SoCs and Graphics

Intel isn’t talking about implementations of Silvermont today other than to say that it will show up in smartphones (Merrifield), tablets (Baytrail), automotive (unannounced), communications infrastructure products (Rangeley) and microservers (Avoton). Baytrail, the tablet implementation of Silvermont, will be available by the end of this year running both Windows 8 (8.1/Blue?) and Android. Silvermont based Merrifield phones will show up early in 2014.

What we know about Baytrail is that it will be a quad-core implementation of Silvermont paired with Intel’s own Gen 7 graphics. Although we don’t know clock speeds, we do know that Baytrail’s GPU core will feature 4 EUs - 1/4 the number used in Ivy Bridge’s Gen7 implementation (Intel HD 4000). Ultimately we can’t know how fast the GPU will be until we know clock speeds, but I wouldn’t be too surprised to see something at or around where the iPad 4’s GPU is today. Given Intel’s recent announcements around Iris and Iris Pro, it’s clear that the mobile team hasn’t yet had the graphics wakeup call that the Core team just got - but I suspect the Atom group will get there sooner rather than later. Intel’s eDRAM approach to scaling Haswell graphics (and CPU) performance has huge implications in mobile. I wouldn’t expect eDRAM enabled mobile SoCs based on Silvermont, but I wouldn’t be too surprised to see something at 14nm.

Penryn-Class Performance

When Atom first came out, I put its CPU performance in perspective by comparing it to older Pentium M based notebooks. It turned out that a 1.6GHz Atom performed similarly to a 1.2GHz Pentium M. So how does Silvermont stack up in PC notebook terms?

On single threaded performance, you should expect a 2.4GHz Silvermont to perform like a 1.2GHz Penryn. To put it in perspective of actual systems, we’re talking about around the level of performance of an 11-inch Core 2 Duo MacBook Air from 2010. Keep in mind, I’m talking about single threaded performance here. In heavily threaded applications, a quad-core Silvermont should be able to bat even further up the Penryn line. Intel is able to do all of this with only a 2-wide machine (lower IPC, but much higher frequency thanks to 22nm).

There’s no doubt in my mind that a Baytrail Android tablet will deliver amazing performance, the real unknown is whether or not a Baytrail Windows 8 detachable/convertible will be fast enough to deliver a good enough legacy Windows experience. I suspect it’ll take Airmont before we really get there by my standards, but it’ll be close this round for sure.

What’ll really be interesting to see is how Silvermont fares in smartphones. Max clock speeds should be lower than what’s possible in a tablet, but not by all that much thanks to good power management. When viewed in that light, I don’t know that there’s a more exciting mobile architecture announced at this point. The ability to deliver 2010 11-inch MacBook Air performance in a phone is insane.

The Silvermont Module and Caches Tablet Expectations & Performance
POST A COMMENT

174 Comments

View All Comments

  • GTRagnarok - Monday, May 06, 2013 - link

    We'll miss you :'( Reply
  • Pheesh - Monday, May 06, 2013 - link

    I don't think the Anandtech site will miss your views and contributions. I appreciated the article and explanations for what's to come. Not everyone cares only about 'specs'. Reply
  • Krysto - Monday, May 06, 2013 - link

    That's what I hate most about Intel's press releases, too. They are typically HIGHLY misleading, in some way or another. Remember when they made people believe the 22nm IVB would be 37% faster AND 40% more efficient? - when it was actually one OR the other, but because of the way they phrased it in the PR...and the way tech sites cloned it in their articles, that's what most people got - that they get BOTH those benefits. In the end it was HALF of each.

    But the worst part about it is the "churnalism" that comes after it. Since most tech sites are either too ignorant or too lazy, they just rewrite what Intel said in the PR - no questions asked.

    Please remember EXACTLY what Intel says Silvermont is capable of - and I PROMISE you when this chip gets reviewed, it won't even seem close, and if you remember everything, you'll be disappointed.
    Reply
  • Khato - Monday, May 06, 2013 - link

    Actually, I don't recall where Intel stating that IVB would be 37% faster AND 40% more efficient. I believe you're confusing technical information regarding the 22nm process capabilities with a product using that process... and incorrectly interpreting that process information to boot. Reply
  • Homeles - Monday, May 06, 2013 - link

    The only thing that I can think of is the chart comparing voltage and gate delay on the first page of this article, which was circulating a lot before the launch of Ivy Bridge. There's nothing misleading about it though, unless you don't know how to read the chart. Reply
  • Homeles - Monday, May 06, 2013 - link

    "That's what I hate most about Intel's press releases, too. They are typically HIGHLY misleading, in some way or another."

    Intel, AMD and Nvidia are all guilty of this, and you know it.
    Reply
  • t.s. - Tuesday, May 07, 2013 - link

    Exactly! Hate it. It looks like all of them have an inferiority complex. They have to boast. Not confident enough with their true offerings. Reply
  • AssBall - Monday, May 06, 2013 - link

    You know I have been coming here for 15 years and thought exactly the same thing. This article is the worst. Terrible PowerPoint graphs and gibberish. No real information. Kind of a sadness. Reply
  • VivekGowri - Monday, May 06, 2013 - link

    The "PowerPoint" graph actually came from Excel, but it came from an article from earlier this year: http://www.anandtech.com/show/6536/arm-vs-x86-the-... (see page 3). And I dare you to find a more thorough or more comprehensively researched and tested article on SoC power consumption literally anywhere else. Reply
  • AssBall - Tuesday, May 07, 2013 - link

    I dared, "literally", but couldn't find much. Sue us for asking for better standards. Reply

Log in

Don't have an account? Sign up now