Bay Trail Power Considerations and Closing Thoughts
Due to the limited time with our Bay Trail testing platforms doing the standard rounds of battery life testing and comparisons was impossible, but I was able to get some solid information on what you can expect going forward and how the new SoC compares the previous Clover Trail options. You might be surprised to learn that even according to Intel, Bay Trail does not use less power than Clover Trail.
During part of our briefing we were taken to a room where a Clover Trail and Bay Trail system were working side by side and connected up to advanced power monitoring hardware and software. Measuring power at the battery we found that both BT (Bay Trail) and CT (Clover Trail) systems were using approximately 2.6 watts at idle in Windows 8. That is not the SoC power draw, but the entire system including display and accessories. When we fired up a web browser and loaded up CNN’s website, BT used 2.9 watts while CT drew a bit more at 3.2 watts.
Direct hardware monitored power consumption testing
Video playback was tested both streaming and with 1080p local content and Bay Trail used about half of a watt less power consistently. That doesn’t sound like much, but where milliwatts matter on tablets and smartphones, these kind of improvements are a big win for Intel. Couple that with the documented performance advantages you get with the Atom Z3770 and the overall package the company is offering now is head and shoulders above last years’ options.
While gaming or running other tasks we continued to find that BT and CT were using very similar levels of power though performance varied pretty dramatically. Take a typical run of SunSpider where the Bay Trail platform SoC pulled about 1.4 watts (CPU + GPU) while the Clover Trail SoC was slightly lower at 1.25 watts (CPU + GPU). Score results though favored Bay Trail with a 346 ms result compared to 720 ms with Clover Trail. The result is much less consumed power to complete the task – Bay Trail has an advantage of 79% (~900 mWs vs ~501 mWs). For applications and workloads that can “hurry and stop”, Bay Trail can offer both better performance and better battery life than Clover Trail.
There are some areas where this method won’t work of course – video playback is consistent and can’t be “sped up”. In that case, Intel has tried to lower total platform power consumption to improve the story over Clover Trail.
We still need more time with Bay Trail to really make a final assessment of the battery life improvements, but that will require some retail, consumer-ready devices to be evaluated.
Closing Thoughts
Bay Trail is an impressive design and the Silvermont architecture looks like it could really move Intel into the tablet markets previously dominated by ARM/Android and Apple. Intel has been claiming that this is its “Conroe moment,” comparing this redesign of the Atom processor to the shift Intel made on the notebook markets after finally migrating from the power hungry Pentium 4 CPUs to the efficiency of the Core architecture. I cannot argue with the fact that Bay Trail has all the qualities necessary to make that kind of forward movement, but I really need to see design wins from the major players to confirm it.
When I reviewed a reference AMD Kabini APU notebook this past spring I came away very impressed with how much the company had progressed in a short time. Clover Trail was slow and AMD’s APUs promised performance and graphics prowess in a smaller form factor. Though AMD didn’t send us any Temash-based tablets or notebooks I did purchase the HP TouchSmart 11z to see how much of our excitement was warranted. As it turns out, Bay Trail is on another level of performance and efficiency than either Temash or Kabini from AMD. Look back at our Windows tests and see how many times the 3 watt Atom Z3770 matches the performance of the 15 watt A4-5000; or how much faster it is than the AMD A4-1250 at 4 watts.
Admittedly, part of AMD’s power concerns are platform dependent, but in a world where OEMs are more likely to depend on a complete solution from the SoC vendor, raw stats on processor cores won’t get you very far. Intel could really shut down the competition from AMD by getting tablets like the reference model we looked at today in the $400-500 range but even at $600 the performance and battery life advantages of Bay Trail will be hard to beat.
On the Android front there is still a lot of work to do for Intel. Today’s results don’t show the same kind of performance advantage over the competing ARM-based designs that we saw under Windows 8, but we were told that should actually improve as we near the slightly delayed Android Bay Trail platform. Based on our tests and experiences this week, I think Intel has put itself in a great position to offer up high performance tablets in a market that they had basically zero presence in with Clover Trail. Yes, Qualcomm and Samsung have higher performance parts due out very soon, so things will likely fluctuate before the end of the year.
When Clover Trail tablets were announced I was excited about thin and light tablets using Windows rather than Android. As that turned out, the performance and experience on CT was pretty bad and Intel knew it. Bay Trail is a completely different story and even our reference models made it obvious that these machines can be used as a dedicated PC while leaving the door open for the “all day” battery life goals we have had for devices like this for years.
Silvermont and Bay Trail have changed Atom, taking it from a product line that was often ignored in Intel’s portfolio to what is likely the most important architecture and launch for the company in some time. Market growth is in tablets (and smartphones) and with the Atom Z3000 series of SoC Intel has not only a competitive solution for both Windows and Android environments, but product that can finally put some pressure on the likes of Qualcomm and ARM.
Expect to see Windows-based Bay Trail systems before the end of the month and Android designs in early 2014.
What is the clock speed of
What is the clock speed of the z3770 you are testing?
How could let Intel not talk about TDP… is it that bad?
It’s not great. The Z3770
It's not great. The Z3770 bursts up to 2.4 GHz with a base clock of 1.8 GHz.
Given that all the reviewers
Given that all the reviewers are benching and testing what is top of the line Atom Z3770 (with dual channel mem and turbo of 2.4Ghz), it is reasonable to expect that the current OEMs using the Clovertrail Z2760 in pricey tablets/hybrids would utilize the new top line Baytrail in similar pricey products?
Another thing to consider is how well or not well Intel has supported the current Atom lineup with buggy crash prone drivers that has ticked off a large number of owners for the past few years.
To top it off, Intel claims that the new Atoms will be found in low priced tablets by the end of the year and if so, those Atoms will be very low spec low clock single channel CPUs.
As you say, the proof of viability of Intel’s new Atoms will be in the price of their product pudding.
And eat healthy food that
And eat healthy food that will make you fit and fine.
Many nutritionists believe that the best time to eat food is when a person feels hungry since
this is arguably the body’s natural way of signaling that our
energy supply needs replenishment. The cardinal sin committed by many endurance athletes is complete disregard of strength training for fear of slowing
them down by adding muscle.
Would be nice to see this on
Would be nice to see this on say a low power server system, perfect for File/Media Storage/Streaming as an option to a 35W TDP CPU.
maybe get rid of the h.264
maybe get rid of the h.264 encoder/decoder and make the GPU even smaller?
Poor AMD it looks like they
Poor AMD it looks like they have lost this round already. 🙁 Well if we look at the bright side, it at least appears that their new chips seem to be producing decent performance numbers in graphic intense applications.
This is a stunning upgrade
This is a stunning upgrade from clovertrail .. Intel has finally arrived in this space.
interesting to see the higher
interesting to see the higher 3dmark performance on android!?
also, ST performance on cinebench looks equivalent to a A64 3000+ from 2003, but obviously 4x cores and the power efficiency can justify the 10 years difference.
Considering that this isn’t a
Considering that this isn’t a desktop chip and has much lower power consumption, I think your comparison is a bit ridiculous.
U can’t really compare a 2003
U can’t really compare a 2003 desktop cpu with this
Architectures have changed and in 20003 u still
Had ddr ram now it’s ddr 3.
Besides I have compared a core u duo 1.8 ghz cpu with this this baytrail and the baytrail seems
To do Everything faster
Is it possible to do a NAS
Is it possible to do a NAS benchmark for the CPU?
I currently use a DIY NAS and it can easily max out a gigabit connection, but it is using a a CPU (AMD APU) that pulls 65 watts. Handling NAS loads and media streaming, and VPN access does not put high load on the CPU.
I don’t like premade NAS devices because there are many that cost $1000 with no hard drives and still perform worse than a $550 NAS including 6TB of storage.
It would be awesome if this CPU can work well in a low powered and near silent NAS build
What he said. I am also very
What he said. I am also very interested in the prospect of this technology being used in this way
What kind of test would you
What kind of test would you like? How could I get a benchmark that is relavant?
Isn’t the temash HP laptop
Isn’t the temash HP laptop based on HDD? Did you change the Temash to a SSD to compare against BayTrail?
Also, how about PCMark / Vantage results for comparison?
Thanks.
untill i dont know die size
untill i dont know die size of this Atom and its max TDP I cannot apreciate it. Intel avoided to give those numbers, so something is strange.
lets consider that in this moment AMD Jaguar A4-5000 get a die schrink at 22 nm . Then will beat hard the ass of Intel in power, graphics, performance and whatever you want.
Beat Intel?really?better hope
Beat Intel?really?better hope Intel doesn’t offer this with 3.5 w.if Intel does I’m pretty sure they could up GPU frequency.at 3.5 they would beat and in all but heavy gaming .nobody do hardcore gaming on tablet.
Are your benchmarks using a
Are your benchmarks using a single core, or all available cores? This is very important, as you directly compare to AMD’s competing Temash 1250, which is only a dual core part. You also fail to specify that it’s dual core in the spec breakdown, though you give Clovertrail that privelege.
Not getting my tinfoil hat out or anything, and my point is moot if the benchmark only tests a single core, but you should probably highlight (and bold, and underline) core counts as they play a huge role in benchmark results if multithreaded…
Sorry, I should revise my
Sorry, I should revise my comment to say this is only based on reading the first page – though the points are still valid for that page.
Wow, apologies. I meant the
Wow, apologies. I meant the first page with benchmarks (SiSoft Sandra). On a roll tonight..
Aw, this was an incredibly
Aw, this was an incredibly nice post. Finding the time and actual effort to generate a superb article… but what can I say… I hesitate a lot and don’t manage to get anything done.