Nice (throttling) housefire you got here Nvidia, at least

File: 1356313994810.png (210KB, 486x480px) Image search: [Google]

210KB, 486x480px

>>55902183
>Total System Power Consumption

>>55902271
Yeah, the CPU uses at least 200W

File: 1469959892166.gif (848KB, 500x281px) Image search: [Google]

848KB, 500x281px

>>55902297
>at least 200w
WHERE YOU FAGGOT

So it uses a little more than a 290X and screeches at 50dB

Which was apparently a Fermi-tier housefire

>>55902183
>throttling
>housefire
>posts a graph of power consumption instead of temperature or clock speeds

>>55902560
>missing the point

File: TITAN-X-345-86.jpg (114KB, 640x356px) Image search: [Google]

114KB, 640x356px

>>55902858
Nobody cares about temps, even 2W CPUs can hit 90C, temps are irrelevant.

>>55902183
So the Titan X uses 88w more than the 1070 which is a 150w TDP card? That would put the TDP around 240w? I wonder how its perf/watt ratio is now.

See this is why system power consumption figures are unreliable, nobody ever mentions that on more powerful GPUs the CPU does more work and consumes more power which skew the results quite a lot.

Honestly the only way to properly test GPU power draw would be methods AMD and Nvidia use or oscilloscopes.

>>55902892
Eh, the Titan X's boost clock is 1531MHz, so it looks like on that graph it was always above that.
Is there a problem with this?

>>55903047
Boost clock(for Nvidia) is just an average clockspeed figure, the GPU can and will go higher if the temperature, fanspeed or TBP limit wills it, like what's shown at the start of the graph.
Ergo it's throttling.

>>55903137
It's the same deal with Intel processors. You have your basic clock, the turbo clock, and it may go even higher if the thermal room.
The Nvidia cards have a target clock speed but will boost further as long as the thermals allow. So long as it is delivering at least their rated boost clock I don't see what the problem is. It just shows that the card has greater potential with a better cooling solution on it.

>>55903277
The problem is that Titan X only comes with reference, Nvidia doesn't allow IHVs to make their own versions so the card is left with stock power delivery and stock cooler that isn't adequate to make the GPU run at its full potential.

>>55902271
>>55902858
>>55902932
>Higher power consumption than a Fury X
>Not a housefire

>>55902297
Total system power consumption is retarded. Although Titan cards always suck, because nvidia does not allow 3rd party heatsinks. Same with AMD fury X. Fucking shitbirds.

>>55903431
>240w > 275w
Fury X has a TDP of 275w. If the calculation (and measurement) is true the Titan X at 240w how do you come to the conclusion the Titan X has "higher power consumption?"

>>55903449
>nvidia does not allow 3rd party heatsinks
You can always (though it voids warranty) set up your own water cooling. A lot of vendors release water cooling blocks after Titans are released.

>>55903388
Nothing is stopping the slapping of an aftermarket solution on there senpai.

>>55903486
by reading the graph in OP

>>55903534
How do you think I came up with the calculation if I didn't? Are you confusing total system power usage with the gpu's power usage?

>>55903509
No, that's true, but I prefer air coolers and not having to fuck with the graphics card.

>its okay when nvidia does it

>>55903562
Sounds like you wouldn't be buying a Titan anyway.

>>55903509
Cooling is just a part of the problem, you're using stock VRMs, caps and mosfets which aren't all that high in quality.

>>55903554
379 is a bigger number than 361

>>55903585
>Cooling is just a part of the problem
Cooling was the only issue raised in the post responded to.

>>55903584
Of course not. Not only because I'd have to give it a proper cooler, but also because the current version of Titan has some of its clusters disabled. I can't stand cards like that.

To be fair, Titan X is using 12GB of 2500MHz memory, which no doubt alone sucks down at least a good 50W.

>>55903594
Which makes me curious about the source of the graph. The 1070 is 150w card. The Fury X is a 275w card. Yet the 1070 total watts is 291, meaning the rest of the system is using 140w or so with the 1070 but only 90w or so with the Fury X?

>>55903599
You'd gain nothing from the extra 200 shaders, at most some 3-5%

>>55903625
TDP != TBP != Actual power usage

>>55903648
No, it means AVERAGE power usage with typical use. This makes me curious if the measurement of the Fury X at something other than its AVERAGE power usage but a cherry-picked recording at the LOWEST power usage some point after fifteen minutes of gameplay. Thus why I am curious about the source of the graph.

>>55903669
Hardwarecanucks

>>55903669
>AVERAGE power usage with typical use
Just to address any pedantry before it is raised (hopefully it isn't) TDP is specifically about heat management but that's related to power used.

>>55903578
because amd can't do anything

>>55903633
It's not about the gain, I just don't like that.

>>55903687
>In order to stress the GPU as much as possible we used 15 minutes of Unigine Valley running on a loop
>graph states RotTR was used
I think I'll hold off on making any further comments until TPU, Anandtech, or Guru3d publish their review.

>>55903669
TDP = Thermal design power
Notice the 'thermal' part.
Means a guideline for the heat management, which in this case is 250W meaning the guideline is to expel 250W of heat on average.
TBP = Typical board power
This is strictly a GPU measurement alone and doesn't count in the VRAM, VRM and other assortment of stuff on the PCB, this figure is rarely actually stated for some reason.


Now measures like this are fucking pointless because Intel, AMD and Nvidia calculate TDP differently.

>>55903625
>>55903669
Each manufacturer defines TDP differnetly because there is no official standadized definition.
AMD defines TDP as maximum power consumption. Nvidia defines it as maximum power consumption under typical loads.
That's why Nvidia cards throttle when you run Furmark on them but AMD cards don't.

>>55903747
Obviously it means preheating the GPU before running benches.

>>55903749
>Notice the 'thermal' part.
My hopes were dashed.

>>55903760
>Each manufacturer defines TDP differnetly because there is no official standadized definition.
So AMD is using a definition of TDP contrary to the common usage?
https://en.wikipedia.org/wiki/Thermal_design_power

>>55903773
>heat
>measuring watts
Even more reason to question that journalistic value of the author's work.

>>55903749
>Notice the 'thermal' part.
Just out of morbid curiousity. Does your post mean that the Fury X, with its closed loop cooling solution, produces more heat that the actual watts used would normally produce?

>>55903815
Yes, Fiji actually has measurably lower power consumption with smaller temperatures than other GPUs.

>>55903783
For GPUs AMD uses typical board power.

Matter of factly TDP is not one thing, it is an adjustable target. If I take one single 10w chip and say it can run up to 100c then I can get away with a 10w cooler on it. If that chip can only safely run at 50c then it needs more cooling, and its TDP will be rated higher even though its the exact same chip drawing the exact same power.

Intel is the nastiest with TDP figures, especially involving the low voltage stuff going from 4W-17W

File: kek.jpg (115KB, 940x621px) Image search: [Google]

115KB, 940x621px

>>55903783
>Actual source discusses that AMD and Intel use different definitions of TDP
>Wikipedia author only cites Intel's definition
This is why you you never use Wikipedia as a source

>>55903848
>Yes the Fury X produces more heat than would normally be produced by the watts actually used.
I have a hard time believing AMD would do something like this as it is very inefficient.

>>55903859
>For GPUs AMD uses typical board power.
[citation needed]
Even if it is true it still makes the OP's graph questionable in regards to the measurement of the Fury X. We know, given all the recent reviews, that the total board power of the 1070 is 150w. AMD rates the Fury X TDP (which by your post is also equal to TBP) at 275w yet the power difference between the two in the OP's graph means there is a 50w different in watts used by the system.

It is almost like you two are not reading from the same AMD memo.

>>55903888
>AMD and Intel
We are not talking about Intel. We are talking about AMD and Nvidia, more specifically about the 1070, with its confirmed 150w power draw to match its 150w TDP and the Fury X with a 275w TDP.

>Wikipedia author only cites Intel's definition
Going by the citation, this is wrong. The definition is cited to silentpc not Intel.

File: cant out bid the yid.jpg (93KB, 1024x575px) Image search: [Google]

93KB, 1024x575px

>>55903883
Oy vey! Dontcha know its okay to exceed TDP for brief periods? Totally fine. I swear on my muddah! So what if a 5w Broadwell CoreM pulls over 12w!? Whaddaya gunna do? Holocaust me?

File: Screenshot from 2016-08-03 23-38-28.png (26KB, 989x302px) Image search: [Google]

26KB, 989x302px

>>55903926
All ICs use less power on lower temperatures, Fiji is just a extremely visible display of it.

http://www.anandtech.com/show/9390/the-amd-radeon-r9-fury-x-review/5

>>55903927
>CoreM pulls over 12w!?
>over 12W
You mean over 20W

I remember those Ivy Bridge 15-17W ULVs drawing almost 40W

>>55903926
>[citation needed]
Its called the PowerTune whitepaper.
Even Anandtech has directly referenced it.

>>55903974
Speaking about Anandtech. The review of the 1070 by Anandtech measured the Fury X using over 100w more power (though with a different methodology). I wonder what the difference is to create a 30w disparity between the two?

>>55904011
Different hardware, different workload, different measuring tools, different software versions, different drivers.

>>55903003
oscilloscopes measure voltage not current...

>>55904032
I assume neither hardwarecanucks nor Anandtech changes their testbed system, workload, measuring tool between testing cards. Otherwise the measurement is not reliable.

>>55904078
http://www.tomshardware.com/reviews/graphics-card-power-supply-balance,3979.html

The moar you know

>>55904097
They stay the same during the review, but changing the hardware and workload for different reviews involving the same cards is something that happens quite often.

>>55904133
I find it hard to believe the power used by the Fury X changes depending on the testbed.

>>55904138
That's because Anandtech power measuring tools are shit and they only calculate system power and not GPU power.

>>55904178
Hardwarecanuck's review only calculated system power as well.

>>55904198
I didn't say their power measurement tools are good either, they all do the same lazy wallmeter shit and call it a day.
Only tom's does actual measurements for the PEG/rails/PCie

>>55904241
>I didn't say their power measurement tools are good either
Did you miss the point? Comparing the total system power measurements between hardwarecanuck and anandtech in regards to power usage of the Fury X?

>>55903926
Nvidia sets the TDP as typical gaming power consumption. Hence why the 1070 consumes 150W while gaming and if you throw Furmark at it it just throttles and stays at 150W.

AMD sets the TDP much higher. Hence why a Fury X uses only ~220W while gaming but if you throw Furmark at it it consumers well above 300W.

>>55904286
Well, for starters the Anandtech review is a year old while the hcanucks one is 2 days old.

>>55904318
U wut m8? Anandtech's review of the 1070 is a year old? Yeah, you clearly missed the point.

>>55904305
Furmark on Nvidia hardware is pointless, the drivers downclock to fucking half the base clock.
Nvidia hardware uses more power running fucking Crysis 3 than furmark

>>55904305
>AMD sets the TDP much higher.
[citation needed]
It seems rather questionable to claim AMD determines TDP on a value different from maximum or average heat production relative to watts (since watts is the measurement).