Andrei F.

@andreif7

Luxembourg
Joined June 2012
187 Photos and videos Photos and videos

Tweets

You blocked @andreif7

Are you sure you want to view these Tweets? Viewing Tweets won't unblock @andreif7

  1. Dec 9

    Transparent Huge Pages set to default enabled on a production server? Yay or nay?

    7 replies 2 retweets 1 like
    Undo
  2. Dec 7

    I'm wrapping up testing of an Ampere Altra Q80-33 vs Xeon 8280 vs EPYC 7742 review. Taking requests on some any additional *meaningful* benchmarking that could be run on it - I don't have a setup for server/client tests due to networking reasons, so local tests only.

    10 replies 1 retweet 29 likes
    Undo
  3. Dec 4

    I also can get to a similar efficiency on the M1 cores - 48500 CoreMarks per Watt, not that it means anything of significance.

    Image
    25 likes
    Show this thread
    Image
    Show this thread
    Undo
  4. Dec 4

    "The M1 chip is roughly the equivalent of 10,000 CoreMarks in EEMBC terms; divide this by eight cores and 15W total, and that is less than 100 CoreMarks per Watt." They're only off by two orders of magnitude, but hey, it's the same 1st digit number.

    Image
    5 replies 3 retweets 65 likes
    Show this thread
    Image
    Show this thread
    Undo
  5. Dec 3

    It's such a pity that absolutely none of the big camera vendors talk about their camera SoCs. The way that these overheat they must be using some ancient process node with some old CPU IP.

    4 replies 2 retweets 39 likes
    Show this thread
    Image
    Show this thread
    Undo
  6. Dec 3

    This talk about phone cameras makes me wish for one thing: An actual ILC mirrorless which actually employs the latest advancements in ISP and sensor technology, computational photography, and what not. Stick a Snapdragon SoC into a proper camera.

    6 replies 2 retweets 48 likes
    Show this thread
    Image
    Show this thread
    Undo
  7. Nov 22

    This might be a wee too close for the 850W PSU...

    Image
    14 likes
    Undo
  8. Nov 21

    Thought: Apple years ago stated that they design their CPU&GPU µarch with desktop workloads in mind. Meanwhile Arm, Samsung and etc were designing cores with mobile in mind. Those designs scaled badly with newer workloads. Statement no longer valid with Arm Neoverse now ofc.

    3 replies 7 retweets 74 likes
    Undo
  9. Nov 19

    Hey look. We now debunked that Apple devices aren't ahead just because of a better Javascript engine.

    Image
    7 replies 12 retweets 100 likes
    Undo
  10. Nov 19

    Not sure how to feel about the M1 core having 3x the load bw and 2x the copy BW of my 3700X.

    6 replies 2 retweets 81 likes
    Undo
  11. Nov 18

    Now I know what I have sitting on my desk and can't open up 😅

    2 replies 4 retweets 23 likes
    Undo
  12. Nov 17

    By the way, there's *actual* TDP, as in the thermal envelope of the device, and actual chip power consumption. The M1 seems to have no issues going from 21W to 32W (CPU+GPU fully loaded), even after longer usage at 21W.

    Image
    5 replies 6 retweets 61 likes
    Undo
  13. Nov 17

    Aztec High

    Image
    2 replies 14 likes
    Show this thread
    Image
    Show this thread
    Undo
  14. Nov 17

    Heh, of course libquantum DRAM and core power is off the charts.

    Image
    1 reply 17 likes
    Show this thread
    Image
    Show this thread
    Undo
  15. Nov 17

    21W in SPEC 511.povray MT 15W in R23 MT 5.4W in SPEC 511.povray ST 3.8W in R23 ST (!!!!!) 4.8W to 5.4W in 502.gcc ST I had no idea Cinebench had such a low core utilisation. That's pretty absurd. That's a huge red flag there.

    6 replies 6 retweets 38 likes
    Show this thread
    Image
    Show this thread
    Undo
  16. Nov 17

    TIL about Mac powermetrics. Unfortunately I wasn't that familiar with these toolings, never actually having a Mac device before. There you go, 5W "power virus" on singe-core.

    Image
    Image
    Image
    Image
    2 replies 12 retweets 79 likes
    Show this thread
    Image
    Show this thread
    Undo
  17. Nov 17

    🤣

    Replying to @TheRealVeedrac
    I hate to break it to you, but Cinebench is a real workload for Cinema 4D, a real application that someone may run on a 16-core desktop processor. Spec2006 is a synthetic benchmark from 14 years ago when Intel's best desktop CPU was a dual-core
    5 replies 1 retweet 35 likes
    Undo
  18. Nov 13

    Lots of accusations of me being an Apple shill 😂 I literally use a Windows machine and Android as my daily devices, don't even own any Apple devices besides the review samples, which I personally dislike because of iOS. Their silicon is another story.

    22 replies 4 retweets 195 likes
    Undo
  19. Nov 12

    1 reply 1 like
    Show this thread
    Image
    Show this thread
    Undo
  20. Nov 12

    These are the actual numbers of a few subtests without, and with *actual* PGO, this time on the A14, with *even better* branch mispredict numbers than Intel.

    Image
    4 replies 2 retweets 25 likes
    Show this thread
    Image
    Show this thread
    Undo

@andreif7 hasn't Tweeted yet.

Loading seems to be taking a while.

Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.

    You may also like

    ·