AMD Bulldozer Eng. Sample leaked, benched

[cadaveca]

So? Whatever power the VRM pulls needs to be accounted for. Like I hear what you are getting at(and I am not denying any of it, either), however, the same could be said of VGA power draw, which most reveiwers compare in the same way I do my motherboard compares. I'm not focused on CPU-specific draw, I am focused on the CPU and VRM together, as this is what needs to be looked at for PSU choices.

When reviewers report VGA power draw, they are not testing the GPU or memeory power draw seperately, and they are not soldering wires to the PCB to measure current....they measure at the point the PSU terminates, and the component begins, as I do. Of course, the caveat here is W1zz, who actually measures PCIe power draw as well(which is powered via the 24-pin, or a molex, if the board is equipped with one).

When it comes to PCIe power draw, you can simply google "24pin burnt out", and you'll find many instances of the 24-pin motherboard connector being burnt out from excessive PCIe power draw from multiple VGAs, as well as finding mods to boards to prevent this issue.

 

[Thatguy]

So? Whatever power the VRM pulls needs to be accounted for. Like I hear what you are getting at(and I am not denying any of it, either), however, the same could be said of VGA power draw, which most reveiwers compare in the same way I do my motherboard compares. I'm not focused on CPU-specific draw, I am focused on the CPU and VRM together, as this is what needs to be looked at for PSU choices.

When reviewers report VGA power draw, they are not testing the GPU or memeory power draw seperately, and they are not soldering wires to the PCB to measure current....they measure at the point the PSU terminates, and the component begins, as I do. Of course, the caveat here is W1zz, who actually measures PCIe power draw as well(which is powered via the 24-pin, or a molex, if the board is equipped with one).

When it comes to PCIe power draw, you can simply google "24pin burnt out", and you'll find many instances of the 24-pin motherboard connector being burnt out from excessive PCIe power draw from multiple VGAs, as well as finding mods to boards to prevent this issue.

well if the vrms are sweating 10-20% of the power that your measuring, it could push up reported power consumption, for both intel and AMD.

 

[cadaveca]

well if the vrms are sweating 10-20% of the power that your measuring, it could push up reported power consumption, for both intel and AMD.

Yes, I am well aware of that. And I do not care. I am not reporting JUST CPU power draw. I report 8-pin power draw, which includes VRM efficiency, CPU draw, and possibly other components(that I've yet to find in current implementations).

The chokes are going to shed a bit, the MOSFETs, the input driver, the capacitors; every component is going to affect power consumption in some way, and if the component gives off heat, naturally this is going to translate into power consumed.

And that 20% number is acutally pretty close to real-world numbers for just about every board on the market, too.

But like I said, just like a VGA is considered the full PCB, GPU memory, VRM and all, I look at CPUs in the same fashion. The only difference is that in CPUs, memory is most commonly powered via the 24-pin.

Note that my tables are not labelled CPU power consumption:

Of course, once Bulldozer is out, these numbers will change for the 9-series board, but again, because this is dependant on CPU installed, I consider both the VRM and the CPU as a single power-draw source.

It's not about being technical..its about explain things in an easy-to-understand manner. But as you can see, when questions do arise, I am ready and waiting with explanations as to why i do what i do.

 

[Mr McC]

I didn't always own an auto shop. The ISO standards are a farce and little to no corporations pay anything to meet those standards. This isn't Canada, this is the US where corporations do what they want regardless of law or whatnot.

In Europe, some poor office chargehand gets lumbered with the task of inventing evidence that everybody is doing what they should be doing, and have been doing so, consistently, since the last time the ISO-guy came round, to get a rubber seal, coveted for tenure and contracting purposes, or so I'm told.

 

[cadaveca]

contracting purposes

ISO certification is primarily used for contract negotiations. Usually it denotes a specific level of quality within either manufacturing, or record-keeping, and sometimes in safety.

My wife has ensured ISO certification for quite a few companies now, and now does patient health and safety in a hospital, where things like ISO standards aren't something to negotiate a contract...meeting those levels ensure lawsuits do not follow any sort of incident.

 

[Thatguy]

Yes, I am well aware of that. And I do not care. I am not reporting JUST CPU power draw. I report 8-pin power draw, which includes VRM efficiency, CPU draw, and possibly other components(that I've yet to find in current implementations).

The chokes are going to shed a bit, the MOSFETs, the input driver, the capacitors; every component is going to affect power consumption in some way, and if the component gives off heat, naturally this is going to translate into power consumed.

And that 20% number is acutally pretty close to real-world numbers for just about every board on the market, too.

But like I said, just like a VGA is considered the full PCB, GPU memory, VRM and all, I look at CPUs in the same fashion. The only difference is that in CPUs, memory is most commonly powered via the 24-pin.

Note that my tables are not labelled CPU power consumption:


http://tpucdn.com/reviews/ASUS/M5A97_EVO/images/vrm_load.gif

Of course, once Bulldozer is out, these numbers will change for the 9-series board, but again, because this is dependant on CPU installed, I consider both the VRM and the CPU as a single power-draw source.

It's not about being technical..its about explain things in an easy-to-understand manner. But as you can see, when questions do arise, I am ready and waiting with explanations as to why i do what i do.

well aside from basic device operational consumption for the vrms, chokes etc and the thermal loss, what it really paints is a different picture. If you measuring 280w and we know 20% of that is thermal shedding maybe another 5% is used in operational uses and conductive losses and RFI emission that not thermal in nature we can figure easily that at least 50watts of your measurement are radiated.

so then what we see is that a really bad cpu is only 230watts not 280 and it says that intel is a good bit less leaky to.

I'd really like to know exactly what a cpu does consume to.

 

[Thatguy]

ISO certification is primarily used for contract negotiations. Usually it denotes a specific level of quality within either manufacturing, or record-keeping, and sometimes in safety.

My wife has ensured ISO certification for quite a few companies now, and now does patient health and safety in a hospital, where things like ISO standards aren't something to negotiate a contract...meeting those levels ensure lawsuits do not follow any sort of incident.

Having done some ISO work, ISO is essentialy a process guarentee. It promises that you will repeatdly do the same over and over and over again, regardless of quality. The only quality ISO does grant is that you can consistently build a good product or a crappy one. ISO make no dstinction on that end.

 

[cadaveca]

well aside from basic device operational consumption for the vrms, chokes etc and the thermal loss, what it really paints is a different picture. If you measuring 280w and we know 20% of that is thermal shedding maybe another 5% is used in operational uses and conductive losses and RFI emission that not thermal in nature we can figure easily that at least 50watts of your measurement are radiated.

so then what we see is that a really bad cpu is only 230watts not 280 and it says that intel is a good bit less leaky to.

I'd really like to know exactly what a cpu does consume to.

Sure, of course. I mean, to me, that's a given. However, my CPU will pull OVER 300W @ 4 GHz, no matter the VRM efficiency. So will many others, depending on how far you push. Like the video I posted...35a x 12v = 420W.

However, if i used your suggested method for testing, those numbers above, like the compare of 80W vs 73W, for the same clocks, highlights VRM efficiency between the two products, that wouldn't be shown using your method.

My testing also eliminates differences in VRM frequency.

I do not review CPUs, so I have no real interest in specific CPU power draw numbers. But what i can say, 100%, which started us on this conversation, is that AMD's TDP, is not the MAXIMUM CPU POWER DRAW, as shown by AMD themselves in the document I posted earlier.

The only quality ISO does grant is that you can consistently build a good product or a crappy one. ISO make no dstinction on that end.

Yes, of course. I think the industry it is applied to affects how ISO certification is approached, too. Kinda a fair bit off topic, even more so than our other ongoing discussion.

 

[seronx]

Guys, GUYS!

AM3+ is 145A that is with an 8pin so two 8 pins should give off 290A...(3480 Watts can be provided to Bulldozer at max)

TDP is maximum/worst possible case for stock

TDP isn't watts of energy but watts of heat

125w TDP = 125 J/s of heat

Heat is the excess of work, which the other watt provides for

(3480 Watts of energy, 125 Watts of heat for stock)

Check out intel propaganda for 2x8pin power oooooohhhh

Output is only 1200 Watts because there is only 12 Chokes

 

[cadaveca]

TDP isn't watts of energy but watts of heat

You got it.

Which also highlights why I test the way I do, as TDP of a CPU does not mean that PSU power budgets for the CPU are the same as TDP.

I gotta say though, I do not think your wattage numbers are accurate there...the amperage lsitings are not based off of 12V, AFAIK, but off of default CPU voltage.

 

[theeldest]

Guys, GUYS!

AM3+ is 145A that is with an 8pin so two 8 pins should give off 290A...(3480 Watts can be provided to Bulldozer at max)

TDP is maximum/worst possible case for stock

TDP isn't watts of energy but watts of heat

125w TDP = 125 J/s of heat

Heat is the excess of work, which the other watt provides for

(3480 Watts of energy, 125 Watts of heat for stock)


Check out intel propaganda for 2x8pin power oooooohhhh

Output is only 1200 Watts because there is only 12 Chokes

I thought the EPS 8-pin was rated at 28a / 12v (336 watts). So 2 would provide up to 672 watts to a CPU?

Where are they getting 1500 watts? That would suggest that a Radeon 6990 or nVidia 590 would each only require a single 8-pin or 6-pin power plug...


EDIT: Yes, I realized the CPU and PCIe plugs are different, but power ratings will be similiar given pin count.

 

[cadaveca]

I think they are talking hardware capability, not actual usage. For example, each choke can handle 50A...and there are mant chokes, the total of which gives the 1500W figure. That doesn't mean you can actually supply 1500W to the CPU.

 

[seronx]

You got it.

Which also highlights why I test the way I do, as TDP of a CPU does not mean that PSU power budgets for the CPU are the same as TDP.

I gotta say though, I do not think your wattage numbers are accurate there...the amperage lsitings are not based off of 12V, AFAIK, but off of default CPU voltage.

I thought the EPS 8-pin was rated at 28a / 12v (336 watts). So 2 would provide up to 672 watts to a CPU?

Where are they getting 1500 watts? That would suggest that a Radeon 6990 or nVidia 590 would each only require a single 8-pin or 6-pin power plug...


EDIT: Yes, I realized the CPU and PCIe plugs are different, but power ratings will be similiar given pin count.

I think they are talking hardware capability, not actual usage. For example, each choke can handle 50A...and there are mant chokes, the total of which gives the 1500W figure. That doesn't mean you can actually supply 1500W to the CPU.

+12 volts - 4 - 28 amps - 336 watts

EPS: 4pin = 336 Watts
4-pin provides 336 Watts so technically 2x8 pin provides 1344 Watts
ATX PCI-E: 6pin = 75 Watts, 8pin= 150 Watts

looked it up and finally found a place that talks about this

-------

The choke amount is 24 which equals 1200 Watts(I counted wrong in the previous posts)

1344 Watts -> 1200 Watts

The rest of the 1500 Watts are from the mainboard

24pin ATX Mobo
+3.3 volt: 79.2 watts
+5 volt: 150 watts
+12 volts : 144 watts

http://www.playtool.com/pages/psuconnectors/connectors.html

 

[theeldest]

EPS: 4pin = 336 Watts
4-pin provides 336 Watts so technically 2x8 pin provides 1344 Watts
ATX PCI-E: 6pin = 75 Watts, 8pin= 150 Watts

looked it up and finally found a place that talks about this

-------

The choke amount is 24 which equals 1200 Watts(I counted wrong in the previous posts)

1344 Watts -> 1200 Watts

The rest of the 1500 Watts are from the mainboard

24pin ATX Mobo
+3.3 volt: 79.2 watts
+5 volt: 150 watts
+12 volts : 144 watts

http://www.playtool.com/pages/psuconnectors/connectors.html

Do you have a source for the wattage limit on the EPS/ATX plugs? I'd assume that the amp limitation is based on the wire gauge with a minimum requirement. And the EPS12V 8-pin is going to use gauge very similar to the PCIe 8-pin. I have a hard time believing a CPU 8-pin can carry 4 TIMES the power of a PCIe 8-pin.

 

[seronx]

Do you have a source for the wattage limit on the EPS/ATX plugs? I'd assume that the amp limitation is based on the wire gauge with a minimum requirement. And the EPS12V 8-pin is going to use gauge very similar to the PCIe 8-pin. I have a hard time believing a CPU 8-pin can carry 4 TIMES the power of a PCIe 8-pin.

EPS
http://www.playtool.com/pages/psuconnectors/connectors.html#eps8

PCI-E
http://www.playtool.com/pages/psuconnectors/connectors.html#pciexpress8

ATX
http://www.playtool.com/pages/psuconnectors/connectors.html#atxmain24

If you notice

the EPS is basically a PURE 12 Volt cable no grounds

while the PCI-E is mostly grounds

While ATX is a variety pack

 

[theeldest]

EPS
http://www.playtool.com/pages/psuconnectors/connectors.html#eps8

PCI-E
http://www.playtool.com/pages/psuconnectors/connectors.html#pciexpress8

ATX
http://www.playtool.com/pages/psuconnectors/connectors.html#atxmain24

If you notice

the EPS is basically a PURE 12 Volt cable no grounds

while the PCI-E is mostly grounds

While ATX is a variety pack

Hey that's awesome. That's the best/most comprehensive listing of power plugs and ratings that I've seen. It says the EPS 8-pin is rated to 336 watts ...

 

[seronx]

Hey that's awesome. That's the best/most comprehensive listing of power plugs and ratings that I've seen. It says the EPS 8-pin is rated to 336 watts ...

I got mixed up xD

Ya, 336 Watts is the max for an 8pin

When I did the posts I was under the influence of staying up for 32-36 hours

 

[Heavy_MG]

The best way to test real consumption would be to grab the socket pins. Yeah I have touch probe and current meters that might be workable but nothing with beat the accuracy of a current bridge to get the most solid numbers to within a very small window of error.

I don't have time to build anything like it, but I bet some large company has it.I also know I don't have a top bin part either.

Testing how much power it takes from the wall with a kill-a-watt meter is a better measure.

 

[cadaveca]

Testing how much power it takes from the wall with a kill-a-watt meter is a better measure.

Um, no. That will tell you full system power only...harddrives, fans, graphics cards, etc...also, unless you already know exactly how efficient your PSU is, there's no way to get real accurate numbers for individual components.

 

[Horrux]

Um, no. That will tell you full system power only...harddrives, fans, graphics cards, etc...also, unless you already know exactly how efficient your PSU is, there's no way to get real accurate numbers for individual components.

...and because PSU efficiency varies according to power draw that's all bound to be a real headache.

 

[cadaveca]

Exactly. yet another reason why i do what i do in reviews.

 

[cdawall]

I got a 6 core running at 4 ghz on a 450watt power supply with a 5770 video card. Max load the system is pulling 380 from the wall. not really seeing it. I don't have time right now to make a isolator socket with a current bridge.

went well over that on mine with an old Crosshair III formula and quad core the chips suck power. i smoked a pair of 650W CWT units running a 4850X2 was running the PSU's in tandom.

 

[Horrux]

OK so *THAT* is why you need, say, a 750w PSU if you have 2 x 200W graphics cards and one "125W CPU" which is actually a 300W CPU...

 

[Heavy_MG]

Oh, look, here's OVER 400W going through an AMD CPU(Crosshair V Formula, 8-pin + 4-pin):

Oh, and look...a meter over the 8-pins. I guess some are really just not aware of what happens when you overclock to the extreme. 300W is NOTHING, and ASUS is more than aware of it(otherwise there'd be no need for both 4-pin and 8-pin)!!

Facepalm.JPG
You'll find anything to prove your point won't you?
That is on dry ice,with the chip clocked at what? 5Ghz? 6Ghz?
Any normal OC on air will NOT consume 300W.

 

[cadaveca]

Are you done yet?