Kabini Rumors

[monstercameron]

Isn't the Nexus 10 an 8w TDP device?

Its 4w but under heavy like load it will eat 8w before being throttled

 

[Gideon]

Isn't the Nexus 10 an 8w TDP device?

Almost certainly not! That's one of the easiest misconceptions to make. TDP != max possible power usage. In layman's terms it's average power usage over time. Considering that Nexus 10 throttles until it reaches ~4W (which it does pretty soon) the actual TDP is probably closer to 4W.

Edit:
Ok it seems Anand thinks differently:

http://www.anandtech.com/show/6536/arm-vs-x86-the-real-showdown/13

The SoC is allowed to reach 8W, making that its max TDP by conventional definitions, but seems to strive for around 4W as its typical power under load.

It tends to go against what I've heard before, but Anand has a degree in Electrical Engineering and I don't, so he's probably right. Idontcare care to specify ?

Regardless TDP tells you very little about real-life workload power usage of the processor let alone the entire device. According to notebookcheck:

Nexus 10 has a 22.75 Wh battery and a peak power usage of 9.4 watts on max brightness
Z-60 has a 36 Wh battery and a peak power draw of 18.7 watts on max brightness

While their battery runtimes were:
Z-60
Idle: 13h 33min
Web: 4h 48min
Load: 2h 48min

Nexus 10
Idle: 17h 10min
Web: 10h 53min
Load: 3h 32min


If Z-60 is indeed 4.5W and Nexus is 8W, then even if we take all the other components and OS differences into account the difference is still too much in favor of the latter.

Sources:
http://www.notebookcheck.net/Review-Google-Nexus-10-Tablet.86358.0.html
http://www.notebookcheck.net/Review-Fujitsu-Stylistic-Q572-Tablet.91078.0.html

 

[monstercameron]

Almost certainly not! That's one of the easiest misconceptions to make. TDP != max possible power usage. In layman's terms it's average power usage over time. Considering that Nexus 10 throttles until it reaches ~4W (which it does pretty soon) the actual TDP is probably closer to 4W.

Edit:
Ok it seems Anand thinks differently:

http://www.anandtech.com/show/6536/arm-vs-x86-the-real-showdown/13


It tends to go against what I've heard before, but Anand has a degree in Electrical Engineering and I don't, so he's probably right. Idontcare care to specify ?

Regardless TDP tells you very little about real-life workload power usage of the processor let alone the entire device. According to notebookcheck:

Nexus 10 has a 22.75 Wh battery and a peak power usage of 9.4 watts on max brightness
Z-60 has a 36 Wh battery and a peak power draw of 18.7 watts on max brightness

While their battery runtimes were:
Z-60
Idle: 13h 33min
Web: 4h 48min
Load: 2h 48min

Nexus 10
Idle: 17h 10min
Web: 10h 53min
Load: 3h 32min


If Z-60 is indeed 4.5W and Nexus is 8W, then even if we take all the other components and OS differences into account the difference is still too much in favor of the latter.

Sources:
http://www.notebookcheck.net/Review-Google-Nexus-10-Tablet.86358.0.html
http://www.notebookcheck.net/Review-Fujitsu-Stylistic-Q572-Tablet.91078.0.html

components make a huge difference, the model of the display, supporting chipsets, wireless radios, ram (ddr3 vs lpddr3?), the os (android vs windows), sdcard reader(grr google!) the fan to cool the cpu etc.
they all add up. That said the exynos does use less power.

http://www.notebookcheck.net/Review-Asus-Eee-PC-1215B-Netbook.55843.0.html
http://www.notebookcheck.net/Review-Acer-Aspire-One-722-Netbook.56413.0.html

2 netbooks with supposedly higher tdp amd c-50 at ~17W, HDD and ddr3 (non-lp) so the fujitsu q572 with z-60 -with 4.5W tdp- seems a bit high!

 

[erunion]

8W is for the whole tablet when undocked and 15W when docked.
Contrary to your viral marketing motivated claim , this is not Temash that is docked or undocked but a whole tablet....

The person who posts solely to pump up AMD accuses me of viral marketing?

No were did AMD say 8w was system power. If they did you would have quoted it.

So either what AMD is saying about the reference design is wrong or the slide is incomplete/out of date. Since it doesn't even mention turbo dock, obviously its the latter.

 

[parvadomus]

Some more info about big cores link.
Jim Keller says they are on track to catch up on high performance core, a function of design improvements. "We couldn't pin down a timeline for this, but with a time scale of two years core design and one year build and test, it's not going to be immediate. My expectation is 2015."

 

[Evleos]

Some more info about big cores link.
Jim Keller says they are on track to catch up on high performance core, a function of design improvements. "We couldn't pin down a timeline for this, but with a time scale of two years core design and one year build and test, it's not going to be immediate. My expectation is 2015."

FYI: That quote is not from Jim Keller, I believe that's the journalist's interpretation.

 

[parvadomus]

Thats right

 

[Hitman928]

AMD are on track to catch up on high performance cores

This is an actual quote from Keller. The rest of it, timeline and such, is speculation from the journalist.

 

[Ventanni]

That's a pretty bold statement for AMD to make. I'm sure SR/EX will bring good performance improvement, but they have a lot of work to catch up to the performance/watt of Intel.

 

[frozentundra123456]

That's a pretty bold statement for AMD to make. I'm sure SR/EX will bring good performance improvement, but they have a lot of work to catch up to the performance/watt of Intel.

Well, it is a pretty ambiguous statement. Does it refer to absolute performance, performance per watt, performance in certain applications?

I can see the way AMD is going, they might catch up in absolute performance just by using a brute force, more cores, more clockspeed approach. Whether they can catch up in performance per watt on the big cores seems very questionable, since they are well behind now and intel is focusing most of their energy on improving in this area, so by 2015 or so, who knows what the target will be.

 

[CTho9305]

Some more info about big cores link.
Jim Keller says they are on track to catch up on high performance core, a function of design improvements. "We couldn't pin down a timeline for this, but with a time scale of two years core design and one year build and test, it's not going to be immediate. My expectation is 2015."

Wow, that "article" is crappy! It says Chekib was hired as part of Rory's reset, but Chekib has been at AMD since 2008 and Rory didn't start until 2011. The whole thing reads like it was written by somebody who doesn't really understand the business and just lapped up everything AMD gave them.

 

[monstercameron]

Temash tested!!!

http://ultrabooknews.com/2013/05/10/acer-aspire-v5-122p-unboxing-amd-temash-a6-1450-first-tests/

The cinebench 11.5 scores hold up ~0.8 battery 0.9 plugged in...oddly no turbo score...

 

[Vesku]

So quad 1GHz Temash Cinebenches like a single core Phenom II or Core 2 at 3GHz. Not bad for the wattage.


To compare to full desktop scores: http://www.tomshardware.com/charts/x86-core-performance-comparison/Cinebench-11.5,2753.html

(1 core @ 3GHz of a whole procession of CPU families)

 

[sontin]

Only if your application can use all 4 threads...
That is very low. A 1,6 Jaguar quadcore would be around 1,44 points or a little bit faster than a Single-Core SB processor with more than 3GHz...

I mean the PS4 only gets the speed of a DualCore SB processor with 3,6GHz...

 

[frozentundra123456]

Temash tested!!!

http://ultrabooknews.com/2013/05/10/acer-aspire-v5-122p-unboxing-amd-temash-a6-1450-first-tests/

The cinebench 11.5 scores hold up ~0.8 battery 0.9 plugged in...oddly no turbo score...

Seems expensive for the base model with only 2gb ram, since you can already get an 11.6 inch touchscreen asus with a ulv i3 and 4gb ram for less than 500.00.

Also, if you are going to make the sacrifice for a low performance processor, battery life needs to be a lot better.

 

[sontin]

I think it's clear that Kabini is not a low power design. With 8w for the lowest Kabini version with 1GHz and Quadcore you can't really win anything. And Temash is not really better with 5,9W.

 

[Vesku]

Only if your application can use all 4 threads...
That is very low. A 1,6 Jaguar quadcore would be around 1,44 points or a little bit faster than a Single-Core SB processor with more than 3GHz...

Somehow that is a bad result out of AMD's equivalent of Atom? Pentium G620T scores 1.69 : http://anandtech.com/bench/CPU/345

The Temash GPU being roughly equal to the HD3000 is telling. The current Intel mini-itx gap filler Celeron 847 has low clocked HD2000. A sign AMD will maintain its GPU strengths in this market segment.

 

[Khato]

Also, if you are going to make the sacrifice for a low performance processor, battery life needs to be a lot better.

That was my first thought as well. The estimate of 4 hours with a 30wh battery under light usage would be quite poor. Hopefully that's a result of the screen being too bright or the estimate being skewed by the benchmarks that had been done recently. (For comparison, the core i5 based v5 with a 37Wh battery is somewhere around 6 hours under light usage.)

 

[monstercameron]

The most amazing thing is that this soc beats the brazos 2 e2-1800 0.8 to 0.6 respectively for cpu, both tied for GPU perf, less board space and components and most importantly nearly half the power consumption. This is truly impressive...the ultimate netbook chip.

 

[mikk]

How is this amazing? Brazos has two cores on 40nm while Temash 4 cores on 28nm. If A6-1450 is really a 8W SoC this result is terrible.

 

[SiliconWars]

How is this amazing? Brazos has two cores on 40nm while Temash 4 cores on 28nm. If A6-1450 is really a 8W SoC this result is terrible.

The E2-1800 is an 18W chip without southbridge. This is a 6W chip with southbridge. This is better performance for less than 30% of the TDP!

 

[Vesku]

If that's terrible then I guess what Intel did with Atom going from 45nm to 32nm is criminal?

 

[erunion]

its a clamshell, not a tablet. It turbos to 1.4 so presumably has a fan and is running with cTDP at 15w.

 

[monstercameron]

How is this amazing? Brazos has two cores on 40nm while Temash 4 cores on 28nm. If A6-1450 is really a 8W SoC this result is terrible.

Compared to brazos 2 c-60, they doubled CPU perf, nearly doubled GPU perf, integrated fch yet they kept the same -a little lower- power envelope...its even somewhat competitive with ulv sandy bridge and celerons -beats current atoms- and will cost less...how is that not an amazing product?

 

[Vesku]

Compared to the preview it seems the retail model is running a bit under the 6W tdp in tablet mode: http://ultrabooknews.com/2013/02/21/amd-turbo-dock-teaser-and-temash-performance-figures/

That or Acer's implementation lacks some of the oomph of AMD's reference design.