Preliminary benchmarks have VIA's Isaiah besting Intel's Atom
You knew this day would come: Intel positioned Atom perfectly to compete with VIA's low-power offerings, and VIA is trying to stay one step ahead in the low power game with its Isaiah processor. Who will be the winner? Well, we'd say it's still a little early to call it, but German site Eee PC News did some quick and dirty benchmarks that show Isaiah on top by a decent margin. At this point the numbers are just in "ALU" and "FPU," but hopefully some real world benchmarks from some retail products can clear this up before long.



















I would expect the VIA to stay ahead because that's been their stratagem all along. They design processors with low-power usage in mind first, then speed them up. Intel's ploy has always been brute force, regardless of how much power it uses. It's only because of the recent surge in popularity of laptops that they've given power consumption any consideration at all.
UP AND ATOM!
Rainier Wolfcastle(Radioactive man) "UP AND AT THEM"
The thing is, Intel got the most chips, it will win ultimately. How many rounds of all-in can VIA or AMD toss at Intel for trying to gain a little more competitive edge?
Via does have the experience and performance edge, but Intel has mass-market appeal. This will be an interesting competition.
Via is a lot bigger than Intel in the embedded world. They're probably second only to Motorola.
As for Intel's power consumption problems, most people think this was due to a brute force approach to speed simply because power draw increased over time as Intel's chips got faster. The actual cause of this was power loss due to quantum tunneling -- as the silicon dioxide approached a few atoms in thickness, the electrons in the circuit were tunneling right past the gate at an alarming rate under Netburst's design.
In fact, this was one of the major hurdles for the Penryn project, where their new hafnium (I think) gate material allows for a thicker barrier (thus cutting down on power loss due to tunneling) while facilitating a better conduction (which of course makes the circuit faster).
It's fascinating stuff. There's a detailed description of the Penryn innovation out there (I think I found it on Ars Technica). If you've suffered through some university-level chemistry or physics courses (and remember them, of course ^~), you'll probably have a fairly good idea of what it's talking about.
I, for one, am more excited about Larabee than I ever could be about Atom. I'm not even sure there's much point to Intel reaching into this market, given how determined Via is to hold onto it (it's really their only business after all) and how strong that hold is.
Of course, all this needs to be taken with a grain of NaCl -- these are first silicon, after all.
You can't compare low-power chips without measuring POWER CONSUMPTION.
this and the nvidia integrated card they posted a while back in the revision of the hp 2133 mini note? i hope...
The C7 chips can frequency scale in one clock cycle. If you measure performance per watt, VIA's offerings slap any Intel around like a bitch.
Intel all the way!! Just having that Intel sicker on my machine makes me feel so much more comfortable.
Hey if you like a buggy machine, "Intel all the way, yeah!".
Quite frankly I think Intel has lost their view, they're all about brute force, it's all a Ghz game with Intel, and my personal experience has shown that they've forgotten about all the rest of it.
I'm sure i'm going to get flammed by all the Intel fan boys, i'm not saying AMD is that much better, in some respects they are but not in all.
VIA will maintain the low power market I believe, does anyone checked out www.mini-itx.com frequently? They've been in the market for a long time is all i'm trying to say.
Let the flames be thrown!
i sure hope that's satire
I does checked it out.
If this were the age of P4 Prescott and Pentium D a couple years back, I would be right with you. but suddenly you don't see intel chips hitting 3.8ghz with the 67x and nor do you see the current trend with graphics cards, which is just the constant increase in power consumption.
After Intel ditched Pentium and started basing everything off the Core architectures, they really have started looking at power consumption, even for desktop machines. AMD recently released a line of power saving chips that competed directly with Intel's standard line.
Like AMD and the Athlon 64+ series, Intel went back and started making sure each clock cycle could do as much as possible, ending their GHz racing streak, in my opinion for the better. Yes, the clock speeds are rising once again (the inital Core desktop CPUs topped out under 3ghz) but that is not to say that Intel is just going for brute force, especially considering that their current desktop processor has a processing step pipeline closer in length to an A64 or mobile Dothan core than the Prescott/Cedar Mill Pentiums.
Even since Intel started focusing on mobile CPUs as an individual asset (and that does NOT include "Mobile Pentium 4") power consumption has been a very important focus on their mobile chips, which has indeed carried over to their desktop products, which you now see in the form of 45nm die sizes, speedstepping, etc.
Now, an intel sticker on my machine doesn't go as far as to make me "feel secure", but there's nothing about hardware that's buggy either... i didn't think there were CPU drivers. But at this point i think Intel does have the best all around product set around. Props to them for venturing into the ultra mobile, ultra low voltage/power consumption field. Even if at first they don't succeed (seems their success in mobile, desktop, and server products has got their reputation hyped quite high) they'll hopefully improve and push the field as they do.
*sighs* back to linear algebra
I don't think a single bug I've ever had could be traced back to a processor error... and if it routinely happens to you, vacuum tubes are no longer the modus operandi for the rest of us.
I'm pretty sure intel HAVE dropped the pentium architecture in all CPUs apart from the atom, which has a weird "out of order" execution thing going on, like the origional pentium and no intel CPU since. The only reason for this is simplicuty and crucially low power consumption.
The Pentium dual cores Exxxx are real fast , they're cut down models of the Core 2 duo. Mine's humming along at 2.835GHz, 1Ghz above normal. :)
bethel03... that a reasoning boy! Don't you feel like sheep when you write these kind of stupid posts? Let Joe Sixpack be comfortable with an Intel sticker, but you? If you read engadget I was hoping you were so much better than that.
Do you need fairly low power requirements and very decent number-crunching performance? Isaiah wins here.
Is it stellarly low power requirements and average computing performance? Atom is your horse.
I would love to see the Atom at what ever clock it will be set at in the 8.9" EeePC benched against the Celeron M ULV at what ever clock it is going to be set at in the Eee PC.
It has been benched by the same site I think and the celeron in the eee was faster than the atom....
Isn't power consumption the real issue...?
That burn the rope game sucked. And go Via!
ATOM: 2.4W TDP @ 1.86GHz
Isaiah: 25W TDP @ 2.0GHz
I'm not missing a decimal point here... at 10 times the power consumption I don't think Isaiah is besting Atom.
Different companies use different ways of calculating TDP. I'd wait for benchmarks to come out before you start comparing power usages.
TDP=! Power Consumption, it is a measure of theoritical maximum of heat energy the cpu will have to dissipate. My turion 64 as a higher tdp than my core 2 duo yet it consume less power.
Isaiah>Atom
Response:
http://farm3.static.flickr.com/2327/2423035895_c58722e81a_o.jpg
Its going to be really hard to compair them when the mobo's they come on will be different. Performance isnt just in the CPU, but also the surounding hardware.
the nvidia/via team-up is looking pretty hot.
Paris?
durrr
Those are effectively real-world numbers. ALU stands for the Arithmetic Logic Unit on the CPU, and FPU for the Floating Point Unit. ALU performance is key for most everyday applications such as word processing and web browsing, and FPU performance for more complex mathematical tasks such as Photoshop filters and whatnot.
What a horribly skewed benchmark. The VIA Isaiah has a 25 Watt TDP while the max 2.5 Watt. What a terribly inefficient chip the Isaiah chip is in comparison. You should be comparing to the Celeron-M. So VIAs latest processor can keep pace with the Celeron-M, which is Intel's old processor that is being replaced by Atom in 2 months. I wouldn't call that much of a win.
Should have been:
"The VIA Isaiah has a 25 Watt TDP while the Atom has a max 2.5 Watt."
Isaiah has a min TDP of 3.5W and a max of 25W. This is inline with intel's other current CPUs.
Inline with Intels other mobile processors, like Celeron-M, but not Atom which has a TDP of 0.6 to MAX 2.5 W. The Isaiah is an order of magnitude more.
That is because Atom and Isaiah were designed to compete against different products, not against each other.
Isaiah was designed to compete against Pentium/Celeron/Core Solo in laptops/desktops.
Atom was designed to compete against ARM cpus in small mobile devices.
Doesn't matter. The rest of the VIA chipset doesn't have linux drivers for 3d acceleration. That means no compiz. That means I'd rather have an atom with the gma950
The atom chipset uses a PowerVR graphic accelerator, it will not be cross compatible with any other Intel graphics driver code. I dont know the status of linux PowerVR drivers, but I know there are no open source ones. I've been told PowerVR are kind of dicks about it actually.
There were always linux drivers for the power discrete graphics cards back in the day. The graphics chipset suppourts directx 10.1 and opengl ES 2 though so 'code compatability' will not be an issue if the platform is put together right.
Whoa rektide...PowerVR LIVES?!?!
I'm having flashbacks to the Dreamcast now.
Waiting for real-life tests like encoding high-res video, rendering 3D scene in 3D Max Studio 9 or maybe... Crysis? xD
Guys, get real, this thing is supposed to be able to get your web browser up and running, as well as some simple video playback (I don't have doubts it will even chew up high-def video), some Excel, Word ect. Its not supposed to replace your PC! What kind of real-world test you wait for?
Get real! The less power this thing eats while playing vids the better it is. As for performance - as long as nothing stutters, who cares what processor makes more FPS in crysis?! You're not going to play it on 7-10 inch screen anyways!
Real world benchmarks would still be a great deal more interesting. There are office suite and web browser based benchmarks that would be far more interesting than these. Video and audio encoding/decoding benchmarks would be equally relevant.
I think you might be suprised at the performance achievable by the graphics core on the atom. It will easily exceed that of the sega dreamcast.
It's funny to see Isiah winning at anything these days, so props to him i guess
atoms good but these names need more creativity. Isaiah is wha you name a kid...imagine someone named 'conroe'
Err...That's my son's name.....Conroe
I did, and then I beat him up for his lunch money. I'm about to pound Northwood next. Little bitch.
Kelly Nair, why does your name remind me of "we wear short shorts!"?
To give you a bit of perspective for that reason, VIA's other chips are named Esther (C7), Samuel, Ezra, Nehemiah (all C3). Sound familiar? I think it's kind of interesting, at any rate. I personally think Esther is a nice name, although I don't know if I could name my kids that.
Nothing is more pompous sounding than 'Prescott'. "My name is Prescott, la dee da." *gets killed*
lol, coming from England I have to say its hard imagine Prescott as a pompous name, we had John Prescott as deputy prime minister for like 10 years. He was a union man from up (the industrial) North, who liked to eat pies.
++ maybe its just cos I'm English, but I would'nt call my son Isiah, i think I probs prefer Conroe.
but we have to accept that CPUs are really, really jewish.
These sort of synthetic benchmarks are entirely useless in the real world. They don't represent real-world workloads, and Atom's architecture is different enough from typical x86 processors that you can't really draw any conclusion from these figures.
Sure, this might tell you how fast you can crunch numbers, but if that's all we cared about, the Cell would be the most powerful consumer processor on the market. In practice, the limitations of the Cell's SPs mean that they're well suited for some tasks, and ill suited for others.
Sorry, I meant SPEs, not SPs.
On these links Celeron 900 seems faster than Atom 1.6:
http://laptoping.com/intel-atom-benchmark.html
http://www.uberpulse.com/us/2008/03/chip_shot_intel_atom_is_half_the_performance_of_celeron_video.php
even if Celeron 900 has only Alu: 3420 and FPU: 4111 much less than Atom 1.6.
What's the meaning of this test? Which one is the faster for everyday applications?
I was wondering the same thing. Intel's Atom seems to be a step backwards in processing power. The question is, will its power saving make it worth the downgrade in speed. How many more minutes of battery life is the Atom going to gain a user? My guess is not much. CPU speed still matters even when web browsing or watching/decoding Internet video. But the great news is that they'll probably be many CPU options.
There are 2 entirely separate Atom CPUs. The only one thus far tested and not going into the 900, is the way slower of the 2. Look for performance figures on the Atom Diamondville soon.
You are right. There are 2 Atom: http://en.wikipedia.org/wiki/Intel_Atom#Atom_architecture
Anyway, seems that the Diamondville is more hungry power than the Silverthorne (and the dual core more than celeron-m ulv) so I doubt we will see it on the eee.
Now lets see power consumption comparisons
Could someone tell me which one of Atoms it is? Silverthorne or Diamondville? One is solo and the other one is dual core and of course dual core is for those small laptops, which are so popular nowadays. And solo is more for smaller mobile devices.
And the power consumption, that's a huge gap, how can they even be compared? VIA requires a lot more power with Atom you would have additional 1-2 hours.
Silverthorne is for Mobile Internet Devices and super small things.
Diamondville is for Low End laptops and Desktops.
Diamondville will be 1 or 2 core
Silverthorne appears to be 1 core with a new version of HyperThreading.
The problem with the comparison is that they're comparing the WRONG thing.
It's easy to make a processor to best the ones listed here. AMD phenoms are way more powerful.
It's also easy to make a processor to use less power than the ones listed here. I happen to have one that consumes ZERO watts. That's right, ZERO watts. (However, it also performs zero apu's and fpu's per year).
The CORRECT comparison for these "low power" parts is NORMALIZED FPU's and APU's. When you divide the number of operations by the power consumed to generate those operations, you get a MUCH different picture and a clear winner; Atom.
Yes when you look at nothing but performance per watt, Atom is the clear winner by a huge margin. But all people care about is USABILITY, who cares if the Atom has 10x the performance per watt ratio when compared to this or that CPU if Atom is still to dirt slow to make decently running device. These benchmarks are performance oriented, which helps establish the usability of a device because you can compare the scores to a device that is already on the market. These benchmarks help you draw an approximate conclusion to the performance of the same device when equipped with the Atom CPU and helps in the buying decision.
Oh and i'm not saying Atom actually is dirt slow, because it actually performed quite well, much better then the C7-M, but we only know that because of these performance benchmarks.
... and if the ratio of operations per unit power used by those two processors isn't that much different, Isiah's got my vote.
What's 5x battery life if applications run 5x longer?
Care to share how you got to that result? Sources?
Everyone waiting for their $750 HP MiniNote should commence crying. Seppaku is an advanced option.
Yep. I've been holding off on the HP Mini-Note to wait till it gets Isaiah, then I will order it in a heartbeat. Hopefully they offer Isaiah in it sooner then the 6 month refresh cycle that HP has mentioned. Since Isaiah is pin compatible, I hope they offer it as an option when it debuts.
Also from the internal pics I've seen of the Mini-Note, the CPU appeas to be soldered to the M/B so the early adopters can't upgrade, or I'd order a mini-note now!
Ok, if you can't beat 'em then buy 'em!
Tried and tested by all the big IT companies. VIA will have to watch out for the men from Intel ...
These benchmarks are meaningless without power consuption data. As far as I've read, they're compating a 2.5 watt Atom to a 3.5 watt (the lowest powered) Isaiah, so although these are syntetic benchmarks and TDP estimates, it's safe to say that at this point the two CPUs are very similar in terms of performance per watt. Neither Via nor AMD have ever come close to besting Intel in terms of an overall mobile platform.
Don't forget that Intel will most likely be making components the platform as well. I'm sure they will have things communicating efficiently.
God, I feel humiliated to have AMD chips in my PC and laptop.
I'll just go and sit in the corner over there. :(