PDA

View Full Version : New chip architecture VS Ghz numbers?



Johnny
03-23-2009, 09:13 AM
This bit caught my eye in a review of the new Nehelem Mac Pros:

"Most routine apps perform better with fewer, faster processors"
^^^ This in reference, I guess, to the idea that having many cores is a great idea on paper, but in reality, most apps aren't there right now in terms of using all cores to their max, so, you're better off with fewer chips which are faster, for the best bang/buck ratio.

One concept I remember in LW discussions going way back is that render times are directly proportional to clock speed...this was true when the G4 came out with altivec, when the G5 came out with its supposed space-warping speed.

So, is it still true with these new Xeons?

and if it is, then a 3.06Ghz C2D iMac would be comparable in 3D rendering to a C2D Nehelem-based MP, on a core-to-core basis, correct?

Speaking for myself, my Mac upgrade dollars have to get me something right now, today, not in the theoretical future when apps may or may not be written to fully utilize all cores to their maximum, which is the whole entire reason for having an 8-core machine.

From my reading so far, one of the only apps that DOES fully utilize multiple cores is some benchtesting doo-dad!

Nobody drops that kind of coin so that they can run benchtesting doo-dads!

I'm shying away from the MP iron partly bcs of cost, partly because of the power consumption. Call me nutty, but the amount of electricity used in the production of eye-popping graphics matters greatly to me.

In past comparisons I've seen, the Mac Pros are the fastest mac rigs, but the iMacs are in the same ballpark, often turning in negligibly slower times in certain tests.

The top iMac now costs about 2700 with RAM maxed; the top MP, more like $6K with only 1/4 of max RAM,

To a guy like me, that's a BIG difference.

any feeling about the 3D rendering/buck ratio? or are most folks here willing to pay a premium to get a machine which will be way faster in a few situations?

thank you for any honest opinions on this all.

J

Lightwolf
03-23-2009, 09:32 AM
So, is it still true with these new Xeons?
No. And to be honest, it never has been between CPU generations either.


and if it is, then a 3.06Ghz C2D iMac would be comparable in 3D rendering to a C2D Nehelem-based MP, on a core-to-core basis, correct?
The Nehalem isn't a C2D... that's a different generation. So, no, again ;)


From my reading so far, one of the only apps that DOES fully utilize multiple cores is some benchtesting doo-dad!
Well, pretty much all CG rendering apps, compositing apps more or less, some audio apps. For some it depends on the kind of workload you throw at them.


any feeling about the 3D rendering/buck ratio? or are most folks here willing to pay a premium to get a machine which will be way faster in a few situations?
I won't say a thing... toby would :twak: me ;)

But looking at bang for the buck, and Apple only: Get the cheapest mini ;)

Cheers,
Mike

Chilton
03-23-2009, 11:25 AM
Personally, I prefer multiple small systems with large screens over a single faster system with a normal screen. However, my bottleneck is usually me, and not the host system. I wonder if a small array of minis would outperform a single mac pro, for a somewhat lengthy render.

-Chilton

Johnny
03-23-2009, 11:30 AM
Personally, I prefer multiple small systems with large screens over a single faster system with a normal screen. However, my bottleneck is usually me, and not the host system. I wonder if a small array of minis would outperform a single mac pro, for a somewhat lengthy render.

-Chilton


OK..the larger screen making your chair time more productive bcs you can see what you're working on in a larger pane, have more tool palettes or windows open?

I like the idea of a larger number of "smaller" machines but I can't find any data wherein that configuration is proven to be better than the single mambo machine.

One thing I know is a benefit: with my 2-mini setup, if I need one machine's full attention during a screamer session, I can take that node off line, then put it back to SN work when I'm done.

THAT is very functional to me.

J

Lightwolf
03-23-2009, 11:32 AM
I wonder if a small array of minis would outperform a single mac pro, for a somewhat lengthy render.

On the previous gen 8 core MacPro, 2.8GHz, OS X 10.5, Cinebench 10 (which scales very well) gives you a speedup of 588% using 8 cores vs. using one core (630% on the same machine running Vista x64, but it's roughly the same region). That's an efficiency of 75% per CPU.
The efficiency using two cores is 89% on OS X in that task.
Do the math... ;)

Cheers,
Mike

jwiede
03-23-2009, 12:19 PM
Right, but efficiency's a curve either way, and a cluster arrangement of mini's won't be 89% per CPU for 8 CPUs either. If anything the cluster will do worse in a 4x2CPU benchmark cetarus parabus, due to the overhead of network communications versus QPI.

Lightwolf
03-23-2009, 12:24 PM
Right, but efficiency's a curve either way, and a cluster arrangement of mini's won't be 89% per CPU for 8 CPUs either. If anything the cluster will do worse in a 4x2CPU benchmark cetarus parabus, due to the overhead of network communications versus QPI.
What network communication? Loading the scene is about it if you use a bog standard render farm arrangement (and even then that's usually only at the start of the sequence), and a simple farm scales extremely well. And that's all you need for animations. No fancy cluster...

Cheers,
Mike

avkills
03-23-2009, 11:05 PM
I just setup a render farm of 2 8 Core 3.2 Ghz Mac Pros. I can honestly tell you it renders more than twice as fast as doing them on a single machine from layout.

Basically I have my main edit machine (also an 8 core Mac Pro 3.2Ghz) running layout without the hub controlling the other 2 Mac Pros via ScreamerNet II.

There is no network overhead; if you are using Gigabit Ethernet then the only thing that would slow down the process is a slow *** hard drive.

Basically in OS X 10.5, you mount share points on the node machines and go to town.

My node setup is running 2 nodes on each machine with 4 threads each (which makes sense since each machine has 2 quad core chips).

If I had to, I could use my laptop to control and use all three machines; but I like being able to render and still work on a fast machine.

The specs on all three machines are the same -- 3.2 Ghz 8-core, 8GB RAM, GeForce 8800GT 512MB. They pretty much kick *** for rendering.

-mark

P.S. a big kudos goes out to Mike S for his excellent tutorial on setting up network rendering on the Mac. (and yes Mike I will probably be buying your ScreamerNet app thingie (SNUB) when we have some more $$ comes rolling in. :D )

avkills
03-23-2009, 11:10 PM
Oh one more thing. Lightwave uses every core available. So does After Effects. So can Compressor using Qmaster.

-mark

Lightwolf
03-24-2009, 02:15 AM
Oh one more thing. Lightwave uses every core available. So does After Effects. So can Compressor using Qmaster.

That doesn't necessarily mean that it uses them effectively though. ;)

Cheers,
Mike

Johnny
03-24-2009, 04:53 AM
has anyone with these machines used Menu Meters or Activity Monitor's CPU usage/history tools to get a sense of just how busy the many cores of their machines are during LW renders?

And if they do, just how scientific can that be? Lots of activities will "peg" those meters, at least visually...

J

Johnny
03-24-2009, 05:03 AM
What network communication? Loading the scene is about it if you use a bog standard render farm arrangement (and even then that's usually only at the start of the sequence), and a simple farm scales extremely well. And that's all you need for animations. No fancy cluster...

Cheers,
Mike

I'm not sure what you're saying, Mike...I reads like your saying that a dual-core machine achieves greater per-core efficiency than an 8-core machine, but what do you say accounts for this?

Is it that, while doing 3D rendering, there is overhead in "managing" the cores, and that this overhead increases with the number OF cores needing to be managed?

J

Lightwolf
03-24-2009, 05:06 AM
And if they do, just how scientific can that be? Lots of activities will "peg" those meters, at least visually...
Even if they're all at 100% doesn't mean that the process is scaling at 100%.
In the CineBench (I like to look at CineBench numbers because it's the closes thing to LW as far as cross platform rendering benchmarks are concerned... and because it's used all the time) example, with an efficiency of 75%, the CPUs will still be at 100%, but 25% of that performance is going toward managing the multiple threads and shared ressources.

Cheers,
Mike

Lightwolf
03-24-2009, 05:07 AM
Is it that, while doing 3D rendering, there is overhead in "managing" the cores, and that this overhead increases with the number OF cores needing to be managed?
Yup, precisely.

Cheers,
Mike

Lightwolf
03-24-2009, 05:11 AM
As an exercise: Try to find out the CineBench scores for the machines you're interested in (making sure it's the same CineBench release) and then make a CineBench points per US$ table.
That should get you a decent estimation of bang per buck... for rendering.

Also, on the same machine you can usually use the GHz of the CPU as a ballpark figure to estimate the performance of a different clocked CPU.
I.e. a 3.2 GHz CPU in the same machine is likely to be ( 3.2 / 2.8 ) = 114% the speed of a 2.8GHz one.

Cheers,
Mike

Johnny
03-24-2009, 05:13 AM
OK..so, to me that sounds like a beautiful example of diminishing returns, and the 8-core machine shines when the largest amount of brute rendering power is needed to bear down on a task that must be done on ONE machine, but that for distributed renders, many machines of fewer cores per machine will each render more efficiently than the same number of machines each with more cores.

wierd choice, as I think we ALL would like to tap the most power possible while we're sitting in the chair working, but would also like that after we hit F10 and go do something fun while our animation blasts away...

"having your cake and eating it, too" comes to mind...

J

Lightwolf
03-24-2009, 05:16 AM
OK..so, to me that sounds like a beautiful example of diminishing returns, and the 8-core machine shines when the largest amount of brute rendering power is needed to bear down on a task that must be done on ONE machine, but that for distributed renders, many machines of fewer cores per machine will each render more efficiently than the same number of machines each with more cores.
There's a lot more things that come into play if you look at it from an even closer angle (maybe even too many things ;) ):

Multithreaded programming tools and techniques are getting better all the time (heck, they have to nowadays).

You still need to take the cost for the base machine into account, RAM, HD space, power usage etc.

A heavy workstation will also help you when running separate tasks at the same time (i.e. render in the background while you composite up front).

...to name just a few.

Cheers,
Mike

Captain Obvious
03-24-2009, 05:20 AM
But looking at bang for the buck, and Apple only: Get the cheapest mini ;)
I don't know if it applies anymore, but about a year or two ago, I had a look at various rendering solutions. The Mac Pro actually gave the highest performance per money spent out of everything I looked at -- including home-built PCs! The price advantage of fewer, more powerful machines comes in when you buy memory. Fitting out twenty machines with 8 gigs of memory each is a hell of a lot more expensive than fitting out five more powerful and expensive machines with the same amount, and this becomes all the more pronounced the heavier your scenes are.

Lightwolf
03-24-2009, 05:33 AM
The price advantage of fewer, more powerful machines comes in when you buy memory.
Oh, absolutely. The same is also true if you need to actually pay for render nodes.
I assume you also didn't calculate with memory from Apple ;)

It is quite easy to tip the scales depending on the actual needs. Bang for the buck per CPU doesn't always help either (current leader: Pentium Dual-Core E5200, 5727 Cinebench points... but the CPU only costs 58€) as it doesn't take the whole system into account.

Cheers,
Mike

Scazzino
03-24-2009, 11:21 AM
I assume you also didn't calculate with memory from Apple ;)

I get all my Mac RAM from OWC... MacSales.com (http://eshop.macsales.com/shop/memory/Mac-Pro-Memory) Great prices, lifetime no hassle warranty. :thumbsup:

Captain Obvious
03-24-2009, 11:33 AM
Oh, absolutely. The same is also true if you need to actually pay for render nodes.
I assume you also didn't calculate with memory from Apple ;)
Of course I didn't! I didn't use Dell's memory pricing, either, or Boxxtech's, or anybody elses -- the pricing for all hard drives and memory was based on minimum spec in the OEM machine plus third party stuff.



Bang for the buck per CPU doesn't always help either (current leader: Pentium Dual-Core E5200, 5727 Cinebench points... but the CPU only costs 58€) as it doesn't take the whole system into account
Exactly. Getting a twice as expensive CPU that's 50 % faster might only increase the TOC by 10 % or so.

Chilton
03-24-2009, 12:35 PM
Exactly. Getting a twice as expensive CPU that's 50 % faster might only increase the TOC by 10 % or so.

I suspect that four Core 2 Duo based Minis would render slower than one new MacPro 4x2, right? And that eight C2D based Minis would be slower than one MacPro 8x2, right?

(The following is American money, as that's all they'll let me spend at the Apple store here...)

C2D Mini: ($600) x 4 = $2400
Mac Pro: $2500 x 1 = $2500

But each Mac Mini at that price has 1Gb of RAM, and the Mac Pro at that price has 3. BUT BUT, the Mac Minis aren't easily upgradeable, while the Pro is. So maxing out the RAM in the Pro would possibly cost less. And you can also cram more in, but it's less overall. 8GB in the Pro vs. 4Gb per machine in the above scenario. Hmm.

Mac Minis with 4 GB each would cost $750, bringing the total to $3000.

I further assume that an '8 banger' (a term that cracks me up) Mac Pro would cost around $3300. That system has 6Gb of RAM by default, and a max of 32Gb.

I'm not discounting the need for a 64-bit Mac version of LightWave, but most of the obscenely huge scenes that our users render come in just under that 4Gb limit. I suppose one could argue it's because they have to, at this time. But my guess is that for most of our users, while the 4Gb version of the Mac Mini would be almost necessary, the $3300 8x2 Mac Pro would have more memory than you'd probably use, most of the time.

And I'm not POSITIVE, but I bet that the $3300 8x2 Mac Pro would shred the Minis in a render test, like my limited edition Boba Fett under my dad's lawnmower.

But let's be real. Most of you are filthy stinkin' rich, having made it to the big time in the 3D world, and you're contemplating building a farm of maybe a dozen 8x2 Macs, and loading them up with their maximum 32Gb of RAM each.

Comparatively, you're thinking that for just the base price of that (12 x $3200), before RAM, you could buy around 54 Mac Minis. But you'd have a maximum RAM ceiling of 4Gb before you start thrashing the hard drives...

Why do I feel like I'm devising tactics for an RTS game every time I shop for a new Mac?

-Chilton

Captain Obvious
03-24-2009, 01:27 PM
I suspect that four Core 2 Duo based Minis would render slower than one new MacPro 4x2, right? And that eight C2D based Minis would be slower than one MacPro 8x2, right?
Yes, very much so.

Lightwolf
03-24-2009, 02:13 PM
Yes, very much so.
Hm, not quite... I've got CB (R10) scores across most of the current range of Intel compatible CPUs here - unfortunately not _exactly_ the one that's in the mini though.
However:
C2D E4500 (2,2GHz, 2MB Cache, FSB800) : 4805 points
Core i7-9920 (2.66 GHz): 16161 points
In this case the 4 minis (which have a compareable CPU) would be as fast as an i7 at 3.2 GHz (19220 points for the 4 minis vs. 19146 for the i7 965).

(these are Vista x64 scores, but the relations should be compareable on OS X, even if the individual scores might differ).

Mind you, this is rendering, pure CPU horsepower and doesn't take anything else into account (power usage might be interesting though).

Cheers,
Mike

Captain Obvious
03-24-2009, 04:34 PM
Hm, not quite.
Surely 8 x 2.26 GHz worth of Nehalem is faster than 8 x 2 GHz worth of... whatever was before Nehalem. Especially at rendering, what with the SMT and all.

Lightwolf
03-24-2009, 04:36 PM
Surely 8 x 2.26 GHz worth of Nehalem is faster than 8 x 2 GHz worth of... whatever was before Nehalem. Especially at rendering, what with the SMT and all.
A single socket i7 is only 4+4 though, and the extra +4 add roughly 15% each.

Just look at the numbers...

Cheers,
Mike

Captain Obvious
03-24-2009, 05:16 PM
I was thinking 2x4 as in two quad-cores. With RAM, four Mac minis will cost you about the same as a single eight-core Mac Pro, so that seems the logical comparison.

Lightwolf
03-24-2009, 05:27 PM
I was thinking 2x4 as in two quad-cores. With RAM, four Mac minis will cost you about the same as a single eight-core Mac Pro, so that seems the logical comparison.
I suppose it depends on the amount of RAM. If you look at the baseline numbers Chilton posted a single quad vs. 4 minis seems reasonable though (I also used the slowest single core i7 in my comparison, which is still 2.66 GHz, not 2.26).
It also depends on how well the dual quad scales within a single rendering task.

Cheers
Mike

avkills
03-24-2009, 06:21 PM
Now that the minis can have 4GB of ram in them, it makes for a good argument on whether to get 4 of those or 1 MacPro. Hmmm...

-mark

Johnny
03-24-2009, 08:02 PM
Now that the minis can have 4GB of ram in them, it makes for a good argument on whether to get 4 of those or 1 MacPro. Hmmm...

-mark

pretty interesting comparisons and "what-ifs" in this thread.

I think it would be FAR more gratifying and a better use of time to do modelling, scene construction, surface testing, scene development, etc, on one dang fast machine than on one not too bad machine.

J

dwburman
03-28-2009, 02:30 AM
I think the lower-end macs are relatively expensive. The MacPro workstations are a good value when compared to equivalent Windows machines but $600-$800 (USD) for a 2.0 GHz dual core box without a keyboard or mouse seems high to me. Sure, the mini looks cool, doesn't take up much space, and comes with all the Apple goodness that Macs come with these days but if you're just looking to use it as a render box you won't be using many of those features anyway.

My computer platform fanaticism died with the Commodore Amiga so I'm not a 100% Mac guy. :)

Lightwolf
03-28-2009, 04:53 AM
The MacPro workstations are a good value when compared to equivalent Windows machines...
I wish. The dual socket ones are... the single socket MacPro isn't.

Cheers
Mike

Captain Obvious
03-28-2009, 08:47 AM
I wish. The dual socket ones are... the single socket MacPro isn't.
Well yeah.

BigHache
03-28-2009, 09:44 AM
I think the lower-end macs are relatively expensive.
I tend to agree with that, but I think it also just depends on what you're doing. I'd say their major benefit is space-saving, which can be a major benefit.

Since I went with AMD the Minis became a non-option though.

Johnny
03-28-2009, 10:32 AM
one other thing is different, but which doesn't get much talk, is that a mini running balls to the wall consumes much less electricity than a Mac Pro does, with all 19 fans going..or however many fans...

there's no doubt that the Mac Pro is faster...but how much more juice per bit of work done does it require to get that speed?


J

Chilton
03-28-2009, 10:38 AM
And, while your mini is rendering, you'll have time to plant a tree.

;)

-Chilton

Johnny
03-28-2009, 11:13 AM
And, while your mini is rendering, you'll have time to plant a tree.

;)

-Chilton

that is a fact...renders that take me days or weeks prbly take 8-core owners hours or days.

J

Chilton
03-28-2009, 01:44 PM
This is a fascinating thread!

I know that the Mac Mini CPU is faster now, by a long shot, than my G5. But I wonder if the disparity between the Mac Mini and the Mac Pro is more dramatic than the high end vs. low end CPUs of the past.

-Chilton

Captain Obvious
03-28-2009, 05:04 PM
The CPUs they use in the Mac Pro are very good at rendering, what with the HyperThreading and all.

Johnny
03-29-2009, 05:17 AM
This is a fascinating thread!

I know that the Mac Mini CPU is faster now, by a long shot, than my G5. But I wonder if the disparity between the Mac Mini and the Mac Pro is more dramatic than the high end vs. low end CPUs of the past.

-Chilton


what I find very confusing in pursuit of an answer TO that most worthy question, is those new comparisons of Nehelem MPs with G5..some of them show the new machines to be appx 2 times as fast as the fastest G5.

Yet, I have heard others say that the very first MacIntel Mini was twice as fast as the G5.


and whichever the case is, I don't find "twice as fast as a G5" to be very compelling in 2009!

The G5, as we all remember, came out in mid-2003! In computer terms, that's the Ice Age!


J

Captain Obvious
03-29-2009, 05:13 PM
The new highest-end Mac Pro is probably five to ten times faster at rendering than the highest-end G5 ever released.

Weepul
03-29-2009, 05:51 PM
and whichever the case is, I don't find "twice as fast as a G5" to be very compelling in 2009!

The G5, as we all remember, came out in mid-2003! In computer terms, that's the Ice Age!

Keep in mind that you're comparing a several-thousand-dollar workstation to a sub-thousand-dollar low-end computer, and that comparison suddenly seems a lot more impressive. :hey: That's progress!

(Of course, I have doubts about whether that "twice as fast as the fastest G5" claim is even accurate.)


The new highest-end Mac Pro is probably five to ten times faster at rendering than the highest-end G5 ever released.
Based on the numbers that are available, I'd say about 5 times at best.

Captain Obvious
03-30-2009, 09:00 AM
Based on the numbers that are available, I'd say about 5 times at best.
Depending on which software you're looking at, of course. The benchmarks I was looking was not Lightwave, and it seems that it runs better on Intel's CPUs than on the G5, even when they've got reasonably comparable specs. I don't have any LW benchmarks to compare, so I don't know about that one...

Johnny
03-30-2009, 03:32 PM
Depending on which software you're looking at, of course. The benchmarks I was looking was not Lightwave, and it seems that it runs better on Intel's CPUs than on the G5, even when they've got reasonably comparable specs. I don't have any LW benchmarks to compare, so I don't know about that one...

that gets at the kernel of what an individual might want to know: given the same Lightwave render file, and 2 Mac computers, using whatever chips they use, which one's faster at a set of tasks that they can both perform, that ANYbody can perform on their rig right in front of their own eyeballs.

the benchmarks that I've seen make it difficult to cook down into a convenient "how does this affect ME?" form.

I don't play quake. I don't sit there ripping mp3 files all day.

I do lightwave, video, some photoshop, some 2-D apps, internet and mo def, I'm almost ALWAYS using the Finder.

then there are the ones which are supposed to be a good bench for any computer bcs they tax all aspects of a box, but then it matters which video card you have.

"3D graphics" isn't the same as "doing 3D" in the sense that Lightwavers do it..we create, model, texture, animate and render 3D.

We're not trying to save the cyber world from extra-galactic shape-shifters.

J

dwburman
03-30-2009, 04:08 PM
http://www.blanos.com/benchmark/


that gets at the kernel of what an individual might want to know: given the same Lightwave render file, and 2 Mac computers, using whatever chips they use, which one's faster at a set of tasks that they can both perform, that ANYbody can perform on their rig right in front of their own eyeballs.

toby
03-30-2009, 04:47 PM
one other thing is different, but which doesn't get much talk, is that a mini running balls to the wall consumes much less electricity than a Mac Pro does, with all 19 fans going..or however many fans...

there's no doubt that the Mac Pro is faster...but how much more juice per bit of work done does it require to get that speed?


J
What makes the difference to me is the interactivity. Working in lw on a mini must be punishingly slow compared to a mac pro, in ogl and test render time. This is a huge time difference, esp. when you add in the time saved not having to network. Setting up a net render on 1 8core is also an option. And one need not speng 6 g's either, i'm planning on getting a 2.8 for less than 2800$. I'm betting the power difference is not too bad, since it finishes 4-5x faster. Guess you'd want to consider idle time too, not sure which is better there.

Glendalough
03-30-2009, 04:53 PM
one other thing is different, but which doesn't get much talk, is that a mini running balls to the wall consumes much less electricity than a Mac Pro does, with all 19 fans going..or however many fans...

there's no doubt that the Mac Pro is faster...but how much more juice per bit of work done does it require to get that speed?


J

I don't think that the Mac Pros use that much electricity, judging by the amount of heat that they give off, the G5 may use more. I use them for heating my office in warmer weather and also keep a sharp eye on the electricity bill.

My intel mac mini (06) seems to get very hot in summer but it certainly was much faster rendering than the old G5 single which was all I could afford in 2003. 2 weeks after i bought it, the price came down almost 500 and annoyed cause this would have paid for the 2 processor model. Wondering is this going to happen to these new N. Macpros? Same thing happened to the 'cube' in the late 90's.

Lightwolf
03-30-2009, 05:00 PM
I'm betting the power difference is not too bad, since it finishes 4-5x faster. Guess you'd want to consider idle time too, not sure which is better there.
11.7W idle for a current generation mini, 29.7W at full load.

You'll be hard pressed to find a single socket workstation that uses less than 60W in idle and 250-300W at full load.

Cheers
Mike

Captain Obvious
03-30-2009, 06:02 PM
The old dual quad-core Mac Pro had a power consumption of up to 318 Watts while maxing out the CPUs, according to this Apple article (http://support.apple.com/kb/TA25202). That means the same power consumption as ten Mac minis, but probably lower performance when you're able to use it to its full advantage.

JonW
03-30-2009, 09:14 PM
http://www.intel.com/performance/workstation/xeon5000/dcc.htm?iid=perf_ws_body+5000_dcc


Configuration details
SPECapc* for Lightwave* 9.6
Benchmarks: SPECapc for Lightwave 3D 9.6

Baseline Configuration and Score on Benchmark:- Intel pre-production system with two Intel® Xeon® processor X5492 (3.4 GHz, 2x6MB L2 cache, 1600MHz system bus), 8GB memory (4x2GB FB DDR2-800), 74GB SATA 10000RPM HDD, NVIDIA* Quadro* FX3700, 175.51 GFX Driver, Microsoft Windows XP* Professional x64-Edition SP2 OS. Source: Intel internal testing as of March 2009. Score: Interactive 2.22, Render 6.15, Multitask 5.12.

New Configuration and Score on Benchmark:- HP Z800 system with two Intel® Xeon® processor W5580 (3.20 GHz, 8MB L3, 6.4 GT/s, quad-core, 130W TDP), 6 GB (6x1GB), NVIDIA Quadro* FX 4800, Microsoft Windows Vista* 64 OS. Source: Submitted to www.spec.org for review at score - interactive 2.92, render 7.58, multitask 5.38.

JonW
03-30-2009, 09:33 PM
http://www.bit-tech.net/hardware/cpus/2009/03/30/intel-xeon-w5580-nehalem-ep-review/9

LW 9.6 test.
W5580 66sec
X5482 89sec

Johnny
03-31-2009, 08:16 AM
11.7W idle for a current generation mini, 29.7W at full load.

You'll be hard pressed to find a single socket workstation that uses less than 60W in idle and 250-300W at full load.

Cheers
Mike

the comparison of those sets of idle/full load power figures are pretty dramatic...

I suppose there is also some power lost thru heat in the Mac mini's power brick, but prbly nowhere near even the lower figure in the Mac Pro.

J

Lightwolf
03-31-2009, 08:29 AM
the comparison of those sets of idle/full load power figures are pretty dramatic...

I suppose there is also some power lost thru heat in the Mac mini's power brick, but prbly nowhere near even the lower figure in the Mac Pro.

Well, the numbers are including the power brick...

As a comparison, 2009 Mac Pro, 8 Core, 2.26GHz (default config): 129W/258W (idle/load).

From the same source, Cinebench:
2009 Pro 8 Core: 20160
2008 8 [email protected]: 18661 (it also uses 30W more than the current 8 core under load).
current mini @2GHz: 4395

If my math is right then 5 minis would have a combined score of 21975 and use 148,5W.

Fun :D

Cheers,
Mike

Johnny
03-31-2009, 10:10 AM
I know that this is idle wishing, but it'd be nice if Apple would make a nehelem-based mini..4 cores @ 2.93 running in that little box booting to an externa 7200 rpm drive would be a nice piece of kit...

..and apple has mo def considered this, and the impact upon sales of spendier rigs.

I spoze heat would be an issue, but I think many would accept a larger form factor if it meant being able to have a quad-core Xeon in their mini.


J

Lightwolf
03-31-2009, 10:20 AM
I spoze heat would be an issue, but I think many would accept a larger form factor if it meant being able to have a quad-core Xeon in their mini.

130W of heat ain't no sucker... well, actually: would need quite a sucker ;)

The current quad in a smaller case (drop the Xeon, use an i7 which is pretty much the same silicon), 6 RAM slots, max. 1250US$ (which still makes it more expensive than PCs and allows for the Apple headroom)...

*sigh*

Cheers,
Mike

Johnny
03-31-2009, 10:24 AM
that config and that price work for me...I'd write a check this minute, if only...


this CPU power/watts/heat dissipation is reminding me of something..when I was in school, I was interested in aerodynamics and learned some principle which stated that, above a certain velocity, it takes a huge amount of power more to go a bit faster.

IOW, going twice as fast doesn't require merely twice the power, but multiples more of power required at the slower speed, even if the vehicle is shaped like a needle..

...or the thinner air found at higher altitudes..

the air can move out of the way only so fast...there must be a similar principle at work on the electron level in theses chips.

J

Lightwolf
03-31-2009, 10:25 AM
there must be a similar principle at work on the electron level in theses chips.
There is. The higher the clock the more energy you need to pump into them for a relatively small gain in performance.

Cheers,
Mike

Johnny
03-31-2009, 10:38 AM
what is also true, and I saw this back in my print days, is that, as computer power increases, it is almost immediately slurped up by users creating larger and/or more computationally-intensive work.

the new mac pros or xeon-based PCs with RAM to burn will encourage users to create scenes in 3D which will take those machines to their limits, rather than keeping models and scenes modest, using time-saving rendering and modelling/shading techniques.

for my 2 cents, there's a lot to be said for not taking for granted or squandering that new computer power you just bought, and concentrating on the User half of the equation in making files which can render much faster on the much faster hardware..

but I know the demand for everything is "always more."

J

toby
03-31-2009, 01:39 PM
what is also true, and I saw this back in my print days, is that, as computer power increases, it is almost immediately slurped up by users creating larger and/or more computationally-intensive work.

the new mac pros or xeon-based PCs with RAM to burn will encourage users to create scenes in 3D which will take those machines to their limits, rather than keeping models and scenes modest, using time-saving rendering and modelling/shading techniques.

for my 2 cents, there's a lot to be said for not taking for granted or squandering that new computer power you just bought, and concentrating on the User half of the equation in making files which can render much faster on the much faster hardware..

but I know the demand for everything is "always more."

J
EXACTLY - then complian that it takes a 'whopping' 15 min. to render 12 bounce gi with hv and fibrefx!

Captain Obvious
03-31-2009, 03:14 PM
Basically, the warmer a CPU gets, the higher the resistance is. The higher the resistance, the more power you need for the same "throughput," which in turn increases the temperature even more. It's kind of a downward spiral.



what is also true, and I saw this back in my print days, is that, as computer power increases, it is almost immediately slurped up by users creating larger and/or more computationally-intensive work.
Ain't that the truth. It used to be that all my scenes fit into 3GB of RAM, because that's all we had. Now I have scenes that can barely load with 16GB...

Lightwolf
03-31-2009, 03:17 PM
It used to be that all my scenes fit into 3GB of RAM, because that's all we had.
3GB? Young whippersnapper... 4MB! (quickly expanded to 24 MB though).

Cheers,
Mike

Johnny
03-31-2009, 03:19 PM
Ain't that the truth. It used to be that all my scenes fit into 3GB of RAM, because that's all we had. Now I have scenes that can barely load with 16GB...

why is that? are your clients demanding more whiz-bang?

are there things you would do to reduce the size and resource demands of your scenes?

or, is it the case in which, since the computer with more RAM and power will accept more demanding scenes, you tend to create scenes which are ever more demanding?

in an earlier day, computers just COULDN'T do more than XYZ, so you took that reality.

J

Captain Obvious
03-31-2009, 06:58 PM
3GB? Young whippersnapper... 4MB! (quickly expanded to 24 MB though).
I've only been doing 3D for... uh... five years, I guess. Four, maybe. Maxing out a 32-bit machine with RAM has been achievable for as long as I've been using Lightwave. Of course, our first machine only had 4MB, but I never did 3D on it.



why is that? are your clients demanding more whiz-bang?

are there things you would do to reduce the size and resource demands of your scenes?
I'm but one stage in a pipeline, I can't control the other parts. If someone gives me a 40 million polygon mesh with 10 gigabytes of images, then that's what I'm stuck with.

Johnny
03-31-2009, 07:05 PM
I'm but one stage in a pipeline, I can't control the other parts. If someone gives me a 40 million polygon mesh with 10 gigabytes of images, then that's what I'm stuck with.



jeebus! who's your boss? Brad Bird??


J

JonW
04-02-2009, 01:51 AM
Abit more info on W5580 v X5492 & E5450, no LW but it should help.

http://techreport.com/articles.x/16656/10

JonW
04-02-2009, 02:01 AM
[QUOTE=Lightwolf;860839]3GB? Young whippersnapper... 4MB! (quickly expanded to 24 MB though).

My first Mac was a 7300 with 272 mb ram 2 x 8mb + 4 x 64mb (64mb cards were about AU$700 ea) & 2 x 2 mb of video ram. 2 x 2 gb HD

Used it for doing photo montages, it was a very good beast!

toby
04-02-2009, 03:28 AM
My first was a handmedown - a mac se with 2mb ram upgraded to 4. Not gigs, megs.
Great for 2 bit b&w graphics and chess games...

Johnny
04-02-2009, 08:26 AM
heh...I seems my first machine was a ferrari compared to you guys's machines...

it was a Mac IIsi with 16MB of RAM...20 mhz machine, and when I put that NuBus card in, floating point calcs speeded way up.

don't think 3D was really possible on that machine..

that didn't happen until I had the 7600 with 24 whoppin' MB of RAM...I got Ray Dream as a freebie when I bought Painter 3D...

I remember those first exciting moments when I made a chrome airplane, watching that image appear 1 line of pixels at a time..

heh...


J

Chilton
04-02-2009, 08:35 AM
Hi,



it was a Mac IIsi with 16MB of RAM...20 mhz machine, and when I put that NuBus card in, floating point calcs speeded way up.

don't think 3D was really possible on that machine..


Sure it was! There was an app by Silicon Beach called Super3D. And they did a color version of it right before Aldus bought them. And it did actually kinda kick ***. It had several ideas in it that I haven't seen in 3D software since then. I might need to drag out my IIsi and play with that again...

-Chilton

Captain Obvious
04-02-2009, 10:22 AM
It had several ideas in it that I haven't seen in 3D software since then.
Now I'm curious. Did it do differently?

Chilton
04-02-2009, 10:59 AM
Hi Capt'n,

There were a few things that stood out.

Keep in mind this was around the time Bruce Tognazzini (AskTog.com (http://www.AskTog.com)) published one of his first books on human-computer interaction, and he brought up the arguments for why they ditched the multi button mouse found on the Apple II, and went with a single button for the Mac.

So UI designers at that time were struggling with some concepts we take for granted now. One of them was mouse navigation in 3D space.

Context sensitive manipulation of points and faces was an obvious one, though you couldn't make really high poly objects, so it's nothing like it is today. Clicking off of an object, in space, would deselect.

Keyboard chording wasn't as popular then as it is now, so they had a rotating scroll wheel thing that would pop up for rotating objects in 3D space. Think of a normal scrollbar circling around an object. Only it was gigantic and only lines. When you moved it though, it would behave like a scrollbar.

It had multiple scrollbars on the viewports. 3D viewports obviously needed to move in more than just up/down and left/right. The two scrollbars found in most Mac apps at that time just didn't cut it. So there was a third scroll bar for navigation along the z axis.

Lastly, the concept of a scroll bar works great in finite dimensions, like on a document. But 3D space is infinite ( or at least limited to the size of an integer at that time ;) ) so the scroll bar thumbs sat at the center of the scroll track, and as you moved it up or down, you were increasing your velocity as you (the camera) travelled along that axis.

Neat tricks.

-Chilton

Johnny
04-02-2009, 11:02 AM
that is pretty interesting, esp. that scrollbar device wrapping around an object.

Wasn't LW first released as a freebie with Toaster?

LW's methods for scooting around a scene or model seem to me to reside on the instinctual level, and other apps imitate the LW way...

was it always that way? I got my copy of LW back in 2002...


J

Scazzino
04-02-2009, 07:08 PM
Hi,



Sure it was! There was an app by Silicon Beach called Super3D. And they did a color version of it right before Aldus bought them. And it did actually kinda kick ***. It had several ideas in it that I haven't seen in 3D software since then. I might need to drag out my IIsi and play with that again...

-Chilton

Ah yes... I remember Super3D... was there another one called Easy3D? Both black and white on my old 1BM Mac Plus. My first color 3D program on the Mac was Swivel3D for modeling and Mac Renderman for rendering... on the "Wicked Fast" Mac II fx... I had to hand code the shadows in Mac Renderman's RIB language... ;)