TPU singlepcb 295 review is up! 20-55W higher power consumption

Printable View

Show 100 post(s) from this thread on one page

07-05-2009, 05:22 AM
perkam

Quote:

Originally Posted by OBR

http://i40.tinypic.com/24yybt3.jpg

Testbed: W3540 @ 4GHz, EVGA Classified, 2x GeForce GTX 295 (new and old)

Temps Idle:

old: 54C, new: 50C

Temp Load: (2nd test 3dmark vantage at Extreme preset)

old: 85C, new: 69C!

Temp Load: (Furmark)

old: 100C, new: 78C!

Power Consuption

idle old: 257W, new: 254W
load old: 430W, new: 432W

TechPowerUp says BS about Power Consumption, like always ...

Here's his post again:

Quote:

Originally Posted by W1zzard

ah hi OBR. good to see you commenting on our reviews not only on tpu.
how do you measure power? wall? card power? whats the resolution, repeatability, calibration of your device? how often do you sample power draw? how do you calculate your average? how do you load the card? which app? which resolution? whats the rest of your test system like?

I'll summarize it for you:

1. How do you measure power? wall? card power?
2. What's the resolution?
3. What's the repeatability, calibration of your device?
4. How often do you sample power draw?
5. How do you calculate your average?
6. How do you load the card? which app? which resolution?
7. Whats the rest of your test system like?

As far as I can tell, you have only answered ONE of those questions.

Perhaps you would like to elaborate?

Perkam
07-05-2009, 05:32 AM
W1zzard

one more question. WHEN do you measure power consumption? card installed right into the pc and booted up? card sitting idle for x minutes? right after load run? i noticed pretty big differences between those scenarios.

my guess is that with temperature of the _whole_ card some resistances/electric properties of components change that affect current draw. any EEs to comment on this ?
07-05-2009, 05:51 AM
Lightman

Quote:

Originally Posted by W1zzard

one more question. WHEN do you measure power consumption? card installed right into the pc and booted up? card sitting idle for x minutes? right after load run? i noticed pretty big differences between those scenarios.

my guess is that with temperature of the _whole_ card some resistances/electric properties of components change that affect current draw. any EEs to comment on this ?

Yes, that is the case for every silicon. With temp. increase increases resistance of the circuit. That's one of the reasons why superconducting so far only happened in very low temps. That's also why colder means better clocks.
07-05-2009, 08:40 AM
G0ldBr1ck

Quote:

Originally Posted by Lightman

Yes, that is the case for every silicon. With temp. increase increases resistance of the circuit. That's one of the reasons why superconducting so far only happened in very low temps. That's also why colder means better clocks.

That does not necessarily apply when your dealing with semi-conductors. Most semi-conductive materials have a negative temperature coefficient where temperature decrease results in lower conductive ability/higher resistance.

am I wrong?
07-05-2009, 02:39 PM
Lightman

Quote:

Originally Posted by G0ldBr1ck

That does not necessarily apply when your dealing with semi-conductors. Most semi-conductive materials have a negative temperature coefficient where temperature decrease results in lower conductive ability/higher resistance.

am I wrong?

Well, that probably was the problem with AMD's cold bug on 90/65nm process. Depends on what type of doping was used for silicon.
Anyway by my own observations I've concluded that at least in temperatures above 0C higher temp. equals more power consumption. Of course PWM is affecting this, but here we are discussing power consumption fluctuations of the test systems used in the tests. Therefore we can agree that temperature has influence.

My own system in sig using exactly the same test environment under load had close to 30W higher consumption after 30min. burn-in.
07-05-2009, 10:48 PM
largon

Quote:

Originally Posted by OBR

TechPowerUp says BS about Power Consumption, like always ...

Coming from the guy that would win title of "The most BS XS member", if such a vote was called, that's pretty funny.
07-05-2009, 11:11 PM
dinos22

lol OBR
why not answer some questions there
they are pretty important on how you draw your conclusions
07-05-2009, 11:21 PM
spajdr

Im afraid and pretty sure you wont get answers you want from OBR :)
07-05-2009, 11:35 PM
OBR

why not to answer? Yes i do it, but i have lot of work ... i cannot to sit whole day on forums, i am working.
1. How do you measure power? wall? card power?
2. What's the resolution?
3. What's the repeatability, calibration of your device?
4. How often do you sample power draw?
5. How do you calculate your average?
6. How do you load the card? which app? which resolution?
7. Whats the rest of your test system like?

1. Wall - (only rig, without monitor) - using Metrix TRMS PX 120
2. Resolution: 2560x1600
3. Yes that pretty expensive device has been calibrated ... :) :)
4. Whole 2nd test 3DMark Vantage ...
5. Device measuring average in predefined time (during test)
6. Idle - after Windows starting up + when consuptin is +/- stabilized
6. Load - 3DMark Vantage Extreme, 2nd test (Calico)
7. Test bed: Core i7 @ 1,3V, EVGA Classified, 6GB Corsairs, 1x common HDD SATA, 1x fan on RAM, 2x fan on CPUs, no sound card ... look at the picture above, this is actual testbed
largon: funny you are and your funny AMD/ATi craps ... HD4870X2 is parody to performance for noobs only and Phenoms II too ...
07-05-2009, 11:51 PM
BulldogPO

Dudes stop that whinging before someone makes you stop it and locks this thread.
07-06-2009, 12:15 AM
W1zzard

are you measuring active, reactive or apparent power ? what differences do you see between those three measurements?
07-06-2009, 12:31 AM
OBR

Power consumption in reviews:

Zol.com.cn (first review):
Idle

295old: 181W
295new: 176W

Load

295old: 427W
295new: 423W

Guru3D
Idle

295old: 239W
295new: 239W

Load

295old: 406W
295new: 411W

PCGHW
Idle

295old: 62W
295new: 59W

Load

295old: 213W
295new: 211W

My measuring
Idle

295old: 257W
295new: 254W

Load

295old: 430W
295new: 432W

TPU!
Idle

295old: 213W
295new: 237W

Load

295old: 352W
295new: 381W

Only on TPU is Power consuption higher ...
Next time, measure better Wizzard :)

BUT maybe i know, where is problem ...
Wizzard measured OLD GTX 295 with GeForce 181.20 driver and newer with GeForce 186.18. I had both card here with last driver measured.
07-06-2009, 01:16 AM
W1zzard

Quote:

Originally Posted by OBR

BUT maybe i know, where is problem ...
Wizzard measured OLD GTX 295 with GeForce 181.20 driver and newer with GeForce 186.18. I had both card here with last driver measured.

oh yes. that is correct. could you test power consumption of old and new driver? maybe check if the 2d clocks are different
07-06-2009, 01:21 AM
Coldon

Quote:

Originally Posted by OBR

2. Resolution: 2560x1600

LOL! I might be wrong but i think he meant your measuring resolution not your display res.

As for the fluctuations in different cards, are all the cards made at the same factory? IE same brand capacitors, resistors, etc? I would hazard a guess and say that if various manufacturers used different brand components that they would account for more of the fluctuation that the actual silicon of the GPUs.

I might talking BS here, its been around 5yrs since I changed my major from engineering to comp sci.
07-06-2009, 01:37 AM
W1zzard

Quote:

Originally Posted by Coldon

LOL! I might be wrong but i think he meant your measuring resolution not your display res.

As for the fluctuations in different cards, are all the cards made at the same factory? IE same brand capacitors, resistors, etc? I would hazard a guess and say that if various manufacturers used different brand components that they would account for more of the fluctuation that the actual silicon of the GPUs.

I might talking BS here, its been around 5yrs since I changed my major from engineering to comp sci.

afaik they are all made in the same factory. but even identical components have tolerances
07-06-2009, 03:00 AM
crotale

Quote:

Originally Posted by W1zzard

my guess is that with temperature of the _whole_ card some resistances/electric properties of components change that affect current draw. any EEs to comment on this ?

True. Leakage, which is a large contributor to power draw today, is greatly dependent on temperature.

As for the topic as a whole, I wouldn't really care too much about the power consumption. I have measured power draw of a bunch of CPUs over time and the numbers can vary a lot even between chips from the same batch.
07-06-2009, 07:28 AM
saaya

well yeah, but whether a vga pulls 200 or 250W is quite a diference robert :D
if my card would pull 50W more than everybody elses and runs hotter and hence clocks a lot worse, id actually return it to the shop and ask them to change it as id expect something to be wrong with it or the heatsink not making good contact etc...

btw, that could be another cause for diferent results, if you take off the heatsink and remount it not 100% perfectly, the gpu will run hotter, hence draw more current, and consume more power... right? w1zz, did you take the card apart before or after the measurements?
07-06-2009, 07:35 AM
W1zzard

of course i took the card apart after testing everything so that when i break it during the photo session i can still post a complete review
07-06-2009, 10:25 AM
crotale

Quote:

Originally Posted by saaya

well yeah, but whether a vga pulls 200 or 250W is quite a diference robert :D

I can't find the 200W vs 250W number you're mentioning in the same test. The largest difference I can see is 213 vs 237, and I've seen larger discrepancies in CPUs than that. In relative terms, I don't find it to be any big deal.
07-06-2009, 08:40 PM
saaya

Quote:

Originally Posted by crotale

I can't find the 200W vs 250W number you're mentioning in the same test. The largest difference I can see is 213 vs 237, and I've seen larger discrepancies in CPUs than that. In relative terms, I don't find it to be any big deal.

427W load
http://www.techpowerup.com/reviews/I...atinum/27.html

382W load
http://www.techpowerup.com/reviews/Z...TX_295/28.html

it seems meassuring power consumotion is quite tricky and numbers from diferent articles cant be compared directly...
07-06-2009, 10:16 PM
crotale

Quote:

Originally Posted by saaya

427W load
http://www.techpowerup.com/reviews/I...atinum/27.html

382W load
http://www.techpowerup.com/reviews/Z...TX_295/28.html

it seems meassuring power consumotion is quite tricky and numbers from diferent articles cant be compared directly...

That would be around 10% difference, not too bad IMO.

And that is between two different types of cards, I got the impression you meant it could vary 50W between two of the same type of card.
07-07-2009, 01:24 AM
OBR

Every card (same model from same vendor) has different Power consumption, only ONE parameter must be the same - max TDP. One card from the same batch can have +/- 20W different power consuption, its very depends on temps of card too! Try measure power consumption on that same card with AIR and Water cooling, there is a HUGE difference!

I agreed, your card can be "worst" and has higher power draw then another, but not 40W ... your measuring causes HUGE Flame on many sites and in many countries, but only because wrong numbers ...

We can end this boring discussion, and for ppl: read many reviews, not only ONE and start thread like this :(
07-07-2009, 02:29 AM
Sn@ke:~

So if I understood this thread about the single PCB gtx295.

It as a bigger power consumption because a single PCB gpu runs hotter than a dual PCB gpu?

Even thought the new gtx295 as better cooling to make it even compare to the old ones?

It too bad :flame:
07-07-2009, 02:37 AM
OBR

You are wrong, new single-pcb is pretty damn cooler the old version ... i inspected 20-12C difference in Idle/Load

next proof ... today expreview review:

http://en.expreview.com/img/2009/07/04/PowerC.png
07-07-2009, 09:55 AM
saaya

Quote:

Originally Posted by crotale

That would be around 10% difference, not too bad IMO.

And that is between two different types of cards, I got the impression you meant it could vary 50W between two of the same type of card.

it doesnt really matter what its percentage wise in this case imo, 50W is too much of a fluctuation, i doubt cards vary that much...

maybe earlier silicon compared to later one... sure... but not the same batch...

Quote:

Originally Posted by OBR

Every card (same model from same vendor) has different Power consumption, only ONE parameter must be the same - max TDP. One card from the same batch can have +/- 20W different power consuption, its very depends on temps of card too! Try measure power consumption on that same card with AIR and Water cooling, there is a HUGE difference!

I agreed, your card can be "worst" and has higher power draw then another, but not 40W ... your measuring causes HUGE Flame on many sites and in many countries, but only because wrong numbers ...

We can end this boring discussion, and for ppl: read many reviews, not only ONE and start thread like this :(

yes, i agree...
i dont think there has been a huge flame war about this though...
and yes, i should have checked more reviews before starting this thread.

it doesnt change too much though, even with the same power consumption the new card is still dissapointing cause it costs the same, clocks the same, and has no software voltage controls imo...

Quote:

Originally Posted by Sn@ke:~

So if I understood this thread about the single PCB gtx295.

It as a bigger power consumption because a single PCB gpu runs hotter than a dual PCB gpu?

Even thought the new gtx295 as better cooling to make it even compare to the old ones?

It too bad :flame:

no, there must be something wrong with the card tpu received or the results... they didnt do a direct comparison so its possible there was some variation, driver for example...

the single pcb cards seem to consume the same or slightly less power, run slightly cooler and less noisy... yet they cost the same, overclock the same, and dont allow software voltage adjustments which is a bit let down to enthusiasts who this card is actually aimed at. the power connectors also make pot mounting a pita on the single pcb card, and since the mounting holes are very asymetric mounting a custom waterblock or heatsink is pretty much impossible, so watercooling is much much easier than with the dual pcb card, but you need a specially made block which will cost a lot...

i think nvidia could have done better... mounting holes, power connectors, software voltage adjustments... those are things that dont cost any money, those are all stupid design mistakes at least for overclockers...

Show 100 post(s) from this thread on one page