VIA's attempt to regain lost chipset marketshare

  • Thread starter Thread starter Yousuf Khan
  • Start date Start date
Y

Yousuf Khan

X-bit labs - Hardware news - VIA Makes Another Attempt to Re-Enter
Chipset Market.
http://www.xbitlabs.com/news/chipsets/display/20051122145105.html

VIA has lost a lot of ground to Nvidia and ATI in the chipset market.
It's probably not surprising, given their poor reputation for quality.
Known more for their compatibility glitches, I think they're even behind
SIS nowadays in marketshare. So they've upgraded K8T890 to K8T900.
 
X-bit labs - Hardware news - VIA Makes Another Attempt to Re-Enter
Chipset Market.
http://www.xbitlabs.com/news/chipsets/display/20051122145105.html

VIA has lost a lot of ground to Nvidia and ATI in the chipset market.
It's probably not surprising, given their poor reputation for quality.
Known more for their compatibility glitches, I think they're even behind
SIS nowadays in marketshare. So they've upgraded K8T890 to K8T900.

I don't know how nVidia gets away with what they do - the complaints are
think & fast and they have no end user support at all. If you have a
problem there is nowhere to even make it known.... corruption of disk
transfers... a NI which just doesn't work right: checksum offload has
errors. I'm getting a little pissed with them.
 
George said:
I don't know how nVidia gets away with what they do - the complaints are
think & fast and they have no end user support at all. If you have a
problem there is nowhere to even make it known.... corruption of disk
transfers... a NI which just doesn't work right: checksum offload has
errors. I'm getting a little pissed with them.

What are you talking about?

Yousuf Khan
 
It's kind of strange that most windows user's like Nvidia, and most
Linux users like VIA. It seems that VIA has a better open source
approach then Nvidia. For example how long has Gnu/Linux users been
waiting for SLI, its been out for how long? Not to mention that VIA has
usually had good support from the Linux kernel developers.

I use Nvidia GPU's but I would think twice about getting a Nvidia based
Motherboard to use in Linux. I would probably opt for VIA. One thing is
for sure I would not get an ATI based machine.

Gnu_Raiz
 
It's kind of strange that most windows user's like Nvidia, and most
Linux users like VIA. It seems that VIA has a better open source
approach then Nvidia. For example how long has Gnu/Linux users been
waiting for SLI, its been out for how long? Not to mention that VIA has
usually had good support from the Linux kernel developers.

I use Nvidia GPU's but I would think twice about getting a Nvidia based
Motherboard to use in Linux. I would probably opt for VIA. One thing is
for sure I would not get an ATI based machine.

Gnu_Raiz
 
X-bit labs - Hardware news - VIA Makes Another Attempt to Re-Enter
Chipset Market.
http://www.xbitlabs.com/news/chipsets/display/20051122145105.html

VIA has lost a lot of ground to Nvidia and ATI in the chipset market.
It's probably not surprising, given their poor reputation for quality.
Known more for their compatibility glitches, I think they're even behind
SIS nowadays in marketshare. So they've upgraded K8T890 to K8T900.

Actually k8t800 AGP chipset was quite popular and quite good - I have
a dual opteron board with k8t800, and can't complain about either
performance or stability. But looks like they missed PCIe bus.

NNN
 
Yousuf Khan said:
X-bit labs - Hardware news - VIA Makes Another Attempt to Re-Enter
Chipset Market.
http://www.xbitlabs.com/news/chipsets/display/20051122145105.html

VIA has lost a lot of ground to Nvidia and ATI in the chipset market.
It's probably not surprising, given their poor reputation for quality.
Known more for their compatibility glitches, I think they're even behind
SIS nowadays in marketshare. So they've upgraded K8T890 to K8T900.
Technologies aimed at AMD64 platform, such as K8T890-series products,
were adopted massively. With the K8T900 the company may have hopes to
regain positions. However, the designer did not mention of a single
mainboard maker to be planning products that utilise VIA K8T900
core-logic.

VIA just announced sampling on Nov 22. Which suppliers are expected
to announce they are planning products even before receiving, much
less evaluating, those samples? ;-)
 
What are you talking about?

nVidia has problems - check http://forums.nvidia.com/. The recent nForce4
has had corruption problems with some hard disks on SATA. nVidia still
denies it and says Maxtor has fessed up but Maxtor denies it - similar
probs have been seen on Western Digital drives; either way there *is* a
problem and one could hope for better discussion of the issue than posting
on nVidia's forum, with the hope that just *maybe* an nVidia employee
*might* happen by and notice and *might* feel inclined to do something.
Their silence is aggravating.

The nForce chipsets with Gb NIs have a hardware bug which seems to only
manifest in a few situations. I've seen it when working as a nForce3 & 4
client to some mail servers: 1) with Eudora as a mail client, sendmail just
eventually fails with a 10053 or 10054 Eudora error code (a Google search
on "Eudora 10053" will turn up the complaints); 2) trying to upload HDTach
results eventually fails silently. In both cases the fix is to disable
checksum offload in the network driver which, of course, cancels out their
wild performance claims. Actually I'd appreciate it if anyone who has an
nForce3/4 chipset could try an HDTach results upload and post their
experience, just to confirm reproducibility. The error only seems to
happen with large e-mails (HDTach just e-mails its results to Simpli's
mailserver) so it could the length or it could be content sensitive.

Also check out the table of "compatibility" here:
http://www.nvidia.com/object/motherboards.html. It's a bloody mess and is
apparently not the whole story: nVidia's Network Access Manager rarely
works at all - the hardware firewall, is basically unuseable for most
people. In the above table, the "Hardware Network Engine" does not work
with any nForce3 mbrd and several nForce4 mbrds but nowhere on their site,
that I can find, does nVidia tell what "Hardware Network Engine" really
does. There's some suggestion that it's the mechanism which allows
bypassing the OS network stack and performing DMA directly to the
application memory space but it's hard to pin down and impossible to figure
what (not) to do to avoid problems or which software to (not) use.

On top of all the above, the nVidia drivers are going from bad to worse -
every new iteration brings new problems and may or may not fix old ones; it
reminds me of the way video drivers were rigged to wring out extravagant
performance numbers at the expense of system stability.

Quite honestly I'm getting fed-up with this, especially nVidia's refusal to
even listen to end users. I hope VIA makes a comeback - at least they
listen.:-)
 
George said:
nVidia has problems - check http://forums.nvidia.com/. The recent nForce4
has had corruption problems with some hard disks on SATA. nVidia still
denies it and says Maxtor has fessed up but Maxtor denies it - similar
probs have been seen on Western Digital drives; either way there *is* a
problem and one could hope for better discussion of the issue than posting
on nVidia's forum, with the hope that just *maybe* an nVidia employee
*might* happen by and notice and *might* feel inclined to do something.
Their silence is aggravating.

Is this in a server environment?

Yousuf Khan
 
Is this in a server environment?

What's the difference? I've no experience with this particular one myself
- I have a nForce4 mbrd in our server but I stuck with my hot-swappable
Promise-based mirror for that. I believe the reports are on desktops
though judging by the people posting on the forum.
 
X-bit labs - Hardware news - VIA Makes Another Attempt to Re-Enter
Chipset Market.
http://www.xbitlabs.com/news/chipsets/display/20051122145105.html

VIA has lost a lot of ground to Nvidia and ATI in the chipset market.
It's probably not surprising, given their poor reputation for quality.
Known more for their compatibility glitches, I think they're even behind
SIS nowadays in marketshare. So they've upgraded K8T890 to K8T900.


If interested in more details:

http://www.techreport.com/reviews/2005q4/via-k8t900/index.x?pg=1

NNN
 
George said:
What's the difference? I've no experience with this particular one myself
- I have a nForce4 mbrd in our server but I stuck with my hot-swappable
Promise-based mirror for that. I believe the reports are on desktops
though judging by the people posting on the forum.

Well, usually server environments stress out disk environments much
more, so I would imagine the chances of seeing this problem are higher
in the server side. So what you're saying is that this happens mainly
when using the Nforce internal RAID configuration?

I wonder how this is going to affect Sun Microsystems? They're using
Nforce in their newly announced Galaxy servers. Mind you, Sun has its
own software RAID drivers for Solaris, so it isn't going to be affected
as long as you use the Solaris-included Logical Volume Manager software.

Yousuf Khan
 
Sorry all for the double post, I was using a system that had a funky
internet connection. I did not know you could double post's though
google, anyway sorry all.

Yes I know I am taking up more bandwidth, and maybe should of just let
it fly, but I thought a little reason why it happened would clear
things up for those who do care about this type of thing.

Gnu_Raiz
 
Well, usually server environments stress out disk environments much
more, so I would imagine the chances of seeing this problem are higher
in the server side. So what you're saying is that this happens mainly
when using the Nforce internal RAID configuration?

I wonder how this is going to affect Sun Microsystems? They're using
Nforce in their newly announced Galaxy servers. Mind you, Sun has its
own software RAID drivers for Solaris, so it isn't going to be affected
as long as you use the Solaris-included Logical Volume Manager software.

Well hey, maybe it'll get fixed - I'd imagine that nVidia listens directly
to Sun.

The network bug, while easy to get, around bothers me: in particular the
fact that nVidia makes such extravagant claims about their network
performance, which is bound to take a serious hit if checksum offload is
disabled.
 
George said:
Well hey, maybe it'll get fixed - I'd imagine that nVidia listens directly
to Sun.

The network bug, while easy to get, around bothers me: in particular the
fact that nVidia makes such extravagant claims about their network
performance, which is bound to take a serious hit if checksum offload is
disabled.

Well, this article seems to indicate that there were some problems with
Nvidia chipsets in the early revisions:

"In October 2004, AMD made the transition to PCI Express technology with
NVIDIA's newest chipset, the nForce 4. Once again, the platform got a
rocky start, with retail boards only hitting the channels some time
after the launch. Also, several board revisions were necessary before
the motherboard makers finally got a grip on the chipset. However, today
the nForce 4 is considered the most stable AMD desktop platform, a view
that was supported by the lack of problems during our live stress test."
http://www.tomshardware.com/cpu/20051121/the_mother_of_all_cpu_charts-12.html

Of course, the "live stress test" that Tom is referring to is the one in
which the Intel system suffered several lock up and reboots, and the AMD
system suffered none (other than the occasional Tom tripping over its
power cord, oops). :-)

Yousuf Khan
 
Well, this article seems to indicate that there were some problems with
Nvidia chipsets in the early revisions:

"In October 2004, AMD made the transition to PCI Express technology with
NVIDIA's newest chipset, the nForce 4. Once again, the platform got a
rocky start, with retail boards only hitting the channels some time
after the launch. Also, several board revisions were necessary before
the motherboard makers finally got a grip on the chipset. However, today
the nForce 4 is considered the most stable AMD desktop platform, a view
that was supported by the lack of problems during our live stress test."
http://www.tomshardware.com/cpu/20051121/the_mother_of_all_cpu_charts-12.html

Of course, the "live stress test" that Tom is referring to is the one in
which the Intel system suffered several lock up and reboots, and the AMD
system suffered none (other than the occasional Tom tripping over its
power cord, oops). :-)

A "live stress test" does not have to deal with an irate boss who cannot
send his urgent e-mails.:-) The network bug is till present and has been
around since the nForce3 with no comment from nVidia or any of their OEMs
that I know of. According to Eudora support it even predates nForce and is
present in 3COM Gb chips; IIRC nVidia used a 3COM design, either through
emloyees or sub-contracting.

It's easy to verify - just upload HDTach results.
 
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

I use Nvidia GPU's but I would think twice about getting a Nvidia based
Motherboard to use in Linux.

Why? I have a couple of nForce2 boards running Linux, one in a MythTV box
at home and another in a desktop box at work (that dual-boots Win2K and
Gentoo). For video, the home box uses a GeforceFX 5200 and the work box
uses a Geforce4 MX440.
I would probably opt for VIA.

I've never had much reason to complain about VIA, though the 3D part of the
integrated video in my old notebook (which used a KM133, IIRC) sucked colon.
One thing is for sure I would not get an ATI based machine.

My new notebook is an HP Special Edition L2000. It uses the ATI Radeon
Xpress 200M chipset. Getting it working with Linux was mostly painless.
(wpa_supplicant doesn't want to work right with the built-in (Broadcom)
wireless NIC, and USB support needed to be compiled as modules for
high-speed devices to connect at the correct speed. I've not tried using
the modem yet, and most likely won't bother as I have a controller-based PC
Card modem I can use.)

_/_
/ v \ Scott Alfter (remove the obvious to send mail)
(IIGS( http://alfter.us/ Top-posting!
\_^_/ rm -rf /bin/laden >What's the most annoying thing on Usenet?

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFDjJyxVgTKos01OwkRAlzzAJ4wnZLo9rhpdI+HocoKqcroWrmVOwCfYrr3
LuAd8HNA+1Oq9ryJykJsUp0=
=gbjG
-----END PGP SIGNATURE-----
 
Back
Top