R9 295x2 on a 650W?

[DXMember]

Yes, on low load stock clock cpu... And stock gpu. Also like Prysin said you need two 8 pins that can deliver 28 amps. 2x28 is already 700w.

mind looking at the tomshardware topic linked above?

there's no way that card can pull 700w

 

and if a psu is rated for 650w continious load it better be delivering 650w continous or get all kinds of false advertisement issues and bad reviews from jonnyguru

and I still think you are confusing AC wattage from the wall with DC wattage delivered to the system, power supplies are rated for their dc output

[ZetZet]

mind looking at the tomshardware topic linked above?

there's no way that card can pull 700w

 

and if a psu is rated for 650w continious load it better be delivering 650w continous or get all kinds of false advertisement issues and bad reviews from jonnyguru

and I still think you are confusing AC wattage from the wall with DC wattage delivered to the system, power supplies are rated for their dc output

There is no way? Are you kidding? They ran unigine. Try running furmark? 

Techpowerup got it to pull 646W 

http://www.techpowerup.com/reviews/AMD/R9_295_X2/22.html

 

guru3d got 507 for card ALONE http://www.guru3d.com/articles_pages/amd_radeon_r9_295x2_review,12.html

[Prysin]

There is no way? Are you kidding? They ran unigine. Try running furmark? 

Techpowerup got it to pull 646W 

http://www.techpowerup.com/reviews/AMD/R9_295_X2/22.html

 

guru3d got 507 for card ALONE http://www.guru3d.com/articles_pages/amd_radeon_r9_295x2_review,12.html

actually, if you use an amp-meter on the wires. it rarely if ever exceeds 500w at stock.

 

ive done that. ive done thermal imaging of my 295x2....

 

if you add up stuff you get higher wattages, but!

 

mobo is around 15w for  your average ATX board

RAM is around 15-20w per 2 sticks

HDD 20w/SSD 5w

fans come in addition to this. usually around 5-10w for stock fans.

typical i7 is around 90w, reviewers use X99 parts which is like 130-140w

 

so prior to any GPU, you got between 130-180w from the other parts...

 

then add a 295x2, total sum for MY system with stock i7 4790k is around 590w during valley but i have multiple fans, radiator, pump, 2 HDDs, 3 SSDs, lights etc....

[DXMember]

There is no way? Are you kidding? They ran unigine. Try running furmark? 

Techpowerup got it to pull 646W 

http://www.techpowerup.com/reviews/AMD/R9_295_X2/22.html

 

guru3d got 507 for card ALONE http://www.guru3d.com/articles_pages/amd_radeon_r9_295x2_review,12.html

I've seen those reviews, i can discard guru3d immediately because they measure at the wall for total system draw and then they do some voodoo 3rd grade maths to roughly estimate numbers

as for techpowerup they do not provide any information on how they get their numbers - if you can find one please link,

but from looking at the numbers they are giving it seems to be some voodoo magic as well

 

toms uses actual $10 000 worth professional grade laboratory equipment

and here's a quote where they refer to number reports just like you linked:

We measure just under 430W in while gaming, which is in line with the company's specifications and a lot less than the >500W figure we've seen thrown around.

[...]

The stress test is next. AMD's Radeon R9 295X2 stays just under 450W there as well, which goes to show that if you're not collecting data quickly enough, your results will be way off.

as for the data colletion they record a 2 minute sample at 2ms intervals

and they measure the actual DC power at PCI-E connectors and motherboard which what power supply ratings are

 

and the stress test they are reffering to is coin mining which pretty much is furmark

[Necrodead]

Listen to @Prysin on this one, he was the only one to mention it has to be 2x separate 28amp connections, and also they cannot be daisy-chained. You could probably get away with a 600w psu if the rest of your system has low power draw and you don't do anything like run benchmarks on your card, just stick to gaming. 

[ZetZet]

 

 

then add a 295x2, total sum for MY system with stock i7 4790k is around 590w during valley but i have multiple fans, radiator, pump, 2 HDDs, 3 SSDs, lights etc....

there you go buddy, run that on 500w.

[Prysin]

there you go buddy, run that on 500w.

the 295x2 itself CAN run off a 500w.

and depending on the PSU, some PSUs can be "overloaded" by as much as 25% without shutting off. Naturally, overloading a PSU is like overclocking a CPU or GPU in terms of IT WILL CAUSE DETERIORATION.

[DXMember]

the 295x2 itself CAN run off a 500w.

and depending on the PSU, some PSUs can be "overloaded" by as much as 25% without shutting off. Naturally, overloading a PSU is like overclocking a CPU or GPU in terms of IT WILL CAUSE DETERIORATION.

nobody cares the psus are rated for what they are rated and most manufacturers give you 7 year warrany

exploit the hardware or else what's they point in spending all the Money in it

[Prysin]

nobody cares the psus are rated for what they are rated and most manufacturers give you 7 year warrany

exploit the hardware or else what's they point in spending all the Money in it

meh, just buy a bigger PSU

 

something like what i did...

get a AX1500i for the price of a AX1200i

[DXMember]

meh, just buy a bigger PSU

 

something like what i did...

get a AX1500i for the price of a AX1200i

my seasonic platinum 1000w melted down after 10 months and now I'm just sitting here and waiting on the RMA, what do you have to say about that

[Prysin]

my seasonic platinum 1000w melted down after 10 months and now I'm just sitting here and waiting on the RMA, what do you have to say about that

shoulda bought a Titanium PSU

 

#platinumplebs

#titaniumtriumf

[DXMember]

shoulda bought a Titanium PSU

 

#platinumplebs

#titaniumtriumf

#shitsonic

 

should've gone with corsair cx would last just as long and cost two times less

[quan289]

be it single or multi-rail PSU

 

due to most PSU manufacturers still following the outdated "max 20 amps" standard, that was removed in march 2007. Many PSUs actually CANNOT power the 295x2. Because whilst they technically can, their 8pin PCIe connectors are limited to 20 amps.

 

meaning the moment you fire up the 295x2, it will trigger the OCP and the PSU will shut off

 

If a power supply has 50A on the +12v rail and there's an OCP that limits a PEG connector under that 50A rating to like 20A, then it is multi-rail unit. What you said doesn't apply to single rail, as there isn't any sort of OCP limiting the 8-pin to 150w specs. Meaning a device could, in theory, pull the entirety of the +12v off of a single cable until the OPP trips or something burns.

Also, the 150w 8-pin PEG specification is set by PCI-SIG on March 2008, and PSU manufacturers is still following it because it's still the standard being used today (technically, they don't follow it 100% either as two of the +12v wires are supposed to be v-sense wires). It's just that AMD want to violate it, and then suggested a 28A (336w on 12v) rating per connector in multi-rail units.

 

 

as for techpowerup they do not provide any information on how they get their numbers - if you can find one please link

 

TPU stated how they had measured it.

"For this test, we measure the power consumption of only the graphics card via the PCI-Express power connector(s) and PCI-Express bus slot. A Keithley Integra 2700 digital multimeter with 6.5-digit resolution is used for all measurements. Again, the values here only reflect the card's power consumption as measured at its DC inputs, not that of the whole system."

 

The average and peak was measure during gaming load of Metro: Last Light, and maximum is Furmark.  Either way, I don't find Furmark to be a representation of power consumption, as it's a power bug application that draw far beyond what you typically see in the real world.

[DXMember]

If a power supply has 50A on the +12v rail and there's an OCP that limits a PEG connector under that 50A rating to like 20A, then it is multi-rail unit. What you said doesn't apply to single rail, as there isn't any sort of OCP limiting the 8-pin to 150w specs. Meaning a device could, in theory, pull the entirety of the +12v off of a single cable until the OPP trips or something burns.

Also, the 150w 8-pin PEG specification is set by PCI-SIG on March 2008, and PSU manufacturers is still following it because it's still the standard being used today (technically, they don't follow it 100% either as two of the +12v wires are supposed to be v-sense wires). It's just that AMD want to violate it, and then suggested a 28A (336w on 12v) rating per connector in multi-rail units.

 

 

 

TPU stated how they had measured it.

"For this test, we measure the power consumption of only the graphics card via the PCI-Express power connector(s) and PCI-Express bus slot. A Keithley Integra 2700 digital multimeter with 6.5-digit resolution is used for all measurements. Again, the values here only reflect the card's power consumption as measured at its DC inputs, not that of the whole system."

 

The average and peak was measure during gaming load of Metro: Last Light, and maximum is Furmark.  Either way, I don't find Furmark to be a representation of power consumption, as it's a power bug application that draw far beyond what you typically see in the real world.

okay, then with TPU I am only concerned with what 6.5-digit resolution means... how many readings per second they record for that?

what exactly is peak power - because in the graphs it's less than tourture test, I would normally associate a title "peak power" with the power spikes modern GPUs have

 

and furmark would be pretty much very representative for cryptocurrency mining actually or distributed computing

 

EDIT: kek, looked up the Keithley Integra 2700 digital multimeter - it's just $1.5k on amazon, toms uses equipment 5 times more expensive and their data in graphs and tables is more detailed instead of showing just the final arbitrary number

[quan289]

okay, then with TPU I am only concerned with what 6.5-digit resolution means... how many readings per second they record for that?

what exactly is peak power - because in the graphs it's less than tourture test, I would normally associate a title "peak power" with the power spikes modern GPUs have

 

and furmark would be pretty much very representative for cryptocurrency mining actually or distributed computing

 

EDIT: kek, looked up the Keithley Integra 2700 digital multimeter - it's just $1.5k on amazon, toms uses equipment 5 times more expensive and their data in graphs and tables is more detailed instead of showing just the final arbitrary number

 

A $1500 piece of equipment will be far more accurate than most reviews that only used a $30 wall meter like the Kill-A-Watt, where not only it does not isolate the entire computer from the graphic card, it's may not be as accurate at higher loads.

 

Either way, I don't see how getting a more costly piece of equipment would immediately invalidate the findings of another site, if that's the point you are trying to get at. Sure, it may be more accurate, but TPU findings had shown the reference 295x2 to have an average of 430wDC which is within range of Tom's reading, and the peak (Highest single reading during the test. Basically, the peak power draw of the card under a more real world application - not synthetic) is 500w (30w difference) which can be explain by the difference in testing methodology (running different applications) as well as samples (two different samples of the same card would not yield 100% exact same results).

 

I do agree, however, that Toms' setup gives you a much bigger picture of the actual fluctuation of the power draw in real time, as the transient spike and drops in loads can give you an idea how certain card would behave in certain PSU's design, but that isn't exactly relevant atm. Don't you find it strange that the 20w average difference between mining (supposedly representative for Furmark) and gaming in Toms' figure would yield a ~200wDC difference in TPU?

 

Here's an example at Toms (that is also review by Igor) on the Fury X:

http://www.tomshardware.com/reviews/amd-radeon-r9-fury-x-power-pump-efficiency,4215.html

 

The Fury X had shown to draw 347wDC during Furmark - a 92~145wDC difference on average depending on the resolution during gaming load of Metro: Last Light (TPU test at FHD). While it doesn't show the results of mining (Seriously Tomshardware? Consistency), comparing the findings with the Fury X and TPU's 295x2, the results between gaming and Furmark seem to be more inline than the results between the gaming and mining results of Toms' 20w difference.

[DXMember]

@quan289

http://media.bestofmicro.com/R/V/509035/original/24-Power-Consumption-All-Gaming.png

http://media.bestofmicro.com/R/Y/509038/original/25-Power-Consumption-All-Torture.png

both graphs are from the article linked by you

 

295x2 has a board power limit, the power delivery keeps track of how much power it's pulling and it knows it's limits and under no circumstances will the power delivery exceed this limit, instead it will drop clock speeds or insert idle times to counter the load

so does every other card, for 295x2 this limit happens to be 450W that is in line with AMDs specification of the card, for Fury X this power limit is 350W which again is in accordance to AMDs specification, toms tourture tests show that the power consumtion indeed is in accordance to that

[quan289]

@quan289

http://media.bestofmicro.com/R/V/509035/original/24-Power-Consumption-All-Gaming.png

http://media.bestofmicro.com/R/Y/509038/original/25-Power-Consumption-All-Torture.png

both graphs are from the article linked by you

 

295x2 has a board power limit, the power delivery keeps track of how much power it's pulling and it knows it's limits under no circumstances will the power delivery exceed this limit, instead it will drop clock speeds or insert idle times counter the load

so does every other card, for 295x2 this limit happens to be 450W that is in line with AMDs specification of the card, for Fury X this power limit is 350W which again is in accordance to AMDs specification, toms tourture tests show that the power consumtion indeed is in accordance to that

 

The "Torture" graph displayed the results of the Fury X during Furmark, while the 449w figure for the 295x2 is reused with the bitmining figure; rounded up.

 

So that doesn't really tell me if Furmark can really be used as a representation of cryptomining or other GPGPU compute applications. Even going off of the figure one of my colleagues had measure a few years back on his 7970 had shown to be a bit higher at DC output. Maybe I should go about testing it myself when I have the time.

 

As for power limits, I'm aware of them and how it should throttle down, but seeing how Toms states "AMD tells us that the Radeon R9 295X2’s TDP is 450W" (I'm pretty sure its 500w btw) and how it "isn't just a rough estimate either, but a real limit." in which bitmining had comply to that, the Fury X had shown to draw up to 347wDC during Furmark and exceeding the rated 275w TDP. Since the mining value was average, the test done for the Fury X during Furmark should be average as well. If that's the case, TPU data on the Fury X is a somewhat more understandable @ 436wDC as that is "highest single reading after a short startup period".

 

Anyways, I got too much into this, and I need to get going (another long business trip). All I wanted to show was that TPU and THW findings during gaming loads wasn't that far off (setting aside Furmark). I do wish THW set up an actual format as a template on how they do / setup these reviews though.