The Intel Atom Thread

[ShintaiDK]

Moving the goalposts?

Is that why you took the Bay Trail with the shortest listed batterytime?

ASUS: 10.1-inch "Bay Trail" Transformer Book Trio T100TA. Battery: 12 hours. Price: $329.

Temash is simply rubbish for tablets. And why there are no products so long after its release.

The A4-1200 is also a weak 1Ghz dualcore that got no chance of competing with Bay Trail in any way.

 

[podspi]

Isn't intelligent turbo a good thing? Are we surprised when a thermally limited SoC has a lower performance than a less thermally limited one? Are we really going to call Jaguar's current lack of turbo a plus?!

Keeping in mind the power consumed by BT and the other CPUs/APUs/SoCs, I think BT's showing was very impressive.

 

[monstercameron]

Isn't this what you want from an efficient chip? Shut down what isn't needed? Doesn't it say more that the chip can run a pretty standard android game using only 100-150mW for the CPU? Also, with those numbers that means BT was pulling a total of 1.35W running the game. There is plenty of wiggle room for a more demanding application.

no one is discounting baytrails prowess, as for wiggle room...there wont be much tdp headroom left when gpu load rises for it to keep its cpu perf advantage[read turbo] compared to temash/kabini or s800/tegra 4.

 

[monstercameron]

Is that why you took the Bay Trail with the shortest listed batterytime?



Temash is simply rubbish for tablets. And why there are no products so long after its release.

The A4-1200 is also a weak 1Ghz dualcore that got no chance of competing with Bay Trail in any way.

the gpu is faster.

 

[ShintaiDK]

the gpu is faster.

Are you sure? The GPU is not running at 500Mhz like the A4 5000, but 225Mhz. And maximum memory speed is 1066Mhz, unlike the 1600Mhz from the A4 5000. Not to mention far from finished drivers for the BT.

 

[Abwx]

Keeping in mind the power consumed by BT and the other CPUs/APUs/SoCs, I think BT's showing was very impressive.

Showing is the term , reality will be different once
real reviews will be done..

The current numbers enginered by intel are clearly rubbish ,
the same 22nm producing a BT that has 6 times the perf/watt
of Haswell is not realist , two times would be logical but not 6x....

 

[Khato]

From Techreport, THIS IS THE MOST INTERESTING THING and the one to look closely,

http://techreport.com/review/25329/intel-atom-z3000-bay-trail-soc-revealed/4

CloverTrail almost use half of the TDP for the CPU and Half of the TDP for the GPU. On the other hand, BayTrail uses 89% of the TDP for the GPU and only 12% for the CPU.

100-150mW for the CPU in Gaming ??? Well, that was an Android Game (i believe they used the Epic Citadel). At that low voltage for the CPU its impossible to have turbo, not to mention that i dont expect to even use all 4 cores even at base frequency of 1.45GHz. In Windows gaming, the CPU cores will need to work more and the GPU will not be able to use all the TDP it needs, the result will be even less GPU performance.

And now when you run CPU intensive apps,

Baytrail uses two thirds of the available TDP for the CPU by closing GPU parts. By doing so, it has free TDP to give to the CPU and it can Turbo higher.​

The one flaw with the above logic being that 1.2W for GPU + 1W for running a single thread at full throttle is still 'only' 2.2W. It's only if there were a game that required 2 or more cores running at full speed that it would potentially surpass the ~2.6W seen in the multi-threaded cinebench run. And it appears that it wouldn't go past that, instead lowering CPU frequency by a small amount in order to keep power in check. (Recall that multi-threaded cinebench was roughly 3.7x faster than single-threaded despite only using 2.5x the power, which would imply that running at ~92.5% performance only requires ~62.5% power.)​

 

[blackened23]

the gpu is faster.

I don't really think so, but there is no clear data on this. None of the machines with the 1200 have been released as far as I can tell..(nor have they been benchmarked..) Besides which, at low resolutions,3d applications are often CPU limited. Bay Trail T is around 3 times faster in terms of CPU performance. So that is another factor.

Of course, I would welcome any clear GPU data on the 1200. I have not seen any as of yet aside from CPU tests (which is definitely in favor of BT-T). If you find any, let us know.

On the other hand, I have to give you credit for arguing to the strengths of the A4-1200 instead of restoring to intel conspiracy theories. Kudos on that. It really is getting old to see the anti intel nonsense such as cheating, anand was lied too, compiler nonsense, and other such garbage. I would love to see direct comparisons between A4-1200 and the BT-T, but none such exist aside from a few CPU benchmarks.

 

[Erenhardt]

Are you sure? The GPU is not running at 500Mhz like the A4 5000, but 225Mhz. And maximum memory speed is 1066Mhz, unlike the 1600Mhz from the A4 5000. Not to mention far from finished drivers for the BT.

Why 225MHz? Disable all CPU cores but one, underclock it to 200MHz and boost GPU clock by 75%... Just like BT testing was done... SDP 2.5W*

CPU:0.2 W
GPU: 2.3 W

Now you see where it is going?

 

[dahorns]

Showing is the term , reality will be different once
real reviews will be done..

The current numbers enginered by intel are clearly rubbish ,
the same 22nm producing a BT that has 6 times the perf/watt
of Haswell is not realist , two times would be logical but not 6x....

Why? One chip is primarily designed for high performance, the other for primarily for efficiency. Seems perfectly logical that BT can be more efficient at the low end, but perfectly incapable of matching the performance of Haswell at the high end. They are different designs targeting different markets.

 

[Erenhardt]

Why? One chip is primarily designed for high performance, the other for primarily for efficiency. Seems perfectly logical that BT can be more efficient at the low end, but perfectly incapable of matching the performance of Haswell at the high end. They are different designs targeting different markets.

Last time haswell was designed for low pawer consumption. They redesigned it apparently...

 

[ShintaiDK]

Why 225MHz? Disable all CPU cores but one, underclock it to 200MHz and boost GPU clock by 75%... Just like BT testing was done... SDP 2.5W*

CPU:0.2 W
GPU: 2.3 W

Now you see where it is going?

You might want to lookup the specs for the A4 1200.

 

[blackened23]

Last time haswell was designed for low pawer consumption. They redesigned it apparently...

Yeah, but it isn't designed for 2W tablets either. Haswell mobile chips are core i7 and core i5 products so obviously efficiency plays a factor, yet so does performance. If you value performance mobile Haswell is the obvious choice.

If you are willing to sacrifice the high performance of Haswell in exchange for more mobility and smaller tablets, BT-T is the obvious choice.

I'm certain you know this, as its very obvious, but they're both "efficient" for what they do. Haswell gets high battery life given the high performance that it achieves. Haswell prioritizes high performance while the other (BT-T) prioritizes super small form factors and tablets. There is also the obvious price differences between the two since they're designed for different market segments. But they are both efficient.

 

[Sweepr]

Is that why you took the Bay Trail with the shortest listed batterytime?



Temash is simply rubbish for tablets. And why there are no products so long after its release.

The A4-1200 is also a weak 1Ghz dualcore that got no chance of competing with Bay Trail in any way.

With less than a 1/3 of Atom Z3770's MT CPU performance I wonder if that thing even beats Clover Trail. It wouldnt surprise me if an Atom Z2560 was faster in MT. I doubt that 225MHz GPU would be faster than BT, maybe at 1 or 2 GPU-only synthetic tests, but with 2 Jaguar cores @ 1GHz you are going to be CPU-bound all time in actual games, especially at those ~1366x766 resolutions. I hope Jaguar 2.0 improves on this, cause this 3.9W APUs really fall short on the performance side (or maybe Bay Trail-T is too good).

 

[Shivansps]

Still ive been looking, and it seems the IGP on Bay Trail performs a bit better than NVIDIA ION2 dgpu that has been used on Asus 1215N on ICE STORM.
http://www.3dmark.com/is/419545

Considering that, i can guess that Bay Trail its gona get better windows gaming results than the ION2 gets here, there is a huge CPU advantage over a D525
http://www.tomshardware.com/reviews/amd-e-350-atom-d525-nvidia-ion-2,2905.html

I really expecting BT to at least match the E-350..

 

[escrow4]

The battery life in my Vivotab smart reading ebooks/Calibre with the screen @ 40% brightness cracks just over 8hrs, would 12 really be possible to squeeze out? I'd say the bigger problem would be the poky eMMC storage, it took like 6-7hrs to install the 70 Win 8 updates when I bought it.

 

[SiliconWars]

The E-350 is about 30% faster in Cinebench 11.5 OpenGL, which is usually a good indicator. I'm sure I saw game benchmarks somewhere but can't seem to find them now.

 

[Enigmoid]

No it is not,
The only numbers we have is from 3D Mark, lets have a look.

As you may have read, BT shares the TDP between the CPU and GPU. When you run a CPU application like Cinebench(not OpenGL), portions of the GPU shut down so the CPU has more TDP headroom and it can Turbo higher with all cores. It is the reason that BT has that high MultiThreading scaling vs Single Core (Turbo) in CB.

But what happens when you run a GPU intensive application ??

From Techreport, THIS IS THE MOST INTERESTING THING and the one to look closely,

http://techreport.com/review/25329/intel-atom-z3000-bay-trail-soc-revealed/4

CloverTrail almost use half of the TDP for the CPU and Half of the TDP for the GPU. On the other hand, BayTrail uses 89% of the TDP for the GPU and only 12% for the CPU.

100-150mW for the CPU in Gaming ??? Well, that was an Android Game (i believe they used the Epic Citadel). At that low voltage for the CPU its impossible to have turbo, not to mention that i dont expect to even use all 4 cores even at base frequency of 1.45GHz. In Windows gaming, the CPU cores will need to work more and the GPU will not be able to use all the TDP it needs, the result will be even less GPU performance.​

Correlation != causation. We do not know if it is unable to maintain CPU and GPU loads or simply doesn't need to in this case. Intel's strength has always been its CPUs and that kind of drop is reasonable. Honestly I don't care and would actually prefer for the CPU load to be as low as possible when running a game as long as its not going to affect my performance. This is just smart turbo and as AT and other reviews measured up to 2.5W Soc power usage I doubt you are running into problems when your Soc power usage is only sub 1.5W.

Baytrail uses two thirds of the available TDP for the CPU by closing GPU parts. By doing so, it has free TDP to give to the CPU and it can Turbo higher.

3D MARK is a nice example,

Have a look at the Graphics Score in 3D Mark, A4-1250 is 68% faster in Ice Storm and 69% faster in Cloud Gate. That clearly shows that BayTrail GPU performance in way slower.

I believe it will be the opposite, In real Windows games the CPU has to do a lot of work, if BayTrail cannot lower the CPU to give TDP headroom to the GPU then the performance will not be higher but lower, like the Graphics score in 3D MARK.

End of Q1 early Q2 AMD will release Jaquar 2.0, i believe they will do the same thing they did with Trinity and Rithcland. Lisa Su also said they are going to bring lower power APUs for 7-8 inc Tablets.

http://seekingalpha.com/article/167...lobal-technology-conference-transcript?page=5

You need to conclusively show that performance under a GPU only load is significantly better than performance under a CPU + GPU load. Also there is an AMD slide floating around (saw it on this site) showing that 1 jagaur core at 1 ghz requires 1w and at 1.4 ghz requires 2 watts. And I doubt windows games will really be played on this thing.

I'm still only seeing a max of ~1.2 watts on the igp and 2.5 watts on the CPU (actually including the gpu under cinebench). Its still sub 4W tdp and will fit in any tablet.

Yep and 'improved jaguar' could be a minor refresh like richland (clock and power improvements for the most part). GPU will still likely be GCN (possible sea islands). I do not expect to see anything like the jump jaguar was over bobcat.

Still ive been looking, and it seems the IGP on Bay Trail performs a bit better than NVIDIA ION2 dgpu that has been used on Asus 1215N on ICE STORM.
http://www.3dmark.com/is/419545

Considering that, i can guess that Bay Trail its gona get better windows gaming results than the ION2 gets here, there is a huge CPU advantage over a D525
http://www.tomshardware.com/reviews/amd-e-350-atom-d525-nvidia-ion-2,2905.html

I really expecting BT to at least match the E-350..

In games it will be better due to a better CPU and better bandwidth.​

 

[AtenRa]

Well, we are speculating as of now because we havent seen any windows games. We have to wait for Q4 for retail BT products to be reviewed.

 

[Khato]

Showing is the term , reality will be different once
real reviews will be done..

The current numbers enginered by intel are clearly rubbish ,
the same 22nm producing a BT that has 6 times the perf/watt
of Haswell is not realist , two times would be logical but not 6x....

Two times would be a demonstration of incompetence on the part of Intel's CPU design teams. Or has the fact that Kabini offers roughly twice the performance per watt of Haswell despite being produced on a vastly inferior process node been completely forgotten so soon in favor of discrediting Intel? (Kabini is around 4.5 watts per multi-threaded cinebench point while Haswell is around 8.) If Silvermont was merely as efficient as Kabini on the same process then it would already be at roughly 4x the performance per watt of Haswell thanks to the superior power characteristics of Intel's 22nm SoC process. Hence it's not much of a stretch to believe that Intel's designers could come up with an even more efficient design than AMD did with Jaguar given that they were specifically targeting this power envelope.

 

[Shivansps]

Well, we are speculating as of now because we havent seen any windows games. We have to wait for Q4 for retail BT products to be reviewed.

Yes, i know, but as a owner of a E-350 netbook i know its limits, and i can asure you that the IGP is not fully used in most games, while the IGP is adecuate, its mostly CPU limited all the time, thats something you cant check with a benchmark tool.

 

[Sweepr]

Fall 2013 Windows Tablet Preview

Microsoft and its hardware partners will unleash a new generation of Windows tablets for the 2013 holiday season, in a variety of price points, sizes, and form factors. The new choices I see this year are far more interesting than the 2012 entries, I think, and benefit from updates to the underlying hardware and software platforms.

What we see here are several mini- and full-sized tablets running on the Intel Atom "Bay Trail" platform, plus Microsoft and Nokia devices running on ARM. What's most interesting about Microsoft's Surface lineup for 2013, perhaps, is that they have three devices. And two of them are running Windows RT 8.1 on ARM.

Acer W3-810

Codename: Cheetah
Platform: Intel Atom "Bay Trail"
Screen: 8-inch, 1280 x 800
Battery: 8 hours
Weight: 1.2 pounds
Thickness: .45 inches
Price: $349
Key features: Smallest Windows 8 x86 tablet with one-handed use. All-day computing. Docks with full-sized Bluetooth keyboard.
Peripherals: Anti-shock back cover, capacitive stylus ($14.99), protective screen film, protective covers ($24.99 and up), micro-HDMI-to-VGA ($14.99), protective bags, mice, portable full-sized Bluetooth keyboard dock ($49.99).

ASUS Transformer Book Trio T100TA

Platform: Intel Atom "Bay Trail"
Screen: 10.1-inch, 1366 x 768
Battery: 12 hours
Weight: 2.2 pounds
Thickness: .9 inches
Price: $329
Key features: Slim keyboard docking with large clickpad, full-sized USB 3.0 on docking, ASUS exclusive apps, Office 2013 Home & Student included.

Dell Venue

Codename: Bellair
Platform: Intel Atom "Bay Trail"
Screen: 8-inch, 1280 x 800 IPS
Battery: 10+ hours
Weight: .88 pounds
Thickness: .35 inches
Price: $299
Key features: Lightweight design, HD display with 10-point multitouch, pen and wireless keyboard accessories, variety of color choices.
Peripherals: Capacitive Active stylus, Bluetooth wireless keyboard with stand.

Dell (unnamed)

Codename: Midland
Platform: Intel Atom "Bay Trail"
Screen: 10.8-inch, 1920 x 1080
Battery: 9 hours, replaceable battery
Weight: 1.65 pounds
Thickness: .4 inches
Price: $399
Key features: Optimized for 4G connectivity, HD with touch screen and pen, great accessories.
Peripherals: Pen, touch keyboard, travel keyboard with 10 hour battery, productivity dock.

Lenovo Miix 8

Platform: Intel Atom "Bay Trail"
Screen: 8-inch, 1280 x 800 IPS
Battery: 8hours
Weight: .77 pounds
Thickness: .31 inches
Price: $249
Key features: Dual cameras with auto-focus, HD video calling, Lenovo face recognition, includes a detachable keyboard, 3G version available, two color choices (black, silver).
Peripherals: Pen.

Lenovo Miix 2

Platform: Intel Atom "Bay Trail"
Screen: 10.1-inch, 1366 x 768
Battery: 8 hours
Weight: 1.3 pounds
Thickness: .4 inches
Price: $449
Key features: Smart dock with seamless connection, detachable folio case with built-in keyboard

Toshiba Encore

Codename: Malta
Platform: Intel Atom "Bay Trail," up to 1.8 GHz
Screen: 8-inch, 1280 x 800 IPS
Battery: 6-7 hours
Weight: 1 pound
Thickness: unknown
Price: $329
Key features: AutoBrite touch display; Skype certified with high quality dual-array mics; speakers, mics, and cameras placed outside hand grip zone; InstantGo to keep the device connected

More here: http://winsupersite.com/windows-8/fall-2013-windows-tablet-preview

 

[dahorns]

I really can't see why anyone would buy the W3-810 or the Encore given the competing devices and their prices.

 

[monstercameron]

Are you sure? The GPU is not running at 500Mhz like the A4 5000, but 225Mhz. And maximum memory speed is 1066Mhz, unlike the 1600Mhz from the A4 5000. Not to mention far from finished drivers for the BT.

atom z3770 12100 overall and 11,000 graphics.
[ http://www.pcper.com/reviews/Proces...rail-and-Silvermont-Arrive/3DMark-Ice-Storm-a ]

amd touts 11920 icestorm for the a4-1200
[ http://www.amd.com/us/press-releases/Pages/amd-rejuvenates-mobile-2013may23.aspx ]

the a4-1250 [higher gpu and memory clock] has 13500 overall and 20,200 graphics
[ http://www.pcper.com/files/imagecache/article_max_width/review/2013-09-10/3dm13-icegraphics.png ]

cpu, atom has[had] it beat but gpu, intel cant even touch 3 year old brazos c-50...
[ http://www.3dmark.com/is/187221 ] 10,000 overall and 12,400 graphics

anands e-350 score is pretty low again.
[ http://www.3dmark.com/search#/?mode...ch/gpu/is/P/666/500000?minScore=0&gpuName=AMD Radeon HD 6310 ]

 

[mikk]

End of Q1 early Q2 AMD will release Jaquar 2.0

Available half a year later as usual when Airmont is around the corner.