Thanks for the info on some of the programming stuff. Brief but informative. That’s how i like things anyway. More than that and I would have dozed off to sleep again. It’s first thing in the morning for me after all. I was kind of aware of the limitations, but that clears a lot of things up. Cool initiative to get programming started by nVidia too, with the LAME encoding. Rewards are always a good way to get something going.

Would you actually recommend ATI for max then? I’ve just always been much happier with the nVidia drivers. And since I know how to optimize them already I like to stick to it, but I’m always open to check something else out. I’d have to also check with AfterEffects, Rhino, and ProE but I’d be interested if ATI can do a better job.

I still think something is wrong with your system that is getting a 90x speed difference. I’ve talked to a few friends. They haven’t seen a 90x speed increase from computers 5 years ago and computers now. Also, none of us have ever seen max render in 1 second, just because it takes about 1 second for preprocessing, another couple seconds to save, and usually 1 or 2 seconds to render. I don’t think I’ve ever seen a render under 2 seconds, even on a blank scene. What renderer are you using? I’m very interested to try it. Anyway, my point was that I have a single core laptop at home at half the Ghz at work where I have dual processor. The renderings at work are only 3 or 4 times faster. That’s with 2 times the Ghz, hyperthreading, and 2 processors. I did that test using Mental Ray with GI and about 400k polys.
That’s why I have a really hard time believing that you are getting 90x one processors that are the same speed with just a different architecture. I’d guess 4-16x MAX if things are set up correctly.

Would you actually recommend ATI for max then?

I really would , if not for one small detail. Normaly I do recomend it, but i have to say this the one bad thing.

The problem oviously isn’t the latest and by the looks of it, I don’t see it getting solved any time soon. Pretty lousy on ATi part, but what can you do…

They haven’t seen a 90x speed increase from computers 5 years ago and computers now. Also, none of us have ever seen max render in 1 second, just because it takes about 1 second for preprocessing, another couple seconds to save, and usually 1 or 2 seconds to render.

From the one sec render time it should be obvious it was a test render. I use mental ray with final gathering only (no gi). Its extremely annoyingly long on the old P4. 1min 30sec is no exaggeration.

don’t think I’ve ever seen a render under 2 seconds, even on a blank scene.

My old p3 rendered blank scenes in 0sec.
Today, my blank-to-simple-test scenes render from 0 to 2 sec. Big previews take up to 10sec. Again, if u have massive geometry, preprocessing(converting to MR) alone will eat up your time. (this fact made me look into XSI as it has MR as the default renderer so no geometry conversion time… but im getting sidetracked here)

That’s with 2 times the Ghz, hyperthreading, and 2 processors. I did that test using Mental Ray with GI and about 400k polys.

That’s why I have a really hard time believing that you are getting 90x one processors that are the same speed with just a different architecture. I’d guess 4-16x MAX if things are set up correctly.

If you’d looked up the mentioned P4 Prescott 320, you would have noticed it’s actually a Celeron class CPU with 256k L2 cache. C2Q 6600 has 8mb. Than if you look that I had 1gb of ram and that I use 8gb now… All rest being equal (which it isn’t)… It adds up. I don’t know exactly how, but the numbers don’t lie and they’re consistent too.

To be fair, my old AMD 2500+ renders the same scene in ~30-40sec. So its only 30 to 40 times slower (I don’t remember the exact number in the tests). Even so, the new is much more than 4x2 times faster.

And don’t get me wrong, because it might seem like I’m complaining. I’m just happy it is that way ;)

EoDEo

Ideas drift like petals on the wind. You have only to lift your face to the breeze.

Maybe that’s why I don’t see that much of a difference. Our renders are nearly always huge amounts of data, so processors aren’t always the bottleneck. It probably has to cache a lot of the data to the HD or something, which can really slow things down. We’d go 64bit and add mem, but we’re stuck with dells (and you have to buy a 300 dollar fan , no joke, to put in more than 4gigs of memory), and some of our software doesn’t work at all with 64bit in this corporate environment.

I’ll definitely see if we can get a really good gamer card and compare, but I remain a little skeptical with their compatibility with all the different software we use. Although, our current cards aren’t that great either. lol
We ALWAYS get overlay problems.

I found this link (http://www.tomshardware.com/de/G...stberichte-240063-22.html), and this table () shows impressive data: under SPECViewPerf 10 benchmark, the 3870x2 score was about twice of the gtx280. I bought cg hardware for more than 15 years. So I like a lot the nv initiatives, like the future MentalRAY on hardware one - truly gold for cg professionals. But those numbers show very deceptive performance for the gtx280, and we all know that it is a card with a very impressive gpu. If anyone could post links or info to viewperf 10 and other benchmarks with the gtx280 I will be very glad of. What’s the point? Well, nv could boost a lot their selling numbers if leave this professional versus gaming boards nonsense. Today, all the technological differences between the “gaming” and the “quadros” are vaporware - and we all know this. The results are clear: nv is losing a lot of key and loyal customers - I’m one of them.

Zilon X 02 September 2008 04:17 PM

I found this link (http://www.tomshardware.com/de/G...stberichte-240063-22.html), and this table () shows impressive data: under SPECViewPerf 10 benchmark, the 3870x2 score was about twice of the gtx280. I bought cg hardware for more than 15 years. So I like a lot the nv initiatives, like the future MentalRAY on hardware one - truly gold for cg professionals. But those numbers show very deceptive performance for the gtx280, and we all know that it is a card with a very impressive gpu. If anyone could post links or info to viewperf 10 and other benchmarks with the gtx280 I will be very glad of. What’s the point? Well, nv could boost a lot their selling numbers if leave this professional versus gaming boards nonsense. Today, all the technological differences between the “gaming” and the “quadros” are vaporware - and we all know this. The results are clear: nv is losing a lot of key and loyal customers - I’m one of them.

You need to realize couple of things here:
1) the x2 card acts as x1 card as no “pro” program can use sli/crossfire/x2 cards. they will act as single GPU here. This means that the old gen 3870 is 2x faster than gtx 280. This can be confusing at first, until you realize:
2) gtx 280 has 240 unified shaders. hd 3870 has 320.

ATIs next gen card, out to compete with gtx 280 is HD 4870. It has 800 (!) unified processors.

I think you got my point as math is very simple.

Even so, those unified processors are not directly comparable on ati/nvi cards, and seeing how 9800gtx has 128 and is not 2x slower than gtx 280, just goes to show that not all of them are being utilized and that SPECheatTest prefers fewer faster processors (9800gtx) over more slower (gtx 280)- altho, it does show that ati card uses its arsenal far better.

to read more: http://www.tomshardware.com/reviews/FireGL-Quadro-Workstation,1995.html and scroll down to comments and find my name “eodeo”. its a bit wordy… be forewarned.

EoDEo

Ideas drift like petals on the wind. You have only to lift your face to the breeze.

Hi Eodeo,

I completely agree with your point - if an Autodesk app. can have full acceleration on mainstream hardware, why not the apps written exclusively for OpenGL? This attitude, today, is a shame from hardware manufacturers part, if not illegal at all. ATI have some interesting features on higher end cards, and this can have a significant value added to the professional - but those numbers of various times more for the same or even inferior hardware is unjustifiable. It`s nonsense. I sincerely hope to see this kind of manners vanish in the coming months - one way or another.