Intel & AMD: Rated TDP vs. Actual TDP?

[piesquared]

Many reviews and comments around the net have been pointing to the fact that intel's advertised TDP is well short of actual power consumption. Articles and comments are based off these numbers, claiming AMD is uncompetitive because of TDP, yet inte's advertised TDP seems to be a lie or false advertising at the very least.

What are the real TDPs and power consumption of these chips by intel? Are they cheating consumers?

Here is the 10WATT intel j1900 vs AMD 25W Athlon 5350:

http://uk.hardware.info/reviews/533...l-d-review-cheap-desktop-platforms-conclusion

All while handily outperforming the j1900, significantly in many cases, and while being on 28nm vs 22nm finfets

 

[positivedoppler]

I dunno really. Perhaps Intel uses a number that reflects the cpu performing tasks such as web browsing or watching a movie(less gpu intensive) while AMD who advertise more of the APU estimates the TDP with the system performing tasks that stresses both the cpu and gpu. I'm really just reluctantly guessing. Offering an opinion here risks getting caught in the crossfire of Intel white knights and AMD zealots. It's scary dangerous.

 

[Enigmoid]

I see cherry picked benchmarks.

And a rather high idle power of 17W.

I see a max of 21W on a 10W tdp soc. Which is likely within 10W when you look at idle power, RAM, disk, and mobo power use.

Every intel chip I've personally used (mobile notebook chips) follows intel's tdp spec.

i7-3630qm

1. Prime 95 runs at max turbo (3.2 ghz), package power ~40-42W.
2. Furmark on igp. Max turbo CPU + GPU for ~10 seconds. 56W
3. Turbo on CPU disabled (2.4 ghz), full turbo on igp. ~45W. Runs indefinitely this way within thermal constraints.

Intel limits their CPU's to the power allowed (unless short term turbo or configurable tdp). HW monitor or other applications will show that under heavy and protracted load the CPU is using approximately the tdp.

Max power could easily be a turbo power spike.

 

[witeken]

How TDP is determined is different for every company, as far as I know, so it isn't exactly a 1:1 comparison, although it shouldn't be a huge difference. Secondly, TDP is determined for the chip only, not the whole system. Anand measured power consumption of Silvermont at IDF, and 4 cores consume 2.5W under load, so a 10W TDP for the whole chip seems reasonable to me. I don't know about AMD, but there's probably a reason why the chip has a 25W TDP.

Do note, though, that TDP isn't about power consumption specifically, but rather the maximum heat the cooling system needs to dissipate.

 

[Sweepr]

Idle / CPU Load difference:
- Celeron J1900: 5W
- Athlon 5350: 16W

Idle / CPU Load difference:
- Celeron J1900: 5W
- Athlon 5350: 15W

Idle / CPU Load difference:
- Celeron J1900: 4,4W
- Athlon 5350: 16,7W

Now Cinebench 11.5 performance:
Celeron J1900: 1.8
Athlon 5350: 2.0

http://us.hardware.info/reviews/533...ap-desktop-platforms-benchmarks-cinebench-115

What did you say again?

Do you really want to learn something here or are you just causing trouble so you can go link to this thread for your buddies?

Please stop intentionally littering this forum.

Mods have been very tollerant with him, hardly adds something useful and usually derails threads and gets involved in flame wars.

 

[Erenhardt]

I think the measuring time is very important here.
As shown by PcLamb.pl, J1800 has a delta of 4 watts only between idle and throttling point which (knowing pclab) only occurred during power measuring run, not when benchmark score was recorded

 

[positivedoppler]

Do note, though, that TDP isn't about power consumption specifically, but rather the maximum heat the cooling system needs to dissipate.

I'm dusty on all the technical stuff after years out of school. But doesn't the heat dissipated directly translate to amount of power consume? Other circuit element such as capacitance and inductance does not burn power but stores it.

 

[piesquared]

I see cherry picked benchmarks.

And a rather high idle power of 17W.

I see a max of 21W on a 10W tdp soc. Which is likely within 10W when you look at idle power, RAM, disk, and mobo power use.

Every intel chip I've personally used (mobile notebook chips) follows intel's tdp spec.

i7-3630qm

1. Prime 95 runs at max turbo (3.2 ghz), package power ~40-42W.
2. Furmark on igp. Max turbo CPU + GPU for ~10 seconds. 56W
3. Turbo on CPU disabled (2.4 ghz), full turbo on igp. ~45W. Runs indefinitely this way within thermal constraints.

Intel limits their CPU's to the power allowed (unless short term turbo or configurable tdp). HW monitor or other applications will show that under heavy and protracted load the CPU is using approximately the tdp.

Max power could easily be a turbo power spike.

The Kabini system also had RAM, disk drive and a mobo.

The 5350's rating is 25W, J1900 is 10W. The j1900 blew threw the 10W TDP, while the 5350 1W or so above TDP would be explained by peripherals.

 

[piesquared]

I see cherry picked benchmarks.

And a rather high idle power of 17W.

I see a max of 21W on a 10W tdp soc. Which is likely within 10W when you look at idle power, RAM, disk, and mobo power use.

Every intel chip I've personally used (mobile notebook chips) follows intel's tdp spec.

i7-3630qm

1. Prime 95 runs at max turbo (3.2 ghz), package power ~40-42W.
2. Furmark on igp. Max turbo CPU + GPU for ~10 seconds. 56W
3. Turbo on CPU disabled (2.4 ghz), full turbo on igp. ~45W. Runs indefinitely this way within thermal constraints.

Intel limits their CPU's to the power allowed (unless short term turbo or configurable tdp). HW monitor or other applications will show that under heavy and protracted load the CPU is using approximately the tdp.

Max power could easily be a turbo power spike.

We aren't talking about the 3630 here, we are talking about the j1900. Power consumption is much more significant at these lower TDP's, and the j1900 seems to be way above the rated TDP.

 

[piesquared]

Idle / CPU Load difference:
- Celeron J1900: 5W
- Athlon 5350: 16W

Idle / CPU Load difference:
- Celeron J1900: 5W
- Athlon 5350: 15W

Idle / CPU Load difference:
- Celeron J1900: 4,4W
- Athlon 5350: 16,7W

Now Cinebench 11.5 performance:
Celeron J1900: 1.8
Athlon 5350: 2.0

http://us.hardware.info/reviews/533...ap-desktop-platforms-benchmarks-cinebench-115

What did you say again?



Mods have been very tollerant with him, hardly adds something useful and usually derails threads and gets involved in flame wars.

Could you link to the source of the first graphs so we can see the power supplies used?

 

[Erenhardt]

The Kabini system also had RAM, disk drive and a mobo.

The 5350's rating is 25W, J1900 is 10W. The j1900 blew threw the 10W TDP, while the 5350 1W or so above TDP would be explained by peripherals.

Tubo.
Mobile parts especially will use available thermal margin to boost clocks so it can go to idle faster. Helps in benchmarks and IMHO should be ditched because the device doesn't work the same every time. I hate turbos.

Also, gives another way to manipulate data in reviews:
-take performance data during boost state that exceeds sustainable clocks,
-put power consumption under thermal throttling

mix it in perf/watt chart and call it a winner

 

[piesquared]

Tubo.
Mobile parts especially will use available thermal margin to boost clocks so it can go to idle faster. Helps in benchmarks and IMHO should be ditched because the device doesn't work the same every time. I hate turbos.

Also, gives another way to manipulate data in reviews:
-take performance data during boost state that exceeds sustainable clocks,
-put power consumption under thermal throttling

mix it in perf/watt chart and call it a winner

Agreed turbo should have been ditched from reviews,a long time ago, but that is what got intel all those glowing reviews over the years, so it'd be blatantly biased to ditch them suddenly when AMD designed a much better implementation with STAPM.

Besides that, I don't think that is the issue here. Intel is blowing past their rated TDP, while AMD is not.

 

[ninaholic37]

Interesting that so far, no one has been able to prove the graph wrong. A 10W CPU getting 16.8W at Idle does not look good. Hopefully the TDP was just false advertisement and Intel really meant 25W, which would be better than people's computers starting to blow up and Intel having to recall the chips or get sued.

 

[witeken]

I'm dusty on all the technical stuff after years out of school. But doesn't the heat dissipated directly translate to amount of power consume? Other circuit element such as capacitance and inductance does not burn power but stores it.

Yes, but things like turbo boost are designed to consume a lot more energy than the TDP for a short period of time to improve performance.

 

[witeken]

Besides that, I don't think that is the issue here. Intel is blowing past their rated TDP, while AMD is not.

How can a 10W TDP chip blow past its rated TDP when the difference between idle (aka ~0 power consumption) and load is only about 5W?

 

[dave_the_nerd]

Interesting that so far, no one has been able to prove the graph wrong. A 10W CPU getting 16.8W at Idle does not look good. Hopefully the TDP was just false advertisement and Intel really meant 25W, which would be better than people's computers starting to blow up and Intel having to recall the chips or get sued.

Or it's total system power, not just CPU draw.

 

[OCGuy]

Interesting that so far, no one has been able to prove the graph wrong. A 10W CPU getting 16.8W at Idle does not look good. Hopefully the TDP was just false advertisement and Intel really meant 25W, which would be better than people's computers starting to blow up and Intel having to recall the chips or get sued.

It is measuring total system consumption, which includes DDR3.

Power draw from the wall is not what a chip is actually using. A 150w power supply running @ 10% is going to be insanely inefficient.

TDP can be exceeded for short periods of time (turbo), as stated in Intel's definition of Max TDP.


There ya go.

 

[piesquared]

How can a 10W TDP chip blow past its rated TDP when the difference between idle (aka ~0 power consumption) and load is only about 5W?

How can it be idling at 0 when the compared chip is idling 7W lower?

 

[piesquared]

It is measuring total system consumption, which includes DDR3.

Power draw from the wall is not what a chip is actually using.

TDP can be exceeded for short periods of time (turbo), as stated in Intel's definition of Max TDP.


There ya go.

AMD's system was for total measured power consumption also.

 

[OCGuy]

AMD's system was for total measured power consumption also.

What does that have to do with the measurements not representing the actual chip power draw?

DDR3 Power Dissipation

Power Dissipation for reference only.
A DDR DIMM needs 5.4 watts, a DDR2 DIMM needs 4.4 watts and a DDR2 FB-DIMM needs 10.4 watts.
DDR3 provide a 30% reduction in power consumption

http://www.interfacebus.com/Memory_Module_DDR3_DIMM.html



Somehow I doubt you will update your thread title to reflect this though. I am by no means a PSU/component wattage expert, so if I can figure this stuff out, I am sure others can as well.

 

[piesquared]

What does that have to do with the measurements not representing the actual chip power draw?

Really? One draws significantly less using the same parts (actually, the RAM on AMD's system was even clocked higher) so that leaves chip power draw for the balance. Kabini is significantly lower than bay trail.

We provided the platforms with 8GB DDR3-1600 memory (clocked at 1333 MHz on the Intel platform). In addition, the test systems were outfitted with an OCZ Vertex 4 128GB SSD. We used a PicoPSU 150W as our power supply.

 

[BSim500]

Total system power consumption is obviously going to be higher than just the CPU's advertised TDP. LOL. From linked article : "Note that this chart shows the power consumption of the complete test systems (including memory, SSD, etc.)" Obviously, the motherboard + memory + HDD + peripherals, etc, are going to add to that. Intel's "10w" is for the CPU alone:-
http://ark.intel.com/products/78867/Intel-Celeron-Processor-J1900-2M-Cache-up-to-2_42-GHz

And they use different motherboards : "The AMD Athlon 5350 was tested using the MSI AM1I motherboard, and the Celeron J1900 using the Gigabyte J1900N-D3V board". That alone could make up 10w difference. Examples of different consumption between even boards within the same platform & CPU:-
http://media.bestofmicro.com/R/X/387213/original/image023.png
http://hothardware.com/articleimages/Item2103/z87-roundup-power.PNG

And for high end CPU's - I thought it was the exact opposite. I know my "77w" i5-3570 uses more like 50w load (Prime on all 4 cores) and total system idle's at 26w (that's inc 1x 2.5" HDD + 1x Samsung 830 SSD + 1x 5.25" BD-RE burner + Z77 motherboard + 3 or 4 USB devices). Likewise, my "55w" HTPC's i3 uses about 36w CPU by itself (4 thread Prime again). Real world apps & games are 5-15w lower.

Edit : Just seen OCGuy's post. OK, that explains a lot... :whiste:

Edit 2 : Can't people trim their posts when quoting to not repost the same nested pictures over and over again?

 

[OCGuy]

Really? One draws significantly less using the same parts (actually, the RAM on AMD's system was even clocked higher) so that leaves chip power draw for the balance. Kabini is significantly lower than bay trail.

So are we comparing Intel/AMD performance/watt, or are we discussing TDP?

:biggrin:

 

[piesquared]

Still waiting for Sweepr to post links to the first slides so we can all see what power supplies were used....

 

[Sweepr]

Techspot. Go search yourself.
12W lower power consumption under CPU load than the 25W Athlon 5350 (15W less @ CPU+iGPU load). Probably a different MB+J1900 combo than the one used in the AMD-biased review you picked.