Give me a reason to go AMD... I WANT TO BELIEVE.

[Abwx]

"Law of physics"?

Hm. 3800 points against 3200 points with 5W is better with 1/3 of the power. 😵

3 x the perf/watt is impossible , moreover with both
chips using TSMC 28nm process , think a little...

 

[sontin]

How did you calculate Tegra 4's TDP ? AFAIK ARM SoC's don't reveal these numbers, Nvidia might be an exception,

nVidia published the number a few months ago.

Also 5W is not under load, in that SDP analysis on AT I saw a fully loaded Exynos 5 pulling well over 8W(it was dual core I believe) so 5W is not a number you should be waving around ! Better do your homework next time 🙄

But you believe the 15W number from AMD?! :sneaky:

3 x the perf/watt is impossible , moreover with both
chips using TSMC 28nm process , think a little...

It is possible: Two different architectures with different characteristic.

 

[Abwx]

It is possible: Two different architectures with different characteristic.

Such claims should be considered with a load of cautiousness
as this is the best way to lose any credibility with people
who have a basic understanding of electronic design ,
though you could keep on spreading this kinds of mythology
for thoses who are not in the know , assuming they have
the willingness of not being too critical of extraordinary claims.

 

[sontin]

Or:
Jaguar is not competitive with ARM.

 

[Mallibu]

It seems you are mentioning the figure that appears in the top right side. That is a net applications estimate. Below in the same page you can find a table with different estimations based in web clients. Contrast the 1% with the 2.73% given by W3Counter. It is only 0.5% behind Windows 8 share in the same row. Now look to the text before the table (bold mine):

Now take a look to

http://en.wikipedia.org/wiki/Linux_adoption

for other estimations of share. You can find from the high 93.8% of share for supercomputers to a low 8% on desktops:

Check also

http://en.wikipedia.org/wiki/List_of_Linux_adopters

Nobody is saying that all windows benchmarks are biased, only some are: the names of some of those benchmarks were given.

Finally, look to this AMD announcement

http://www.pcworld.com/article/2040...exclusivity-adopts-android-and-chrome-os.html

I think that the era of measuring performance using only windows-based benchmarks has gone.

I'm using Linux from 2007. It's market share among us, regular users and not servers or supercomputers has always been 1% to 2%. The 8% figure you are fantasizing yourself would be pretty nice if it was true, since it's a great OS, but unfortunately it's not true. If Linux had wider adoption, most review sites would include it in their benchmark suite ( don't you think they want more hits and popularity?)

The 98% of regular users, use a Windows platform and/or Mac. That's where the vast majority of benchmarks should be run, because that's what matters. Noone cares if FX can peform in 1 benchmark, in a OS that 1% of users use it. Cherry picked scenarios that matter only for fanboys of cpu x-y-z.

 

[R0H1T]

nVidia published the number a few months ago.

But you believe the 15W number from AMD?! :sneaky:



It is possible: Two different architectures with different characteristic.

Umm no I don't but hey you're the one claiming 3x performance/watt not me, so I think it's up to you to prove that Tegra4 uses 5W vs AMD's A4 at 15W running geekbench.

Or:
Jaguar is not competitive with ARM.

Intel isn't any better in this regard D:

 

[sontin]

I'm using the official TDP number from both companies.
So it's more up to you to prove that the actual power consumption while running Geekbench is not close to it.

 

[galego]

The 8% figure you are fantasizing yourself would be pretty nice if it was true, since it's a great OS, but unfortunately it's not true.

The 8% figure is not from mine. I gave you the full quote from the wikipedia link in my previous post:

In estimating true worldwide desktop adoption and accounting for the Windows-distorted environment in the USA and Canada he indicated that at least 8% of the world desktops run Linux distributions and possibly as high as 10–12% and that the numbers are rising quickly.

That 8% has been repeated by others:

So what is Linux real market share on the desktop? The best estimate for present sales is around 8%

http://broadcast.oreilly.com/2010/09/debunking-the-1-myth.html

From my own experience that 1% is a myth. Steam statistics already gives about a 2% of share and linux is not precisely popular for gaming.

http://www.phoronix.com/scan.php?page=news_item&px=MTMxNTQ

My point is simple. As many people, I consider both windows and linux benchmarks. Nobody will convince me to ignore linux benchmarks.

 

[Abwx]

Intel isn't any better in this regard D:

With stretched numbers it s not easy to etract the real datas,
Anand just made an article about low power HW , there s
a direct comparison of the SDP and real TDP of the SKUs.

TDP 11.5W 11.5W 13W 13W 13W SDP 6W 6W 7W 7W 7W

http://www.anandtech.com/show/7021/introducing-the-dualcore-haswell-skus

 

[galego]

You are right. And thx to that we see how bad AMD's performance level really is. Even nVidia is beating AMD's x86 SoCs with 1/3 of the power consumption:
Tegra 4 - 3800 points in Geekbench 2
A4-5000 - 3200 points in Geekbench 2

The first link uses "Geekbench 2.4.2" on Android 4.2.1. The second link uses "Geekbench 2.4.3 Tryout" on Windows 8 Enterprise. You cannot compare scores using two versions. Moreover, Android is a much more lightweight operative system.

Can you give a link to your "1/3 of the power consumption"?

 

[R0H1T]

I'm using the official TDP number from both companies.
So it's more up to you to prove that the actual power consumption while running Geekbench is not close to it.

And we all know TDP isn't power consumption 🙄

 

[joshhedge]

The 8% figure is not from mine. I gave you the full quote from the wikipedia link in my previous post:

You should never reference Wikipedia.

 

[jihe]

Dangerous question on a very Intel minded forum. Here we have a saying: shooting the ball in an open goal(from socker). For me its obvious: I prefer my 8350 over my 3770K. But thats personal:biggrin:

Forum was a lot less intel minded in the a64 vs netburst days.....

 

[Idontcare]

And we all know TDP isn't power consumption 🙄

http://ark.intel.com/products/75123/Intel-Core-i7-4770K-Processor-8M-Cache-up-to-3_90-GHz

System and Maximum TDP is based on worst case scenarios. Actual TDP may be lower if not all I/Os for chipsets are used.

 

[Abwx]

http://ark.intel.com/products/75123/Intel-Core-i7-4770K-Processor-8M-Cache-up-to-3_90-GHz

True over a given time but peak power can reach ,
well , 153W with Haswell faster chips , although
it wont be for no more than 10ms....

 

[galego]

http://ark.intel.com/products/75123/Intel-Core-i7-4770K-Processor-8M-Cache-up-to-3_90-GHz

The key here is on what does Intel mean by "worst case scenarios". Intel determines a processor's TDP by running a few dozen workloads on the product and measuring thermal dissipation/power consumption. Intel doesn't specify what applications they run neither for how many time.

The footnote that you quote was already available on Sandy Bridge description

http://ark.intel.com/products/61275/Intel-Core-i7-2700K-Processor-8M-Cache-up-to-3_90-GHz

But as Intel explained when introduced that generation, TDP doesn't consider turbo

Moreover, Intel has a historic record of confusing the market when claimed 7W figures for its successor, Ivy Bridge, but forgetting to mention that they had invented a new SPD metric

http://www.fool.com/investing/general/2013/01/11/did-intel-just-mislead-chip-investors.aspx

I am still waiting some link from the poster who claimed that Nvidia chip to have a third of the power consumption of Kabini 15W. Next is a real power consumption of Kabini 15W against intel chip with a claimed maximum TDP of 17W

http://ark.intel.com/products/65697/

The Intel-based system consumes near the double than the Kabini-based

 

[Enigmoid]

Next is a real power consumption of Kabini 15W against intel chip with a claimed maximum TDP of 17W

http://ark.intel.com/products/65697/

The Intel-based system consumes near the double than the Kabini-based

Couple things to remember.

There is 14 watts separating the two chips.

Intel's 17 watt ULV chips do not include the PCH (3.6 watt tdp).

tdp != power consumption

I also suspect the 15W TDP is perhaps a bit conservative, total platform power consumption with all CPU cores firing never exceeded 12W (meaning SoC power consumption is far lower, likely sub-10W).

http://www.anandtech.com/show/6981/...ality-of-mainstream-pcs-with-its-latest-apu/2

Kabini doesn't really use 15 watts so it wouldn't be wrong to assume that the 17 watt intel chip + 3.6 watt PCH would use quite a bit more power than the sub 15 watt kabini chip.

With that said perf/watt is similar; 69% more power for 50% better performance on different systems isn't that different (you can find just as great a difference on notebooks with the same cpu).

 

[galego]

Couple things to remember.

There is 14 watts separating the two chips.

Intel's 17 watt ULV chips do not include the PCH (3.6 watt tdp).

tdp != power consumption

The absolute difference is unimportant. And the 3.6 do not account for the observed relative difference in performance.

Yes, the figure that I gave above is precisely showing that TPD != power consumption.

http://www.anandtech.com/show/6981/...ality-of-mainstream-pcs-with-its-latest-apu/2

Kabini doesn't really use 15 watts so it wouldn't be wrong to assume that the 17 watt intel chip + 3.6 watt PCH would use quite a bit more power than the sub 15 watt kabini chip.

I doubt that Kabini TDP is overestimated. At contrary, I believe that it is Intel who underestimated the TDP of its ULV, because avoids its aggressive turbo.

With that said perf/watt is similar; 69% more power for 50% better performance on different systems isn't that different (you can find just as great a difference on notebooks with the same cpu).

The low model A4-5000 is targeting Pentiums performance. It makes no sense their comparison with an i3. Other models from AMD will be competing with the i3.

Moreover, those scores were obtained in a kabini prototype. Wait improvements with final hardware and the new drivers.

In any case a 69% more power consumption for a mere 50% gain in performance for the i3 gives a mere 0.72 factor. And finally F1 2012 seems to be one of those games optimized for intel.

 

[Enigmoid]

The absolute difference is unimportant. And the 3.6 do not account for the observed relative difference in performance.

Yes, the figure that I gave above is precisely showing that TPD != power consumption.

I doubt that Kabini TDP is overestimated. At contrary, I believe that it is Intel who underestimated the TDP of its ULV, because avoids its aggressive turbo.

The low model A4-5000 is targeting Pentiums performance. It makes no sense their comparison with an i3. Other models from AMD will be competing with the i3.

Moreover, those scores were obtained in a kabini prototype. Wait improvements with final hardware and the new drivers.

In any case a 69% more power consumption for a mere 50% gain in performance for the i3 gives a mere 0.72 factor. And finally F1 2012 seems to be one of those games optimized for intel.

I highly doubt that anand is wrong.

It doesn't matter where amd positions their chips. The market decides where it will compete and right now there are a low of i3 ULV notebooks on sale for $400.

Your math is wrong (have to include the base performance).

amd: 30 fps for 20 watts or 1.5 fps/watt

intel 45 fps for 34.3 watts or 1.3 fps/watt.

Using those numbers kabini is 15% more efficient per fps. (i3 is 87% as efficient 1.3/1.5=0.87 not 0.72).
Then again its the difference between playable and unplayable.

F1 2012: i3 is 50% faster
Skyrim: i3 is 54% faster
Tomb raider (720p) i3 is 76% faster
Metro LL: i3 is 66% faster

I think with other games it would have been worse from a fps to watt perspective.

 

[SolMiester]

I am ignoring the 99% of posts like this. But I will explain my signature.

I can build a top gaming PC using only AMD parts (e.g. FX and Radeon), but I cannot do that using only Intel (lacking dGPUs) or only Nvidia (lacking X86 CPUs). You need Intel plus AMD or Intel plus Nvidia.

Join date April 2013, posts pro AMD, anti Intel....I know, you are one of the new AMD advocates aren't you!...

 

[galego]

I highly doubt that anand is wrong.

It doesn't matter where amd positions their chips. The market decides where it will compete and right now there are a low of i3 ULV notebooks on sale for $400.

Your math is wrong (have to include the base performance).

amd: 30 fps for 20 watts or 1.5 fps/watt

intel 45 fps for 34.3 watts or 1.3 fps/watt.

Using those numbers kabini is 15% more efficient per fps. (i3 is 87% as efficient 1.3/1.5=0.87 not 0.72).
Then again its the difference between playable and unplayable.

F1 2012: i3 is 50% faster
Skyrim: i3 is 54% faster
Tomb raider (720p) i3 is 76% faster
Metro LL: i3 is 66% faster

I think with other games it would have been worse from a fps to watt perspective.

They tested a prototype. The market did not decide anything. The A4 is not aimed to competing with the i3, that is a task for the A6.

We are talking about different stuff, you are comparing the relative efficiency A4/i3 or i3/A4, I was obtaining the efficiency factor for the i3 alone.

Apart from the driver issue mentioned before, the gaming selection does not look well-balanced. E.g. Skyrim is one of those games optimized for Intel graphics. Moreover, if you look to Tomb Raider (768p) the gap is reduced to 21%.

 

[Abwx]

Join date April 2013, posts pro AMD, anti Intel....I know, you are one of the new AMD advocates aren't you!...

A safe discussion would be to attack , if ever you can ,
his sayings rather than making personnal attacks that
only magnify your lack of arguments , so far , since
i see none in your post.

 

[Abwx]

If you want ~200ish cpu/mobo combo, fx6300 is better than i3, at $300 mobo/cpu combo the i5 is better.

At $300 the i5 is no match for a 8350 currently
and even more in a half year , because our new
bought PCs are axpected to reach at least
this lifecycle , arent they.....

 

[galego]

Apple and Club 3D believe...

http://www.zdnet.com/apples-updated-mac-pro-is-another-win-for-amd-7000016619/

http://hexus.net/tech/items/graphics/56621-club-3d-goes-exclusively-amd/

 

[Enigmoid]

They tested a prototype. The market did not decide anything. The A4 is not aimed to competing with the i3, that is a task for the A6.

We are talking about different stuff, you are comparing the relative efficiency A4/i3 or i3/A4, I was obtaining the efficiency factor for the i3 alone.

Apart from the driver issue mentioned before, the gaming selection does not look well-balanced. E.g. Skyrim is one of those games optimized for Intel graphics. Moreover, if you look to Tomb Raider (768p) the gap is reduced to 21%.

Sure its a prototype and perhaps we may notice some difference with OEM systems but the silicon is final (NOT and ES), bios and drivers could be updated but don't expect more than -/+ 15% difference in anything. It doesn't matter where AMD aims their chips if market conditions change (just like with the release of the 770 nvidia is forcing amd to drop the price of the 7970 ghz). If the market supplies the consumer with plenty of comparable i3 ULV ivy systems for the same price as an A4 kabini that that's what kabini is competing with.

No your calculations were completely flawed. You must look at complete numbers to compare raw efficiency.

Skyrim is not optimized for intel graphics at all (and note the difference between optimized and highly single threaded).

Pretty much where you would expect it (BF3, WOW performed better relative to trinity compare to skyrim).

I'm picking 1280 x 720 for a reason (its the closest to native res of the systems these chips will ship it).