[sodamnregistered]

Got my new dual800 in last week and I've finally gotten a chance to use it some.

Impressions? A pretty fast Mac, or is it?



First, what I care most about: I have a C4d render farm comprised of a PIII850, dualG4/450, Athlon 1100, 9500 G3/400 and the dual 800, and some more junk which I did not fire up for this experiment. The render farm was controlled from a Pismo 500.

The Athlon took my dual 800 to school- nearly 50% more rendered frames. The Athlon was rendering with power and authority.

Next, I found this page: at xlr8yourmac.

And this was my result:



I know GaugePro is old, but most folks think that the memory test is still valid since RAM has not changed much in the last couple of years. See the web page.

What gives?

The ram came from The Chip Merchant and from what everybody says, CL3 or CL2 doesnt not make a measureable difference, but this result is half of my G4/450mp and is about what my Pismo 500 is capable of.

The Chip Merchant claimed no knowledge of suitable 2-2-2 or CL2 512MB RAM for the Quicksilver models, and their little (3) DIMM CL3 1.5GB kit was decent at $250. Hmm.

All I know is Athlons have it goin' on and my Mac is the pretty computer. Kinda like a friggin Camaro don't you think? All show and no go?

Proof is in the pudding, Athlon put the frames out, my dual 800 was marginally faster than my dual G4/450.

PS here's a sample frame from the 10sec movie that was rendered.



Just the same thing over and over. Pretty fair for all computers- all visible lights.

[TNproud2b]

Don't feel bad. I have an Athlon 1.33 (266DDR) running slightly overclocked at 1.43GHz on a crappy maxxed-out ASUS A7-E KT133 chipset (200MHz FSB)...and it absolutely hammers every machine I've ever owned in terms of raw performance. I've owned some mighty fast stuff, like dual big-cache Xeons, dual P3's @ 900MHz, and a dozen others. I'm impressed by my $100 investment in a used Athlon 1.33 and a display sample el cheapo motherboard. I did splurge on the RAM, though by using Kingmax TinyBGA which runs happily below spec at 147MHz 2,2,2 - got it used for $24/shipped.

I've seen what dual Athlons @ 1.6GHz can do, and it's staggering. You would need to spend a ton of money to get something faster...
That's with MMX and 3Dnow(lol), only.

Raw power. Seemingly regardless of how sloppily the app is coded.

Can you imagine what SSE instructions could do? Athlon MP to the rescue.


I dunno. My last AMD machine was a k6-2/400 which didn't impress me at all. In fact, I haven't touched an AMD product since then, maybe 2 years or so.

Scary indeed. You are not alone in feeling like the Athlon makes your fast computers look slow.

[Leonis]

Just wondering what version of Cinema you are using.

If you are using version 7 go to Cinema's user forum http://www.postforum.com/forums/list.php?f=6

and you will see lots of people complaining about the rendering speed in version 7.

In some cases version 6 renders a lot FASTER and in other cases version 7 renders slightly faster

here are the benchmark I made a while ago:

Cinebench 2000 V1.0 Performance
************************************************** **

Processor : AMD K7 950Mhz
Number of CPUs : 1
Physical Memory : 512MB
Operating System : Win 98 SE

Graphic Card : ATI Rage II+ 4MB
Resolution : 1024 x 768
Color Depth : 32-bit

************************************************** **

Shading (CINEMA 4D) : 8.49 CB
Shading (OpenGL) : 5.66 CB
Raytracing (Single CPU): 12.73 CB
Raytracing (Multiple CPU): --- CB

OpenGL Shading is 1.50 times slower than CINEMA 4D Shading!


================================================== ===========

000 V1.0 Performance
************************************************** **

Processor : Power PC G4/500 DP
Number of CPUs : 1
Physical Memory : 1.5GB
Operating System : MacOS 9.2.1

Graphic Card : ATI Radeon 32MB AGP
Resolution : 1280x1024
Color Depth : 32-bit

************************************************** **

Shading (CINEMA 4D) : 4.74 CB
Shading (OpenGL) : 6.22 CB
Raytracing (Single CPU): 5.58 CB
Raytracing (Multiple CPU): 11.44 CB

OpenGL Shading is 1.31 times faster than CINEMA 4D Shading!
2 CPUs are 2.05 times faster than 1 CPU !

************************************************** **


Athlon is right now the fastest and cheapest processor out there for 3D rendering anyway....why not just get one of these just for rendering and then you have the best of both world

[ 09-02-2001: Message edited by: Leonis ]

I thought the notion was to use the G4 for design and other hardware for rendering. I believe that is Apple's approach to this issue. They know their chips suck, but does it really matter when you're building the scene, since you're not rendering anyway.

Apple wants to be the front end interface for 3D design. I don't think they care or can compete on the rendering farm aspect.

[l008com]

Well obviously if you dual 800/133 is going jsut as fast as your dual 450/100, then something ELSE is wrong. I mean, the 800 should be TWICE as fast.

[sodamnregistered]

TNpround2b:

I did spend a ton if money, and it's not faster!

Leonis:

C4d 7, except all machines were Net rendering C4d 7, so the playing field was fair. Also, there are no textures or anything complicated. Just visible lights. The render time itself was alright, it just that the Athlon was completely faster.

Another concern, I was getting "Out of Memory" on my Macs until I spaced the lights out some more. The Macs have 1.5GB and 1GB of RAM each. The PIII850 has 512MB and the Athlon1100 has 756MB. Not a peep from the W2k machines during rendering, plus the RAM useage seemed to level out on both W2k machines at about 290MB.

Is W2k memory management _that_ superior to MacOS9? I know OSX is a whole other ballgame. I can't farm render with OSX yet anyhow.


Something:

Why wouldn't I (or any other logical cold hearted ex-Mac user) just move the whole shebang to Athlons? It's not like the C4d interface is pure Mac anyhow.

I can't believe Apple is happy selling me a dual 800 to model on, just so I can go buy 6 Athlons to render on.

1008cpm:

The dual 800 was faster than the dual 450. But not twice, more like 50%.

----

To everybody, what about the paltry 148.7 MB/sec memory performance? I know the lights in this example render were a little heavy on memory useage. Don't dual Athlons move over 1 GB/sec in these memory throughput tests? 150MB is so not even close.

Also, I thought that the G3 and G4 chips were killers, or at least solid for FPU stuff? Why this huge gap?

Plus it's a single Athlon 1100 puttin' the rap on (2) G4/800 chips. Where was this test at MWNY?

If you ask me, AMD has the real Megahertz problem. Their chips stomp all comers, but Intel markets at 2GHz. I hate Intel more than ever now.

----

I don't hate my Mac, I do some video stuff as well and the DVD-R will pay for the whole thing when it's all said and done, but, I almost bought a dual 1.2 Athlon instead of this dual 800 Mac. This only justifies that line of thinking.

[Codename]

To everybody, what about the paltry 148.7 MB/sec memory performance?

Guauge PRO's last revision was released before the advent of the PPC7450. It does proprietary diagnostics of moving memory throughput we know nothing about. It is not a valid test.

Maximum main memory bandwidth without the use of Altivec on QuickSilver G4's is 1GB/sec(64bits * 133.33(...)MHz).

Also, I thought that the G3 and G4 chips were killers, or at least solid for FPU stuff? Why this huge gap?

Bottom line, Altivec.

For example, a 1.2GHz Athlon decoding RC5 keys will achieve about 6MKeys/sec. A dual 800MHz G4 achieves over 14.5MKeys/sec. This is one case of several where the G4 utterly stomps the competition.

C4d has very limited use of Altivec in rendering.

[ 09-02-2001: Message edited by: Codename ]

[Amorph]

Yes, OS 9's memory management is that bad. Win2K's is in another league entirely, and so is OS X's. On top of that, OS 9's file system code is dog slow, which further impacts the (already lame) VM system.

OS 9 is also a horrible choice for showing off a machine whose performance depends on multiple processors.

Run OS X 10.1 on that baby when it comes out, and if the single-proc Athlon still wipes the floor with it then you'll know that Apple has a serious problem - or your new G4 has a serious problem, and I don't think anyone would blame you for switching to the Athlon machines full time.

Oh, and you can send me the G4.

[TNproud2b]

Did you mention memory bandwidth?

dual Athlons @ 1.6GHz - who needs RDRAM when PC2400 DDR looks this good?

[Leonis]

AMD's K7 (Athlon)
Super cheap. Strong FPU makes it the best choice for 3D rendering

Motorola's G4
Relatively expensive. Good on media compression, DSP, 2D. But absolutely sucks on 3D rendering

Intel's P4
Nothing really special about it beside Intel's lame marketing and super high Mhz

Anyone want to add your opinion to make this list valid?

[TNproud2b]

Cinebench 2000 V1.0 Performance
************************************************** **

Tester : Athlon 1.33 @ 1.43

Processor : K7 Thunderbird
Number of CPUs : 1
Physical Memory : 128MB
Operating System : Win98SE

Graphic Card : V3/3000 AGP
Resolution : 800x600
Color Depth : 32bit

************************************************** **

Shading (CINEMA 4D) : 13.52 CB
Shading (OpenGL) : 12.92 CB
Raytracing (Single CPU): 19.26 CB
Raytracing (Multiple CPU): --- CB

OpenGL Shading is 1.05 times slower than CINEMA 4D Shading!

************************************************** **


I think I have an OpenGL problem. All my friends are getting 17-19
on OpenGL shading using similar machines. Wonder if it's because I use a
Voodoo3 and they all have GF3s and Radeons. Hmm.

Used Athlon 1.33 = $40
Floor sample ASUS A7-E motherboard = $55
128MB Kingmax Tiny BGA (150MHz 2,2,2)= $24
used V3/3000 = $38
New Western Digital 30GB = $79
Used Pioneer 10X DVD = $40
new SONY 12x8x32 CDRW = $120
cheap ATX case = $35
10/100 ethernet = $9

Total = $440

[Leonis]

My K7 box costs $600 CDN....but only with 950Mhz K7 (oc'ed to 1.1Ghz a few weeks ago but haven't benchmarked it yet), and 512MB RAM. The CD-ROM, CD-RW, Zip, floppy drive are carried over from the dead Pentium 200.....

BTW, Vodoo has really bad OpenGL support......as far as I know

[pherein]

I use a dp800 for work as well, and can not see any validity in the test you ran sodamn. If you ran C4d cocoa version (in OSX) or a version that supported the motorola AV, your test might be worth a look, or if gauge pro was cocoa and supported the 7540.
I know my dp800 is faster than the intel dual p3/1GHz or any of the anthlons I use at work in ps6, actually allot faster.
But it really does not mean much because my testing method is probably flawed. As flawed as about 90% of all the benchmarks out there. If I remember college correctly, the source and method of any scientific study is much more important than the result, which is why I do not give home or commercial benchtesters much validity.
Waiting for OS X to be version 10.1 (so that it is clearly out of the beta stage), and then waiting for a C4d to produce a cocoa version that supports AVec will give you a better platform to test on I would think. I my self am very please with the speed increase over the g4/dp533.

[TNproud2b]

You miss the point.

Any processor would be fast if it only ran apps that took advantage of hardcoded instructions.

The beauty of the Athlon is that it was designed to be a good all-around performer - not just when the planets come into alignment and OSX is fixed and the particular app you want to run is finally optimized to run on your platform.

I see it this way. One company believes it's better to minimize the performance penalty incurred by badly written applications - while the other company believes that apps should be optimized to run on their hardware. Both ways will get good performance...but one gives it to you now instead of tomorrow.

*edited to add that even the P4 is blindingly fast when running apps that utilize SSE2 instructions (for the few apps that use SSE2, that is)*

[ 09-03-2001: Message edited by: TNproud2b ]

[crh]

I've been testing Maya 3.0 for OS X (pre-R v's) on a DP 533 (1.5 gig RAM, AGP Radeon, UW SCSI RAID, yada yada) for some time, and can tell you, at least with Maya, that the story's very different using OS X (BOTH 10.0.4 and 10.1's later builds). I know several folks who've worked on the Maya OS X port, and have extensive experience rendering with Maya, Cinema, LightWave, Hash, Form�Z, ElAS, etc. on G4's, P3's, P4's, and Athlons. They agree that Athlons rock renderwise. However, they say that OS X gives back what 9.1 has witheld; i.e.- the memory management and multithreaded speed that allow (dual) G4 PowerMacs to run past and, in many cases, plain stomp Athons IN RENDERING. Don't ask for published reports 'cause there ain't none- just anecdotal conversation, and that's what I'm presenting those claims as. Take it with a grain of salt- I'm sure they have no idea what they're talking about .

That said, I'd like to see an Athlon Palomino ddr'd duallie crunch frames in all the above apps and see how a DP 800 running OS X 10.1 stacks up.

I personally hope that, eventually, either Apple buys out Moto's DT chip division or goes to IBM soon, because AMD's not standing still, and competition can only improve the breed when both breeds are still 'breeding', if you know what i mean...

-crh

Occam's Razor- never dull!

[ 09-03-2001: Message edited by: crh ]

Although the AMD Athlon is very fast, I should know...I have dual 1.2GHZ MP's, SiSoft Sandra as a benchmark is misleading, unaccurate, and can be faked without much effort. I'am not saying you're benchmarks are, but take other benchmarks with a grain of salt.
http://www.overclockers.com/articles444/

[iCartman]

Well obviously if you dual 800/133 is going jsut as fast as your dual 450/100, then something ELSE is wrong. I mean, the 800 should be TWICE as fast.

Not at all. The 7450 chip just isn't as efficient as the older chips. Even apple claims the DP800 is only 20% faster than the DP500 in FCP.

[crh]

I heard that, Yea yea- 'there are lies, damned lies, and statistics'...

One more thing about the 411 I paraphrased: the results these folks accumulated were for internal corporate tech comparison/research purposes. I'm familiar with their research methods and have found them to use sound, contextually correct methods of R&D over the years that produced accurate specs. In their case, they'd be wasting their time doing it at all if they "faked" their own internal research, although I'm well aware of those at some corporations and academic institutions that do give certain projects a little "Kentucky Windage" (to insure funding on pet projects, etc.).

Since they don't have axes to grind per se about which platform/chip/app runs fastest, but instead are attempting to gather info about relative "field" rendering (pardon the pun) performance comparisons, etc. as well as absolute lab specs, I put stock in what they tell me. No one else should, necessarily. I just put it out there as info I believe to be far more accurate than much of the "published" stuff (such as some reports using SiSoft Sandra) I've seen over the years and up to now.

But since I'm often prone to daydreaming about a PowerMac quad G5 DDR'd six PCI slot tower for $2k this year, why believe a partison, right?

I guess I'll have to toddle along with the DP 533 awhile longer...

[ 09-03-2001: Message edited by: crh ]

[The Ancient One]

iCartman wrote:

Not at all. The 7450 chip just isn't as efficient as the older chips. Even apple claims the DP800 is only 20% faster than the DP500 in FCP.

Uhhh... the real-world tests tell a more complex story. Check out http://www.barefeats.com/pm01.html - less than 20% faster than a 533DP on a Photoshop suite, but 37% faster on CineBench 2000 and 63% faster (!) on Alivec Fractal Carbon.

We all agree that OSX rules, no matter what.

The CPU issue is hugely debatable. No doubt Apple is doing something about this. Maybe Apple is working on an OSX-of-CPUs ... the near perfect CPU with great performance.

Asta-la-babe!

Macs needs an Athlon with Altivec!! ouch! that sucker would burn rubber!

[PMG4DP]

Don't quote me on this but,

According to Matt Chrisopher(Lead programmer for the mac version of lightwave) the MacOS X Math Lib's are about 2 to 10 times slower then the ones on OS 9. The rendering in OS X with lightwave is really slow. They even went so far to write their own lib's so they can get decent render times. I'm not sure if this goes for Maya and C4D but it could be the problem.

[Turbocharger]

Of course that's the problem. OS X isn't even really completed, muchless the libs. It's going to take awhile. Meanwhile, there's a smorgashbord of compilers with libs to choose from for the wintel machines, including some from Intel.

[Nimisys]

Originally posted by <ultimateChip>:
<STRONG>Macs needs an Athlon with Altivec!! ouch! that sucker would burn rubber!

</STRONG>

scew altevec, get the next gen SUN SIMD unit... yumm 256bit goodness

but i agree the athlon needs better SIMD, adding SSE and SSE2 (in the hammer) helps but remember that 3Dnow and SSE came before AlteVEc and as such are a gen behind. i t needs the ability to run single 128bit calculations instead of daul 64bit... better yet get a 256bit unit capable of daul 128bit instructions and a far more important 4 and 8 sequential 64bit and 32bit instructuions.

[Turbocharger]

You CAN'T compare SSE/2 to Altivec, they're two totally different technologies.

[Leonis]

Originally posted by crh:
<STRONG>I've been testing Maya 3.0 for OS X (pre-R v's) on a DP 533 (1.5 gig RAM, AGP Radeon, UW SCSI RAID, yada yada) for some time, and can tell you, at least with Maya, that the story's very different using OS X (BOTH 10.0.4 and 10.1's later builds). I know several folks who've worked on the Maya OS X port, and have extensive experience rendering with Maya, Cinema, LightWave, Hash, Form�Z, ElAS, etc. on G4's, P3's, P4's, and Athlons. They agree that Athlons rock renderwise. However, they say that OS X gives back what 9.1 has witheld; i.e.- the memory management and multithreaded speed that allow (dual) G4 PowerMacs to run past and, in many cases, plain stomp Athons IN RENDERING. .............
-crh
</STRONG>

Really? I have a DP 500 and I run the Classic version of Cinema under OS9 and the Carbon verison under OSX. Render the same file. No difference.

However, the app launches twice as fast under OSX. The cpu monitor says both processors go 100% when launching Cinema

[ 09-03-2001: Message edited by: Leonis ]

[crh]

Leonis-

Don't know; could be several things- mathlibs yet to be optimized by Maxon, OS X, or that the Carbon Port's rendering engine is yet to be tuned. Dunno- ya got me.

Let me re-phrase, here, all: I didn't mean to imply that all apps for Mac 3D rendering engines are screaming past both Athlons and their OS 9.x versions at present, but that, from what I've been told, some are, and that it looks very likely all will by year's end.

Maya, btw, cranks very well indeed. But, it too will be on an up ramp as OS X matures.

-crh

[Nimisys]

Originally posted by Turbocharger:
<STRONG>You CAN'T compare SSE/2 to Altivec, they're two totally different technologies.</STRONG>

bullsh1t... they are both SIMD technologies and AlteVEc being the newer/more advanced of them. Go read at ARS in there CPU & Praxis Theorum section... the article is 3 1/2 SIMD technologies

[real]

You know whats weird that fact that I hear people saying we cant compare this with that because it different technology. !#*(. Im I wroing in saying that the whole mhz battle is comparing different tech to the AMD's chips the G4's the intels. If you can't compare the suns and the SGi machines to AMD or G4's, who says( those machines scream they blow all of us out of the water. So if you really want the fastest nachine for 3D rendering get an SGI or SUN.Stop the complaining and get the computer you want it sounds like you want a AMD speeds with the mac OS WELL you can't have it. either you run windows or you run OS 9 or OS X. get over it. high visual effects houses buy the machines they need for the job.And so should you. I feel that that I can get my work done with time to spare on my DP 800 and I get to run the Best OS ever and thats only going to get better. If they render tons of 3d then it's PC's (windows, or SGI'S, if it's 2D then its mac. If they are running Flame,Fire, Inferno, smoke then they are runnning SGI's, with more killer stuff than must of us at home will ever get to play with at home.I dont see alot of people running SGI's to work in word or surf the web.
Sorry about the novel but this type of thing is just crazy
OH! yeah one more thing
MY G4 is better than your Windows boxes.
real

[Randycat2001]

Originally posted by sodamnregistered:
<STRONG>TNpround2b:
Another concern, I was getting "Out of Memory" on my Macs until I spaced the lights out some more. The Macs have 1.5GB and 1GB of RAM each. The PIII850 has 512MB and the Athlon1100 has 756MB. Not a peep from the W2k machines during rendering, plus the RAM useage seemed to level out on both W2k machines at about 290MB.

Is W2k memory management _that_ superior to MacOS9? I know OSX is a whole other ballgame. I can't farm render with OSX yet anyhow.</STRONG>

I hope you bothered to up the memory allocation on the rendering app on your Mac. If you are technically-inclined to carry out all of these benchmarks, you should be similarly capable of typing in a few numbers in your Get Info window. So what if W2k "does this for you"? That point is not exactly a new revelation to us- big deal. Get over it and move on to optimize your tools to do the tasks you ask of them. Monitor how much RAM your app will soak up and then allocate that amount (and maybe a bit more) in the Get Info window. Simple as that. If you've bothered to get 1.5 GB of RAM, might as well put it to work, wouldn't you agree? Look on the bright side- if you want to give a particular app 1 GB of RAM to work with, you can do exactly that in Mac OS9 (and others)- none of this paging-to-disk crap because the OS "knows better" than you, all the while 75% of RAM is sitting there doing nothing. I run a Win2k setup at work with 1 GB of RAM, and it is just as fallible as it is "intelligent" about memory management when you run it hard enough. It's all a matter of how well you set up your tools and how you dish out the tasks.

[Leonis]

I did talk to couple of Cinema experts regarding memory issue with all the dual processor machines (not just G4, but also Dual PIII, and Dual Athlon)

They said on a dual machine the total amount of memory will split into half to each processor. In other words, if you machine has 2 processors and 1GB of RAM. Each processor will only have 512MB max. If using a Quad-cpu machine, the max amount of RAM each processor has will be 256MB...and so on.....

So when getting a DP or QP machines make sure to have TONS of RAM

[sodamnregistered]

Loads and loads of arrogant crap from Randycat2001:
<STRONG>

I hope you bothered to up the memory allocation on the rendering app on your Mac. If you are technically-inclined to carry out all of these benchmarks, you should be similarly capable of typing in a few numbers in your Get Info window. So what if W2k "does this for you"? That point is not exactly a new revelation to us- big deal. Get over it and move on to optimize your tools to do the tasks you ask of them. Monitor how much RAM your app will soak up and then allocate that amount (and maybe a bit more) in the Get Info window. Simple as that. If you've bothered to get 1.5 GB of RAM, might as well put it to work, wouldn't you agree? Look on the bright side- if you want to give a particular app 1 GB of RAM to work with, you can do exactly that in Mac OS9 (and others)- none of this paging-to-disk crap because the OS "knows better" than you, all the while 75% of RAM is sitting there doing nothing. &lt;"W2k sux too" crap snipped&gt; It's all a matter of how well you set up your tools and how you dish out the tasks.</STRONG>

Oh, I only had 640k assigned to C4d. I better run my little test again!

I did not run any benchmarks. I sent a job to my farm, and noticed that my W2k machines with half the ram of my Macs ran everything without complaint. I noticed that the Athlon was ruling class, and that the case on my dual800 and very pretty.

The 1.5GB of RAM was certainly not bought to impress you, and god help the Mac user that can't set RAM useage by now. I'm not sure one could even surf the web with the default ram settings that come with a browser.

Take a nap child, and when you have something _useful_ to contribute, give it another go.

I mean if you are socially-inclined enough to make such a snide post, you should be similarly capable of realizing that that there was nothing relevatory in your words. Simple as that, don't you think? Aww, get over it. Big deal, right?

Oh boy, here we go again... I'm sure I'm not the only one here to recall that this very same subject has been rehashed to death in a similar thread very recently. Is there really a need to start this again?

Really "sodamnregistered", you make a post complaining about poor performance on a fast machine, then you get upset when someone questions if your machine was configured properly?!?! If you're honestly looking for tips and suggestions, then post them on Cinema 4D forum. This forum is little more than a "mine is better than yours" type of forum. Considering this topic has been rehashed over and over, you come off looking like an AMD zealot and Mac basher. If that's not the case, then you truly should seek helpful advice in a technical forum on how to configure your machine properly.

Steve