Slashdot Mirror


NVIDIA's nForce Professional and Tyan's Words

CoffeeJunked writes "There's a lot of buzz about dual-core CPUs and with the release of the nForce Professional chipset from nVidia, there's a lot of buzz about the future of SMP machines as we know them. LinuxHardware.org has just published a couple of articles that get to the heart of the new chipset and what board manufacturers will be doing with them. The first article covers the chipsets and boards, while the second article is an interview with Tyan about what to expect from them this year. It's a good read all around."

17 of 138 comments (clear)

  1. The boards look great, except... by RebelWebmaster · · Score: 3, Insightful

    Where are the SATA connectors?!?!?! I find it amazing that the K8WE only has 2 and the K8SER 4. While we're on the topic, having at least 1 PCIe x1 slot would be nice. These high end server boards are being outclassed by nForce4 SLI motherboards. (And for the record, using more than 4 SATA ports is very doable)

    1. Re:The boards look great, except... by lachlan76 · · Score: 3, Insightful

      High end servers sure aren't gonna be using SATA...

    2. Re:The boards look great, except... by jred · · Score: 4, Insightful

      CPU load. SCSI puts much less of a load on your CPU than (s)ata does.

      SATA rocks, though.

      --

      jred
      I'm not a mechanic but I play one in my garage...
    3. Re:The boards look great, except... by Anonymous Coward · · Score: 3, Informative

      Sorry but 10K rpm SCSI drives (not to mention 15K) pound even Raptors extrememly hard in the performance arena. Look it up. No, here you go:

      http://techreport.com/reviews/2005q1/maxtor-diamon dmax10/index.x?pg=4

    4. Re:The boards look great, except... by dasunt · · Score: 3, Insightful

      CPU load. SCSI puts much less of a load on your CPU than (s)ata does.

      Depending on the task, the CPU for PATA/SATA isn't that bad.

      For fileservers, on a price/capacity ratio, SATA will kick SCSI's ass to the curb and back. While SCSI is faster, and, on average, more reliable, SATA is often 'good enough'.

      Or imagine a webserver with huge amounts of memory. For performance, SATA and SCSI will be roughly equal, since most files will be cached in the memory.

      What about a DNS server: Again, the performance of the system should be dependent on memory, not the hard drive speeds.

      Don't forget firewalls. SATA is fast enough for log files, and the CPU shouldn't be a bottleneck unless your firewall rules are extremely complex.

      I wouldn't use SATA in a database server or in any other application with a lot of random disk reads/writes, but it has its uses, even in servers.

    5. Re:The boards look great, except... by NeuralAbyss · · Score: 3, Informative

      I've got a SuperTrak (the 12 drive version) running on a Linux 2.4 server running Debian. Make sure you enable the non-Windows OS option on the card (changes it to I2O operation), and turn on I2O options in kernel config. Works like a charm.. appears as one device.

  2. Talk about useless. by Anonymous Coward · · Score: 5, Insightful

    So, they are designing a chipset for servers, which will run linux or bsd, but they refuse to provide docs or hardware to linux and bsd developers, meaning their shit is always poorly supported. Hooray.

    1. Re:Talk about useless. by Anonymous Coward · · Score: 5, Interesting

      I'm a bit concerned about the nforce4, from what i read already, there are 3 models, a "normal" nforce4, a "ultra" nforce4, and a "sli" nforce4. But altough you can't use SLI on the non-sli models, there are ways to enable SLI, on at least the ultra model ( http://www.anandtech.com/printarticle.aspx?i=2322/ ).
      To quote the article, "Just as quickly, we learned that nVidia was not happy with this "SLI hack" and they changed their drivers quickly so that "semi-SLI would not work with current and later Forceware drivers." It appears that the later Forceware drivers check the chipset ID and if the driver sees "Ultra", then SLI is not enabled. MSI decided to kill the "semi-SLI" board because it would be a nightmare supporting a board that would only run with older nVidia SLI drivers."

      So, how will this be (un)supported by the opensource community? Is nvidia doing to chipsets what they did to graphic cards? Everyone remembers how they locked out rgb overlays and unified front+back buffers from the geforce4 cards, altough the chips had the funcionality built-in, the drivers would disable these features, and save them for the more expensive quadro cards (there were some quick fixes for this, for windows, mainly rivatuner and softquadro4).
      Does this means that now they're going to lock-out funcionality available on the chipset to maximize profit? I can't imagine how (linux) kernel developers will support a chipset which relies on closed drivers to enable or disable a specific funcionality, and judging by nvidia's attitude in the graphic cards department (which has a point, up to a certain extent nevertheless), i can't imagine nvidia releasing the specs for opensource drivers for this chipset, therefore loosing the income from the sli model, which would become redundant.
      Do we now have to taint the kernel with chipset drivers? If so, i'm out of it, this is certainly a chipset to avoid.

  3. The New Tyan Boards using the Nvidia Chipset by rchatterjee · · Score: 5, Informative
  4. Free Drivers by gustgr · · Score: 4, Interesting

    Finally, NVIDIA's SLI has been a hot topic here because, as of yet, we haven't seen Linux drivers that support this hot new feature. When we talked to NVIDIA about this we were finally given a time-line which stated that it may be a couple of months still.

    If the drivers were free software someone skilled enough would hack the missing features. Isn't about time to nVidia change its mind and release the sources?

    1. Re:Free Drivers by Jah-Wren+Ryel · · Score: 3, Interesting

      If the drivers were free software someone skilled enough would hack the missing features. Isn't about time to nVidia change its mind and release the sources?

      Tell that to David Kirk nvidia's chief scientist whose, "sense is that developers on those platforms are quite happy with our efforts" as a justification for not going open source. Plus some totally bizarro bullshit about "hackers tak[ing] bad advantage of raw hardware interfaces."

      It is telling that he did not pull out the old, tried and true "competition sensitive" bullshit that so many hardware vendors have been hiding behind since day one.

      --
      When information is power, privacy is freedom.
  5. The K8WE has 4 by attemptedgoalie · · Score: 3, Informative

    The picture doesn't label the other two. They're down by the SCSI controller pointing forward instead of up. They're also on the RAID with the ones in the picture.

    Trust me.

    (I have one of these boards at my desk.)

    --
    My mom says I'm cool.
  6. Anandtech article was quite interesting... by Bob-o-Matic! · · Score: 4, Informative

    Normal view: http://www.anandtech.com/cpuchipsets/showdoc.aspx? i=2327

    All in one page/"print" version: http://www.anandtech.com/printarticle.aspx?i=2327

    Lots of intersting possibilities. Seems to me that given a motivated/visionary motherboard maker, the only real limits are based on the form factor. Is there a super-ATX out there that would allow for say 8 PCI-e slots, 16+ hard drives, and all the rest of the goodies, all in one case?

    Some will ask if there really is a need for this. Anandtech's Derek Wilson points out that having all the onboard disk controllers could add up to substantial savings-- apparantly expansion card controllers are quite pricey.

    Now, if only those Opteron 8XX processors didn't cost $8XX... (or thereabouts... you get the idea!)

  7. I'd at least respect honesty. by Anonymous Coward · · Score: 4, Insightful

    "We use our drivers to cheat on benchmarks, and if we released info for people to write a driver, it would show our hardware's not as good as we pretend."

  8. nvidia and Linux drivers by jgarzik · · Score: 4, Informative
    As a for-what-it's-worth from a Linux driver author...

    nvidia SATA status and other Linux SATA info.

    nvidia wrote the SATA driver that's current in the Linux kernel, and has generally been helpful in addressing problems that arise in it.

    Although the ethernet driver ("forcedeth") was indeed reverse-engineered, nvidia eventually lent their support behind the effort: they contributed gigabit ethernet support to the driver.

    The video stuff is still closed, of course.

    Jeff
  9. Re:brains on the side by ckaminski · · Score: 3, Informative

    Many operating systems have a concept of processor affinity, whereby due to caching issues wish to ensure that thread/process migration from CPUs does not occur. WindowsNT once had a problem whereby you could ensure that after every context switch your thread had migrated to a new processor, invalidating it's cached data and killing performance. Some applications require this sort of thing, and if you want to ensure that this migration behavior occurs as little as possible, then you can set affinity flags that clue the OS into this fact. It will then attempt to migrate your thread only as a last resort.

    So, Windows can do this, even though it's only a guideline, as opposed to a true enforcement. I understand Linux has this capability, however I'm not positive.

  10. Re:Nvidia Taking a Stand by MojoStan · · Score: 3, Interesting
    However, you forget that Nvidia hasn't actually integrated a GPU in their core logic since the nforce2 chipset... Perhaps Nvidia found that IGP sales hurt their discrete solutions?

    Perhaps. But I think another possibility is that the nForce3 chipset was not meant for "budget/mainstream" users, but for "enthusiasts." As we all know, enthusiasts don't want integrated graphics that share memory with the system.

    The nForce4 chipset, on the other hand, does look like it's aimed at budget/mainstream users as well as enthusiasts. But with PCI Express and TurboCache, NVIDIA might have a cheap solution that's better than integrated graphics.

    PCI Express x16 has more bandwidth than AGP (4 GB/s upstream and downstream) and allows writes directly from the GPU to system RAM. This allows a non-integrated graphics card to share memory with the system without the huge performance hit that AGP would have caused.

    Instead of integrated graphics, maybe NVIDIA is planning to "bundle" their cheap TurboCache cards with nForce4 motherboards. That seems cool to me.

    --
    TO START
    PRESS ANY KEY

    Where's the 'ANY' key? I see Esk, Kitarl, and Pig-Up...