Announcement

Collapse
No announcement yet.

SiFive HiFive Unmatched Hands-On, Initial RISC-V Performance Benchmarks

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

  • coder
    replied
    Originally posted by GruenSein View Post
    The only big player I can actually see investing in RISC-V is Apple
    Well, in terms of governmental entities, China and the EU seem very interested in the independence RISC-V affords them from either foreign corporations or governments. In fact, both Baidu and Alibaba have been working hard in this area:
    As for the EU, their EPI program is still progressing on the RISC-V front:
    Last edited by coder; 29 September 2021, 02:56 AM.

    Leave a comment:


  • GruenSein
    replied
    Originally posted by igxqrrl View Post

    Rumors are that Intel is trying to buy SiFive for $2 billion, so that's a bit of an investment.

    My long-held contention is that Risc-V needs an equivalent to the raspberry pi. A sub-$100 SBC that can run Linux. At that price point performance isn't critical, but it gives college students and hobbyists an avenue to actually have a computer that runs the same ISA they're learning, which potentially opens up the floodgates for further development.

    At this point SiFive seems to be the only company with the IP portfolio to make that possible, but they have very little interest in it. The problem of course is that it's not a money maker. But I do think it would go a long ways to increase mindshare.
    My fear is that Intel might buy SiFive to make it go away...

    Regarding the cheap dev-platform, I am all for that. The problem is that this is only helpful on the software side giving people a cheap device to tinker with and potentially leading to more distros etc. picking up RISC-V support similar to what happened with the rpi becoming widely available and popular. And still, we have a bunch of ARM-ready distros but hardly any actual ARM devices you could run them on apart from the rpi. Qualcomm doesn't seem to think it is worth releasing an mITX-Snapdragon 8cx gen ... board, for example, which they easily could, because it wouldn't hold a candle to AMD or Intel. There is the rare snapdragon laptop here and there which gets great battery life but that is mostly it and doesn't seem to all that promising. The problem is that switching to anything new causes some headaches at first. To make it worth it, the new solution must be significantly better than what we have. Even though decent cores are available, nobody seems to really think that ARM can accomplish this except for Apple (and that is partly because they have the fewest problems transitioning to new architectures due to their fully integrated ecosystem).
    For RISC-V, it's the same except that there are no high performance cores and college students and hobbyist are usually not the ones driving high performance processor core design forward. The hurdles of entry into that type of development are much higher. For RISC-V to become an actual option for real users, e.g., home users and professionals, a high performance core with significant advantages over existing solutions is needed. Currently, I don't see who'd develop this and bring it to mass market. And even when there are good cores: Just consider the Power-based workstations. Those can be bought but they are mostly an oddity with no widespread use case.

    Leave a comment:


  • igxqrrl
    replied
    Originally posted by GruenSein View Post
    And even though people are going to get mad when I say this: The only big player I can actually see investing in RISC-V is Apple (because they control everything that is needed and might have an interest to become independent of ARM at some point) and they just switched their entire ecosystem to ARM.
    Rumors are that Intel is trying to buy SiFive for $2 billion, so that's a bit of an investment.

    My long-held contention is that Risc-V needs an equivalent to the raspberry pi. A sub-$100 SBC that can run Linux. At that price point performance isn't critical, but it gives college students and hobbyists an avenue to actually have a computer that runs the same ISA they're learning, which potentially opens up the floodgates for further development.

    At this point SiFive seems to be the only company with the IP portfolio to make that possible, but they have very little interest in it. The problem of course is that it's not a money maker. But I do think it would go a long ways to increase mindshare.

    Leave a comment:


  • GruenSein
    replied
    Originally posted by igxqrrl View Post

    I agree with everything you're saying.

    On the other hand it seems very slow progress in this space. It's been nearly 4 years since the HiFive Unleashed. I suspect many of us were hoping for a system that was at least "usable" by now, but clearly this isn't it.
    I'd like that, too. However, someone has to finance the progress of RISC-V and apart from micro-controller designs larger companies don't seem all too interested. The big CPU manufacturers either have their own cores in which they have invested decades of work (Intel, AMD) or they just buy ARM designs. To them, leveling the playing field by starting fresh isn't very appealing. Other companies either lack the knowledge, funds or business model to pick up RISC-V. And even though people are going to get mad when I say this: The only big player I can actually see investing in RISC-V is Apple (because they control everything that is needed and might have an interest to become independent of ARM at some point) and they just switched their entire ecosystem to ARM. SoC makers focusing on Android devices for example need to stay compatible and have little control over the direction Android and the apps will take.
    Last edited by GruenSein; 28 September 2021, 10:50 AM.

    Leave a comment:


  • brucehoult
    replied
    Originally posted by PerformanceExpert View Post

    Here are some you can still buy: https://www.deviceranks.com/en/platf...o-p25-mt6757cd (there seem to be 2 bins of that SoC, some are 2.6GHz, others are 2.5GHz)
    That's a 16nm chip, not 28nm.

    https://en.wikichip.org/wiki/mediatek/helio/mt6757cd

    Leave a comment:


  • rene
    replied
    Originally posted by agd5f View Post

    No offense, but I think everyone would like to tinker around with stuff that they find interesting and get paid for it regardless of whether that project is open source of not. Unfortunately, there are not a lot of people that want to pay you to do that. I would take that job if I could get it.

    Open source is a community, that is where everything comes together. Everyone that participates has a seat at the table. In the case of companies that support open source, like AMD or Intel, they are providing millions of lines of code to make something like Linux work at all on your platform. Without that work, you wouldn't have a baseline to even debug these issues in the first place. Plus, if you do find an issue or have a question about the hardware or driver, etc. you can ask the relevant developers directly. Sometimes it's better to start there to save debugging time. A few good questions to the relevant mailing lists could have saved you hours of debugging time. I could have told you that DCN requires FP support for example.
    I plugged in the RDNA2 card into the RISCV board and it had no output. By the time I analyzed why ,I already found it needs in-kernel FP support. What should I have asked? Hey, I got an RDNA2 card and it does not work on this shiny new RISCV board please help? This is a bit oversimplified and not how it usually works. Beside in the time waiting to maybe get an answer I already had a working prototype ;-)

    Leave a comment:


  • agd5f
    replied
    Originally posted by rene View Post
    Well, having some contribution bounty program is cheaper than hiring. Also filling out application paperwork likely takes longer than finishing the patch, beside I have no plan to change my life like this. If I'd wanted to work in a big cooperation I would have worked at IBM, Siemens, SAP, Volkswagen, Bosch or Apple already since a decade or two... IMHO there is a huge imbalance between companies profiting from open source and given back to individuals in the community who regularly contribute useful work or change.
    No offense, but I think everyone would like to tinker around with stuff that they find interesting and get paid for it regardless of whether that project is open source of not. Unfortunately, there are not a lot of people that want to pay you to do that. I would take that job if I could get it.

    Open source is a community, that is where everything comes together. Everyone that participates has a seat at the table. In the case of companies that support open source, like AMD or Intel, they are providing millions of lines of code to make something like Linux work at all on your platform. Without that work, you wouldn't have a baseline to even debug these issues in the first place. Plus, if you do find an issue or have a question about the hardware or driver, etc. you can ask the relevant developers directly. Sometimes it's better to start there to save debugging time. A few good questions to the relevant mailing lists could have saved you hours of debugging time. I could have told you that DCN requires FP support for example.

    Leave a comment:


  • rene
    replied
    Originally posted by agd5f View Post

    If you are interested in getting paid to work on AMD hardware, consider applying to work at AMD. We are hiring.
    Well, having some contribution bounty program is cheaper than hiring. Also filling out application paperwork likely takes longer than finishing the patch, beside I have no plan to change my life like this. If I'd wanted to work in a big cooperation I would have worked at IBM, Siemens, SAP, Volkswagen, Bosch or Apple already since a decade or two... IMHO there is a huge imbalance between companies profiting from open source and given back to individuals in the community who regularly contribute useful work or change. Instead of hiring a handful of people, IMHO we should also work towards normalizing such a program similar to what Infosec has long established in terms of security bounties. We don't we simply have this for non security code work? But if AMD insists, they can make an offer, my email and LinkedIn and such are open ;-)
    PS: I also finally tracked down and patched why enabling x2apic usually (or always) disables P-states on AMD Zen* this weekend (tested on x570 and x470 w/ 5950x and 3950x, but it probably simply affects most BIOS' ACPI tables ;-): https://www.youtube.com/watch?v=BTI78pG2V80
    PPS: Did I mention I have a Patreon? https://patreon.com/renerebe
    Last edited by rene; 27 September 2021, 06:37 PM.

    Leave a comment:


  • agd5f
    replied
    Originally posted by rene View Post

    but I have plenty https://www.youtube.com/watch?v=tFNKXSZGyEo of https://www.youtube.com/watch?v=RXXBFcQ0I-M AMD https://www.youtube.com/watch?v=pC4JCPb4v1Q CPUs! ;-) (and obviously more)

    It's a bit ironic to start with "not aware, where did you ask", just to follow up with "doubt there is an easy path to funding this sort of work" though :-/ If it helps, I was eying an TR 5990x next, maybe AMD has one to spare?

    PS: This is exactly why I started the YT channel, to have means to finance independent OpenSource work without begging for money for each change and month.
    If you are interested in getting paid to work on AMD hardware, consider applying to work at AMD. We are hiring.

    Leave a comment:


  • SavageX
    replied
    Originally posted by PerformanceExpert View Post
    However dual-issue is rarely the bottleneck once you look beyond Dhrystone or CoreMark. Branch prediction, caches, TLB handling, prefetchers etc matter a lot more, particularly on in-order cores.
    You're very likely completely right.

    I tried to come up with some C code example where there would be considerable back-to-back load/stores and/or branches, but even then there's usually e.g. address calculation stuff (incrementing pointers and/or offsets and such) that can make good use of the dual-issue capabilities. Integer divisions usually are too rare to make a real difference.

    So yeah, it should mostly come down to caches, branch-prediction, TLBs and such.

    Leave a comment:

Working...
X