Announcement

Collapse
No announcement yet.

NVIDIA's Jetson TK1 Is Being EOL'ed Next Month

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #11
    Originally posted by coder View Post
    You can get a new Lenovo Thinkpad 13 Chromebook with that amount of RAM & resolution for $330. Better yet, you can run a mainline distro and get working wifi and GPU drivers.
    https://www3.lenovo.com/us/en/commer.../p/22TP2TX132E
    Interesting
    Battery Runtime
    Idle (without WLAN, min brightness) 16h 29min
    NBC WiFi Websurfing Battery Test 1.3 9h 40min
    Load (maximum brightness) 1h 53min
    So you can finally get 1920x1080, 4 GB and ~10 hours battery runtime for a similar price I paid over 2 years ago. I actually don't see the acer chromebook anywhere in germany anymore so I have no direct comparison but by now it should be a good bit cheaper.

    Btw i assume you refer to this model?
    ThinkPad 13 Chromebook - Black
    Part Number: 20GL0006US
    Web Price: $479.00
    After Instant Savings: $329.99

    As far as I can tell this has never been sold anywhere in germany.

    Comment


    • #12
      Originally posted by haagch View Post
      So you can finally get 1920x1080, 4 GB and ~10 hours battery runtime for a similar price I paid over 2 years ago. I actually don't see the acer chromebook anywhere in germany anymore so I have no direct comparison but by now it should be a good bit cheaper.
      Depends on how much you value working wifi and GPU drivers. I have v1 of Thinkpad 13 w/ i3 CPU, and the wifi + GPU have been flawless.

      Comment


      • #13
        Originally posted by coder View Post
        No.

        You can build a better NUC or SFF-based PC, for the money. The CPU will be faster and the iGPU is utterly comparable.
        Except you can't learn CUDA on your version.

        Comment


        • #14
          Originally posted by vegabook View Post
          Except you can't learn CUDA on your version.
          LOL at seeing this from someone with "vega" in their username.

          In case that wasn't sarcasm: I don't want to learn a vendor-specific API, when I can use OpenCL (which runs quite well on the Intel iGPU in mine).

          The Tegra SoCs do not support OpenCL, for absolutely no good technical reason. So, if you didn't want to be stuck using CUDA, then Tegra leaves you SoL.

          ...at this point, someone will chime in with a comment about an API emulator that runs atop CUDA, or an open source Tegra driver that does support OpenCL. While I applaud such efforts, why go down this road if you don't have to?

          Comment


          • #15
            Originally posted by coder View Post
            LOL at seeing this from someone with "vega" in their username.

            In case that wasn't sarcasm: I don't want to learn a vendor-specific API, when I can use OpenCL (which runs quite well on the Intel iGPU in mine).

            The Tegra SoCs do not support OpenCL, for absolutely no good technical reason. So, if you didn't want to be stuck using CUDA, then Tegra leaves you SoL.

            ...at this point, someone will chime in with a comment about an API emulator that runs atop CUDA, or an open source Tegra driver that does support OpenCL. While I applaud such efforts, why go down this road if you don't have to?
            vega is a common 'greek" letter in mathematical finance. It wasn't invented last year by AMD.

            I like, and indeed prefer, OpenCL but your point doesn't invalidate mine which is that, if for whatever reason (and there are many) you might want to learn CUDA, this board is great.
            Last edited by vegabook; 28 March 2018, 09:04 AM.

            Comment


            • #16
              Originally posted by vegabook View Post
              vega is a common 'greek" letter in mathematical finance. It wasn't invented last year by AMD.
              Relax, I was just pointing out the irony.

              Originally posted by vegabook View Post
              if for whatever reason (and there are many) you might want to learn CUDA, this board is great.
              I think a PC with a low-end card like GTX 1050 or something 900-series is better. Performance is higher, price is lower, and you can do much more with it (i.e. games, most of which are x86, OpenCL, and using it for desktop graphics with CPUs lacking an iGPU).

              Best of all, it's still supported! I think even the latest SDK supports CUDA capability 2.0+, which means everything from Fermi (launched in 2010), onward. If you want to use older GPUs, just stick with CUDA SDK v8.

              Comment

              Working...
              X