OpenBSD Journal

New C7 CPU from VIA

Contributed by jolan on from the omg-via-c7-sha-rsa-rng-nx dept.

Thanks to Ryan M. Graham for pointing out:
VIA issued a press release on May 27th announcing their new C7 Processor. Noteworthy PVR related features include its low power consumption (2.0GHz at 20W, idle at 0.1W) and support for SSE2 and SSE3 (in addition to existing MMX and SSE support). In the crypto department, they've expanded their PadLock™ hardware crypto/security suite with this CPU by adding in full hardware SHA-1 and SHA-256 (20BGps), hardware assisted RSA (hardware Montgomery Multiplier), a beefed up RNG that feeds directly into the SHA units, and the NX bit.

Work has already begun to support the C7 even though it hasn't been formally released yet.

(Comments are closed)


Comments
  1. By RC (71.105.37.129) on

    GHz numbers mean nothing... Particularly when coming from VIA.

    With AMD (finally) comming around and making some very, very low powered AMD64 processors (even their almost-top-of-the-line CPUs), I think VIA may lose their CPU market soon.


    [Not-so-happy owner of a few (new) VIA and (old) Cyrix CPUs]

    Comments
    1. By Anonymous Coward (64.231.52.56) on

      I've always wondered how OpenBSD performs on the VIA EPIA boards like this one for example. The only reason I came to hear about them was because of the hardware AES accleration feature. Otherwise for the most part Soekris boards seem to be the most popular for running OpenBSD as a router on small hardware.

      Any VIA EPIA users that would like to share their experiences with the rest of us? How well does that hardware perform with OpenBSD when trying to route traffic on a T1 and have 10 VPN tunnels?

      Comments
      1. By Anonymous Coward (63.251.255.223) on

        I also really like the Soekris boards, but they can be underpowered for some applications. What'd be really nice is if the next-gen Soekris board used a C7 to get the hardware crypto acceleration, though that would likely cut into their sales of PCI hw crypto accelerators.

        An ideal system would be something like the existing net4801+lan1621 with a C7 CPU, and an HGST E-series (eg E5K100, E7K60, E7K100) drive. And, yes, I'd really like it fanless/quiet please, so no need to run the C7 at 2GHz.

        Comments
        1. By Brad (204.101.180.70) brad at comstyle dot com on

          I would consider a next-gen Soekris box if Soren uses a decent NIC chipset too. The NS chipset (sis(4)) just doesn't cut it.

          Comments
          1. By marco (149.169.52.82) on

            soren asked for comments a few months ago. probably too late now though. he tried getting sk(4), but they didn't seem interested in selling to him in small numbers

      2. By Anonymous Coward (68.63.157.203) on

        I've been looking for an alternative to the Soekris boards. I've been less than impressed by their customer service. It will be nice if this C7 winds up on a more mass-market mini-ITX motherboard.

    2. By djm (203.217.30.86) on

      AMD don't have AES or SHA instructions, until they do they are in quite a different market.

      Comments
      1. By sthen (81.168.66.229) on

        Geode LX (as the net5501 is meant to be using) apparently does AES 128 ECB/CBC.

      2. By RC (71.105.192.160) on

        > AMD don't have AES or SHA instructions, until they do
        > they are in quite a different market.

        Nonsense. Thousands of people aren't buying VIA processors to get AES hardware encryption support, particularly since they only recently got that AES support.

        Besides, a PCI crypto card only costs $100, and can do much, much more than just AES. It will also be vastly faster than just about anyone would need (and if not, they probably need a CPU faster than a C7 to support what they are doing anyhow).

        Comments
        1. By sthen (81.168.66.229) on

          A PCI crypto card has a lot more overhead than instructions on the CPU itself. Cards can be quite useful where larger buffers are involved, but where only small amounts can be encrypted at a time, e.g. voip packets, typing into ssh sessions etc, low-overhead instructions run on the CPU can be a lot more useful.

  2. By Anonymous Coward (208.252.48.163) on

    Where can I see the source for VIA's PadLock suite?

    Comments
    1. By Ryan M. Graham (208.181.115.2) on http://ryan.rsgraham.net

      Either you don't understand what padlock is, or I don't understand the question, but just in case it's the former, the padlock suite is a set of features in the cpu.

      I'm not sure if it's accessed through an extended instruction set, or through a separate device interface, but in the end, you're interfacing with hardware, not software, so it doesn't really have source code.

      Unless you're refering to microcode, or possibly the benchmarking software? If you want to see source code for actually using padlock, take a look in the OpenBSD source tree, I'm sure it's existing support for the C3's will give you what you want.

      Comments
      1. By Anthony Roberts (68.145.103.21) on

        I think it's extra instructions. Drivers have been written that can use them when available but they're also available in userspace (I think OpenSSL uses them).

      2. By Daniel (213.6.117.129) on

        Programming Guide for RNG in PDF format is here.

        Programming Guide for ACE (Cryptographic Engine) is here.

  3. By Anonymous Coward (213.89.220.64) on

    Is it possible to run this CPU without a fan ?

    Comments
    1. By RC (71.105.192.160) on

      > Is it possible to run this CPU without a fan ?

      Whether a fan is necessary or technically "optional" depends upon how much power the CPU uses, and how well your case can transfer heat away from hot components like the CPU.

      At 30watts, I'm sure the 2GHz C7 isn't considered fanless by anyone's standards, though that's not out of realm of possibility with a very well-designed system.

    2. By Achoice (212.214.217.4) on

      I don't think so. I run a EPIA PD6000E at 600Mhz - it is certified to run fanless, but good case flow is recomended - i.e. a case fan. I admit I've switched to Debian for better desktop performance and nice packet-upgrades. It works fine as a fanless desktop - only HD noise, and no nois at all when it spins down. Some air flow is requierd at 100% CPU for more than 10 min - like compiling.

      Comments
      1. By sickness (82.58.141.186) sickness@tiscali.it on http://www.sickness.it

        I've the pd6000e fanless and I use it as my home gateway with openbsd and an internal 2,5 harddrive, 24x7, no problems, BUT i'd like to know if anyone managed to use the serial port, I can't make it work :(

  4. By Anonymous Coward (193.194.84.198) iulian@create.ro on

    Via CPU is a good solution for VPN servers?

    Comments
    1. By djm (218.214.226.34) on

      Probably - anything that uses crypto will benefit, and the benefit is far greater than PCI co-processors because there are no bus bandwidth or latency problems. User-space processes benefit more still, as there is no context-switch required either.

  5. By Anonymous Coward (211.30.32.251) on

    "Any VIA EPIA users that would like to share their experiences with the rest of us?"


    Here's my experience.

    I own a EPIA PD10000 Mini-ITX setup. (The one featuring dual LAN connections). Its identical to the older CL10000 one, but offers upto 4 addition serial connections and uses the newer stepping "Nehemiah" CPUs.

    It features the C3 "Nehemiah" CPU running at 1Ghz. I can overclock this to 1.5Ghz with 100% stability. (tested for 24/7 for 2 weeks).

    VIA is annoying because they don't say which particular CPU is on those EPIA mobos. The reason I'm saying this, is because the PD10000 uses two different CPU steppings for the PD10000.

    I found this out when a fellow OpenBSD user and I compared the CPUs. Mine had AES capability and his didn't...You don't know which stepping of CPU you get until you plug it in and turn it on.

    OpenBSD runs on my PD10000 without an issue, it has been for 6 months now. Installation was pretty uneventful.


    "How well does that hardware perform with OpenBSD when trying to route traffic on a T1 and have 10 VPN tunnels?"

    I'm not sure about T1 line, as I've never own or use that, I'm using 10Mbit Cable, and the PD10000 can easily handle it without breaking a sweat.

    I did some "off Internet" testing...Without VPN, it can handle approx 80MBps to 90MBps. About the same performance compared to a PIII 500Mhz.

    With VPN (not using AES hardware), this drops down to 30Mbps. With AES, this climbs back up to non VPN speeds. (The crypto on this CPU shouldn't have a problem in handling 1Gb speeds...But sadly the EPIA series don't have 1Gb, and if they do, its shared with PCI bus). :(

    I hope that helps.



    I agree that the crypto card offers more flexibility, BUT, AES abilities in the CPU completely craps on a PCI crypto card. So I don't know why people are saying a PCI crypto card is faster, when its not in the AES department.

    The C7 should be a much better CPU than the current C3 line, offering much more in crypto and overall performance (using the Pentium-M or P4 bus compared to the older PIII one).

    I suspect it will be closer to last of Pentium IIIs. (The mobile C7 is marketed against the Celeron-M, the budget variant of the Pentium-M)...So we can expect alot more than what the C3 currently offers us.

    I'd be glad to get an EPIA mobo based on this newer CPU and do some benchmarks.

  6. By Anonymous Coward (67.174.182.21) on

    I would love to see a small-form factor board w/ following:

    C7 chip, fanless (so maybe not 2.0 Ghz version).
    No regular PCI bus, just PCI Express bus.
    No IDE (ATA) connections, just SATA.
    And wouldn't it be nice if there was a commodity class Serial based memory architecture that could reduce mobo traces further??

    As far as input/outputs, 1/2 ethernets would be nice. I would forgo RS-232 serial and also parallel ports and be happy with just USB 2.0 ports. I am talking about space efficiency here. Of course Audio/Video connectors and keyboard/mouse (PS2 or USB based ??).

    Anyway, it would be nice to see a well designed, not fitting everything but the kitchen sink, small-form factor board like this. At least I'd like to see it.

    Comments
    1. By Terry (216.94.38.98) cyclone@kwic.com on

      I would love to see ANY C7 product at all!!!!! IT IS ALMOST MARCH, 2006 where is this stuff???? There isn't a single board on the VIA site using this processor? (or a prosessor for sale ,for that matter) The VIA site which I have now spent days wondering through (what a mess) does't explain some of the simple obvious things; Like what pin configuration a C7 is? Is it a socket 370? I was asked a month ago to put together a presentation on a Linux/VIA cluster server/client system. With all the improvements in the C7 Notably the Padlock security improvements on the C7; There is no point in considering C3 anymore. But where is this Product! It has been 8 or 9 months! This is approaching "VAPOUR WARE" status! Terry

      Comments
      1. By Anonymous Coward (86.55.238.68) on

        > I would love to see ANY C7 product at all!!!!!
        > IT IS ALMOST MARCH, 2006 where is this stuff????
        > There isn't a single board on the VIA site using this processor?
        > (or a prosessor for sale ,for that matter) The VIA site which I have now spent days wondering through (what a mess) does't explain some of the simple obvious things; Like what pin configuration a C7 is? Is it a socket 370?
        > I was asked a month ago to put together a presentation on a Linux/VIA cluster server/client system. With all the improvements in the C7 Notably the Padlock security improvements on the C7; There is no point in considering C3 anymore. But where is this Product!
        > It has been 8 or 9 months! This is approaching "VAPOUR WARE" status!
        >
        > Terry

        try here: www.logicsupply.com

Credits

Copyright © - Daniel Hartmeier. All rights reserved. Articles and comments are copyright their respective authors, submission implies license to publish on this web site. Contents of the archive prior to as well as images and HTML templates were copied from the fabulous original deadly.org with Jose's and Jim's kind permission. This journal runs as CGI with httpd(8) on OpenBSD, the source code is BSD licensed. undeadly \Un*dead"ly\, a. Not subject to death; immortal. [Obs.]