Temash Tubo Dock demo at MWC

[podspi]

My ATIV 500t (Clover Trail) is doing:

50fps with 100 fish
39fps with 250 fish
27fps with 500 fish
15fps with 1000 fish

This is all in Metro IE10. Turns out I could use my tablet as an aquarium replacement after all, woo.

 

[LogOver]

My ATIV 500t (Clover Trail) is doing:

50fps with 100 fish
39fps with 250 fish
27fps with 500 fish
15fps with 1000 fish

This is all in Metro IE10. Turns out I could use my tablet as an aquarium replacement after all, woo.

I wonder what AMD did to Clover Trail tablet it show 1 fish at 30fps.

 

[cbn]

I don't see what's so revolutionary about lowering max cpu frequency when not attached to docking station. I guess it's just good marketting from AMD for once.

AMD is not underclocking the processor when purely on the tablet battery and then returning it to normal clocks when extra power (from lap dock) becomes available. That would result in a tablet that is thicker than it needs to be for undocked mobile usage.

When the tablet is docked, extra cooling becomes available.

..... which when combined with its Turbo Dock delivers some serious power -- going from 8 W to 15 W, with extra air flow delivered through the connector to keep it cool.

This extra air flow allows the processor (when docked) to run at a higher tdp than it normally would based on the tablet's thermal constraints.

Now, we just need Intel to come out with something like this for their 25 watt cTDPup i7 ULV chips:

http://www.anandtech.com/show/6194/asus-ux31a-putting-the-ultra-in-ultrabooks/8

In fact, you can set the UX31A to 25W TDP, but it appears the cooling solution isn’t actually able to deal with the higher TDP for longer periods of time and thus the CPU ends up dropping back to 17W after a few minutes of heavy lifting. That’s hardly surprising, considering how thin the UX31A is—there’s just not much space for air to flow through.

 

[john3850]

I got 1775+ fish 60fps in ie9. with 1 gpu
I get 2575+ fish 60fps in chrome with 1gpu.

 

[podspi]

I wonder what AMD did to Clover Trail tablet it show 1 fish at 30fps.

Marketing is marketing. Especially in the tech industry, you have to wait for 3rd party reviews to really know what is going on.

 

[Blandge]

AMD is not underclocking the processor when purely on the tablet battery and then returning it to normal clocks when extra power (from lap dock) becomes available. That would result in a tablet that is thicker than it needs to be for undocked mobile usage.

When the tablet is docked, extra cooling becomes available.



This extra air flow allows the processor (when docked) to run at a higher tdp than it normally would based on the tablet's thermal constraints.

Now, we just need Intel to come out with something like this for their 25 watt cTDPup i7 ULV chips:

http://www.anandtech.com/show/6194/asus-ux31a-putting-the-ultra-in-ultrabooks/8

Oh sweet Jesus that is a damn good idea.

 

[Zstream]

My ATIV 500t (Clover Trail) is doing:

50fps with 100 fish
39fps with 250 fish
27fps with 500 fish
15fps with 1000 fish

This is all in Metro IE10. Turns out I could use my tablet as an aquarium replacement after all, woo.

Strange, my friend doesn't get near that on his tablet. Are you sure those numbers are accurate?

 

[podspi]

Here is a screenshot:

 

[jpiniero]

I wonder what AMD did to Clover Trail tablet it show 1 fish at 30fps.

Disable GPU Acceleration? If you don't have Direct2D enabled on Firefox for instance you might only get 1 fps even on Core.

 

[dagamer34]

Oh sweet Jesus that is a damn good idea.

The Thinkpad Helix uses a similar idea, though because it's still using Ivy Bridge CPUs, the chip still uses too much power when idling because it's typical target frequencies are higher.

 

[cbn]

The Thinkpad Helix uses a similar idea, though because it's still using Ivy Bridge CPUs, the chip still uses too much power when idling because it's typical target frequencies are higher.

Thanks for the info.....and regarding the additional cooling when docked ----> http://www.theverge.com/2013/1/6/38...ertible#ooid=RuNDA1ODqHaIHnEWmRxQcAG5_nFOh-_M (the fans are shown at 1:30 in the included video.)

Interesting stuff !!

 

[wlee15]

Here is a screenshot:

That's actually a different test than the one in the article. Here's the test that they used.

http://ie.microsoft.com/testdrive/Performance/FishBowl/Default.html

 

[beginner99]

power setting play a huge role too. I remember doing this test on my laptop and it sucked. Then I changed power settings and the influence was massive.

 

[Abwx]

Seems that CPU/GPU specs are 1Ghz/300Mhz 8W TDP when undocked
and 1.4Ghz/500Mhz 15W TDP when docked.

Given the IPC improvement a single core at 1.4Ghz will be roughly
in par with a Brazos core at 1.6Ghz , the most impressing being
about 100% more throughput CPU wise at equal TDP when all cores
are accounted.

GFx wise the 128 GCN CUs should bring something like 80% better
perfs than Brazos previous generation 80 CUs.

 

[inf64]

I think that Temash ,when docked, will be faster than E350 @stock,in both CPU and GPU benchmarks. CPU part with same number of cores will be much faster in stuff like Cinebench and on par in stuff that does not tax FP units that much(so ~15% IPC increase will amortize the ~15% lower clock). GPU part is good for ~80-95% delta when docked Vs stock E350. All summed up, faster product that fits in 15W TDP bracket Vs 18W for E350 (without FCH power draw counted in!).

 

[Abwx]

E350 is 1.6Ghz , i think your numbers refer to the 1Ghz C60.

There s no way a 1Ghz Jaguar core can be better than
a 1.6Ghz Brazos core unless it is in some short parts
of a code when AVX/128b FPUs paths do a huge difference.

As i pointed , the 1.4Ghz 2C JG will be as good or slightly
better than the E350 in CPU perfs for usual softs.

GPU wise it s likely that Jaguar will be better with the CPU
at 1Ghz than a Brazos GPU being supported by its 1.6Ghz CPU.

The 2C variant TDP is in the 3-4.5W bracket when undocked,
wich is notably better than the Brazos and likely that most
AMD powered tablets will use this dual core version.

 

[inf64]

I specifically said "when docked" . 1.4Ghz Vs 1.6Ghz, Kabini/Temash will win in 99% of the benchmarks, int or fp heavy(same core count). When fp heavy the difference will be more pronounced in favor of Kabini of course. C11.5 result,core for core and clock for clock, is 31% better on Kabini. Just shows how much better the core is.

 

[Exophase]

Hrm, 8W number seems new to me but apparently it has been floating around for a while.. here I thought the TDP for this part ("tablet" oriented quad-core Temash) was 5.9W. Hoping for some official numbers soon :/

4x1GHz + 300MHz GPU @ 8W should be tenable. 128 SPs (that'd be 2 CUs, not 128) goes a long way to explain its performance improvement vs even E-350's GPU while only at 300MHz. Curious about die size. With Brazos I think AMD dedicated a higher percentage of the die to GPU than any other APU on the market, and with Temash this will probably be just as true even for complete SoCs. Using a really dense process helps, I'm sure.. I hear 110mm^2 which affords them quite a lot of space for GPU and other stuff with the CPU part taking up only ~20mm^2 or so.

 

[Abwx]

The numbers for the 4C are between 5.9 and 8W , i took the higher
number since it s more cautious.

The chip is 114mm2 , of wich 12.4 are dedicated to the cores ,
no more than 10% of the area , the caches takes a lot more...

 

[Riek]

I said it in the other thread but it really bears repeating..

You can't do a benchmark comparison between two devices by changing both the load AND the allowed performance. You can't solve for two unknowns in a system of one equation.

Either you fix the load and measure the performance (easy) or you fix the performance and vary the load (hard). Since performance measurements tends to be linearly proportional with capability that makes more sense anyway, there's not necessarily going to be an easy mapping between load and capability.

In other words, keep the damn fish the same.

(this is also directly related to my constant complaint that comparing peak perf vs peak power consumption between two platforms is totally unfair if you don't keep one of the two things fixed, but people keep doing that too)

Isn't this what they did?

They set the test to auto, which increases the number of fish to get 60fps. if the card is faster the number will increase.

The atom was just not fast enough to render a fish at 60fps. (auto setting).

(thats ofcourse besides the reliability of that test)

 

[Exophase]

Isn't this what they did?

They set the test to auto, which increases the number of fish to get 60fps. if the card is faster the number will increase.

The atom was just not fast enough to render a fish at 60fps. (auto setting).

(thats ofcourse besides the reliability of that test)

It doesn't matter if it's because the test automatically does this, it still doesn't make for a useful comparison.

 

[inf64]

Do people seriously think that "new" Atom can compete performance wise with Temash in any workload(be it CPU or GPU,ST or MT)? The only thing it has going for it is lower power draw but with significantly lower performance.

 

[Exophase]

Do people seriously think that "new" Atom can compete performance wise with Temash in any workload(be it CPU or GPU,ST or MT)? The only thing it has going for it is lower power draw but with significantly lower performance.

Are you referring to Clovertrail+ or something later like Merrifield? I'll assume the former.

If you're comparing it with Temash @ 1GHz in tablet mode then 1.8GHz CT+ will have higher single threaded performance at least some of the time. Of course Temash can do better if it can turbo substantially past 1GHz in tablets (if < 4 cores are active, for instance). Beyond that, scenarios are hardly divided into "uses one core only" or "uses all four cores evenly." A wide variety of otherwise MT software won't benefit much or at all from more than 2C/4T, although some will.

In anything GPU dependent it'll be no contest, tablet TDP or not.

 

[inf64]

I was referring to latest CT+ Atom. The OoO design that they should be launching later this year should fare much better,at least in some workloads.

 

[ctsoth]

amd a8-3870k 3.6 ghz cpu, 847mz igpu, 1866 ram


ie10
2000 fish: 12fps
1000 fish: 24fps
500 fish: 50fps
100 fish: >60fps

firefox
2000 fish: 13-14fps
1000 fish: 28-29fps
500 fish: >60fps
etc.