Is the Intel Atom going to beef up mobile performance?

There's a lot of buzz about the June release of the Intel Atom processor.  The Atom is Intel's new mobile platform designed to be used in Mobile Internet Devices (MIDs) and eventually in Ultra-Portable PCs (UMPCs).  I say eventually because according to the blurbs I've seen from Intel the second phase of the Atom, formerly code-named Silverthorne, is the one they claim will run in UMPCs that need more performance  The lower end Atom based on the Diamondville chipset is not recommended for running Vista and Intel has hedged on what it will do running XP.  A lot of excitement is being generated by word that Asus is planning an Atom-based model of the EEE PC later this year and I'm also hearing a lot of similar excitement over the thought that HP will move from the Via processor to the Atom this year too.  While HP has never indicated they would make that switch if they find it in their best interest to do so I'm sure they will.

So what does the Atom buy us in these notebooks over current solutions?  No question the power consumption is much lower with the Atom, that is one of its designed strengths so that's important.  Mobile devices need all the battery life they can get and Atom will sure help out in that area.  A lot of the commentary I am hearing about the inclusion of Atom into these mini-notebooks is about the expected performance boost over current processors and I'm not sure that's what we will get.  I would love to hear from authorities who know about these things but the Atom as I understand it in its initial release is designed for Linux-based MIDs, not Windows-based devices.  I haven't seen any reason to think that the initial Atoms will be able to handle notebook types of devices so I'm not sure that the enthusiasm is a bit misplaced.  Even Intel's own information doesn't mention Vista as a viable OS for the Atom.  I hope I'm wrong about this and the extra battery life of Atom will be great but I don't think we'll see better performance running Vista and maybe not even XP.

Comments

[PJE]

I agree the EEEPC Atom hype seems to be getting a little out of control. My take on the Atom is that it's similar to the C7-M from VIA, but with possibly a lower power requirement.

I think you'll need the dual core version to make a workable Vista machine. You might get XP running OK.

I'm currently more interested in the VIA Isaiah CPU which if dropped in the HP Mini would produce a nice machine. The Isaiah processor should be around twice as fast as a Atom of similar GHz, although it will used more power.

In a (mini) notebook, CPU performance is much more important than in the MID sector.

[vance]

I'm one who values the battery life way more than the performance. I have a dual Xeon rig for my power requirements. When I'm out and about, a long-running machine is far more important.

Just give me the basics, and give them to me for 6-8 hours. And I want the HP keyboard, too.

And less than 20 seconds on boot time.

Oh, Foleo, where art though...

[Nate]

My thoughts exactly. I'm really not understanding all of this Atom hype, not for these mini notebooks, anyway.

[Weezedog]

Yes the Atom has an in-order pipeline like the C7-M, so it's performance will be similar just with much much less power consumption. The Atom does have hyper-threading though, which will increase performance somewhat. I expect atom at 1.6Ghz to perform overall no better then in Celeron 900Mhz that is currently in the Eee PC.
In early tests, VIA's Isaiah processor shows performance comparable to Intel's Pentium-M/Core Solo processor, so I expect it to become the dominant CPU for mini-notebooks, if the price is comparable to the C7-M. It's power consumption is rated at 30% less then the C7 and they have demostrated 1.2Ghz Isaiahs running passively cooled, so it seems to run much cooler then the C7.
In larger devices approaching laptop size, the CPU power consumption is a smaller percentage of the whole power consumption, with the larger LCD screen usually consuming the most power. In these cases, a higher-performance CPU is acceptable because the impact in battery life is less then what would occur in a device where the screen is so small that the CPU is the device consuming the most power.

[peejay]

Interesting that this post should almost coincide with this one:
http://www.umpcportal.com/modules/news/article.php?storyid=1440

...which suggests the Atom will be much faster than expected. Quite a lot faster than the current Eee processor, faster than the OQO...

Interesting.

[James Kendrick]

Note, that is the fastest Atom that will be produced and if I'm not mistaken not the one that will appear this summer. That aside it's still not fast enough to run Vista in my opinion based on those benchmarks. It's the same as the A110 which is only as good as a Celeron which doesn't cut it.

[zunq]

I'm very interested in the battery life of these atoms, more so than the performance. It seems to perform quite well, better than the C7M and if it could provide much lower power consumption I can really recommend the Atomized HP mini note to people. I have a very sluggish '05 Vaio TX which has a 1.2ghz Pentium M, 4200rpm HD and has a smaller keyboard but the Atomized HP seems to trump it at 1/4 the cost so it's time for an upgrade. I know Sony can't be happy about this.

[Geancarlo]

I agree. The Atom isn't going to be a BLAZING fast CPU, it will just provide better battery life for compacts laptops and other similar devices.

HP went with the VIA C7, which is not bad. The 1.6GHz version is known to provide good performance for Win Vista. However, the others that have the 1.2GHz CPU sound like they will have better performance on the GNU/Linux Suse.

If any of you are interested you can join www.eeeuser.com or www.mininoteuser.com. Both good user forums for the discussion of UMPCs like the ones above and the hardware used in them.

[James Kendrick]

You describe your old Vaio as "very sluggish". Do you realize that the Atom is much slower than a 1.2 GHz Pentium M? This is why I am convinced based on actual experience with mobile devices that the Atom will not provide a good performance experience on Windows Vista based devices.

Atom was designed for low-end MIDs running Linux designed to do nothing more than web work. Full-featured laptops or similar devices will tax the heck out of the Atom.

[zunq]

In my case the 4200RPM hard drive is probably the bottleneck, and it shows because it takes around 5 minutes to boot up. It also only has 1gb of ram. I think the 7200rpm will make a world of a difference in boot up.

[Rei]

it's funny to watch James sit here & trash the Atom, A110, Celeron ALL of which outperform the C7-M that he praised numerous times in his HP 2133 video's running Vista "very responsive", "multi-tasking well", "doesnt feel slow".

no doubt the Isaiah is going to be the best performer of the bunch & the obvious choice replacement(pin compatible). the only drawback is the Isaiah uses 3.5w compared to the Atoms 2.4w.

besides, any of the current CPU's would have been a better choice than the lousy C7-M. even the A110 which is in the same performance league but has much lower power consumption (3.5w vs 7.5w).

the HP will receive a dramatic boost in performance AND battery life with either the Atom or Isaiah.

[Mutin]

the fact the Isaiah can hang with the 1.6ghz Celeron is incredible. remember a 1.6ghz Celeron is identical to a 1.6ghz Pentium-M just with less cache & no speedstepping.

HP would be crazy not to use the Isaiah since it's pin compatible & has fantastic performance for just 1watt more power than the Atom.

[James Kendrick]

Rei, I haven't "trashed" any processor, I just want to make sure that realistic expectations are kept with all the buzz surrounding the Atom. I have said many times I wish HP had not used the Via C-7 M for the Mini-Note but the truth is it was all that was available at release time in that class.

The A-110 would have been a good choice although I think the 1.6 GHz Via probably outperforms it in real world testing.

I do suspect that HP will use the Isaiah in future revisions as it makes the most sense from a cost and manufacturing standpoint.

[Steve Paine]

Timely.
Performance figures just in...
http://www.umpcportal.com/modules/news/article.php?storyid=1440

The results look too high to me. Maybe as a result of HT
Steve

[Steve Paine]

oops. somebody already said that!

Ref Atom release. the four versions from 800 to 1.8ghz should all be available in quantities now I think for the oems. big launch in June! Should be great. dont forget the boards that atom come on arfe so small that it could leave space for additional battery capacity. Also, the gpu is very interesting in the area of hardware video decoding.

All in all, lots of choice for designers and lots of choice for customers.

S

[Steve Paine]

oops. somebody already said that!

Ref Atom release. the four versions from 800 to 1.8ghz should all be available in quantities now I think for the oems. big launch in June! Should be great. dont forget the boards that atom come on arfe so small that it could leave space for additional battery capacity. Also, the gpu is very interesting in the area of hardware video decoding.

All in all, lots of choice for designers and lots of choice for customers.

S

[John in Norway]

This is really good news. I'm not interested in these new toy computers that are coming out, it's getting this new Isaiah chip in an OQO that I'm looking forward to.

[Sd]

websites like UMPCportal that post these benchmarks without power consumption numbers are worthless, not to mention amateurish (acting as if performance minus other factors is all that matters). hell, you might as well drop a quad-core desktop CPU as well, because without power consumption numbers whats the difference?

it's also VERY misleading to do ghz-to-ghz comparisons. the Atom needs much higher ghz speeds to achieve similar performance due to design choices, it's not a flaw. once again though, power consumption numbers would have helped negate the misleading ghz ratings.

[JC]

Standard caveat: Please don't confuse TDP with average power consumption. Rei cited TDP numbers. The amount of heat the cooling system has to dissipate in the worst case is not the same thing how much power a processor consumes in the average case. Using TDP as a proxy for actual power consumption is not use, especially across such different designs. (e.g., you'd expect an out of order design to have more going on at once, and thus higher average power, than an in order design, even at equivalent TDP.)

Also, benchmarks tend to be a measure of throughput, even when the unit of measure is time. In that case, it's how much time did it take to plow through a long set of computations. Benchmarks tend not to involve user interaction. Statements about snappiness or responsiveness tend to be qualitative measures of latency, usually in terms of user interaction. Don't mix the two up. Just because a system doesn't deal well with heavy loads doesn't mean that it doesn't deal quickly with light loads (e.g. user interaction.)

In fact, the general consensus across all VIA-based devices is that they benchmark poorly but offer snappy, responsive user interfaces. So if you spend your time calculating a million digits of pi, it's going to take much longer using a C7M than with practically any other x86 processor on the market. However, for whatever reason, it does the relatively lightweight task of dealing with the user in a snappy manner. Anecdotally, it is snappier than the A110 or Celeron M. A bunch of mobile bloggers have said the same thing. I have no idea why. I suspect it has to do with the difference between Via and Intel in how they switch between P-states.

[Unky]

i am waiting eagerly for the Isaiah version, then i will definitely hop on board & buy 1. once the new CPU arrives the HP will have ALL completely modern high performing parts ... at a very low price.

James, when the Isaiah version arrives are you going to get a test unit? i would like to see how much of a performance & battery life increase there will be, i suspect alot.

[Oran]

JC, there is a reason all manufactures have settled on using TDP as a universal form of measurement. yes TDP is under maximum circumstances, so what? if you look deeper into spec sheets they will also reveal average & idle wattage as well. what else would you like them to do?

you could never list actual power consumption at every possible moment under every possible circumstance, because they are infinite. personally, i am glad they decided to list the worse case scenario number as the "standard".

[JC]

Oran, you're absolutely right. Manufacturers do, in fact, list all those numbers. That's because using TDP, by itself, as an indicator of power consumption is misleading. TDP is a design parameter. It lets the system designer know how much heat the system will have to evacuate. Manufacturers are doing exactly the right things. I'm not saying that I'd like for them to do more.

My gripe is with users who behave as if TDP gives you the whole picture with respect to power consumption. It clearly doesn't. e.g., you'll get a better sense of battery life from average consumption.

You seem to think I have it in for the manufacturers. I don't. I have it in for people who speak as if this one number reveals all.

[vance]

So.. JC... given that TDP isn't the whole picture... Would it still be reasonable to say that a CPU that has a TDP of 7.5 "consumes over twice as much power" as a CPU with a TDB of 3.5?

[fab]

in the end the whole discussion boils down to the date when those new CPUs are available in those machines.

Honestly, after reading the reviews of the Mini Note with the 1.6GHz VIA C7 M i'm a bit concerned..ok, for just the browsing session it might be ok, but that UMPC is made for more. so where's the ISAIAH and when will it be available....

[JC]

Vance, your first clause answers your second. Having admitted that TDP is not the whole picture, why do you treat it as if it were?

Just because you have to be prepared to evacuate twice as much heat from part A than from part B doesn't mean that you'll actually need to do this over any consistent period of time. Your question is ultimately meaningless without a context. If you are running some sort of thermal virus, then, sure, the processor with the higher TDP will consume more power. If the system is utterly idle, then what matters is which processor has the better deep sleep power state. There's no reason why this couldn't be the processor with the higher TDP. If that's the case, the processor with the higher TDP could actually have lower power consumption during system idle. (As a side note, most people encounter system idle way more often than thermal viruses.)

TDP is one design parameter out of many. It's an important design parameter since people don't like it when the chips in their computers explode. However, there's a lot about power consumption which TDP, by itself, does not tell you. If you're going to talk about which chips consume more power, you state under which circumstances, and most likely someone will have to do some measuring (or extrapolating from the set of numbers the manufactures give you).

[vance]

Okay, JC, thanks for the reply. I'm just trying to understand here. My thought has always been that the TDP figure scales well when comparing similiar tasks. Meaning this: when doing basic word processing on a 7.5 TDP chip, the cpu will consume twice as much power as doing basic word processing on a 3.5 TDP chip. (CPU only comparison here. No LCD or anything.)