Intel's first Atom-powered tablet does not impress

Performance is among lowest measured, but what's going on with a future chip?

There must have been some jubilation and sighs of relief at Intel HQ when it landed the Samsung Galaxy Tab 3 10.1, finally grabbing a piece of ground from ARM in the tablet market. Well, the initial benchmarks are going to cause some heartburn.

The British blog GSM Arena put the Tab 10.3 through its paces and found that the software layer is good, as you would expect from Samsung, but performance? Terrible. The Tab 10.3 rates at or near the bottom of the scale in every test.

Rather remarkably, they did not go into the battery performance, but given how these things mutate from one design to the next, an apples-to-apples comparison might not be possible. Still, it would have been nice to know what battery life is like because poor power/performance has always been the knock on Atom.

The chip in the Tab is new. Intel released the Z2560 this quarter, and spec-wise, it's comparable to other ARM chips on the market. It's a dual core 1.6GHz processor capable of addressing 2GB of RAM, and the Tab comes with 1GB, which the article notes runs out very fast. That may have something to do with the benchmarks. Like an old friend always said, memory equals performance.

The review does fault the lack of memory as being a reason why some features have been left out and impacting performance, so it could be a lot of this is not Intel's fault. Still, knowing human nature, many people will look at those bar graphs and that's all they will see.

You'll notice that there was one benchmark where the Atom did well: AnTuTu, an Android smartphone/tablet benchmarking application. And just a few days ago, this ExtremeTech story on a Bay Trail benchmark was suddenly everywhere.

The story claimed that a benchmark for Intel's upcoming Bay Trail processor had suddenly appeared on the top of the AnTuTu benchmarks. For those of you not familiar with the benchmarking community, you can upload your results to the site of the benchmark software maker and the top performing machines are listed, giving people bragging rights over their hotrod for all to see. Here's one example from the 3DMark site.

The Bay Trail result meant someone testing the processor – maybe Intel, maybe an OEM partner, we don't know – had uploaded their score, and it blew away the previous best scores by a hefty margin.

Only there's a problem. As Hot Hardware points out, AnTuTu is basically nBench, a mid-1990s benchmark that was compiled by Byte magazine, which has gone out of print and came and went twice online. The new 3.3 version of AnTuTu was compiled using Intel's C++ Compiler, while GCC was used for the ARM variants. The Intel code was heavily optimized for the Intel chips, while the ARM code was not. It's all explained in technical depth in this article

The bottom line is AnTuTu numbers should be viewed at best with suspicion. AnTuTu reportedly has released an updated version of the benchmark in which Intel performance drops back down by 20% to 50% while ARM does very well.

This isn't Intel or Samsung's fault, it's AnTuTu's problem. The rest of the benchmarks for the Tab are pretty clear; Intel still has a gap to close. Bay Trail, a 22nm part, could do it. Some analysts tell me they think Bay Trail will be where Intel finally opens up a lead on ARM, but let's wait to see what the other benchmarks show.

What’s wrong? The new clean desk test
Join the discussion
Be the first to comment on this article. Our Commenting Policies