#3DS benchmark
93 messages in this thread
Trying to find a 3DS-R4 rendering benchmark. Would like to know for examples
how fast a .3DS file like the CHEVY will render on different computers. This is
for evaluating my own machines performance. Thanks,
-Reza
Reza,
>> ..how fast a 3DS file like the CHEVY will render…<<
I did that test with someone a while back. At 1024 x 768 x 1 with the whole
list set to on (Anti-Aliasing, etc.) my Pentium 100 came in at 3:18. That's a
dual but I don't think it matters with R4. I'm also quoting that time from
memory, so if that doesn't sound right, let me know and I'll do it again.
Terry
Terry: Thank you for your reply and the offer to run that CHEVY.3DS test again.
I would appreciate another reading on your render time on your P100 with
CHEVY.3DS. I have recently upgraded my machine with a no-name P100 mother board
and I suspect that I may bought a white elephant! Sincerely,
-Reza
Reza,
With settings at default 1024 x 768 x 1, everything down to and including
Autoreflect on. Rendering to Disk and motion blur shows on during the render,
but I think that has nothing to do with image time. Render time was 3 minutes
12 seconds, and total time until the 21" screen was filled was 4 minutes 7
seconds. We each have P100 boards but may have ram and cache differences. I
have 128 MB RAM with 512K cache. I am also running Liquid Speed, but I assume
everyone with 3DS does also. What kind of times did you turn and how much RAM?
From what I understand, that's where the speed and power is. I'd like to know
how they match up if you have a different amount of RAM. I am debating right
now whether to get a second dual, or just chock this one full of RAM with an
accelerator card when MAX arrives.
Hope this is helpful, -Terry
Do you just want the time it takes to renders the chevy at 1024?
Let me know I have a P120, and I can send back the time
matt 🙂
-Mike Allen
Yes Mike, if you could please render the CHEVY.3DS at 640×480 with all defaults
and no DISPLAY and post your result. Thanks,
-Reza
>CHEVY.3DS at 640×480 with all defaults and no DISPLAY and post your result.<
For your records,
P100, 32MB 512k cache Intergraph TD-30 (neptune chipset) 640×480, no display,
no save. – 1:35.
JNJ
everybody,
you should all make sure you post your pixel size (typically 1.1 to 1.5)
because this really has an effect on speed.
Martin,
Good catch! I checked, and Pixel Size is NOT one of the settings saved with 3DS
files.
Since we're trying to come up with a "standard," and 3DS.SET defaults to a
Pixel Size of 1.2, that setting sounds good to me for these testing purposes.
The more numbers we get, the better! I'm personally interested in results from
those with accellerator boards — That Sept. issue of BYTE wherein they
compared an Intergraph machine with an Omnicomp 3Demon and a Millenium board
piqued my interest!
SCL
Stephen,
>> a Pixel Size of 1.2 <<
I suggest keeping it at 1.1 for the tests, since that is the default anyway.
Mainly I just wanted people to not pixel-size because this has a large impact
on rendering time.
Also, if I want to test processor against processor, I turn off mapping
completely so that you eliminate disk sub-system variations. Instead I raise
the resolution to get a good number to compare. For example: chevy.3ds,
1024×768, everything ON except Mapping OFF, pixel-size 1.1, no display, no save
to disk.
Oops; guess it's been so long since I saw a "virgin" 3DS.SET that my memory's
slipping… <g>
thanx for your input; if we can get large number of responses, the results
should be interesting…
I did the 2 tests on my generic brand Intel chipset m/board Pentium 90 w/32mb
ram:
1024×768 all on EXCEPT mapping, pixel 1.1, no display, no save — 2:33
640×480 all on, pixel 1.1, no display, no save — 1:48 (faster than a
comparable Micronics based system I had by 6 seconds).
It's also interesting to note that you can shave a few seconds off by using no
memory manager at all – not a good idea if you have more than 64mb RAM.
I have a Pentium 100 at work I'll try the test on.
Martin,
The pixel size for my post was 1.1.
JNJ
Reza,
>CHEVY.3DS at 640×480 with all defaults and no DISPLAY and post your
result.<
I tried the test on my 90mhz Pentium w/32mb ram, it's one of the early Intel
mobos (July 94) with the Plato chipset. The test ran 1:50 (110 sec).
Matt (Active Media Graphics)
Y 50 /g fM/
Chevy.3ds at 640×480, all defaults, 1.1 pixel size. 2:30 on a gateway 2000
P-60 with 24 megs of Ram, Diamond Viper PCI.
1023×768 with maps off, 3:39.
Hope this helps,
Jason
Just adding to your chevy.3ds rendering benchmarks:
P120
64 Mb Ram
ATI Graphics Pro Turbo 4 Mb PCI
640 x 480 renders in 1:33 on my machine.
1024 x 768 renders in 2:45
All rendering settings set at default.
>P120
>64 Mb Ram
>ATI Graphics Pro Turbo 4 Mb PCI
>640 x 480 renders in 1:33 on my machine.
>1024 x 768 renders in 2:45
My system: P100, 32MB RAM, #9GXE64Pro
640×480 – 1:37
1024X768 – 2:22
Just for the sake of comparison, I rendered CHEVY.3DS @ 640×480, all defaults
per the 3DS file, on my DX2/66 w/32 megs. It took 5:16 — gotta get a new
box!!!!!
Steve: Yes, you should expect to get about 370% to 400% rendering time
improvement with a P100. Make sure you get the Intel Triton chipset on it and
you can even salvage your 32meg RAM chips. However, the new EDO RAM should up
the performance even more. Thanks for your reply.
-Reza
Yes, I'm well aware of the speed improvements, having worked at places with
Pentiums. Actually, I'm just waiting for the P150 to debut to make my move.
But I'm also elaborating on your idea and trying to get a standard 3DS
benchmark established so anyone can compare their rendering speed with what's
been posted. Chevy.3DS is a good file to use. According to my memory-used line,
it looks like it'll BARELY fit in 16 megs without creating a swap file at
640×480.
IMHO, it's a good idea to render it exactly as the file comes with 3DS (metal
shading, etc.) so we're all doing exactly the same thing. BTW, I think your
renderings are only defaulting to 1024 because that's how you have your display
set in VIB.CFG, not because Chevy.3DS is set that way.
<<<Just for the sake of comparison, I rendered CHEVY.3DS @ 640×480, all
defaults per the 3DS file,>>>
Did the same with my 486DX4-100Mhz w/20Mb ram. time 3:58.
So speed isn't everything, Pentium vs. 486 shows a differance in architecture,
and that counts as much.
Ram is only relevent when you start swapping, which I didn't for this test (
mem used 15mb.)
Bob Kohl
Gidille inc.
-Bob Kohl
Chevy.3ds, 640×480, Null output, Pixel size 1.1, Camera Viewport
Micronics P90 Motherboard, 64 meg RAM
Render Time 1:50
Chevy.3ds, 640×480, Null output, Pixel size 1.1, Camera Viewport
ASUS P120 Motherboard, 32 meg RAM
Render Time 1:31 (91 seconds) for what it's worth.
Murph'
Reza,
Here is my results for the rendering benchmarks:
Pixel Size:1.1 Display:Null Save File:No
Mapping-Off Mapping-On
640x480x1.0 0:59 1:23
1024x768x1.0 1:58 2:50
SuperMicro P55CMS motherboard with 256k pipelined busrt cache
and 32mb ram.
Regards………….Rick – Technical Animations
Rick,
>> SuperMicro P55CMS motherboard with 256k pipelined busrt cache <<
I assume that's a Pentium 133, isn't it? Impressive results. Does it have EDO
ram?
Martin,
>>I assume that's a Pentium 133, isn't it? Impressive results. Does it have EDO
ram?<<
I went back and rendered again to verify my results and got the same number
within 1 second.
Thats a Pentium 100 with 60ns ram and ___no___ EDO ram. Maybe there is
something different about my settings. When I rendered, with mapping ON the
settings in the dialog box were:
Phong | View: Camera03
On | Video Color Check: Off
On | Gama: On
On | Fields: Off
On* |
On |
Off |
Off |
Show |
Tile |
* With mapping turned off this setting would change to OFF.
Here are my results again for the rendering benchmarks:
Pixel Size:1.1 Display:Null Save File:No Shading:Phong
Mapping-Off Mapping-On
640x480x1.0 0:59 1:24 (+1)
1024x768x1.0 1:58 2:49 (-1)
( ) Indicates a change from my first message on rendering times.
Regards………Rick – Technical Animations
Rick,
I think that you should have Metal shading on. Those results are out-of-line
with what one would expect from a Pentium 100.
Martin,
I went back and copied the original 3ds4 set file and tried again with no
change.
When I click on render, I get a message that says "WARNING: Rendering to NULL
without disk output. Are you sure this is what you want?" Is this what everyone
else is getting?
Just to make sure, the file that I'm rendering is "Chevy.3ds – 936,668kb –
1/15/95 @ 12:00am
Any ideas what could make the difference, surely they didn't install a P120 by
mistake, Nah… couldn't happen<BG>
Regards……….Rick – Technical Animations
Rick,
>> Any ideas what could make the difference, surely they didn't install a P120
by mistake, Nah… couldn't happen<BG> <<
must be because I can't explain it <g>. Somebody should come over to your
studio and make sure you're not feeding your cpu steroids <G>.
Rick,
The error when you select no display with without disk output is normal.
The reason that this is used (I think) for these benchmarks is to take the
video card out of the loop. Your system is pretty quick, but try with shading
to metal, changing from phong to metal increases my render times by four
seconds.
-Matt (Active Media Graphics)
Y 50 /g fM/
Matthew,
>>The error when you select no display with without disk output is normal. The
reason that this is used (I think) for these benchmarks is to take the video
card out of the loop.<<
I was trying to confirm that the settings I was using were identical to the
other person. You are right though, it does not make any sense to render with
no display and not save the file<BG>.
>>Your system is pretty quick, but try with shading to metal, changing from
phong to metal increases my render times by four seconds.<<
It seemed to add 4 seconds to the 640×480 images and 8-9 seconds to the
1024×768 images when I went from Phong to Metal.
Regards………..Rick – Technical Animations
-Chevy specs-
Intergraph TD-30, Pentium 100, Neptune chipset,512k cache, 32MB
Camera3- 3D Editor, 1.1 pixel, metal shading, etc.
No Mapping Mapping
640×480 1:08 1:33
1024×768 2:16 3:09
BTW- I don't believe Phong is the default setting, rendering it with Phong
yielded a 11 second improvement over Metal at 1024 here.
JNJ
John,
>>BTW- I don't believe Phong is the default setting, rendering it with Phong
yielded a 11 second improvement over Metal at 1024 here.<<
The difference between Phong and Metal when I render at 1024 with mapping on is
8 seconds.
Regards…………Rick – Technical Animations
Thanks for your most-impressive numbers! Hate to be a pest, but when you've got
time could you run it one more time, with METAL shading? On my machine, that
only adds a few seconds — I wonder what about your machine is making it so
fast — if you've been looking at the other numbers, I'm sure you've seen that
everybody else is running significantly slower — Couldn't be ALL due to that
60-ns RAM, do you think?
Stephen,
I also have 60ns RAM and I can't turn the numbers Rick is churning out over
there. The 640 numbers only changed by 4 seconds from Phong (1:05) to Metal
(1:09) on my system. I wonder if we are using the same pixel sixe, 1.1?
Terry
Terry,
Yes, I'm using the default pixel size of 1.1. There must be something different
about my settings, but I have not figured it out yet. Are you looking at the
numbers with or without mapping? When I turned on Metal shading and rendered
640x480x1 w/o mapping the time went from 0.59 to 1:04. Any ideas?
Regards…………….Rick – Technical Animations
Rick,
>> Any ideas? <<
maybe you hid something <g>.
seriously though, has the file been modified since being installed? The file
date on it (r4c1) should be 1/15/95.
Martin,
>>maybe you hid something <g>.<<
I swear, I didn't hide anything, except the car.<BG>. Once I even set the Hide
or Show button in the dialog box to Show, just to make sure.
>>seriously though, has the file been modified since being installed? The file
date on it (r4c1) should be 1/15/95.<<
Same date that I have, see my reply to your earlier message. Maybe Qemm 6.0 is
faster than 7.5….. who knows.
Regards………Rick – Technical Animations
Rick, I just saw you're using QEMM — did you see the exchange between Martin
and I where we were discussing a separate boot menu item bypassing QEMM (if you
don't have more than 64 megs)? I do that for big renders and gain about an
extra meg available to 3DS,
and knock as much as 10% off my rendering times. Just load a minimal config,
letting it all load low — No EMM386, no HIMEM.SYS; just so long as you've got
about 250K left of low memory, it should give you a bit of an improvement, too!
Steven, please keep in mind that using no memory manager does not *gaurantee* a
shortening of rendering times. It really depends on the machine. After doing
quite a bit of testing on different systems, I advise clients to perform render
benchmarks on 3DS machines with no MM, QEMM, and Himem to see which one their
machine seems to like best, and go with the fastest one. It varies greatly, and
there doesn't seem to be any obvious pattern.
Of course, if you need >64Mb RAM then all bets are off and QEMM is your only
choice.
-Alan
That's interesting… I've only tried it on 3 different machines, but it worked
on all three, so I'd concluded that it was a general rule-of-thumb.
Thanx for the info!
SCL
Stephen,
Thanks for the tip, when I get a chance I will give it a try.
Regards………..Rick – Technical Animations
Rick,
I don't know, maybe something in the options box. Dither 256 or true color,
16 or 24 bit depth? I don't really know enough to say. I tried different
settings with all these things and still could not get times under one
minute. Are you looking at the time at the bottom of the screen, or the last
frame time? The last frame time is always a few seconds faster than the
status line time. If it's none of that, then you are just obviously running
an illegal restrictor plate. <g>
What about Liquid Speed? Doesn't that adjust it's acceleration depending on
what driver you are using? Maybe you just have the very best combo there is.
Whatever it is, Salutes!
Terry
Terry,
>>Dither 256 or true color, 16 or 24 bit depth?<<
I'm sure its set to 24 bit and pretty sure that Dither 256 is on.
>>Are you looking at the time at the bottom of the screen, or the last frame
time? The last frame time is always a few seconds faster than the status line
time.<<
I'm rendering from the 3D Editor and using the times reported at the bottom of
the screen.
>>If it's none of that, then you are just obviously running an illegal
restrictor plate. <g> <<
Tell me where it's at so I can work on it and we will really boggie<BG>.
>>Whatever it is, Salutes!<<
Thanks, but I really owe it all to Mark at MIND COMPUTER CONSULTANTS, he is the
one that put it together.
Regards……………..Rick – Technical Animations
Stephen,
Here are the number you wanted with Metal Shading.
SuperMicro P100; P55CMS w/256k piplined burst cache; 32mb of 60ns ram
File: Chevy.3ds/936,668kb/1-15-95 @ 12:00am
Pixel Size: 1.1 Display: Null Save File: No Shading: Metal
Mapping-Off Mapping-On
640x480x1 1:03(+4) 1:28(+4)
1024x768x1 2:08(+9) 2:57(+8)
( ) denotes additional seconds for Metal shading vs. Phong shading
>>Couldn't be ALL due to that 60-ns RAM, do you think?<<
If it is, it's the best $40 I've ever spent!!
Regards………..Rick – Technical Animations
Rick,
I did some more tests on another machine at my office and your times aren't
that far out of line. I suspect your cache type is reponsible for a slight
4-10% speed increase over my system. I remember Gary Yost saying that that new
cache design could boost performance up to 10%. I think the RAM speed is less
significant. I've tried 70ns versus 80ns with no speed difference in 3ds.
Here's my times next to yours (yours are in parentheses):
Asus P100; P54C w/512k write-back cache; 196mb of 70ns ram
2 cpu's installed (no difference in DOS).
File: Chevy.3ds/936,668kb/1-15-95 @ 12:00am
Pixel Size: 1.1 Display: Null Save File: No Shading: Metal
Mapping-Off Mapping-On
640x480x1 (1:07)1:10 (1:32)1:40
1024x768x1 (2:17)2:22 (3:05)3:23
Interestingly, with mapping off your speeds are only 4-5% less, but with
mapping on they are 8-9% less. I have a fast disk sub-system (Seagate Barracuda
2.1gb Fast and Wide on a Adaptec 2940W PCI controller) but no disk cache
(hardware or software).
Are you using any disk caching software? (smartdrive) Hardware? (eg. on the
controller board). What type of disk sub-system do you have? Caches might help
on repeat renders of the same scene
As a guide to disk access and it's effects, my HD light shows my hardrive is
working for about 6 seconds when I render with mapping on. Course, it doesn't
go on at all with no mapping (no disk access at all).
Food for thought, eh?
Martin
I've noticed that some people are unsure of what the default settings are for
the chevy. So I was thinking that wouldn't it be possible for one of use to
make a chevy.vue (I think .vue is the file were you can render straight from
dos with specififc settings) file. It could the be uploaded and anyone could
download it and just render without having to worry about pixel size or
anything else.
Just a thought
Jason
Martin,
>>I suspect your cache type is reponsible for a slight 4-10% speed increase
over my system. I remember Gary Yost saying that that new cache design could
boost performance up to 10%.<<
Its good to know that some of the new technology is translating into real world
performance gains.
>>I think the RAM speed is less significant. I've tried 70ns versus 80ns with
no speed difference in 3ds.<<
It only cost me another $40 for the 60ns versus the 70ns, I didn't get hurt
anyway.
>>Interestingly, with mapping off your speeds are only 4-5% less, but with
mapping on they are 8-9% less. I have a fast disk sub-system (Seagate Barracuda
2.1gb Fast and Wide on a Adaptec 2940W PCI controller) but no disk cache
(hardware or software).<<
I have the same controller although I don't have the wide Quatum Atlas 2.1MB
SCSI drive, just fast. It seems as though my drive has a 1mb buffer, I guess
thats the same as a hardware cache.
>>Are you using any disk caching software? (smartdrive) Hardware? (eg. on the
controller board). What type of disk sub-system do you have? Caches might help
on repeat renders of the same scene<<
I disabled smartdrive and tested again. The only difference was an additional
2-3 seconds at 1024×768 with mapping ON, all the others were the same within 1
second. I have the Adaptec 2940w without any cache, unless Mark gave me
something we didn't discuss.
>>As a guide to disk access and it's effects, my HD light shows my hardrive is
working for about 6 seconds when I render with mapping on. <<
I tried to time my HD light with a stop watch and got around 4 to 4.5 seconds
with mapping on. This just does not make sense, you have a fast & wide SCSI
drive, it should be faster.
>>Course, it doesn't go on at all with no mapping (no disk access at all).<<
With no mapping my system still hits the HD for 1 second or less, at the very
end of "Transforming Objects".
You mis-read my numbers for METAL shading, the "( )" numbers should have been
subtracted from the adjacent number to get back to phong shading.
The following numbers in ( ) are for my METAL shading times:
SuperMicro P55CMS; P100; 256k Pipelined Burst Cache; 32mb-60ns Ram
File: Chevy.3ds/936,668kb/1-15-95 @ 12:00am
Pixel Size: 1.1 Display: Null Save File: No Shading: Metal
Dither 256: On
Mapping-Off Mapping-On
640x480x1 (1:03)1:10 (1:28)1:40
1024x768x1 (2:08)2:22 (2:57)3:23
Regards……….Rick – Technical Animations
Rick,
wow, we could get really deep into this analysis couldn't we? I'm pretty much
satisfied that you're getting most of your performance from the new cache
design. Something to consider when buying a new system – ask for the super
ruptured pipeline cache <g>.
the next step would be to start comparing memory managers (none, himem, qemm6,
qemm7.5 etc). Oh well, maybe another time I'm in crunch mode at work…..
Martin,
You are right, this stuff does not pay the bills<G>.
Regards……..Rick – Technical Animations
Martin,
My disk subsystem is a PCI Adaptec, with only a 5400RPM type SCSI drive
(connor). I'm suprised your killer Barracuda isn't reflected. Have you had
any experience with Atlas drives?
JNJ
JNJ,
>> I'm suprised your killer Barracuda isn't reflected. <<
I think they excel on larger files. Also it may need a defrag.
>> Have you had any experience with Atlas drives? <<
No. What are they?
Martin,
>I think they excel on larger files. Also it may need a defrag.<
yeah, Chevy isn't that disk map intensive, and those 7200 RPM drives don't do
well at jumping around the disk for small files. I'd gladly trade though. 🙂
>Atlas drives
Quantum makes them, they're similar to your 7200 Barracuda's- but a little
newer. They make a killer four gig with a 2MB cache.
JNJ
JNJ,
>> those 7200 RPM drives don't do well at jumping around the disk for small
files. <<
true, but they excel in swap-file situations. I recently saw this in action. I
have some rendering machines which have the same RAM and CPU, but some have
just a standard 500mb IDE drive, others have 2.1gb fast and wide Seagate
Barracuda's with an Adaptec 2940 wide controller. Yet another machine had 192mb
RAM and had no swap. Here are some frame times:
IDE drive, 128Mb RAM, 80Mb swap file 1 hr 15 mins
Scsi F/Wide, 128Mb RAM, 80Mb swap file 15 mins
Scsi F/Wide, 196Mb RAM, no swap file 7 mins
Kind of shows you, if you are going to get a swap file situation (and we all do
sooner or later) then it pays to have a fast disk sub-system. The above results
show me that the systems with the IDE become practically useless with even a
smallish swap file, while the SCSI fast/wide ones are still somewhat useful.
Martin:
Wow! I knew that I chose SCSI for a reason but I never thought that it would
be so much quicker. Thanks for posting those numbers.
– Bob Ritger
>> IDE drive, 128Mb RAM, 80Mb swap file 1 hr 15 mins
Scsi F/Wide, 128Mb RAM, 80Mb swap file 15 mins
Scsi F/Wide, 196Mb RAM, no swap file 7 mins <<
WOW! That sure proves that swapping "kills" and more ram expenditure is cost
effective in a production environment. Thanks for the benchmarks.
– ej
Ernie,
>> more ram expenditure is cost effective in a production environment <<
well, that's true up to a certain point. RAM is so darned expensive still that
it shows that it's worth spending a bit of a price premium for a decent disk
sub-system – I wish I had now, but fast/wide drives tend to be 2gb and up. The
standard IDE drive just doesn't cut it for any type of swapping.
Martin,
Your times are certainly a significant reason for going with a fast SCSI
subsystem. I noticed the advantage a few years ago when I upgraded my ~12ms
IDE to an 8.5ms 5400 IDE drive.. major improvement at the time. Now, I
certainly need it.
Here at work, We've got a Barracuda 2gb, and Fast (not Wide) system, and it's
well worth it. (but Wide would be better)
I'll be upgrading to one of those 4gb Atlas drives I mentioned real soon (on
my TD) Intergraph didn't bundle one on the package, and it's cheaper buying it
myself.
Do you know if the AV 7200 drives (like Micropolis') are better than a
Barracuda/Atlas type drive? I'm not sure if the t-cal stuff is worth it for
non AV work. (like page files, tex map loads, etc.)
It looks like it's time to upgrade that one IDE system. 🙂
BTW- are your meshes really big, or is most of the RAM needed for film rez tex
maps? (or both? <g>)
JNJ
JNJ,
>> Here at work, We've got a Barracuda 2gb, and Fast (not Wide) system, and
it's well worth it. (but Wide would be better) <<
It'd be interesting to see how much the wide drives compare to merely fast
Barracudas would perform in this swapping scenario.
>> Do you know if the AV 7200 drives (like Micropolis') are better than a
Barracuda/Atlas type drive? <<
I don't know, but I'd guess it wouldn't help much in swapping scenarios.
>> BTW- are your meshes really big, or is most of the RAM needed for film rez
tex maps? (or both? <g>)
the last couple of projects it's been really big geometry, not so much texture
maps. We often have multi-million face scenes now – that's the kind of detail
our clients demand.
Latest 3ds Chevy.3ds benchmark
CYBERIMAGE P133 NT Workstation with Asus Triton board and 8ns pipelined cache
and 60ns EDO ram
640 res with all defaults and mapping on 1:05
640 res with defaults , mapping OFF :45 SEC !!!
P133 overclked to 150mhz 36 Sec !!
Hard Drive is a Quantum FIREBALL EIDE mode 4 dma 2 5400 rpm
very fast on triton !!
video S64 video vram
Cameron
CYBERIMAGE
Hey Cameron,
Nice specs. Certainly the fastest posted. (better be!) With the inprovement
of the overclocking, I'd think that it'd be worth the possible shortened
lifespan to guys like Martin where scenes are huge, deadlines even "huger" and
every second saved is worth good money.
What are the NT HELL specs?
You've got me curious on what a 133 would do on my TD.. It'd be interesting to
tell. Hmm.. 😉
Have you done anything with the Atlas/Atlas2 drives? I hear there's an Atlas 2
out now.. even faster.
Is anyone saving all the #'s posted? If not, I'll go back sometime and compile
them.
JNJ
John,
I'd broached the subject of compiling all these numbers into one reference doc
to Anne when this all started. She replied that she was forwarding the msg to
the FA of Hardware (I think she said Greg Pyros). I haven't heard anything on
the subject since, but I've got about 60 msgs. on the subject, spread across
four or five threads. That's not all of 'em, but it's a good sampling.
BTW, I calculated that Cameron's overclocked 150, if he'd had mapping turned
on, would've done it in :52. I wonder if we can consider this a preview of what
'real' 150s will do?!?
'Faster clock speeds, bigger hard drives, cooler IPASs, more RAM!'
(Remember that old "Faster Horses" tune?–I think it was Tom T. Hall.)
SCL
Cameron,
Very nice times!!!
How do you speed up your Pentium, is there a jumper that makes the change in
clock speed?
Regards……………..Rick – Technical Animations
Hi Rick,
if you have a 2.5clk jumper, you can set the p133 to 60mhz ext speed and
2.5 clk =150mhz
it is not always perfect though, produces errors sometimes in Win 95\nt,
but since 3ds is dos based, timing is not as critical and sensitive.
have not tried to set a p133 at 66 x 2,5 to get 167, but it is on my list.
we got the new ASUS SiS based p55's in ,
they support 1 simm at a time, and have support for AMD k5 and Cyrix M1,
as well as P167
they are FAST !
Cameron
CI
Hi Cameron,
Have you seen my post for my SMP55 100Mghz running at 100 mgz? 640 Mapping
on is 1:22 (very close) and 640 maping off is 0:20!!! Why do you think that
is? HD id Western Digital 1.6 although 3ds is running off of an older conner
1gig. Win95. Non-edo ram. Any ideas?
Timothy
NVision Inc. & Digibotics
Did you forget to plug it in?? <VBG>
Jim…. ComputerWizards/Texas/Blazing on Win95 :)>
Thanx again Rick; I'm sure the pipelined burst cache helps, too, but that 60NS
RAM sure don't hurt! As of this writing, you've got a P100 that runs with the
120s! (Haven't gotten any numbers from a 133 yet) Have you got some kind of
graphic accelerator card in that thing?
SCL
Stephen,
>>Have you got some kind of graphic accelerator card in that thing?<<
I'm using the Hercules Terminator w/4mb of memory(964 chip), don't think that's
going to help with rendering. I'm not sure if it was you that mentioned Liquid
Speed, which I do have, but shouldn't make any difference when it comes to
rendering. If it does I'm sure Vibrant would like to know about it.
Regards…………Rick – Technical Animations
You're right, I was talking about a co-processor-type accelerator, not a video
"accelerator" card. If those had anything to do with rendering speed, I'm sure
EVERYBODY would like to hear about that!!! Liquid Speed doesn't help rendering,
either, in my experience, as I've run the same machine without, then with, and
although it did help screen redraws (somewhat) it didn't appear to have any
effect on rendering times. Maybe somebody did accidentally drop a 120 into the
100s box, and you lucked out! Thanx again for all your trouble.
SCL
More benchmarks:
P133 32MB RAM 512k
CHEVY.3DS All defaults, no display:
640×480 1:20
1024×768 2:43
HH
Harley,
Those times sure got me beat<G>. I assume by "All defaults" you mean that you
used Metal shading and Mapping was on?
Regards…………Rick – Technical Animations
Rick,
>> I assume by "All defaults" you mean that you used Metal shading and
>>Mapping was on?
Yup. Mapping and metal shading were on. I'm happy! 😉
Hey, Harley!
That's a pretty quick box you got there! You said you got it from Micron? Is it
a Millenium?
One of the reasons I got myself involved in this whole benchmark thing is that
I'm in the market to replace my DX2/66 (finally!).
Have you looked under the hood yet? I'm curious if there's enough room for a
PVR. (It needs an unobstructed, full-length PCI slot.)
Any info you can provide (like who your rep was) would be a help.
Thanx,
SCL
Stephen,
>>Any info you can provide (like who your rep was) would be a help.<<
Try going to the PC HARDWARE forum, Micron had two reps there that were very
sharp and friendly. Micron was also looking to open their own forum, but I
don't know if it ever happened.
When I was looking at Micron neither their Mellennium or their standard Pentium
would handle 2 ISA slots or 2 PCI full length slots, but things may have
changed by now. I was looking at Micron very strongly, but decided to go with
MIND COMPUTER CONSULTANTS because the SuperMicro motherboard had 4 ISA and 4
PCI slots with none shared. Out of the 8 slots I have 3 ISA and 3 PCI full
length slots.
I just deleted all the messges concerning the Micron, but I still have some of
the printouts if you would like to have them. Let me know.
Regards………..Rick – Technical Animations
Rick,
Thanx for the info. Where is Mind Computer Consultants? A phone, e-mail address
or CIS number would help. Send via e-mail (my nickel) if you don't want to post
this on the forum (guidelines?).
Thanx in advance,
SCL
Stephen,
>>Thanx for the info. Where is Mind Computer Consultants? A phone, e-mail
address or CIS number would help. Send via e-mail (my nickel) if you don't want
to post this on the forum (guidelines?).<<
I don't think there are any rules against giving out vendor names and contact
info on this forum. Contact Mark DeMazza at MIND COMPUTER CONSULTANTS (203)
669-0895 or CIS# 71203,170. I liked Mark because he never pressures you to
place an order, he just answers your questions and lets you make up your own
mind.
Regards………….Rick – Technical Animations
Stephen,
No, its not a Micron. I went with a local company and had them build
it to my specs. If you want to know all the details and can send them to you.
I bought it with the possibility of getting a PVR in the (probably distant)
future.
Harley
Thanx, Harley, I'd appreciate any info you can provide. Is yours that
SuperMicro mobo, too? It seems to be (one of) the hot setups.
What do you consider local? (I'm in SoCal.)
Thanx in advance,
SCL
Harley,
I would be interested in the specs you used to build your system. There is
a local firm here who could probably do the job if I knew what to tell them. I
have been trying to keep up with all the hardware info, but am not that saavy
on the technical side and probably have missed something. Thanks.
regards,
wayne
Re: Triton chipset.
I am surprised that I haven't seen any discussion on EDO ram when discussing
system speeds. The Micron Millennia Plus ships with EDO ram. In several tests
that I have seen, EDO equiped machines run quite a bit faster than the same
system with standard ram. By the way, I have been running a Micron 90 system
w/32 MB for several months and have found it to be a fine system. Note: I have
only one full length PCI slot that is not being blocked by the CPU.
John,
Actually, there was a short exchange on EDO RAM somewhere in this twisted,
convoluted thread. It turns out that EDO RAM ain't what it's been cracked up
to be… Pipelined-Burst Cache is the real improvement in speed. Evidently, if
you have pipeline cache and substitute EDO RAM for ordinary RAM, the benefit is
barely more than the margin of error in the tests themselves, like 1 to 3%.
That single full-length PCI slot — is it being used? Do you have a total of 4
— 3 PCI only and 1 PCI/ISA?
I'm concerned because I'm considering ordering a Millenium and I want to put a
PVR in it…
Thanx,
SCL
Stephen,
Re: Micron
My system is based on the Powerstation 90. I have 2 PCI, 1 PCI/ISA and 4 ISA
slots. 1 full length is definitely open, 1 is questionable (probably wouldn't
risk it) and 1 is totally obscured. An important note if you have full length
ISA cards: If you populate all six banks of RAM, you will decrease your
available full length ISA slots by 1 or 2 depending on the RAM type {double
sided or single sided).
I would spend the extra $100.00 on the tower case, also. I ended up scraping
the baby-AT case. Depending on which side of the PVR the capture card fits on,
I may need to get a new motherboard for my new case….
I have been waiting for two weeks for the PVR and TBC add-on to arrive.
I will keep you posted.
Disclaimer: The Millenia is a different motherboard design and, as such, will
have a different configuration.
John,
Thanx for the tips.
This stuff just keeps getting more and more complicated, but since it also
keeps getting faster and better, I'm not complaining! <g>
SCL
John,
>>I am surprised that I haven't seen any discussion on EDO ram when discussing
system speeds. The Micron Millennia Plus ships with EDO ram. In several tests
that I have seen, EDO equiped machines run quite a bit faster than the same
system with standard ram.<<
Like most things in life, there are pro and cons to most everything. I would
venture to say that EDO ram is good in some applications and no benefit in
others. If you have it though you can't go wrong. When I bought mine system in
July, EDO ram was only available from the major companies, therefore, when I
went to MIND COMPUTER CONSULTANTS, Mark could not get it.
>>By the way, I have been running a Micron 90 system w/32 MB for several months
and have found it to be a fine system. Note: I have only one full length PCI
slot that is not being blocked by the CPU.<<
When I bought mine system it was between Micron and MIND COMPUTERS. Since I
couldn't get enough full length slots in the Micron and possible second source
components that I couldn't control I went with MIND COMPUTERS.
Regards…………Rick – Technical Animations
Rick,
Looking at the latest numbers you've posted, they make sense. (only a few
seconds faster than mine) If you have a Triton chipset, that'd be a big
improvement over mine, as well as your inline cache vs. my sram cache. Then
add 60ns ram, and they all add up to faster rendering.
Now why our systems beat a couple P120's posted, that's a bigger question.
JNJ
John,
I do have the Triton chipset, I heard that it is faster than the Neptune. But
next week it will be the Super Triton II and time goes on<G>.
I think the P100 talks outside the chip at 33MHz while the P120 talks at 30MHz.
That might explain why the P100 are almost as fast in some situations. If I was
to upgrade the CPU, then I would go to the P133.
Regards………..Rick – Technical Animations
Rick,
Yes, the Triton is faster, but doesn't support dual proc's. (that I know of).
Yes, 33mhz is better than 30mhz in bus speed. But I'm still suprised. 🙂
JNJ
More numbers
Dell XPS 90
P-90 16Mb no EMM386 ram all default settings, no display metal,mapping on,
pixel 1.1 etc… 640×480 1:50(110 seconds) swap 406K.
Generic DX4-100, 48Mb ram, no EMM386, all default settings no display,
metal,mapping on, pixel 1.1 etc… 640×480 3:53 (233 seconds) no swap.
Bob Kohl
Cedille inc.
-Bob Kohl
On my machines
Chevy.
Times Machine
5.10 DX2-66
1.39 Dec Pentium 90
1.51 No-name Pentium 90
2.05 ALR Pentium 66
-Peter
Oops, placed these in the wrong thread initially…
For the render speed list:
Intel 133 (Triton chipset), 70 ns RAM (non-EDO)
Chevy.3ds 640*480, 1.1 AA, Shading Limit Metal
No Display
mapping off: 51 seconds
mapping on: 1 minute 11 seconds
Display
mapping off: 51 seconds
mapping on: 1 minute 10 seconds
John Stetzer
JWS
The times I gave you were for 640x480x24 bit.
All machines had 32MB Ram.
Pentium 90's best effort was 1min39s on a DEC Celebris, other pentium 90's were
anywhere from 1.50 to 2.30