So, how is Bulldozer's gaming power consumption?

[Maximilian]

But don't you understand how much better the AMD tin would be for storing shortbread?

Shortbread is unhealthy D:

 

[frostedflakes]

FWIW here's my 8120 system power consumption at the wall (CPU is OC'd to 4.1GHz currently at stock voltage). Keep in mind that turbo is disabled as well. 5 and 7 threads weren't actually measured but interpolated from the other data points.

Idle - 107W
IntelBurnTest 1 thread - 174W
2 threads - 195W
3 threads - 213W
4 threads - 226W
5 threads (interpolated) - 240W
6 threads - 253W
7 threads (interpolated) - 263W
8 threads - 272W

Nothing exciting, it's just a linear relationship.

 

[Idontcare]

Shortbread is unhealthy D:

So is dying. Some things can't be avoided in life. I count death and shortbread among them.

 

[thilanliyan]

FWIW here's my 8120 system power consumption at the wall (CPU is OC'd to 4.1GHz currently at stock voltage). 5 and 7 threads weren't actually measured but interpolated from the other data points.

Idle - 107W
IntelBurnTest 1 thread - 174W
2 threads - 195W
3 threads - 213W
4 threads - 226W
5 threads (interpolated) - 240W
6 threads - 253W
7 threads (interpolated) - 263W
8 threads - 272W

Nothing exciting, it's just a linear relationship.

What are the rest of your system specs?

 

[sandorski]

Shortbread is unhealthy D:

It overclocks your heart. Keep it reasonable and you'll be ok.

 

[StrangerGuy]

It's painfully obvious OP just want a double standard in power consumption numbers that is favorable to BD, pitting a not-fully loaded BD against a fully loaded SB then claim "BD power consumption is not as bad as we think". Yeah, very honest and objective.

By his same logic power consumption for Sandy Bridge is also going to drop a lot since it is also NOT FULLY LOADED.

 

[Flipped Gazelle]

It's painfully obvious OP just want a double standard in power consumption numbers that is favorable to BD, pitting a not-fully loaded BD against a fully loaded SB then claim "BD power consumption is not as bad as we think". Yeah, very honest and objective.

By his same logic power consumption for Sandy Bridge is also going to drop a lot since it is also NOT FULLY LOADED.

Or maybe the OP has an honest question, and has put forth his theory.

Have some shortbread - it's better than death, overclocks your heart, and will help you forget all this silly CPU business.

 

[frostedflakes]

What are the rest of your system specs?

990FX board, HD 6970, SSD, 2xHDD, and a handful of case fans.

 

[Belegost]

The problem is the two data points are not equal. In Bulldozer's case, the max power shows usage with 8 cores pegged, while the i5 or i7 power usage shows for 4 cores... since the cpu only contains 4 cores.

For example if i5 2500k is x at 2 cores fully loaded and 2x at 4 cores fully loaded, it does NOT follow that fx-8150 would be at y 2 cores loaded and 2y at 8 cores fully loaded... logically you would expect the power usage to increase more when you go from 2 to 8 cores rather than from 2 to 4 cores.

Because we don't have the numbers for bulldozer's power usage in an application loading 4 cores, we can't get the needed information.

My apology, I meant 2600K, which is an 8 thread CPU that performs close enough to BD that the exact architecture used to run 8 threads is a wash. Not to mention that the "8 core" BD is a little bit of a misnomer... The power differences between the 8150 and 2500K are larger.

Also, the power measurements frosted put up are more linear than I might have expected, interesting.

 

[know of fence]

FWIW here's my 8120 system power consumption at the wall (CPU is OC'd to 4.1GHz currently at stock voltage). 5 and 7 threads weren't actually measured but interpolated from the other data points.

Idle - 107W
IntelBurnTest 1 thread - 174W
2 threads - 195W
3 threads - 213W
4 threads - 226W
5 threads (interpolated) - 240W
6 threads - 253W
7 threads (interpolated) - 263W
8 threads - 272W

Nothing exciting, it's just a linear relationship.

Thank you, for sharing this! Again even if scaling is almost linear it still helps to know that single thread consumption is 64% of maximum.
More importantly idle consumption is not part of that line. thus it's impossible to make predictions about that linearity having just idle and load numbers.
Shouldn't turbo disrupt the monotonically increasing consumption at least in lightly threaded tasks and bend the graph the other way?

Is this an "I've got an honest question here" type thread?

Or is it more a "I've got an ax to grind here" type thread?

I have no ulterior motives but curiosity. Certainly a totally new architecture with 4 shared FP-cores warrants attention. Granted AT never does post Gaming Power Consumption, it wasn't a conscious omission. But it would be the nice to see numbers for core scaling, especially as the mainstream moves to 6 and 8 cores.
BD is a revolutionary architecture, something to get excited about, more so than Ivy Bridge, which appears to be just a Tick iteration.

I agree completely with the OP, and I made this point in the days following the initial bulldozer reviews. Alas, nobody really cared.

I suspected that I wouldn't be the first to point that out. More than ever it makes sense to to buy powerful hardware nowadays, because of power gating and performance scaling, looking at just idle and load is a bit negligent in this context.

Exactly... I intended for the OP to realize this (Crysis being dual threaded), and maybe explain why he uses a graph of Crysis Warhead performance after stating, "Then he comes down hard on single threaded performance... "

I picked Crysis because it was the lowest scoring lightly threaded game-benchmarks for BD that shows it at last place and also has both Phenoms as reference. It it really is dual threaded, then it should consume about ~164W playing this game. (AT review load power 229W * Frostedflakes' 2 core load percentage 74%)

 

[GaiaHunter]

Is this an "I've got an honest question here" type thread?

Or is it more a "I've got an ax to grind here" type thread?

When reviewing GPUs reviewers have started to include a few different measure, like a game and a full loading synthetic benchmark.

I don't see a reason why reviewers can't include chart showing consumption during games.

Not that will change much about overall value of a product, but I've in these forums already pointed to others that a certain CPU isn't consuming the amount of power that their full load links show. This is of some relevance due to discussions of GPUs power consumption.

For example from the i7-3960x review.

From the 7970 review.

Is the GTX 580 only using 105W?

 

[Riek]

My apology, I meant 2600K, which is an 8 thread CPU that performs close enough to BD that the exact architecture used to run 8 threads is a wash. Not to mention that the "8 core" BD is a little bit of a misnomer... The power differences between the 8150 and 2500K are larger.

Also, the power measurements frosted put up are more linear than I might have expected, interesting.

Considering he oc'ed all the cores to a fix value the power difference will indeed be very lineair.

But thats probably with disabled power states. (which makes it a pointless test in relation to pwr consumption)

The point topic starter wanted to make is that we do know when 8cores are pegged BD consumes alot but at the same time is equals or bests its competition.

In low threads it is like a fish on dryland, but its power consumption will be 'better' in contrast with the full load comparison.

 

[know of fence]

In Reply to GaiaHunter,

The Sandy Bridge E was tested with a HD 5870 (consuming in the range of 28-188 W) http://en.wikipedia.org/wiki/Evergreen_%28GPU_family). Both Charts that you linked use same i7 3960X CPU at stock frequency (4.3 GHz)

If the video card was idle (at 28W), adjust for PSU efficiency -> (lets assume 90%). Knock this off the 211 W from the wall, highlighted green in the first diagram, you get:

211 W total - 28 W/0.9 = 180 W theoretical power consumption (no graphics, no overclock)

425 W (GTX 580 System) - 180 W(see above) = 245 W consumed by GTX 580 (of the wall), or adjusted for PSU efficiency(*0.9) -> 220,5 W

This is within the expected TDP of 244 W.
http://en.wikipedia.org/wiki/GeForce_500_Series

 

[GaiaHunter]

In Reply to GaiaHunter,

The Sandy Bridge E was tested with a HD 5870 (consuming in the range of 28-188 W) http://en.wikipedia.org/wiki/Evergreen_(GPU_family). Both Charts that you linked use same i7 3960X CPU at stock frequency (4.3 GHz)

If the video card was idle (at 28W), adjust for PSU efficiency -> (lets assume 90%). Knock this off the 211 W from the wall, highlighted green in the first diagram, you get:

211 W total - 28 W/0.9 = 180 W theoretical power consumption (no graphics, no overclock)

425 W (GTX 580 System) - 180 W(see above) = 245 W consumed by GTX 580 (of the wall), or adjusted for PSU efficiency(*0.9) -> 220,5 W

This is within the expected TDP of 244 W.
http://en.wikipedia.org/wiki/GeForce_500_Series

My bad.

I quickly looked for charts and was just assuming they were using the most powerful card available.

 

[Rvenger]

But don't you understand how much better the AMD tin would be for storing shortbread?

Not with a hole in it.

 

[frostedflakes]

Thank you, for sharing this! Again even if scaling is almost linear it still helps to know that single thread consumption is 64% of maximum.
More importantly idle consumption is not part of that line. thus it's impossible to make predictions about that linearity having just idle and load numbers.
Shouldn't turbo disrupt the monotonically increasing consumption at least in lightly threaded tasks and bend the graph the other way?



I have no ulterior motives but curiosity. Certainly a totally new architecture with 4 shared FP-cores warrants attention. Granted AT never does post Gaming Power Consumption, it wasn't a conscious omission. But it would be the nice to see numbers for core scaling, especially as the mainstream moves to 6 and 8 cores.
BD is a revolutionary architecture, something to get excited about, more so than Ivy Bridge, which appears to be just a Tick iteration.



I suspected that I wouldn't be the first to point that out. More than ever it makes sense to to buy powerful hardware nowadays, because of power gating and performance scaling, looking at just idle and load is a bit negligent in this context.



I picked Crysis because it was the lowest scoring lightly threaded game-benchmarks for BD that shows it at last place and also has both Phenoms as reference. It it really is dual threaded, then it should consume about ~164W playing this game. (AT review load power 229W * Frostedflakes' 2 core load percentage 74%)

Turbo was disabled so all the cores were running at 4.1GHz under load, whether only one was loaded or all eight. With turbo the line should get closer to the maximum with fewer cores loaded and then stay pretty flat as number of cores loaded increases.

 

[LoneNinja]

Don't you think it is strange how people lambaste the FX-8150 in unison? Is it possible that the emotional consensus is at least partly groupthink, triggered by the desire to have clarity and simple answers?

Let's not forget a CPU architecture not only consists of transistors but also a million compromises. When at the end of the day two things remain: performance and power consumption, there is a multitude of ways to distort the perspective.

So lets take a look at reviews: Take this-here-site for example. I'll spare you the quotes.
Anand gives praise when pretty much the only mainstream thing that requires a massive CPU - video encoding. So far, so good.


So what is your mental picture after that, let me guess: A CPU that can barely run games and devours 229W while at it.

But if old game performance is bad, due to a lack of parallel threading. Shouldn't power consumption during games also be reduced?
Alas, we are never provided with the numbers Anand chooses to present only min and max. I dare you to find a review that has more numbers on gaming power consumption.

In Conclusion
Wouldn't you agree, that maligning SINGLE THREAD PERFORMANCE while only presenting PERFECTLY MULTI THREADED power consumption may be a little bit misleading?

Power consumption still is a deal breaker compared to the i7 2600k for me, but I really would like to know if low performance in those games at least resulted in lower consumption. Then again it may turn out that the 8 core Bulldozer actually scales pretty well with its tasks, surely we can't have that.

(Review)http://www.anandtech.com/show/4955/the-bulldozer-review-amd-fx8150-tested/11

Here is single thread power consumption, still utter fail compared to Sandy Bridge.
http://www.xbitlabs.com/articles/cpu/display/amd-fx-8120-6100-4100_8.html#sect0

 

[Phynaz]

So is dying. Some things can't be avoided in life. I count death and shortbread among them.

I disagree. I don't know a single dead person who is ill.

 

[Chiropteran]

Here is single thread power consumption, still utter fail compared to Sandy Bridge.
http://www.xbitlabs.com/articles/cpu/display/amd-fx-8120-6100-4100_8.html#sect0

I don't see i7 on that chart. It looks like idle power consumption is 3 watts better in favor of FX-8150, which to me spells utter fail for sandy bridge. 1 thread 100% loaded is 32 W in favor of i5.

Following my usage pattern, 66 watts saved during the 22 hours my computer is basically idle would more than compensate for the extra 64 watts used during the remaining 2 hours. Overall power usage for bulldozer is BETTER than i5, given my usage pattern. Since i7 is absent all we can go by is knowing it uses more power than i5, so the situation would be even worse. I don't really see the fail.

 

[Maximilian]

I don't see i7 on that chart. It looks like idle power consumption is 3 watts better in favor of FX-8150, which to me spells utter fail for sandy bridge. 1 thread 100% loaded is 32 W in favor of i5.

Following my usage pattern, 66 watts saved during the 22 hours my computer is basically idle would more than compensate for the extra 64 watts used during the remaining 2 hours. Overall power usage for bulldozer is BETTER than i5, given my usage pattern. Since i7 is absent all we can go by is knowing it uses more power than i5, so the situation would be even worse. I don't really see the fail.

So in conclusion bulldozer rocks if you like to turn your computer on and leave it alone doing absolutely nothing.

Fantastic then, thats why i bought my comp, just to turn it on and watch the lights flicker, its more of an ambience thing than something that has practical use. :awe:

 

[Idontcare]

So in conclusion bulldozer rocks if you like to turn your computer on and leave it alone doing absolutely nothing.

Fantastic then, thats why i bought my comp, just to turn it on and watch the lights flicker, its more of an ambience thing than something that has practical use. :awe:

Its called the "long idle"...and if you think its cool with Bulldozer wait until you see how awesome it is with your $550 7970

The old mantra was "race to idle"...that was so last year...the new mantra is "just idle".

 

[Phynaz]

I don't see i7 on that chart. It looks like idle power consumption is 3 watts better in favor of FX-8150, which to me spells utter fail for sandy bridge. 1 thread 100% loaded is 32 W in favor of i5.

Following my usage pattern, 66 watts saved during the 22 hours my computer is basically idle would more than compensate for the extra 64 watts used during the remaining 2 hours. Overall power usage for bulldozer is BETTER than i5, given my usage pattern. Since i7 is absent all we can go by is knowing it uses more power than i5, so the situation would be even worse. I don't really see the fail.

If your usage is only single threaded why get an 8 core BD? You would be much better off with a SB dual core. Same idle power, lower loaded power, and higher performing.

I'm also guessing it costs over $100 less to begin with.

Using a BD CPU for what you claim is your usage pattern is dumb.

 

[Don Karnage]

Its called the "long idle"...and if you think its cool with Bulldozer wait until you see how awesome it is with your $550 7970

The old mantra was "race to idle"...that was so last year...the new mantra is "just idle".

I yearn for the first "HD 7970 underperforming with an 8150" thread.

 

[Chiropteran]

If your usage is only single threaded why get an 8 core BD? You would be much better off with a SB dual core.

We were talking about power consumption, performance is irrelevant. Why do you always change the question when the answer isn't the one you want?

 

[Flipped Gazelle]

I disagree. I don't know a single dead person who is ill.

Kim Jong?