rec.autos.simulators

I was wondering if someone can confirm this for me. Is it true that PCI
streaming / latency is disabled in all the new (and supposedly better)
HX/VX chipsets, whilst the old FX has these options. Did Intel leave these
options out of the chipset register?

I have a 3D Blaster PCI and can only get 39.0 for DMAtest. Many other
people are getting 78+. And it makes a HUGE difference to framerates in
Nascar2 etc.

And on a related topic, how come DMA bus mastering is enabled for the video
card in Win 95 but not in dos. What is the BIOS doing to disable it? I can
get around this using vtool but I am still curious.

Is there any way around the PCI streaming / latency issue or is it going to
remain a fact that old FX motherboards are going to be superior for gamers
who use a Rendition Verite based 3D accelerator?

My system:
p166
Tyan Tomcat 1 Rev3 Tomcat Bios 2.22 Award 4.51PG Intel HX chipset
32 Megs EDO
3D Blaster PCI
SB 32
WD 2.5 GB HD

Thanks in advance

Anton Voss

I have quite new QDI 51430HX-T2 and there is option for PCI latency.
It carries AMIBIOS but I know that another board (same model) wich
carries AWARD bios do not have those options. So I think it's upon to
motherboard-  and  bios manufacturer not HX/VX chipset.

: >I was wondering if someone can confirm this for me. Is it true that PCI
: >streaming / latency is disabled in all the new (and supposedly better)
: >HX/VX chipsets, whilst the old FX has these options. Did Intel leave these
: >options out of the chipset register?
: >
: I have quite new QDI 51430HX-T2 and there is option for PCI latency.
: It carries AMIBIOS but I know that another board (same model) wich
: carries AWARD bios do not have those options. So I think it's upon to
: motherboard-  and  bios manufacturer not HX/VX chipset.

Having looked at the Docs for the 430HX chipset, PCI latency seems to be
handled completely different than in the FX series.  I cannot find a
similar register in the HX set for setting the PCI latency timer that
matches the FX set. However, I think it is something that Award seems
to have left out of their BIOS.  I see no mention of a PCI Latency timer
in any of the BIOS docs at Award's web site.  I too have heard of other
HX boards having this option in their BIOS.  I don't think it is
necessarily "disabled", but usually set to a low rate.  I have sent
e-mail to Tyan about this issue and have yet to receive an answer.  And
yes I did send it to "tech support" not the web master...  (It seems
that Jerry Lee is the only person at Tyan to answer my e-mail.)

--
**************************** Michael E. Carver *************************
     Upside out, or inside down...False alarm the only game in town.

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=<[ /./.  [-  < ]>=-=-=-=-=-=-=-=-=-=-=-=-=-=

I don't believe that PCI streaming is disabled on "all" HX/VX boards.
On my ASUS P55T2P4 (HX board) there is an option to enable/disable
PCI streaming in the BIOS setup.  I am pretty sure it has an Award
BIOS, but I don't have the machine in front of me, and I haven't played
with the BIOS for awhile.

A number of people have stated that the Tomcat boards may not be the
best choice to pair with a Rendition video board due to the Tomcat's
lower DMA numbers.

I still feel that the HX chipset is the preferred Pentium chipset for
***, due to its ability to run the higher bus speeds (75/83 MHz).  
With my ASUS bus running at 75 MHZ the system is rock solid in both DOS
and Win95, and it seems to be a good match with the Reactor.  I am
very pleased with this combination, and it was a *lot* cheaper than
buying a PPRO 200!

Of course, this could all change as soon as boards capable of pairing
SDRAM and higher bus speeds become available.  ;-)

--Marcus

Just in case you wanted to know, I just did a quick comparison of my
ABIT VX board with 32megs SDRAM to Mike Schreiner's ASUS HX with 64megs
EDO both with P200MMX CPUs.  Here's a quick run down on what we both
got at 200mhz (66x3) using WinBench97 and Chris Dial's 3D Bench:

             Mine      His
CPUMark16     442      409
CPUmark32     427      389
SVGA(fps)    41.0     40.7

I usually run my system at 225mhz (75x3) which gives me these scores:

           P225MMX
CPUMark16    470
CPUmark32    458
SVGA(fps)    46.2

--

The IWCCCARS Project: Q & A Representative
http://www.racesimcentral.net/
http://www.racesimcentral.net/~ebusch/

If I'm not mistaken, the Asus P54t2p4 motherboard uses the HX chipset.
The award bios I have has a PCI latency adjustment.

Jeff

Nominal fifo = 22.2 mb/sec

I have the bus speed at 68 mhz and the latency timer at 150.

Barry Clay

: Having looked at the Docs for the 430HX chipset, PCI latency seems to be
: handled completely different than in the FX series.  I cannot find a
: similar register in the HX set for setting the PCI latency timer that
: matches the FX set. However, I think it is something that Award seems
: to have left out of their BIOS.  I see no mention of a PCI Latency timer
: in any of the BIOS docs at Award's web site.  I too have heard of other
: HX boards having this option in their BIOS.  I don't think it is
: necessarily "disabled", but usually set to a low rate.  I have sent
: e-mail to Tyan about this issue and have yet to receive an answer.  And
: yes I did send it to "tech support" not the web master...  (It seems
: that Jerry Lee is the only person at Tyan to answer my e-mail.)

Tyan, I am still waiting.  Have not received a reply to my questions sent
to Tech Support via web site.  Why do we not have control of the PCI
Latency timer in our Award BIOS'es?  Can we expect to get such in the
next upgrade?  This question has been asked before and I have never seen
a response (public or private).

--
**************************** Michael E. Carver *************************
     Upside out, or inside down...False alarm the only game in town.

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=<[ /./.  [-  < ]>=-=-=-=-=-=-=-=-=-=-=-=-=-=

Barry Clay

I have the Asus PT55T2P4 with 32 megs edo ram, Intel P133 with the
Intergraph
Intense 3D 100 video card. The book says to set the pci latency timer at
32.
With that  setting I got a reading of nominal dma performance= 67.3 mb/sec.
I bumped the pci timer up to 54 and got a reading of 89.8 mb/sec

PCI Latency is on the individual card, in that cards PCI configuration
space, that is part of every PCI plug in card.  It is NOT part of the chip
set.

Some BIOSs let you set this number, some default to 'some' number.  This
register is only used for bus masters.  Usually only disk controllers and
network adapters provide the bus master function.  Video cards may or may
not.

The PCI revision 2.1 spec says 16 should be the maximum number.  What PCI
latency is, is for how many data cycles a bus master can keep the bus once
it has been granted permission to use the bus.  If you set it to high,
other cards may not get the bus in time, and may loose data.  Honestly, I
wouldn't believe there would be a problem setting it to 32, but I wouldn't
set it to much higher than that.

Austin Franklin

It really depends what you use your system for.  If you are using DOS or
Win 3.1 or even Win 95, and you have no real time requirements, you will
probably not get any effect from what ever number you put in it.  If you
are running a server, or running NT, you may get some slow downs in certain
places.

There is no DMA per se in PCI land.  I don't know what the test you are
refering to does.  Also, be aware that these tests are not any reflection
of any real world situations.  They are dedicated for that one and only
function, and do not show any interactions with other parts of the system
that may be affected by other changes.

If your BIOS set that to 66, then it determined that the requirements of
your system were pretty small, and the large number would not cause any
problems.  You're proably safe with what you did.  If you had a network
card, graphics card, two disk controllers, etc.  then it probably would
have set the number much lower.  It's only an 8 bit regiseter, which means
it can be as large as 255 decimal.

By the way, I believe in the earlier post I said the latency timer was in
units of data phases, it's not, it's in units of PCI bus clocks.

Austin Franklin