#Pentium and 3D Studio!
18 messages in this thread
Ana,
I saw a bunch of AMI Pentium/PCI and Pentium/VLB motherboards at Comdex. Who
cares about cooling if you can get an Icecap active heatsink for less than
$200. I don't know how good AMI's cpu to memory interface is on these
motherboards, but judging from AMI's track record, they probably won't be
slouches. Intel likes to say that the delayed appearance of second-tier
Pentium clones is because of a lack of techinical expertise on the clonemakers'
part, but the real reason is a lack of availability of Pentium chips. Intel
wants to squeeze money out of the 486 line as long as it is more profitable
than the Pentium. I wonder why Intel doesn't just charge (however much it
costs to make the Pentium + (1.10*(their profit on a 486DX2/66 chip)). People
will pay.
Bailey
Bailey,
While I agree that Intel is squeezing the 486 line (and therefore not making
Pentiums as quickly as all of us want) and that AMI can take out Pentium
motherboards (in my opinion AMI is a very big company), I think that there are
lot of problems regarding the Pentium motherboards (and not only me, but
engineers from Digital, Compaq…). And not only Pentiums, but also other
chips, like the Alpha AXP series. The problems doesn't come just from the "hot"
side of the chips, but also from many other things, such as the radio
inteferences and thinks like that. However, I'm not an engineer and just limit
myself to repeat what some people have said about it (hey, I just want a speed
demon that works as well and solid as my old first XT at 8MHz 🙂 BTW, I think
that there is a logic relation between the complexity of a chip and the
complexity of the whole machine. While chips going higher and higher, machines
have to reach new quality levels (just to stay with the chip). I simply said
that probably the Triflex architecture, developed by Compaq and Intel, is the
best to take out all of the Pentium power…for now. I'm sure that there are
other people playing with the silicons to take out the best of them (BTW, I
think that we have to watch over Texas Instruments/Cyrix movements…it seems
that something big is going on there).
Ana
ps.Regarding delayed chips…mmmmh, someone points me out here that in the
presentation of the 386 chip, the chip has a date tag of ONE year BEFORE the
date of introduction. For sure, Intel, as the only chip maker, have played with
us all these years. Fortunately, things have changed, and if they want to keep
with the rest of the market they have to deliver the good things as soon as
they can.
From what I saw at Comdex I'd say the the DEC Alpha is only being held up by
Windows NT release and if Windows NT then DEC Alpha and Intel Pentium will be
left in the dust. Personally, I'm holding off until this three way speed race
takes a pit stop before jumping in on upgrades to my MB. I've put all my
budget for upgrades in software and harddrive (related) dollers for the rest of
the year. Thanks for your benchmark reports and looking forward to future
reports from you.
Don Landis
I agree with you in the "wait-and-see" actitude. I'm wondering if we will see
our Targas and Vistas in an AXP machine soon. For now, I've read that their
EISA AXP uses "six EISA slots, standard SIMM slots for 12MB to 128MB, five
drive bays, Compaq's QVision graphics controller, an Adaptec SCSI
controller…" in June 15 issue of PCMag. Probably if the Yost Group give us a
little hint of future hardware platforms for 3ds (i.e. 3dsNT 🙂 some of us can
plan a long term buy strategy (and save some money for future IPAS routines 🙂
Of course "megas" and hard drives has been always been the best investment.
Ana
I don't belieE you are refering to the US market when you refered to megas and
Hard drives being a good investment. Most computer hardware is not a good
"investment" I consider it a necessary evil to do the job. Since I've been
buying computer hardware circa, 1977, there has been one constant. { Anything
you buy today should be usable today because it will be better and cheaper
tomorrow. Therefore, anything you need for tomorrow should be bought tomorrow.}
Translated to time numbers: Hardware is generally obsolete in 4 to 6 months.
This obsolescence is in speed and power vs. $. eg. I purchased a MB with 0 meg
RAM for $1550.00 in 1990. Today this same MB with a better chipset, better
bios, more compact, and higher cache is available for $135.00.
That same year I bought a 210 meg HD. for $635.00. Today I just priced a 545
meg HD for $535.00 and a 345 meg for $295.00. I will agree that RAM is a good
investment because in the US it is traded as a commodity. The recent price
hike in January we noticed was due to the threat of a retroactive tax on asian
memory chips. Prior to that announcement we saw memory slowly coming down.
The wait and see is for CPU upgrade only. The big boy on the block right now
is the 486-66 as far as $ gross sales. Once a competitive chip such as the
DX3, Pentium, or Alpha/NT hits the market the prices of 486-66 will drop off
fast. Its just repeat of history, Ana. I can buy a MB with DX2-66 as cheap as
I paid for a 486-33 MB last fall just 7 months ago. A friend of mine bought an
EISA machine in 1990 and paid (3 year loan) total P&I $6200.00 He never used
the EISA slots until two months ago. He also just bought a 486-66 for $505.00
to upgrade. I figure if he didn't buy latest fastest out but instead bought 1
year old (obsolescence cycle in 1990) technology he could have saved $4000.00.
Now there are those that will say it doesn't matter what the cost if you can
balance the books with a return on investment in the obsolescence cycle I've
described. My business can't; but, I can by buying 6-9 month old technology.
To make matters worse, our tax structure requires that computers and software
be depreciated in 5 years. The only thing in my computer that's five years old
is a $40.00 serial/parallel card that I could buy today for $9.00.
Regarding future of 3DStudio: Just wait until siggraph because Gary said he
will tell all at that time. I would bet we will see a Windows NT version in
the future (but not rel. 3) though because so many software companies including
Autodesk are betting on Win NT. I was hot on it before Comdex. After Comdex I
really cooled to it. All that hype and no release or ship date announcment
made me suspicious. Then little limitations were leaked such as real speed
benchmarks on the 16 bit apps running in Win NT that were slower than under DOS
and the system crash problems of dos apps and win3.1 apps. I lost my
enthusiasm.
Bottom line, at least for me is upgrade
I'm not talking about obsolescence. I'm just only think that buying now a good
SCSI hardisk or more megabytes for your ACTUAL system(s) (if you need the extra
power NOW. I simply buy things when I need them) and when you buy a new one,
you can allways put your megabytes (hd and mem) to the new one. So they allways
been there (specially hard drives). Of course you can buy them in the future
for less money, but if you need more power now, don't want (or don't need) to
buy a new system, more memory and hd space is a good buy for that reason.
Regarding EISA…well, if I've bought an EISA computer one year ago is because
I know that in the future I can upgrade to new (more powerful and modern boards
than my old ISAs) WITHOUT buying a new computer to hold them (which will extend
its life) I agree with you that now, the correct move is wait. I also think
that for those who need the computer power now (i.e.upgrade a 386) they can buy
a computer that will fit any future upgrade, and a company that really
guarantee you the escalability of your system (and for now, I find the
Triflex/PC to be the best move to do that).
Ana
Yes, you and I agree even on the RAM and HD. Buy now what you need now.
Ana, what do you think about SCSI-2 vs IDE (present state of art)? Your
opinion to include the controllers w cache available.
Don,
I'm running a DX2/66 with a VLB Fast SCSI-2 controller (Ultrastor 34F) and a
Maxtor MXT1280 drive. I can get about 3.5MB/sec off the hard disk without a
cache. The Quantum LP240 on that machine gets about 1.3MB/sec.
In my experience, IDE is a little faster than SCSI on small reads and writes
since there's less handshaking protocol going on between the controller and
drive, but SCSI is much faster on larger operations. Before with an ISA system
(Future Domain TMC1680) I was getting getting about 1.2MB/sec with both drives.
The best I've seen IDE do on ISA systems is around 0.6MB/sec. I understand
that some of the new VLB IDE controllers are getting around 1MB/sec.
I usually run a 2MB cache with smartdrv. This way, small reads and writes
(compiling and linking) go very fast, and large operations (displaying 10MB
Targa files, playing 50MB FLC animations off the disk) go fast too.
The benchmarks I've seen in PC-Magazine all show rediculous speeds like
150KB/sec. Their program just tests small operations and random seeks. I
suppose if all you do is run dBASE on small records and huge files all day long
those numbers might be valid, but I don't think that'a what most users do.
It's certainly not what I do!
Besides with SCSI I can run my CD-ROM and all three hard disks (sometimes four)
off the same controller.
John.
Thanks for the info, John.
I'm currently using 4-IDE drives on a ISA coprocessed 4Meg Cache ctrl. Speed is
about 1.3 meg/sec using special block mode transfer. The next move is to go
to a VLB MB and I'm considering switching to the SCSI -2 route but I've heard
so many horror stories on compatibility and whether to use EISA or local bus so
I thought I'd ask to get some experience/opinions.
Can please explain to me what is VLB MB I keep reading about?
Also whats the difference between EISA and Local Bus?
Thanks JR
VLB MB = VESA Local Bus Motherboard.
A local bus slot is just an extra edge connector past the end of a normal ISA
slot. It provides, for cards designed to use it, a 32-bit data path directly
to the processor, operating at the same speed as the processor (as opposed to
the standard 8Mhz ISA bus speed).
Dave
VLB MB is short lingo for Vesa Local Bus Main Board. This bus is a special
slot that VLB cards plug into that allow the card to access the CPU direct by
way of a 32 bit data path and at CPU speed. These slots are available on MB's
with both EISA and ISA bus slots. The advantage of the VLB is speed for
currently three types of I/O cards; VGA's, Disk controllers, and Network cards.
Also, the cost of a VLB card is lower than the EISA equivalent.
Straight EISA is 32 bit but at bus speed of 8 Mhz. not CPU speed or typical 33
Mhz.
John, there are all kinds of variations of the above and continued new
developments to keep us all confused but the simple VLB has been around for
over a year and is becoming the new standard for speeding up the old ISA bus
technology. At least a standard for the next few months <g>
Don,
1.3MB/sec is about as fast as you can pump data across the ISA bus. The numbers
I qouted you for my SCSI setup were with cache disabled. With SMARTDRV I can
get around 9MB/sec! Of course that's with files that are totally cached — you
would probably get that too.
SCSI CAN lead to some headaches. Here's a few rules (I usually violate them
all unfortunately) to help stay out of trouble:
1) Stay away from old SCSI devices. When I converted from a Future
Domaine ISA controller to the Ultrastor VLB controller one of my
old disk (200MB Fuji– hey, it was only slightly used and cheap)
wouldn't work properly. The Ultrastor couldn't figure out how big
the drive was. It seemed to think it was 1GB! I finally set up
a second machine with the old Future Domaine controller so the
disk came in handy anyway.
2) Stay away from really new SCSI devices too. The latest technology
usually has bugs… I do like the Ultrastor controller though.
Having not tried any others, I can't say how it compares. It is
definitly faster than my old ISA controller when attached to a
drive that can keep up.
3) I've heard from a few friends that have used Corel's SCSI kit
(ASPI?) and have been quite happy with it. It seems to be a
must-have if you plan to hook up lots of odd devices. On my main
machine I'm running 3 hard disks and a CD-ROM under DOS and OS/2
and havn't had problems that required it.
4) I was always unhappy with Future Domaine. But that's a long story.
They have their own SCSI spec which requires an additional layer
to be ASPI compatible. Seems like more stuff to go wrong…
5) I have this other rule about disk drives, although it's not a SCSI
rule: Stay away from anything made by Seagate! My 1239 (200MB) has
a MTBF of about 6 months. It just died again today while installing
OS/2 2.1. I guess I'll just toss it and get another Quantum LP240.
I've always had good luck with Quantum, bad luck with Seagate.
I do have one trick that you can do with SCSI that would be a little harder
with IDE. I have a switch attached to one of my drives SCSI ID jumper blocks.
In one position it's device 0, in the other it's device 4. If I want to boot
OS/2 I just flip it to device 0, otherwise device 1 boots and DOS comes up.
This is simpler and cleaner than OS/2's dual boot option. I just have to be
careful not to switch it in the middle of a session.
I gotta stop writing these long messages…
John.
Thank you for the reply.
With your comment about the 1239 Seagate— Iguess I should watch out because
I've been using them now for over three years. I have three on the controller
now.
You're right about Smartdrive getting very high speed results but then that is
really not accurate if it is really accessing the RAM and not the drive. Since
my files are typically half meg to 1 meg in size I can probably conclude that
the 4 meg cache is adequate with the IDE drives. I don't need to tie in alot
of other devices so SCSI can just wait. I got enough problems without adding
scsi to the list.
How do you measure the transfer rate to come up with your data you quoted?
John,
You're the guy who does VistaPro, right? That's my second favorite program
overall, but from a cost/usefullness perspective, it is the greatest!
I think the PC rags are doing their best to scare people away from SCSI. I
don't understand this since it really isn't that hard. In fact, my system has
been alot easier to maintain with SCSI than my previous IDE system. It's great
to have three devices plugged into one card.
Bailey
I agree 110%, John. ASPI is the winner in the SCSI wars. I also think the same
about Future Domain boards, and also about Trantor (I play a little with their
adapter in the PAS16 and with the ones that come with NEC CDs -if you want to
buy a NEC, buy it without adapter-). Adding layers and layers to the SCSI is
not a good idea. For me, ASPI boards are the only choice: Corel and Adaptec are
really good ones (solid, fast and compatibles).
Ana
Hi Don,
I definetely went for SCSI a year ago. Only two IDE drives are on two of my
systems (how you can throw to trash 500MB? 🙂 but when I move to magneto
optical storage, SCSI was my only choice and I'm very happy with it. I think
that the performance is very good, specially with cache, and have all your
devices connected to an easy-to-set-up chain, including scanners and cd-roms,
makes the only choice for me. Some people said that SCSI on PCs are a
nightmare. However, for me is a dream (a very nice one). I haven't had any
problem setting up peripherals, in our systems or in the magazine lab. If you
are looking for cheap price, good drivers and an easy-to-do scheme, go for
CorelSCSI! If you want outstanding performance, Adaptec is the choice for me
(their highend cached EISA boards are easy to setup too). You can use CorelSCSI
drivers in any ASPI board (I have only tried this with the Adaptecs). The EISA
boards from Compaq (which use the NCR chipset, the same as the Corel boards I
think) are really good too, specially if you want to setup an huge storage
system. We have had a SystemPro/XL with a 2GB disk array in the lab, and it's
an awesome performance.
Hope this helps.
Take care, Ana
Thanks for the reply.
I'm getting closer to the SCSI decision. It's no longer SCSI or IDE decision
since I must have more than 500 meg per drive/directory. Now, which is better?
Quantum, MIcropolis, or Maxtor? What's a good ISA cache ctrl. card ?