Did someone try any dual P4 (socket PGA603) motherboard with QNX6.1? I’m
looking for any ATX format motherboard that supports QNX 6.1 without the
problems.
Adam
Hi Adam,
I have installed QNX 6.2.0 on “Super P4DP6” motherboard (SuperMicro
manufacturer) with two P4 Xeon 2.4 GHz processors and 2 GB RAM. This
motherboard has also two 100 Mb LAN ports or two 1Gb LAN ports in the
“Super P4DP6-G2” model. Installation was very easy and I didn’t have any
problem with it - QNX recognized all hardware devices: processors, both LAN
network cards, graphics card. Moreover I used SMP Kernel to take of dual
processors advantages.
Regards,
Jacek
U¿ytkownik “Adam” <adam.zembala@opal-rt.com> napisa³ w wiadomo¶ci
news:ao6gfj$60m$1@inn.qnx.com…
Did someone try any dual P4 (socket PGA603) motherboard with QNX6.1? I’m
looking for any ATX format motherboard that supports QNX 6.1 without the
problems.Adam
Boy, just go ahead and make me drool why don’t ya. I’d give up two or three
of my girlfriends for a board like that.
(Of course, you could argue that if I gave up a few girlfriends, I could
probably afford a board like that!)
“Jacek Rudnicki” <jacek.rudnicki@quantum.com.pl> wrote in message
news:asnlm8$kj5$1@inn.qnx.com…
I have installed QNX 6.2.0 on “Super P4DP6” motherboard (SuperMicro
manufacturer) with two P4 Xeon 2.4 GHz processors and 2 GB RAM. This
motherboard has also two 100 Mb LAN ports or two 1Gb LAN ports in the
“Super P4DP6-G2” model. . . . .
Jacek Rudnicki wrote:
I have installed QNX 6.2.0 on “Super P4DP6” motherboard (SuperMicro
manufacturer) with two P4 Xeon 2.4 GHz processors and 2 GB RAM. This
motherboard has also two 100 Mb LAN ports or two 1Gb LAN ports in the
“Super P4DP6-G2” model. Installation was very easy and I didn’t have any
problem with it - QNX recognized all hardware devices: processors, both LAN
network cards, graphics card. Moreover I used SMP Kernel to take of dual
processors advantages.
Do you know whether the released SMP kernel also take advantage of
hyperthreading, and if so, how you can tell whether it’s working and
what benefit it’s providing?
lew
Do you know whether the released SMP kernel also take advantage of
hyperthreading, and if so, how you can tell whether it’s working and
what benefit it’s providing?
For all intensive purposes an Intel CPU running in hyperthreading mode
looks like N physical CPUs to the OS. The currently released SMP kernel
will work just fine on them.
chris
\
Chris McKillop <cdm@qnx.com> “The faster I go, the behinder I get.”
Software Engineer, QSSL – Lewis Carroll –
http://qnx.wox.org/
Chris McKillop wrote:
For all intensive purposes an Intel CPU running in hyperthreading mode
looks like N physical CPUs to the OS. The currently released SMP kernel
will work just fine on them.
So a dual Xeon system would appear to have four CPUs?
Thanks,
lew
Yes - that’s right.
Dual Xeon system is visible as four CPUs.
Jacek
Uzytkownik “Lewis Donzis” <lew@nospam.donzis.com> napisal w wiadomosci
news:3DF34D3E.7F57C269@nospam.donzis.com…
Chris McKillop wrote:
For all intensive purposes an Intel CPU running in hyperthreading mode
looks like N physical CPUs to the OS. The currently released SMP kernel
will work just fine on them.So a dual Xeon system would appear to have four CPUs?
Thanks,
lew
Chris McKillop wrote:
For all intensive purposes an Intel CPU running in hyperthreading mode
looks like N physical CPUs to the OS. The currently released SMP kernel
will work just fine on them.
Isn’t it actually N processors (i.e. it doesn’t just appear to be N
processors, it actually is N processors) ?
My understanding is that “Hyperthreading” is a marketing concept to allow
Intel to charge 2 different prices for SMP systems.
btw: It sure looks to me as if in 2 years or so, everyone who buys a new
x86 system will be buying an SMP system.
Rennie
Rennie Allen wrote:
Isn’t it actually N processors (i.e. it doesn’t just appear to be N
processors, it actually is N processors) ?
Not really. It’s a way of maintaining two contexts inside the CPU so as
to better utilize the available resources. So at present, N is always
two.
The performance improvement is not as good as two separate CPU chips.
I’ve heard estimates in the range of 30%, but obviously, it depends a
lot on the specific application.
Intel has some excellent articles explaining HyperThreading on their
website.
lew
Also a good article at Ars Technica on hyperthreading:
http://arstechnica.com/paedia/h/hyperthreading/hyperthreading-1.html
cheers,
Kris
“Lewis Donzis” <lew@nospam.donzis.com> wrote in message
news:3E050CC6.9947344E@nospam.donzis.com…
Rennie Allen wrote:
Isn’t it actually N processors (i.e. it doesn’t just appear to be N
processors, it actually is N processors) ?Not really. It’s a way of maintaining two contexts inside the CPU so as
to better utilize the available resources. So at present, N is always
two.The performance improvement is not as good as two separate CPU chips.
I’ve heard estimates in the range of 30%, but obviously, it depends a
lot on the specific application.Intel has some excellent articles explaining HyperThreading on their
website.lew
Rennie Allen wrote:
Chris McKillop wrote:
For all intensive purposes an Intel CPU running in hyperthreading mode
looks like N physical CPUs to the OS. The currently released SMP kernel
will work just fine on them.
Isn’t it actually N processors (i.e. it doesn’t just appear to be N
processors, it actually is N processors) ?My understanding is that “Hyperthreading” is a marketing concept to allow
Intel to charge 2 different prices for SMP systems.btw: It sure looks to me as if in 2 years or so, everyone who buys a new
x86 system will be buying an SMP system.Rennie
My understanding is that the P4 has had hyperthreading from the
beginning - and that Intel’stest showed that it worked pretty well from
the beginning. Seeing it now is mostly a marketing thing (and probably a
lot more testing).
Dave
Lewis Donzis wrote:
Rennie Allen wrote:
Isn’t it actually N processors (i.e. it doesn’t just appear to be N
processors, it actually is N processors) ?
Not really. It’s a way of maintaining two contexts inside the CPU so as
to better utilize the available resources.
…
The performance improvement is not as good as two separate CPU chips.
I’ve heard estimates in the range of 30%, but obviously, it depends a
lot on the specific application.
It’s an issue for QNX, because the “hyperthreads” contend
for CPU resources, including the cache, register file, and
FPU. If QNX treats a hyperthreaded CPU as an SMP, there’s
the possibility that a low-priority process can steal
CPU resources from a high-priority process and slow down
the high-priority process.
John Nagle
Animats
David Hawley wrote:
My understanding is that the P4 has had hyperthreading from the
beginning - and that Intel’stest showed that it worked pretty well from
the beginning. Seeing it now is mostly a marketing thing (and probably a
lot more testing).
They way I heard it is that P4’s always had the capability, but it was
disabled. So it may have been a marketing decision to disable it
(adding value to the Xeon), but the feature is currently not available
in P4’s slower than 3GHz.
lew