#3DS &Pentium
8 messages in this thread
J. Ruikis,
What (if anything) am I to make of the "Pentium Problem" vis-a-vis 3DS?
I have just sent off a check to a VAR for Release 4 and a P-60. I've been
advised by said salesman that any advanced math that 3DS may call for the
processor to perform out to 7 or 8 decimal places may (or may not) cause a
glitch – possibly hidden, possibly not. At worst, it would be a rare and
intermitant problem, and not necessarialy the Pentium's fault. I realize that
this is early in everyone's awareness of the situation, but any and all
information/advice would be much appreciated.
BTW: Q. How many Pentium engineers to screw in a light bulb? A: 1.8322357454,
but Intel says that 2 will do fine. :>
Thanks,
Dave Knight
Here are some recent quotes from Gary regarding 3DS and then I've included a
statement
regarding ACAD…..
———————————-
The Pentium floating point bug does NOT affect 3DS at all, and I don't even
think it's likely that
anyone would ever run into it with AutoCAD within their lifetimes. Don't
worry about it and get a
P90.
– G
The Pentium glitch should only show up in double-precision floating point
operations. Since 3DS
only does single-precision ops, it shouldn't affect you in the slightest.
– G
———————————
Autodesk, Inc. 12/07/1994
AutoCAD and INTEL's Pentium Floating Point Flaw
Intel has been shipping Pentium chips for almost two years. The majority of
Autodesk CAD
products use double-precision floating point math, including all AutoCAD
releases such as
Release 12, AutoCAD LT, AutoSketch Release 2.0, and all products that run on
top of AutoCAD
like AutoCAD Designer, AutoSurf, AutoVision, and AutoCAD Data Extension.
To date Autodesk is not aware of any customer having a problem due to the
reported floating
point unit flaw in the Pentium chip. And based on the information provided to
us so far by Intel,
we believe it is unlikely that any Autodesk customer will be affected by it.
Although we believe this situation to be receiving more attention than it
deserves, we will
continue to monitor it, and we will take action as necessary to help alleviate
this problem for our
customers.
Cary Fulbright
Strategic Marketing CAD
Autodesk, Inc.
>> The Pentium floating point bug does NOT affect 3DS at all, and I don't even
think it's likely that
anyone would ever run into it with AutoCAD within their lifetimes. Don't
worry about it and get a
P90.
– G >>
Gary,
Just curious, but Stuart Ferguson (on LW prog team) has said on the INTERNET
that Lightwave Windows/NT will most likely be affected by the PENTIUM error
'cause they do dblprec MATH.
Just wondering the Plus/Minus of DblPrec MATH and why they (LW) use it and you
guys don't? And if v5 is Windows will this make a difference for your guys at
YOST?
Thanks,
David.
We don't use double-precision math for two reasons: 1) it forces your models
to take up TWICE the RAM, and 2) it's slower.
– G
Hey, I asked that question on the Internet. 🙂 It's entirely possible that
LW is using doubles for rotation an matrix calculations. Seeing that the models
clearly only use single-precision for points, like 3DS, I doubt they're using
doubles to re-calc points. And this is the 3D Studio forum – how do you expect
the Yost Group to make assessments of the internals of another program?
>Just curious, but Stuart Ferguson (on LW prog team) has said on the
>INTERNET that Lightwave Windows/NT will most likely be affected by the
>PENTIUM error 'cause they do dblprec MATH.
In case you didn't see his correction, Stuart made a typo in his message,
he meant to say it `wouldn't affect it'. If you read the rest of the
message you will notice that that is what he meant.
Btw, he was only making reference to the modeler and not the renderer.
<<< Tony D. Tracer >>>
(Any render time >0 needs acceleration.)
Thanks Anthony,
My Bluewave reader is notorious for hammering messages. I checked again between
the lines and your right on…….
David.
We offer network rendering services to clients using up to eight (8)
Pentiums, 60,66, & 90's (and DX2's) – and have not encountered any
problems attributable to 'math' errors. Over just the last year, this
adds up to thousands and thousands of renderings. Some jobs are 5,000
to 9,000 frames each (compounded by the fact we field render
everything).
In the old days, back when we used the Weitek NDP chip, we'd load
alternate 3DS.EXE and I.P.A.S. routines (optimized for the Weitek's
superior performance). We'd sometimes (not always) choke when pushed
the Weitek past its rated clock speed (putting a 33Mhz Weitek into a
50Mhz DX, etc.) We discovered that we could intermittently 'hang' the
renderer if Phong shading was invoked. The fix was obviously not to
push process the Weitek. But even this wasn't a legit' 3D Studio/NDP
problem.