The AnandTech Podcast: Episode 12

24 Comments

View All Comments

• Paulman - Monday, December 3, 2012 - link

  Another podcast, and so quickly, too! I like.
  
  When it rains, it pours... I managed to score a "Ships in 1-2 weeks" Nexus 4 16GB order here in Canada today, too :) They went on sale at 3PM EST / 12 noon PST today.

• asphyxxiate - Monday, December 3, 2012 - link

  Hey guys, glad you're keeping up with the podcast schedule!
  
  A note about the audio levels: they're really low, at least on the iTunes M4A feed. You should do a little post-production to bump those levels up. I usually have to crank up my volume 4x or so to make it listenable. I think there are some drag-and-drop tools that'll do it for you if you don't want to take the extra time to do it, like Levelator.
  
  Thanks!

• Ken g6 - Monday, December 3, 2012 - link

  First, with Haswell, Broadwell, and keeping the LGA1155 parts around, I'm reminded of the situation with Nehalem where all the low-end processors, Pentiums and Celerons, were still Core 2. I get the impression we'll have Sandy Bridge Pentiums for years to come. I hope that doesn't make LGA1150 parts too expensive, like all the 1156 parts seemed to be.
  
  As far as BGA and sticking the SOC on a board, see also the EOMA-68 standard. http://elinux.org/Embedded_Open_Modular_Architectu... The idea is that one card can power your desktop, laptop, or smartphone - just not all at the same time. Intel could probably get Broadwell into the 10W-with-cooling power profile if they wanted to.
  
  Finally, regarding waiting, obligatory XKCD: http://xkcd.com/989/

• Krysto - Monday, December 3, 2012 - link

  They can't really converge Atom and Core. From a technical point of view, it may make sense, but from a business point of view, it won't, which is exactly why they will not do it. If Core gets integrated into Atom, then they will need to keep pricing them at $20 anyway, which means it would collapse their entire pricing strategy for Core chips.
  
  On the other hand, they don't want to get Atom become as good as Core i3, while it costs $20, because people may start preferring Atom to Core "too much", and against collapse their business, as Intel can only thrive on high margin chips.
  
  Intel is stuck here, because you can bet ARM has no problem trying to make their chips more and more powerful to catch-up with Intel's core, while keeping it at $20 per chip. ARM has no conflict of interest here. Intel does, and that's why it's doomed.

• Krysto - Monday, December 3, 2012 - link

  And my point was that if Intel doesn't learn to compete with ARM at every level for $20 chips - if they can't learn to adapt their business for $20 chips, then Intel will be gone from the mainstream consumer market, because ARM will chase them out, as their chips become more powerful, and entrench more and more on Intel's territory for a much lower price.
  
  So Intel either learns to survive with $20 chips, or they'll be chased away from these markets, with their only escape in high-end servers and supercomputers (at least until ARM and Imagination's MIPS make their play there).
  
  Intel vs ARM is as much of a classic disruption case as you can get.
  
  http://www.youtube.com/watch?v=qDrMAzCHFUU

• peter123 - Tuesday, December 4, 2012 - link

  Hey guys, so I got a little confused by the current/next Krait versions. I thought that the Krait v3 (Fusion3 platform?) was already in the market, inside devices like the Nexus4, Droid DNA or Padfone2 (as per QCM nomenclature S4 Pro tier). And the Krait v2 were the ones with the adreno 225 that debuted early this year (S4 Plus tier). But you said that the V3 will come only in devices next year? What I'm I misunderstanding here?

• smartypnt4 - Wednesday, December 5, 2012 - link

  The APQ8064 is an S4 Pro part - It has four Krait v2 cores and the Adreno 320. The MSM8960 is an S4 part - it has two Krait v2 cores and the Adreno 225 along with an integrated baseband.
  
  Then MSM8960T is an S4 Pro part - it has two Krait v3 cores and the Adreno 320 along with an integrated baseband. The only phone with this chip in it that I'm aware of is the Nokia 920T for China, which is either rumored or confirmed (I forget which) to contain an MSM8960T, giving it a sizeable performance advantage over the US and European Nokia 920, which has an MSM8960.
  
  Personally, I'd love to have a phone with the MSM8960T in it. It seems far preferable to the quad core part for battery life concerns and everyday usability.

• peter123 - Wednesday, December 5, 2012 - link

  Thanks smartypnt4, I got it. So we can expect to see a four core krait v3 in the future, with adreno 320 and integrated baseband? And yes the lumia 920 for china mobile was officially launched today;)
  
  Anyway it will be interesting to which SoC will have the most design wins in 2013. A15 designs to my knowledge do not have integrated basebands (Tegra4, OMAP5, Exynos 5xxx) but apparently can offer superior raw performance over the S4. On the other hand the S4 it's an integrated solution.

• smartypnt4 - Wednesday, December 5, 2012 - link

  Yes, the logical progression is that a 4-core part will surface eventually.
  
  And if the Krait 300 (v3) really does achieve a 15% increase in IPC through architectural improvements, it will be very, very competitive with the current A15 cores from what I've seen. That's difficult to say given that there's only one A15 device on the market (Nexus 10), but I think they'll be very close to each other.
  
  The reason Qualcomm is having yet another record-breaking year is that they've got an integrated solution with a radio that can be used on any of the US carriers, and a fair few of the international markets. It'll be interesting to see if NVIDIA's Tegra 4 with LTE or Intel's next chip that supposedly has LTE will be able to gain any ground against Qualcomm in the coming year.

• smartypnt4 - Wednesday, December 5, 2012 - link

  Is anything known about the Adreno 320's clocks in the Krait 200 vs. the Krait 300 parts? I'd think that if they managed to increase the clock speed of the CPU cores, they'd similarly be able to bump up the clock on the GPU. But maybe not. I'm not sure whether the clock speed increase comes from changes to the CPU core (pipe stage shortening, etc.), or if it's from a different or substantially more mature process at TSMC. Thoughts?