Seeking Alpha

Ashraf Eassa

View as an RSS Feed
View Ashraf Eassa's Comments BY TICKER:
Latest  |  Highest rated
  • Intel: 14 Nanometer Production Looks Like A Go [View article]
    "There are no known plans for a Broadwell desktop chip."

    There will be a Broadwell "K" SKU for desktops.
    Mar 18 01:37 PM | Likes Like |Link to Comment
  • Ambarella Looks Like A Short [View article]
    Oh dear, I guess not...
    Mar 17 09:37 PM | Likes Like |Link to Comment
  • Skullcandy Turnaround Not Likely [View article]
    Charles,

    Superb analysis that illustrates your intellectual honesty and your ability to "tell it like it is".

    Thanks for the article.
    Mar 17 04:22 PM | Likes Like |Link to Comment
  • Applied Micro's X-Gene Proves Uncompetitive [View article]
    Intel already does, and unlike AMCC, Intel is already shipping and has been for months. X-Gene is still vaporware.
    Mar 17 02:58 PM | Likes Like |Link to Comment
  • Sell Intel, Buy Qualcomm? [View article]
    Like I said...shorting QCOM is a BAD idea.

    http://bit.ly/1kCqK3i
    Mar 17 12:15 PM | 4 Likes Like |Link to Comment
  • Sell Intel, Buy Qualcomm? [View article]
    The licensees who pay QUALCOMM royalties aren't the chip vendors, it's the device vendors.
    Mar 16 10:50 PM | 2 Likes Like |Link to Comment
  • Sell Intel, Buy Qualcomm? [View article]
    "INTC has $5billion in cash and $15 billion in short term investments- same as cash. That is $20 billion, like I said."

    Intel has $13B in debt. Net cash of $7B.
    Mar 16 10:50 PM | 3 Likes Like |Link to Comment
  • TSMC's Rebuttal Of Intel's Scaling Advantage Is Just Qualitative, Not Quantitative [View article]
    Moorefield is a 22nm part.
    Mar 16 03:30 PM | Likes Like |Link to Comment
  • TSMC's Rebuttal Of Intel's Scaling Advantage Is Just Qualitative, Not Quantitative [View article]
    "- It's really about the silicon results. The most recent silicon data point we have right now says that 28nm Samsung planar silicon (Apple A7) can accomplish about the the same as Intel 22nm FinFET (Bay Trail Quad) in 102mm^2.
    http://bit.ly/1iaf1en"

    Bay Trail was a crappy design, don't forget this. I would compare Moorefield to A7 as a much better example of what process can do.

    Identical GPU blocks, but the 22nm FinFET can run at 533MHz but Apple's A7 at 450Mhz. ~19% headroom in an apples-to-apples performance comparison. I assume that the Intel will be same or lower power.
    Mar 16 12:43 PM | 3 Likes Like |Link to Comment
  • Sales Of Tablets With Core Processors May Be A Tailwind For Intel In Q1 [View article]
    No problem. Now you understand why the ARM threat is actually far more serious than some give it credit for.

    Intel can't afford another "miss" with its 14nm products. They need to start taking market share (with unequivocally superior products) at 14nm to begin cutting off the R&D air supply to these competitors. Time really is of the essence because, as I've shown, we're already at 2010/2011 laptop performance in our phones and accelerating.
    Mar 16 11:05 AM | Likes Like |Link to Comment
  • TSMC's Rebuttal Of Intel's Scaling Advantage Is Just Qualitative, Not Quantitative [View article]
    @ author

    There's actually a simpler way to think about this.

    Intel's graph was essentially talking about M1 pitch (which is a key determinant of density in a logic device), which TSMC has confirmed does not change in moving from 20nm -> 16FF. TSMC does see a slight reduction in SRAM cell size in going from 20nm -> 16nm so in that sense Intel's graph wasn't accurate, but I don't think that was the point.

    If the chart were talking about SRAM cell size, Intel 22nm would compare much more favorably on density against TSMC 28nm than it does.
    Mar 16 10:59 AM | 2 Likes Like |Link to Comment
  • Sales Of Tablets With Core Processors May Be A Tailwind For Intel In Q1 [View article]
    OK, so the Krait you're looking at is a 1.5GHz dual core Krait 200. The top end Snapdragon 800 today is a quad core 2.5GHz quad core Krait 400 (the Krait architecture has evolved significantly since the 2011 processor your reference).

    So crudely speaking let's assume 30% per-clock performance improvement from Krait 200 -> Krait 400, then assume twice the cores and a move from 1.5GHz -> 2.5GHz.

    Quantifying it:

    1.5GHz -> 2.5GHz brings you from 9880 to 16400
    Per core uplift of 30% brings you from 16400 -> 21320
    Doubling your cores brings you from 21320 -> 42640.

    Peak theoretical performance of the Krait is now 1/3rd of your very best desktop-oriented 95W PC chip. I realize these estimates are crude, but this does illustrate my point that you're not so far off from PC performance as you think with these ARM processors ;-)

    Also, I claimed that a dual core low end Bay Trail (Intel Pentium N2805, for example) is the low end PC chip I'm talking about. The Silvermont core in Bay Trail is roughly equivalent to a Krait 400 found in Snapdragon 800, but there are half of them.

    So, there are actually smartphones with more performance than low end Intel powered notebooks and desktops. In fact, Intel's own Merrifield inside of a mobile phone should be roughly equivalent in performance to its low end Pentium/Celeron PC processors for notebooks.
    Mar 16 12:15 AM | Likes Like |Link to Comment
  • Sales Of Tablets With Core Processors May Be A Tailwind For Intel In Q1 [View article]
    "An ARM based desktop today would be like an old 80486 chip around the turn of the century - not a lot of demand for those today."

    Sorry, but this is nonsense. There are desktops today that ship with dual core Bay Trail-D chips which actually offer lower performance than a quad Cortex A15 implementation found in a phone. These are low end desktops, I grant you, but Qualcomm/NVIDIA/Samsung ARM-based chips - if they weren't restricted by the X86 requirement of full Windows 8.1 - would be quite formidable in the low end where Intel and AMD play today.
    Mar 15 11:43 PM | Likes Like |Link to Comment
  • Intel Is Late, But It Didn't Miss Mobile [View article]
    Really? Because QCOM - the most profitable mobile chip player - generates ~$3.3B in EBIT from chips (including non-AP stuff), so ex-modems it's <$3B.

    Intel's PC business generates $12B in EBIT and its server biz generates $5.3B+.
    Mar 14 04:11 PM | 1 Like Like |Link to Comment
  • Intel Is Late, But It Didn't Miss Mobile [View article]
    I suppose ARM missed the server market by not aggressively pursuing it until recently...;-)

    Some consistency, please?
    Mar 13 10:36 PM | 1 Like Like |Link to Comment
COMMENTS STATS
8,691 Comments
13,356 Likes