I think we discussed this before

Kelpie wrote:

I've seen that chart before. This is not the diamondville processor. The Diamondville is not the Z series, and the Z series is the one tested. If you search for the benchmarks, you will see that a site will tell you it's not the N series, which is the Diamondville processor.

These benchmarks are for the Silverthorne processor.

it was actually the n series diamondville.... 

it is corrected there now

someone posted a celeron 900 benchmark in the comments, obviously the platform affects the benchmark so it's not a scientific comparison but if this claim is correct

Celeron-M 900 Mhz:

Alu: 3420
FPU: 4111

Then the Diamondville Atom @ 1.6GHz is actually superior to the Celeron @900Mhz in this particular benchmark. Which is good news for intel.

Last edited by sideways (2008-04-25 5:29:32 pm)

ah, ok, apologies - I completely missed that discussion (was busy over the weekend)

Ok, I just thought it was significant that the german site posted an update claiming the bechmark was for a Diamondville Atom, not a Silverthorne.

Here's what I expect, for what it's worth:

Intel have basically been honest and the 1.6Ghz Diamondville be just a little better in performance than the 900Mhz Celeron. It will however consume less power which will give a modest battery life improvement (The cpu consumption is a small percentage of the overall chipset consumption). I think the Atom supports speed-stepping which the celeron doesn't - it runs at 900MHz fixed in the bios, the Atom will dynamically adapt. Big business is not completely stupid - why is nearly everyone choosing the atom to build their umpcs around? Hyperthreading, 64bit & Intel Virtualization all make it a far sexier processor than the Celeron @900MHz. That Via Isaiah looks promising though, and its supposed to be pin-compatible with the C7 so i wonder if the HP-2133s will be upgraded?

dd99 wrote:

The Celeron-M supports Speedstep.

Are you sure about that , maybe under windows only? When I try to enable cpu frequency scaling in Fedora 8 it says it's unsupported.

I think we'll see the Atom based Eees this summer.

The Isaiah has a much larger TDP but many peeps will take a 1 hour hit on battery life for performance (I wouldn't)

Well, kudos to Mandriva then, any special configuration needed, kernel customization?

I agree with your point about not having any true idea about the Diamondville's performance. Things are very speculative and untrustworthy at the moment, it will all be made clear over the summer.

IMO, 64 bit and virtualization aren't useful for diamondville.

It's too slow, and who in the right mind will use > 4GB of memory and virtualize anything in a MID or UMPC?

ok again
the benchmark was done on the Intel Mini-ITX Mainboard with the diamondville core!

http://vr-zone.com/articles/Intel_Atom_ … /5721.html

thats the board i tested!
I am pretty sure that the diamondville platform won't bring a big increase in battery life compared to the Celeron-M platform used in the Eee PC right now(i am talking about the cpu+the chipset)

Last edited by eeepcnews (2008-04-26 1:45:01 am)

You got it Marty,

The Atom, especially the diamondville core will be damn cheap to produce for Intel...
And of course, the Celeron-M is running out of life soon, so Intel had to act.
I am pretty sure that they will be able to produce the Atom for around 5$ which will lead to a way higher margin for Intel and also for manufacturers like Asus.
Or does someone really think that Asus will lower the price for the Atom Eee PC 900?

But Intels' marketing machinery seems to work when you look at all these comments. Postings like, that these benchmarks are a fake or that the Diamondville is not finalized are telling me a lot about the background of these users and how much Intels strategy worked out.

I am not saying that the Atom in general and specifically the Diamondville core is a bad cpu! It is just over hyped!

Radek wrote:

Personally I can not wait for those mATX Atom based mainboards.

That would be nice.  I could drop one of these into the Antec Aria case I'm using for my Freevo (HTPC) system.  Since it's up 24x7, a cooler running, lower power system would be advantageous.  I had tried an mini-ITX mobo with a Via C3 in the past, but it was simply too anemic to handle video recording/playback along with music playback.

dd99 wrote:

This has already been discussed here: http://forum.eeeuser.com/viewtopic.php?id=24850
Here is the chart that I made, with a few changes from the original posted in eeepcnews.de:
http://img512.imageshack.us/img512/5102 … arkpm4.jpg
The original chart is meaningless and the data for both the Atom CPU (unconfirmed whether it's a Silverthorne or an Engineering Sample Diamonville; I believe the former) and the VIA Isaiah (ES? Power consumption?) are unverified.
The data for the Core duo (Yonah) CPU is from my Lenovo X-60.
As you can see, the Core duo architecture crushes all the other CPUs, when it comes to performance per MHz.
However, all such comparisons are meaningless unless one defines the exact power envelope for the intended application.
For eeePC class devices, this is around 5W TDP, which the Celeron-M @ 900MHz narrowly complies with.
The VIA C7M @ 1.6GHz is an 8W part and is 60% beyond the required specs.
The Silverthorne @ 1.6GHz is a 2W part and is 60% below the required specs.
The VIA Isaiah probably is a > 10W part @ 1.6GHz and so won't be used in eeePC-class devices. And for subnotebook applications it is vastly superceded by Intel Core-based CPUs.

Gotta remember that there is no standard for TDP ratings (it's almost more of a marketing ploy than a solid specification), and the way this number is determined varies between manufacturers. Intel has had a very bad history of misleading people by their thermal specifications, since they take a lot of liberties when establishing testing parameters. Companies like AMD, for instance, measure theirs as the maximum power the CPU could possibly draw under the worst conditions. Intel rates theirs based on typical power draw in ideal conditions. So I think we have to give companies like VIA and AMD a break here, and wait for third-party testing to determine heat output and power draw of competing chips (so we have consistent testing methods).

I must say that the celeron is amazing. It is old, it is based on 90nm, and it has a tdp of only 5W and underclocked about 3W. That is amazing, the diamondville is brand new, 45nm and has a tdp of 2.4W. That´s not much in difference if you think about that the celly is much older, 90nm and so on. The performance increase doesn´t seem to be so much either. Shrink the celly to 45nm and it will perform better than the diamondville and it will have the same tdp as the diamondville.

Last edited by UsernameWhat´sThat? (2008-04-27 3:40:11 am)