in the motherboard's VRM.

No messing with attempting to approximate measurements from the power plug. Low voltage regulation at high currents is nasty business dominated by I2R losses.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom