The eternal spec bench questions... (sorry so long...)

I have been delving into PC hardware questions this week due to unforeseen
PSU toasting, and am now getting more confused by the day (some say it is
because I am a Mac user...) about what is important and not important in the
eternal question of SW speed and viewing pleasure. And yes, I have trawled
the newsgroup archives and that makes it all the worse

The starting points are these:

1.
I have been tempted to get the FX1000 card (as is is somewhat reduced in
price at the moment for SW users) in order to solve my annoying window
slowdowns with my Ti4600 (which I do like...except for this issue) Note: I
have tried softquatro and it did not work too well for me - I would prefer
reliable....

2.
Also thought while I am at it just plug in a new faster AMD CPU to get a
little kick. Not very expensive, so why not? From an Athlon 1.7 to a 2.6.
(2.6 being the max for this MOBO) However, the MOBO would not be optimal,
but it would work.

3.
I am so pissed about the homebuilt cheapo solutions I got an offer from Dell
for a Precision 360, 3.0Ghz/1Gb RAM, FX1000, 80GB. (strangely it has a 250W
PSU when the Nvidia card says it needs 350... But they certify it!).

4.
I am in the tail end of a project that is very curvy, not many components,
but they are complicated and feeling slow on my machine now (especially
rebuilds in 2004).

So now, not really wanting to run out and buy a new computer when the last
one is only a year old, I was thinking the graphics card and chip would be a
compromise and improve my speed and windowing issue. And for timing this may
be best as the dell takes a few weeks to build.

Now the problem is this. I thought I would be clever and run the SW SPECapc
to see how my workstation measures up to others, specifically the Dell 360.
Now I have always felt specs are not very useful, but now that I have run it
and looked at the Dell SPEC scores, I am really wondering if I understand
them, and how, if at all, they might influence my choices.

I ran the spec on my system, maybe not absolutely as they say, but I
rebooted and ran it on SW2003sp3. My system is: ASUS MOBO, Athlon 1.7Ghz/1GB
RAM, Ti4600 (and now a brand new over spec'd 400w PSU!)
My scores for 5 tests:

Test Total 269
Graphics 47
CPU 137
I/O 85

Now, if I understand, that means the graphic card is fast (excluding the
window slowness issue I assume), that the processor is a dinosaur, and the
I/O could be better. Right?

But I would have assumed that the DELL with the much faster processor and
MOBO and the FX1000 would be MUCH better (precision 360 3.2Ghz / 1GB RAM,
FX1000). But these are the SPEC scores (1 not averaged) from the website:

Test Total 235
Graphics 99
CPU 68
I/O 68

So what looks good is the processor and to a lesser degree the I/O. But the
graphics looks worse than on my score.

After looking at old threads and SPEC scores, I am only more confused. Based
on this I would be most inclined to only upgrade my processor and try really
hard to make the SoftQuatro work. Or the next option of Processor and FX1000
(for more stability peace of mind).

However, the Dell does not look as interesting as I imagined it might.

So
A. am I missing something in my feeble interpretation of what is important?
(even if I do nothing I would like to understand this better)

B. If you were in my shoes what would you do? (and no, drinking is not an
answer

Thanks to all who jump into the fray!

Daniel

 

Looks to me like you want to put in the faster CPU and maybe faster memory
if your MOBO can handle it and skip the Graphics card downgrade. The spec
benchmark is very heavily weighted towards graphics. You might also want to
compare the Dell on some real world problems if you can.

 

I believe the problem is simple... There are two data sets available when
you run the 2003 benchmark - the first called "Standard General Test" does
not take as long to run as the other (I can't remember its name but it's the
second radio button). The published results on the Spec website use the
second, more exhaustive, dataset. Run your benchmark again using that
second dataset and you'll really see how your machine stacks up against the
Dell (or not).

- Eddy

 

Eddy Hicks wrote on 26.9.2003 7:40

Hi Eddy,

Well... Now that I look again.... I think you are right, and I was comparing
my "standard general test" (SGT) to the SPECapc test. I think I did this
because the numbers looked more similar :-> Doh!!!

Really?
If that is the case, then my machine is dogmeat! (which makes much more
sense actually)

But I am still confused. The result format and information output into the
txt file does not seem directly comparable to the online SPECapc score since
there does not seem to be a composite score or the summary shown at the top
of the official results. If that is the case, then this is the actual
comparison... Embarrassingly enough.

My PC: AMD 1.7/1GB RAM, ASUS A7S333 MOBO, 80GB HD, Nvidia Ti4600

SW PC: Dell 360 / P4 3.2Ghz/1GB RAM, Nvidia FX1000


Test 1 My PC Dell PC
Test total 532 235
Graphics 308 99
CPU 132 68
I/0 92 68


So, now, all of a sudden... The option of a new graphics card and a cheap
CPU upgrade makes more sense. If I can improve the CPU a little that is
good, but actually the FX1000 should make a big improvement. Doing this is
about 40% the cost of the Dell.

And even a new machine makes sense. Except I would rather not spend the
money right now.

Now do I understand???

Thanks for jiggling my brain!

 

kellnerp wrote on 26.9.2003 1:55

Love to, but have no idea where I could find one to do that.

But also see my other post where It is clear I was comparing apples and
oranges...

Thanks

 

....snip

but it would work.


If your mobo supports Athlons from 1.7 to 2.6, it probably has a 266 MHz
front side bus. Be aware that there were two different Athlon 2600+ CPUs:
one for a 266 FSB (manufactured for a short time) and a later one for the
333 FSB. I bought a 2600/266 a few months ago as a final upgrade to my mobo.
The CPU was hard to find, and pricey for its performance level.

You might give serious thought to a new mobo with a 333 FSB.

Art W.

 

Yep, now you've got it! I've become sort of a self-proclaimed expert on the
topic of SoftQuadro so if you want to make that work, shoot me an email.
Otherwise any Quadro2 or above is going to be a huge improvement. I
actually have a leftover brand new Quadro4-980XGL that I'll sell you if
you're interested. I only bought it to do benchmarking, which I posted
about on the 15th. There wasn't huge differences to Solidworks between the
various cards I tested, likely because SW has let this area of there
software stagnate. As for a new cpu, I would heed other's advice and stay
away from the 2600 and go with the 2400 since it's easier to find and
guaranteed to be a 266 fsb. But the other advice about a new mobo is
probably the best yet. newgg.com can supply a 333 or 400 fsb and new
matching cpu for less than a couple hundred (or more depending on your
choice). Think MSI, Gigabyte, or Asus in terms of reliability and
performance. I use MSI or Gigabyte myself because Asus is typically a tad
slower in benchmarks.

PC = options. What you do with the options is up to you

- Eddy

 

Eddy Hicks wrote on 26.9.2003 15:30

My options:
1. get another bottle of asparin.
2. Buy a Mac G5 )
3. Disconnect the power and use a pencil. Remember those?


And I'll take you up on the softquadro advice by email. Thanks!

Daniel

 

One man's Mac is another man's doorstop, and one man's pc is another man's
labyrinth.

 

Daniel,

Home built doesn't automatically = cheapo. There are very high quality
components available.

The motherboard manufacturers compete mainly on I.O. performance. This is an
area where the interests of gaming and CAD converge. Fast data through put
(buss speed, band width and efficiency) is what makes both CAD and games
perform.

Companies like Dell, HP, and IBM, compete on price. This can, (and does in
most cases), mean using the cheapest approach to motherboard design. For
this reason, a good custom built system will "ALLWAYS" out perform the best
name brands. If you use quality components, reliability will be better too.

Regards

Mark

 

Eddy Hicks wrote on 26.9.2003 16:31

Hmmmmm.
A labyrinth is complex and designed to be confusing.
A doorstop is very clear in it's purpose, and is intuitive to use.

I like your analogies! Now that I look at it that way, I agree!!! ;-)

 

MM wrote on 26.9.2003 19:27

I would have to agree with all you say. However, what frustrates me, and
probably more when I am under project pressure, is the time it takes to
become an expert to know if you are getting a good deal, good hardware,
etc., or even good timely support. I can certainly get into the process of
selecting and putting together components until I have the perfect high
performance workhorse (to be obsolete in 6 months...). But that is a painful
process and is not billable (I have enough of those hours...).

Part may be being in a country where the freaking immediacy of everything in
the US does not really exist. If I want anyone to look at my PC or install
anything (sure...I can do it too... But....) it will take several days. Not
possible on deadline. Hence my thoughts of off the shelf kit for a degree of
reliability and reasonability for that kit of parts. Hence Dell.

Another thing that factors into all this, and into the specs of my current
machine, is that many of the people around me are more oriented toward tools
like Alias StudioTools, Maya, XSI, and there really is a difference in the
machines they need, vs. SolidWorks or other Solidmodelers. And Each time I
have spec'd a PC (until now) I have been influenced by their logic and
thinking, but from my experience I can see there are different hardware
needs. Next time I will spec a better piece of kit!

But I remain undecided, and my bank is happier for the moment (the weekend
at least)

Regards,

Dan

 

One thing to note is that the division into Graphics, CPU and I/O is rather
arbitrary. In particular, the CPU usually has a strong effect (nearly linear
with speed) on all three. The graphics card and the disk usually have much
smaller effects on the other categories. So the processor is more critical
than the CPU scores might lead you to belive.

Jerry Steiger
Tripod Data Systems

 

Jerry Steiger wrote on 27.9.2003 19:01

I can see that now. Installed the 2400 since it was cheap, and with Eddy's
help a functional softquadro on my graphics card. But it is clear the real
solution is to get a new PC :-( Guess I will time it for Christmas so I can
feel better about it