Under $100: AMD Kabini & Intel Bay Trail-D CPUs put to the test

Steve

Posts: 3,044   +3,153
Staff member
Read the full article at:
[newwindow=https://www.techspot.com/review/806-amd-kabini-vs-intel-bay-trail-d/]https://www.techspot.com/review/806-amd-kabini-vs-intel-bay-trail-d/[/newwindow]

Please leave your feedback here.
 
Not really the tests I wanted to see for systems like this. The winrar and office tests are good, but including games that you will never play, putting photoshop and other applications that have no business being benched is not good.
What I want to know is:
- how well do the systems work with 1080p/4k video playback using hardware acceleration and software acceleration and different video players/codecs. (encoding is nice, but not the focus)
- system/software loading times (video players/browsers/office/etc)
- browser page loading times and video playback
- less demanding games (lol/dota/starcraft/mmo games/other popular or casual games that people play everyday)
- wireless display support and how well it works
- other tests that make sense for low budget systems

Getting something this low budget means that you either want to use it for office work, HTPC, or just simple web browsing.

What I learned from your tests: intel has a better CPU and some better prices in some situations, but I don't know which games can be played, how well it handles video playback, how well wireless displays work with it.
Unfortunately prices for intel motherboards and CPUs vary between regions a lot. where I am I think they are 2x as expensive than the prices mentioned, while the AMD has normal prices.

I would love to see an update with at least the video playback and wireless displays.
 
Last edited:
Not really the tests I wanted to see for systems like this. The winrar and office tests are good, but including games that you will never play, putting photoshop and other applications that have no business being benched is not good.

It’s never easy to please everyone but we do try to include as many benchmarks as we can. I would have thought that as long as all CPU’s are tested under the same conditions in the same programs the data would be valid.

Take the gaming tests that you took issue with as an example. While I agree neither the AMD or Intel budget solutions can play the games we tested, or at least not very well, the percentage differences between the two are transferable to other games (or when lowering the quality settings).

As we pointed out in the conclusion the Celeron G1820 is on average 10% faster than the Athlon 5350 when gaming, that statement will no doubt stand for other games such as DOTA and StarCraft.

You say we have no business benchmarking applications such as Photoshop on these low end systems. By that I take it you mean to say graphic designers/artists on tight budgets have no business doing what they do? :S As our results prove these low-end systems can run the latest version of Photoshop quite well. The Core i3-4130 is very usable in PS, in fact it’s bloody fast. The Celeron J1900 was around 60% slower, that’s a bit but it was still very usable and for those that have to make do will happily do so in this scenario.

Same goes for the encoding benchmarks, I know of people who do this kind of work on much older low-end systems. After all the Athlon 5350 managed 70fps in HandBrake, I remember encoding back in the day when the frame rate was actually lower than the playback.

What I want to know is:
- how well do the systems work with 1080p/4k video playback using hardware acceleration and software acceleration and different video players/codecs. (encoding is nice, but not the focus)
- system/software loading times (video players/browsers/office/etc)
- browser page loading times and video playback
- less demanding games (lol/dota/starcraft/mmo games/other popular or casual games that people play everyday)
- wireless display support and how well it works
- other tests that make sense for low budget systems

The video playback is the first point you have made that I agree with. It would have been nice to include that, I will look at making an update to the article with this info.

I feel your next request was mostly covered in the review already using the a number of applications (namely those you felt shouldn’t have been included) while a browser benchmark was included, did you miss that? There were also two office benchmarks included. Load times are something I don’t like to include in a CPU review because they are heavily influenced by the storage device.

I hear you on the gaming benchmarks, our low-end tests just weren’t low enough in this situation but the results are far from useless. In fact I feel many will find them very useful, at least we did.

Not sure why the wireless display technology needs to be tested here but I might be missing the point of that point.

Other tests that make sense for low budget systems – you will have to be more specific sorry.
 
How is overclocking with these new AMD's APUs?
My Llano NB with A6-3400M is able to run without problems @ 2.6GHz instead of stock speeds @1.4Ghz via K10stats.
 
I have been looking at these new AM1 platform chips with alot of interest. They are cool in the fact its a standard dekstop style chip in a low wattage package that not only cheap but comes with a decent range of features. I would love to get my hands on one of these platforms to see how well they would do on a small machine and try some overclocking on it.

Interesting article, was an enjoyable read. I would also like to see some HD video playback since that would be an important thing for a platform like this to be able to handle 1080p+ video playback.
 
It’s never easy to please everyone but we do try to include as many benchmarks as we can. I would have thought that as long as all CPU’s are tested under the same conditions in the same programs the data would be valid.
Thanks for the really fast and detailed reply.
You are right about the loading times. storage does affect it a lot.
I also use photoshop on a laptop with an I5 480m CPU. Not ideal but it does it's job for website designs.
Encoding is something that I don't have a lot of experience. I like that it's included in the tests though. The more the better.
As for the games, I know that getting the % from the tests is good, but it's not what I was looking for. I want to know in which games can it hit 30-60 FPS and at what level of details and resolution. I might not be able to play BF4, but can I at least play CS:GO or dota2? Testing games that people will actually play on those systems is important in my opinion, because many can and will use them sometimes as console replacements. (albeit cheap and limited in terms of what they can play)

As for the wireless displays, it's just something that I'm interested in.
If I have a laptop that supports WIDI (intel) or AWD (AMD) can I build a cheap PC with these CPUs and use it to stream my videos directly to my TV? I can't find anything good about how the tech works, what I need to make it work. what restrictions apply, is getting a chromecast, or other small ARM device that is just for wireless display better? This can be an entire new article I think ^_^
 
Not really the tests I wanted to see for systems like this. The winrar and office tests are good, but including games that you will never play, putting photoshop and other applications that have no business being benched is not good.
Sorry man, but I disagree. Photoshop performance is a great way to further determine who offers the best CPUs. It brings something different to the table and not just the usual game and synthetic benchmarks.

Great review Steve.
 
I will never ever buy Intel since 1987 to my death, just remember why.
NO, when Intel will be free and AMD price as gold.
Its bad news for Intel.
Intel pay consumers from highest price, just see this compare test, its Intel strategy and this was the test results when AMD was lightyears better and cheaper.
If AMD will close I will use the last AMD product and never buy new anymore.
lol
 
For several weeks now I was on the fence about the 5350 until I saw Microcenter was selling the G3220 for $49 then it became a no brainier which I would pick for my HTPC. I have to say after looking at these tests my choice has been validated. You just get a whole lot more power for those extra 15-25 watts!
 
It would be good; if benchmarks were about the total time, and energy consumed, to complete a particular task.
 
Performance-per-watt is not bound to one architecture. It is a generic/numeric estimate of the computer performance.

The tasks each CPU have to do are not a 1:1 comparison. For an ARM CPU to get the same results as the x86 CPU's in this test would require a massive design overhaul resulting in greater die size, thermals and power consumption.
 
Last edited:
Given the i3 intel NUC is now selling for $260 that may change this comparison markedly
http://www.amazon.com/Intel-Computing-Gigabit-i3-3217U-DC3217IYE/dp/B0093LINVK

To me the mainboard flexibility of AM1, that it's supported in a variety of boards to plug in to existing cases and power supplies, is a major advantage. intel NUC don't actually go into low power mode without the latest power supply anyway and some mainboards simply do not support this Haswell mode. ASrock alone has three AM1 boards
http://www.asrock.com/news/index.asp?ID=2039

I think anyone choosing the AM1 architecture is going to probably max the CPU regardless of the extra $20 just to avoid the bother and risk of changing it after the mainboard is out of warranty. So the solder vs. not just isn't a factor. You don't build low end PCs like this to volunteer to open them up later. You will be giving them away to your elderly relatives, putting them in the guest bedroom, etc..

I don't find either the intel NUC nor the AM1 systems appealing yet for a couple of simple reasons:

1. Just one Ethernet port. If you want to keep the desktop or HTPC silent and small, then you must keep the spinning storage drives far away over the gigabit switch. If you want the really big 1080p files to stream perfectly you may want them on the HTPC. But with just one port the HTPC is a poor NAS so you will end up with media drives attached to the HTPC and other drives attached to a NAS... maybe better for reliability (a flood, fire or jerky cat could decide to take out one set of drives) but not for simplicity. I want one two-gigabit-Ethernet machine near my big TV and another near my desk. I want them connected directly to each other and each to a switch and/or NAS. I want file systems that can handle multiple continuous >100mbit/second loads. So far the only OS that can really do this is FreeBSD/PC-BSD. Fine. I can run Linux at the desktop, Windows at the TV, and BSD on the NAS... but I want one hardware solution and two one-gigabit paths to each drive on each machine at full load.

2. No Thunderbolt yet. I would rather have 10 gigabit powered copper Ethernet get very cheap very fast, just as Firewire made gigabit Ethernet cheap, but until then the 20gbps 100-watt 2x2160p Thunderbolt cable is going to rule big multiple displays. Seriously, 1440p is now under $400 if you look and wait, 2160p ("4K") is under $600. At least *three* not two monitors are required for wraparound gaming, and three 1080p monitors will set you back a whopping $300 if you buy used... so why in the world do I want just two displays? Three is the minimum regardless of the resolutions and at present Thunderbolt is the only connector that is going to do that for me by display chaining.

2a. Three HDMI 1.4a connectors is a fair substitute, though they still require monitor power connectors and therefore are going to be far less efficient than chaining monitors with Thunderbolt, in both cable and power terms. So far the low end systems do not seem to have triple monitor support.

3. Uncertainty about the flash RAM interface. SATA6gbps SSDs just seem like a bad solution as they add a bottleneck and can only be managed as single drives - for RAID0 you need two drives. PCIe SSDs fit in those unused x4 PCIe slots and can be managed better, as multiple drives usually, supporting up to RAID6 on a single device, or splitting in half to accelerate you to RAID0 type speed. RAID0 is a 800mph go-kart anyway so if you aren't backing up this boot image on spinning metal every week, and keeping your data elsewhere, you're a fool - best to keep such images small and local to what the machine is doing, and to keep all the user data in a vault say off on the NAS far from jerky cats.

Now there are Ultra-DIMMs which seem high end but so did PCIe SSDs a few years ago.

I would like to see AMD announce whether there will be an HTX3.1 direct interface for Ultra-DIMMs etc. or whether they see the PCIe slot as being the way for big flash arrays to be connected to the CPU. I think with their deals with OCZ they are signaling the latter.

4. Other port and connector issues. The whole thing seems uncertain right now.

Thunderbolt, USB3, SATA6gbps, mSATA, mini-PCIe, PCIe... there are getting to be competing standards for how to hook up a lot of flash ROM inside or outside the box. I find this very disturbing. I was glad when eSATA died since it never worked. Glad when Firewire died because USB3 and gigabit Ethernet ate its lunch. I will be glad when some of these other connectors die in favor of say GIGABIT POWERED ETHERNET which works now and does everything in VoIP and security. And TEN GIGABIT POWERED ETHERNET which could remove the need for Thunderbolt up to 1440p...

5. Apple hasn't moved yet. Due to its unthinking fanbois and fangrrls, Apple can create an overnight lead market just by announcing the specs of the latest Mac Mini. If it's say an i7 based intel NUC with two Thunderbolt connectors and two gigabit Ethernet connectors (one powered) and two USB3 ports, and a PCIe slot useful for a 10 gigabit Ethernet or PCIe SSD or fiber optic interface, that will tell us a lot. Mainly that Apple is smart. Or it could be the same thing but based on a new AMD socket simultaneously announced, putting AMD boards seriously into the Thunderbolt market also, and giving Apple two viable Thunderbolt optimized CPU suppliers to choose from (what I would do).

If Apple was brilliant of course they would release both an AM1 based low end Mac Mini refresh that can handle 3x 1080p and a high end brand new intel NUC they can have exclusive for say two years that can handle 3x 1440p or 1x2160p+2x1080p displays. And an Apple branded NAS. All at typical Apple inflated price points, but loaded with goodies like

- PCIe SSD configured for RAID0 with automatic image backup and restore from network storage or the cloud. RAID6 for the NAS running a stripped down MacOS/X which is actually really just BSD...

- excellent pricing for the 2160p "4K" display and 1440p "iMac/Cinema" displays refreshed with actual VESA mount points for multi-monitor arrays (right now you have to buy absurd adapters)

- bundling with a new iPhone the size of a Samsung Note, and an Apple iWatch, giving you the walk-around and sit-down experiences you actually need, and eliminating the need for tablets, laptops and making the wait for AR glasses/ Google Glass a lot more annoying as you will not need anything new until then.
 
It is the ceiling power and specific processing efficiency that generalizing performance perwatt useless. Just like INTEL said x86(CISC) is a "jet" while ARM(RISC) is a "propeller plane". That's why ARM (RISC) advancement in PC are slow. Years ago APPLE desktop's MAC series use RISC cpu produced by IBM called "Power" series, now they use INTEL's x86 (CISC) with customization.
 
I agree. The typical configuration for these low power CPUs has to be the HTPC. I mean, the form factor of the AM1 boards (mini-ITX or micro-ATX) indicates that purpose.

Anyway, I think that the most interesting point of this review was the fact that all of the devices are pretty close when it comes to idle power consumption. With that in mind, an Intel 1820 + H81M configuration looks like the best generic, low budget solution
 
Are you sure those PCMark 8 home results are correct? In every other 5350 review with the same benchmarks, the Celeron and Pentium CPUs score considerably better than the 5350.
 
Back