Architectural differences between Core and Atom processors?

[HTay]

Good day,

I'd like to consult you guys on the question, what are the differences in the way the Core CPU's and Atom CPU's are designed? (Over here I'm referring to the ultrabook/tablet line of CPU's like Core i5-4202Y or Atom Z3735F.)

What I mean to say is that, from the published specs, both the Core & Atom chips are 64-bit, have 4 threads, have low power ratings, have HD integrated graphics, etc. etc. but from experience I know that the Core chips are more powerful than the Atom chips. So what I'm asking is, without talking about experience or benchmarking results, how do we explain the differences to the educated layperson on a more 'fundamental' level, and/or why he/she should pay significantly more money for a Core chip when the Atom chip can also perform many of the tasks that the Core chip is able to?

Many thanks!

Hugh

[idata]

Hi Hugh,

I've seen a lot of these questions which no one could really get a simple answer without digging in to the specifications since there are too many differences. For me if I were asked this question from somebody who does not have a lot of knowledge in this area, the best way I could think of is with another question.

Q: Why pay significantly more money for a Core chip when the Atom chip can also perform many of the tasks that the Core chip is able to?

A: Even if the Atom Chip could perform many of the tasks the Core can do. The real question is which one performs it better?

Hope you like my insight on this.

All the Best,

Leon

[HTay]

Hi Leon,

> A: Even if the Atom Chip could perform many of the tasks the Core can do. The real question is which one performs it better?

Hmm, but that would still require the user to have prior experience or benchmarking results for both chips, right? For example, in my case, I don't use any machines with the latest chips from a non-Intel x86 vendor that shall-not-be-named, so how can I answer the "real question" on whether the Intel chips or non-Intel chips do the stuff better?

Even if I were to eventually get my hands on the non-Intel chips for benchmarking, ok, maybe I would know which is better, but I still wouldn't know *why*. So it would be like a magic black box and sometimes Intel chips are better, sometimes not, but each time equally unpredictable which would come out ahead, no?

Anyone else would like to weigh in on this as well?

Thanks,

Hugh

[Adolfo_S_Intel]

Hello

As you can see in the following link:

A quick comparison between the two technologies reveals that the Core i5 has support for more memory, and faster memory, it also has more memory bandwidth, and the Intel HD Graphics 4200 are better than standar Intel HD graphics of the Atom.

As for comparing processors of different manufacturers you can always go to different benchmarking sites like http://bapco.com/products/sysmark-2014 htttp://bapco.com/products/sysmark-2014

Best Regards,

Adolfo.

[HTay]

Hello!

> A quick comparison between the two technologies reveals that the Core i5 has support for more memory, and faster memory, it also has more memory bandwidth...

That's definitely a valid point, that the Core chips can attain higher throughput of data between RAM & CPU, but after the data goes into the CPU, what happens next? Please correct me if I'm wrong (because that's what I'm trying to understand better here) - both Core and Atom architectures have access to the same or highly similar subset of the entire x86-64 instruction set, correct? E.g. ~ same types of registers, stacks, ability to 'look forward' in the instructions, perform out-of-order tasks, etc.? If that is the case how are the Core chips visibly more powerful than the Atom chips?

> Intel HD Graphics 4200 are better than standar Intel HD graphics of the Atom.

That would mean on top of my original question I now have to ask "What are the architectural differences between HD Graphics 4200 and HD Graphics?" =)

> As for comparing processors of different manufacturers you can always go to different benchmarking sites...

Benchmarking is certainly helpful, but as I mentioned before I'm more interested in this question on the 'fundamental' level, because benchmarking is more on the 'observational' level that can only be done after the chips have been designed and fabricated. So what I'm trying to ask is, what sort of 'high-level' technical goals does Intel tell the engineers when they want to design a new CPU? Surely it isn't as simple as, like, "Let's squeeze in more transistors to get more FLOPS" and then do benchmarking after that and call the design with the best results "Core i7", correct?

Would any Intel engineers like to take a shot at this question as well?

Thanks!

Hugh

[FMcNu1 · Accepted solution]

Hugh

Let me add a comment here. First please know that this Embedded Community focuses on embedded applications, not really the tablet sort of products. So my response is in the embedded context.

There are a few tools that you might find useful.

1. www.edc.intel.com select ->hardware -> processors & chipsets. Then expand [+] the Atom and Core sections and you will find a summary table showing the features of the 2 product lines.

2. ark.intel.com here you can compare processors at the detailed level. For example, here's a link to the comparison of the Atom E3845 which I think is the highest end Atom, with the new 5th generation Core i3, which is the lowest end of the new Core family:

http://ark.intel.com/compare/78475,84697 ARK | Compare Intel® Products

You will find numerous differences.

At a very high level, Core targets applications requiring high performance and exceptional graphics performance, with good performance per watt plus advanced features. Atom offers lower power (longer battery life in portables), smaller footprint, with more modest performance and graphics performance.

To round out the Intel line, the new Intel Quark SoC offers much lower power and smaller footprint than Atom but with less performance. At the other extreme, above Core is the Xeon family that delivers server class performance for applications like storage and networking infrastructure.

I hope this helps

Felix

[CSpau]

I actually have a similar question. I'm trying to compare the Intel HD Graphics in the Atom E3815 to the Vivante GC800 and GC2000 but I can't seem find similar specifications. The specs I found for the GC880 so I wanted to see if anyone could tell me similar specs for the Intel HD Graphics in E3815?

Performance SpecVivante GC880 (i.MX6S) Pixel Rate(Mpixels/s)266Triangle Rate(Mtri/s)35Vertex Rate(Mvertices/s)100

[CarlosAM_INTEL]

Hello SAGEE,

Thank you for contacting the Intel Embedded Community.

The information that may help you is stated at http://www.intel.com/content/dam/www/public/us/en/documents/white-papers/atom-e3800-graphics-media-performance-white-paper.pdf Intel® Atom™ Processor E3800 Product Family Graphics and Media Performance in Intelligent Systems White Paper.

Please let us know if this information is useful to you.

Best Regards,

Carlos_A.