News - Intel Silvermont leapfrogs best ARM chips in benchmarks

[HEXUS]

Intel's chip has half the cores and clock speed but trounces the Snapdragon 800.

Read more.

[3dcandy]

About time though to be honest. ARM need someone to keep them honest, wonder what the graphics are like generally

[Biscuit]

Arm need someone to keep them honest?! I think you're looking at it backwards mate

ARM license their technology so there is a lot of competition between ARM chip manufacturers. x86 on the other hand... hmm well there is Intel, AMD and Via?

[kompukare]

About time though to be honest. ARM need someone to keep them honest, wonder what the graphics are like generally

I though it was "Intel needs someone to keep them honest", past form and all that.

Anyway as was already pointed on other forums AnTuTu is a very poor benchmark indeed. The current Atom does really really well in AnTuTu but that's not an indication of it doing well elsewhere. Over on the AT forum krumme managed to dig up this review:
http://www.gsmarena.com/samsung_gala...view-948p4.php

Notice how the old Atom Z2560 trails or is last at almost all the benchmarks but is the second fastest at AnTuTu. For some reason Intel does really well with AnTuTu but the x86 code is not compiled with the same compiler nor is it using the same options (possibly the ARM code is not using NEON or similar).

So until the chip is actually released and properly reviewed this should be taken with a major grain of salt.

[scaryjim]

... until the chip is actually released ... this should be taken with a major grain of salt.

This and more this Are we really surprised that an intel product that won't be shipped for almost 6 months is faster than a product that's already been shipping for more than six months? Intel may get to the punch first, and release a good mobile Atom that's both power efficient and high performing, but ARM partners will be following up in 2014 with A50 parts, and given the huge market lead ARM has Intel need to be both early and excellent to overcome the inertia. And we won't know if they've managed that until devices ship and are benchmarked.

[Agent]

This reminds me of the days people would buy a GPU purely based off it's 3DMark ranking....Only to find out it was actually not that good in games.

[sykobee]

Right, we'll lets wait for it to actually be available, eh? And for benchmarks that aren't AnTuTu or single-threaded Javascript things.

[watercooled]

Don't get me wrong, I'm expecting it to put up a decent fight, but as has been said I wouldn't completely rely on ATT benchmarks, or any single synthetic benchmark package for that matter. I could be wrong, but along the lines of what kompukare said, I think it's compiled separately (natively) for ARM and x86, not using the VM, so they're not directly comparable at all.

I'm also waiting to see what graphics and price will be like.

[3dcandy]

Erm what I meant was having ONE major CPU architecture in phones isn't the best way sometimes to drive innovation etc. I know you can add bits etc. but would be nice for a competitor...

[kompukare]

Well, I do actually expect Intel to put out a strong showing either with Silvermont or their next Atom now that they are taking Atom seriously.

Ironically, Intel is about 3-4 years behind from where Atom could be since for years they were too afraid of offering a decent Atom in case it affected their more expensive CPUs (plus I suspect, Atom and it's poor chipset was designed to keep older fabs busy before they got upgraded). As an enthusiast who laments the amount of crippling of their SKUs which Intel does for market segmentation, I like to point out this irony to Intel's beancounters: sometimes there is a price to pay for these kind of shenanigans.

But if ARM's dominance in phones isn't that healthy the obvious alternative would be MIPS (or would have been - possible a bit late now). Good perf/watt already despite a tiny R&D budget.

While I expect Intel to perform well with future Atoms that is mainly because they have a node advantage and because I expect them to pour $billions into R&D. The history of x86 (from a fairly poor 16-bit design, to outperforming pretty much everything else) pretty much shows that throwing money at problems has worked very well for Intel. But I do not expect Intel to get the kind of value for money for their R&D that the likes of ARM (or even AMD) get. But if since their pot of cash is x10+ the size of their rivals, eventually they should outperform them.

Another consideration is that ARM's business model gives the company using ARM a lot of choice (licence design, ISA, architecture), whereas Intel business model does not. So, a big phone company like Samsung might consider that in the long term tying themselves to Intel may not be that clever.

[watercooled]

Erm what I meant was having ONE major CPU architecture in phones isn't the best way sometimes to drive innovation etc. I know you can add bits etc. but would be nice for a competitor...

ARM is an instruction set, like x86, not a microarchitecture like Sand Bridge, etc - and there are a fair few ARM licensees competing with their own cores, including Qualcomm, Apple, ARM themselves, Marvell, etc, like how Intel/AMD/VIA compete with x86. Not that the competition is a bad thing, but it's not as bad as having a CPU monopoly.

Well, I do actually expect Intel to put out a strong showing either with Silvermont or their next Atom now that they are taking Atom seriously.

Silvermont *is* the next Atom core.

[kingpotnoodle]

Anyway as was already pointed on other forums AnTuTu is a very poor benchmark indeed.

....

So until the chip is actually released and properly reviewed this should be taken with a major grain of salt.

At least it wasn't SunSpider...

We very definitely need to compare shipping devices vs shipping devices - as any fool know mobile SOCs are very dependent on the device in which they run and how the power profile has been set.

I have a feeling Silvermont is going to be swarming all over tablet markets by Christmas, both Android and Windows (handy for manufacturers re-using design elements), I forsee me getting a Windows 8/Silvermont tablet/convertible to upgrade my old Dell ultraportable. I reckon it'll be closer in phones though where the power envelope is tighter and that might require more push from Intel to overcome the ARM inertia.

If Silvermont really is good in Windows 8 x86-64 tablets then it's going to make Windows RT devices look utterly pointless.

[Zimano]

Bare in mind, this being Intel, you are playing wheel of fortune whether or not the drivers will work. If there's a problem with the drivers then they wont get fixed. If a new version of Windows appears you wont be able to use your Atom either. If this sounds far fetched have a read up about Intel GMA 3600.

Intel regarding Windows 8 support for Cedar Trail based Atom CPU's

"At this point Intel® currently has no plans for releasing Windows 8* graphics drivers for GMA 3600. If they may be available, be sure that this will be announced."

I'd be happy knowing they were actually fixing the existing problems with the Windows 7 driver. The irony is I could have went with AMD when I was buying this netbook but was scared the support with AMD would be crap.

[3dcandy]

ARM is an instruction set, like x86, not a microarchitecture like Sand Bridge, etc - and there are a fair few ARM licensees competing with their own cores, including Qualcomm, Apple, ARM themselves, Marvell, etc, like how Intel/AMD/VIA compete with x86. Not that the competition is a bad thing, but it's not as bad as having a CPU monopoly.


Silvermont *is* the next Atom core.

Pedant. lol
You knows what I means....

[Jimbo75]

Just more crap from Intel, they have nothing to compete with S800 so they are back to fixing benchmarks instead. http://seekingalpha.com/user/13480892/comments

[Xzil3]

grabs popcorns