Atom Benchmarks

[CTho9305]

Originally posted by: myocardia

Originally posted by: CTho9305
I would expect it to absolutely demolish Atom. On Intel-optimized 2-thread workloads I would be mildly surprised if Atom wins consistently.

I don't understand how an Isaiah could destroy an Atom, while the Atom wins battle occasionally. Wouldn't those polar opposites?

No. Nothing that matters is perfectly dual-threaded, with the instruction mix optimized for Atom. I qualified my statement because synthetic benchmarks can really suck sometimes (personally I think it's irresponsible to include most of them in reviews - look at the Sandra Arithmetic MFLOPS results here). In real programs, I expect Atom will get trashed.

Also, somebody getting trashed can still win the occasional battle - particularly in the world of synthetic benchmarks. Look at the benchmark suite used to create performance ratings for the Geode GX/LX series and the results compared to the Via processors - there are a couple benchmarks that are significant outliers.

 

[Idontcare]

Originally posted by: bryanW1995
3. Penyrn Shmenryn...Nehalem FTW!

This was my favorite LOL

Honestly though I don't think Intel is worried about, why would they care what they sell us desktop users so long as we buy Intel instead of AMD and so long as the gross margins are nice and juicy. If that's Penryn chips until Nehalem-C comes out then why would Intel care?

Given the supply ramp of desktop Penryns anyways I am willing to bet the first 6-12 months of Nehalem supply is going to see waaaayyyy overdemand from the server segment anyways.

Originally posted by: CTho9305
I still don't think it's reasonable to compare a <20mm^2 die (Atom) to a ~60mm^2 die (Isaiah, roughly what a single core 65nm K8 would be). There's an enormous cost difference... if the two chips really compete, even at the same prices Intel's margins will be drastically higher than Via's. I don't know what the die area is on Geode processors, but I don't think they're likely to be very competitive in any aspect here. That doesn't mean each respective marketing department won't try to play up their individual strengths and Atom's weaknesses.

In this regard you are talking about metrics that matter nothing to the consumer nor to the marketing folks.

I've never seen a metric publicized in the consumer realm relating performance/die-size nor performance/mfg-cost.

The things you speak to here only matter to the manufacturer. Intel wanted a chance to goose up their gross margins, hence the vastly reduced die-size.

If this calculated strategy on Intel's part results in a product that is neither performance/watt nor performance/cost competitive with alternatives in the open market (be it Nano or Geode or other) then the fail is definitely on them and not on the folks who would notice such.

In my book the only criterion for validating a comparison is if the products themselves are marketed for the same product segment. To my understanding this is very much true of Nano and Atom.

 

[Acanthus]

All artificial benchmarks... and they measured power consumption with a friggin raptor in there...

Fantastic job again toms. :thumbsdown:

 

[Idontcare]

Originally posted by: Acanthus
All artificial benchmarks... and they measured power consumption with a friggin raptor in there...

Fantastic job again toms. :thumbsdown:

I stopped visiting them when they markedly became like a CNet review site, clearly pumping reviews for whomever the ad dollars flowed.

Why is a raptor in the review? Want to hedge a guess whether Toms gets ad money for the velociraptor?

There is zero credibility there, in fact you look at this whole review industry and see where it ended up after coming from such noble causes. Now everything is cherry-picked ES samples reviewed over paid vacation retreats.

I remember how Tom Pabst started Toms and his open/published insistence on the philosophy that he buy everything he reviewed from retail paths. AT started out in similiar manner too.

Now you get these dubious reviews where at best you want to give the benefit of the doubt and attribute to stupidity that which could be attributed to greed, but seeings as none of these review sites are non-profit you simply can't be foolish enough to ignore the fact the very existance of these sites (nowadays) is for the money.

 

[myocardia]

Originally posted by: CTho9305
No. Nothing that matters is perfectly dual-threaded, with the instruction mix optimized for Atom. I qualified my statement because synthetic benchmarks can really suck sometimes (personally I think it's irresponsible to include most of them in reviews - look at the Sandra Arithmetic MFLOPS results here). In real programs, I expect Atom will get trashed.

Ahh, gotcha. It makes perfect sense this morning. Last night, when I was too bleary eyed to even form complete sentences, it seemed like somewhat of an oxymoron.

Originally posted by: Acanthus
All artificial benchmarks... and they measured power consumption with a friggin raptor in there...

Yeah, that was the first thing that struck me, too. I came to the same conclusion about why they had used a Raptor as Idontcare did.

 

[Idontcare]

More Atom benchmarks )from fudzilla), still nothing substantial but the impression continues to be "meh".

The first benchmark we ran was "lame", and this time we took a couple of screenshots, because it was unbelievably slow.

http://www.fudzilla.com/index....=view&id=7595&Itemid=1

They do confirm hyperthreading is active. And they confirm performance in anything CPU intentsive will be pretty much suckage on these initial single-core Atoms. (rapidly explains the reasoning behind dual-core Atoms when they say things like rendering and media suckage...exactly where multi-cores help)

I still just simply can't believe the catastrophic foot shooting Intel did by releasing Atom's first platform on this aging 945G chipset that consumes more power than the chip itself. IMO thats one of those critical fatal wounds that could be so massive that it destroys the entire product as it gets out of the gate.

 

[pm]

Why is a raptor in the review? Want to hedge a guess whether Toms gets ad money for the velociraptor?

That's a possibility - but I think I tend to agree more with your second guess. When in doubt, follow Hanlon's Razor (http://en.wikipedia.org/wiki/Hanlon).

I still just simply can't believe the catastrophic foot shooting Intel did by releasing Atom's first platform on this aging 945G chipset that consumes more power than the chip itself.

Yes, I agree, the choice of pairing it initially with the 945G chipset is a shame - in my humble opinion. It results in bringing up the negatives of the processor (performance) while not showing the advantages (low power, small footprint).

 

[Acanthus]

Originally posted by: Idontcare
More Atom benchmarks )from fudzilla), still nothing substantial but the impression continues to be "meh".

The first benchmark we ran was "lame", and this time we took a couple of screenshots, because it was unbelievably slow.

http://www.fudzilla.com/index....=view&id=7595&Itemid=1

They do confirm hyperthreading is active. And they confirm performance in anything CPU intentsive will be pretty much suckage on these initial single-core Atoms. (rapidly explains the reasoning behind dual-core Atoms when they say things like rendering and media suckage...exactly where multi-cores help)

I still just simply can't believe the catastrophic foot shooting Intel did by releasing Atom's first platform on this aging 945G chipset that consumes more power than the chip itself. IMO thats one of those critical fatal wounds that could be so massive that it destroys the entire product as it gets out of the gate.

It does seem incredibly stupid to launch a cpu on a chipset with a TDP 8-9x the cpu...

 

[Lonyo]

http://www.vr-zone.com/article...ption_Test/5802-3.html

20W max power for MSI Wind powered by the Atom. Not too bad, and that's with an LCD.

 

[mooseracing]

Originally posted by: CTho9305
I don't know what the die area is on Geode processors, but I don't think they're likely to be very competitive in any aspect here. That doesn't mean each respective marketing department won't try to play up their individual strengths and Atom's weaknesses.

The geodes are on the old socket A if that helps any...

Originally posted by: CTho9305
Look at the benchmark suite used to create performance ratings for the Geode GX/LX series and the results compared to the Via processors - there are a couple benchmarks that are significant outliers..

Why not compare the Geode NX's, they are the newer ones aren't they?

 

[CTho9305]

Originally posted by: mooseracing

Originally posted by: CTho9305
I don't know what the die area is on Geode processors, but I don't think they're likely to be very competitive in any aspect here. That doesn't mean each respective marketing department won't try to play up their individual strengths and Atom's weaknesses.

The geodes are on the old socket A if that helps any...

Originally posted by: CTho9305
Look at the benchmark suite used to create performance ratings for the Geode GX/LX series and the results compared to the Via processors - there are a couple benchmarks that are significant outliers..

Why not compare the Geode NX's, they are the newer ones aren't they?

I was talking about the Geode GX/LX chips. I don't count the NX's as Geode's since as far as I can tell, they're just Athlons with a different label. Can you currently buy NX's? Are they still manufactured, or are they all inventory?