Intel's Atom doesn't offer nuclear power: first benchmark

Ah, the tradeoff between processing performance and battery life. It's probably the most challenging compromise that we mobile device users have make. Too many horses under the hood and your battery peters out in two hours or less. Slow down the CPU and you can compute for five hours... at a relative snail's pace. My hope for the Intel Atom chipsets was not too much of a performance hit but greater run-time. Although it's way too early to tell, one of the first performance benchmarks has surfaced on the new Atom and it makes the 900 MHz Intel Celeron look good. Note: shorter bars are better.

Again, far too soon to see how this will shake out once we finally see some devices with the Atom, but it's clear that our age-old compromise isn't going away anytime soon. I should also mention that the benchmark test isn't one that I would have chosen, so when we get our hands on an Atom device running Windows we'll give it our own test.

Comments

[ChangBM]

To put things in perspective they should have benched the Via C7m ULV processor with it, since thats what its competing against.

[Dave P]

Any idea where one of the Via C7-M or Intel Stealey chips would fall on this benchmark's scale?

[Nedy78]

Well, from what I can piece together from the web and taking approximately half of the stated times listed, here's how I think they stack up:

1) U1400 1.2GHz core solo - 01:01
2) Pentium M 1.1GHz ULV - 01:18
3) Intel A110 800MHz - 3:11

http://www.tabletpcreview.com/default.asp?newsID=980
http://forum.brighthand.com/showthread.php?p=1393178

[wls]

wow. Thats sad, just sad. We seem to be going in circles and what we really need is better battery technology to fuel the next level of pocketable Intell devices. So after all that's said and done the new Intel chips just run effectively slower to get more battery life. Sounds like a job for CORTEX!

[Shindel]

stop spreading bad info Nedy, or the unknowing enthusiasts like Kevin/CT will take it as fact. 1st, you cant guess a number by cutting in half a previous benchmark, thats ridiculous. leads me to believe you have no understanding of architectural design whatsoever. also previous benchmarks which are not assembled for new architecture are *extremely* misleading as they dont "run through the pipes" right.

these new benchmarks are worthless until the tool is updated. the old days of slow advancing wide adopted desktop architectural designs are over. there mobile market is flooded with new architectures numerous times per year & most tools are not properly updated do to the lack of market penetration. most mobile bloggers are to ignorant to understand this & post benchmarks based on non-optimized tools.

[Mike Cane]

>>>so when we get our hands on an Atom device running Windows we'll give it our own test.

All this is confusing the hell out of me. I wish Intel would still to one frikkin name.

This *isn't* the chip powering MIDs, then? Because that's what the snip seems to indicate. Because if this *is* a MID chip, wouldn't the OS be speedier, lighter than Windows?

[Kevin C. Tofel]

Shindel, as I said in the post, this isn't the test I would have used. It's not necessary to call people "unknowing" and "ignorant" while enlightening folks with your knowledge. This *clearly* isn't the best test, but I think it's important that when folks hear a 1.6 GHz clock speed for the Atom, they don't equate it to a 1.6 GHz Core Duo CPU. This post is meant to reiterate that and the fact that our performance / battery life compromise will continue.

Mike, this is where the naming of UMPCs and MIDs will continue to cause confusion. This chip will be powering many MIDs but not exclusively. If I recall correctly, the new Asus R50 (and possibly the R70) will use it to run Windows Vista. Then again, I'm "unknowing", so I could be wrong. ;)

[ChangBM]

While optimizations haven't taken place it doesn't totally mean this bench is worthless, like Kevin said, it helps differentiate these processors.

Also while not giving a "real world" test, brute force number crunching is the easiest way to gauge performance, so I say, Atleast we have something to go on, as to the performance of what the chip might be.

Also I don't think superPi was assembled to run better on one system from another, wasn't the program made in like 1995?

[fraglez]

my test A110 800mhz (with mp3 encoder in background on idle - because i dont turn out it)
SuperPI 1M = 2:10 (without mp3 probably 1:50 i test when mp3 is done ;-)

Btw

Via C7 1.2ghz SuperPI1M: 7:02

[fraglez]

Ok - now i test in optimal conditions:
Intel A110 800 mhz (Samsung Q1 Ultra)
SuperPi 1M = 1:42
Faster then Atom!

[inomine]

This is a fairly useless test for these CPUs. SuperPI mostly test floating point performance than anything else, so unless you intend to do a lot of scientific computing this benchmark is completely pointless. It was always obvious that the Atom would make for a bad CPU in terms of it ability to do maths, thankfully this is *not* an issue for most applications out there. I suspect that with some realworld benchmarks on applications that people will actually be running on these CPUs they won't be anywhere near as bad.

Anyway, the big draw for these things is the tiny amount of power that they use.

[Ctitanic]

Another Intel FIASCO. And according to what I have seen, it does not come with a chipset capable of running Vista.

[bluespapa]

I'm in the corner of swapping power for battery time. I find that having more ram is more important with all my flavors of computers, though I admit that it was sometimes painfully slow to run early Palm devices and much happier with more powerful processors (and no moving parts).

My students won't spend a fraction on computers what they'll spend on gaming toys.

[Adamantite]

How would the AMD Geode compare here?