CGTalk > Technical > Technical and Hardware
Login register
Thread Closed share thread « Previous Thread | Next Thread »
 
Thread Tools Search this Thread Display Modes
Old 11-10-2013, 10:11 PM   #1
aglick
sushi
 
aglick's Avatar
CGSociety Member
Adam Glick
Sapphire
Austin, USA
 
Join Date: Nov 2006
Posts: 1,281
It's here. The 2013 Pro gpu roundup at CGCHANNEL.COM

__________________
http://adamglick.net
 
Old 11-10-2013, 10:52 PM   #2
cgbeige
Expert
 
cgbeige's Avatar
portfolio
Dave Girard
Opinions are mine. You can't have them.
San_Francisrococo, USA
 
Join Date: Jul 2005
Posts: 7,002
ya, I saw that. Good stuff but wondering why the FirePro W9000 isn't there. I'm considering it because it's a BTO option in the new Mac Pro.
 
Old 11-10-2013, 11:25 PM   #3
ThE_JacO
MOBerator-X
 
ThE_JacO's Avatar
CGSociety Member
portfolio
Raffaele Fragapane
That Creature Dude
Animal Logic
Sydney, Australia
 
Join Date: Jul 2002
Posts: 10,954
It's nice to see the benchmarks as they are rare enough, but I can't say I'm totally enthused by it.
No W9000 or k6000 is a bit of a big gap there, the article is fundamentally benching several cards that are getting sort of long in the tooth now. There's also relatively little benefit in this kind of roundups IME, at least throwing a Titan and a 780 in there would have been nice.

No mention of number of iterations per test or how the results were averaged, or even the drivers version and settings (maybe I missed it?).

When it goes over the differences between professional and gaming cards it also sounds pulled out of the corners of the interwebs rather than matured experience.

Lastly the introductory part doesn't exactly ooze domain knowledge and inspiring confidence. Nearly every other paragraph contains some inaccuracy or common mistake suggesting the info is gathered from the internet's hive mind rather than official sources.
It mentions some borderline myths, but completely neglects mentioning situations when quadro cards are objectively required (hardware stereo, multiple buffers of 10bpc colour in OGL and so on). Again, reading it it seems a rather naive article technology wise, I'd expect a deeper knowledge of hardware from such an article with more facts and less "common knowledge".

And since I don't want this to sound like some random attack, beside the above missing bits I'm referring to inaccurate statements like:
Quote:
Second, the GPU chips on pro cards are usually hand-picked from the highest-quality parts of a production run

Is this a fact? Especially for older cards with long standing production processes nVIDIA never said as much to my knowledge (fermi), and the single most modern line of kepler cards has the same exact Q/C and thermal management and throttling between k5k, Titan and k6k as far as I know.
I'd like to see sources when I see such statements.

Quote:
The first of these is that pro cards typically carry much more RAM than their consumer equivalents: important for displaying large datasets in 3D applications, and even more so for GPU computing.

Since when? Every single card in the nNVIDA ranks in that test carries the same or less ram than the cost equivalent GTX.
The k5k carries less than a titan costing 60-70%, the k4k carries the same of a cheaper 780 and so on.
The only ONE card out there currently out doing the GTX line is the k6k, which is absent from the test anyway.
This is factually inaccurate and pretty easy to verify, it reads like writing for the sake of getting over it with no fact checking.

Quote:
Most people know that using Nvidia’s SLI technology (or AMD’s equivalent, CrossFire) to run two or more graphics cards side by side can increase 3D performance in games. I am often asked if it has the same kind of benefits for DCC applications. Until now my answer has always been, “I’m not sure: I’ve never benchmarked it.”

We could have told him years ago, not that the info coming from AD employees themselves is hard to find.

Just to hand pick a few.
Hopefully the article will be updated.
__________________
"As an online CG discussion grows longer, the probability of the topic being shifted to subsidies approaches 1"

Free Maya Nodes
 
Old 11-11-2013, 12:27 AM   #4
aglick
sushi
 
aglick's Avatar
CGSociety Member
Adam Glick
Sapphire
Austin, USA
 
Join Date: Nov 2006
Posts: 1,281
hehehhh Raffaele -

You sure you're not just upset that Quadro got totally smacked around in a few important tests?

Jason knows what he's doing. His methodology seems sound enough. His results back up what I've seen in my own testing and others I've seen. I think that for the very most part, his observations and analysis' where it matters are pretty much right on. There may be a few editorial details of the article which may be hard to substantiate (hand-picking ASICs, etc) but i'd be interested to know what you found in his work that materially missed the mark or drew misleading conclusions about the data recorded.

If you can make gaming cards work for you and your pipeline, great! But your experience doesn't/can't de-legitimize the market for pro-class graphics hardware.

Your assertion that AMD FP cards can't competently provide stereo, 10-bit color, etc is also not well-supported by the facts. There are many high-end visual simulation and digital media solutions leveraging these features on FirePro cards with perfectly acceptable results.

Come on, man. Why not give credit where credit is due? AMD seem to have been doing their homework. Looking at straight up viewport performance and price-performance it looks to me like FirePro has the strongest pro gfx offerings for ADSK 3D and a few other toolsets.

But Individual mileage may vary, right?
__________________
http://adamglick.net

Last edited by aglick : 11-11-2013 at 07:30 PM.
 
Old 11-11-2013, 12:29 AM   #5
aglick
sushi
 
aglick's Avatar
CGSociety Member
Adam Glick
Sapphire
Austin, USA
 
Join Date: Nov 2006
Posts: 1,281
Hi Dave-

I'm personally kinda glad he skipped ultra-high-end tests.

Honestly -and I can say this because i don't work for AMD anymore- the W9000 seems to be way less compelling (for the money) than the W7000 and W8000 for a lot of 3D M&E workflows.

If money is no object, I think NV have a very strong product in their QK6000 today.
__________________
http://adamglick.net
 
Old 11-11-2013, 12:47 AM   #6
ThE_JacO
MOBerator-X
 
ThE_JacO's Avatar
CGSociety Member
portfolio
Raffaele Fragapane
That Creature Dude
Animal Logic
Sydney, Australia
 
Join Date: Jul 2002
Posts: 10,954
Quote:
Originally Posted by aglick
You sure you're not just upset that Quadro got totally smacked around in a few important tests?

Huh? Why would I? I'm very far from being a fan of nVIDIA. I use CUDA over OCL for convenience and context, but I hardly prefer either company to the other, and have a mild dislike for both in terms of corporate and marketing policies.
I've also been a long time non-fan of quadros. So what I believe you're trying to imply there (I imagine that'd be that I'm some nVIDIA fanboy and therefore resent some numbers on a website) has no foundation whatsoever.

Quote:
Jason knows what he's doing.

Nowhere I said he doesn't, but the article doesn't show that at all, see my notes, which are all pretty factual and easily verifiable.
I deal with facts, and I take measure of demonstrated competence (which is different from making an assumption about the person himself) by those. The article doesn't show much, regardless of the data presented (which again as mentioned I'm grateful to see available).

Quote:
His methodology seems sound enough. His results back up what I've seen in my own testing and others I've seen. I think that for the very most part, his observations and analysis' where it matters are pretty much right on. There may be a few editorial details of the article which may be hard to substantiate (hand-picking ASICs, memory capacity comparisons, etc) but i'd be interested to know what you found in his work that materially missed the mark or drew misleading conclusions about the data recorded.

Again I'm not entirely sure what you're getting at.
Where in my post did I say I thought the numbers inaccurate?
I said the journalism is sloppy, important cards are missing, the "facts" are unverified and in some cases plain wrong, and there is VERY important information missing (such as why you would actually HAVE TO pick a quadro/firepro in place of a gtx/radeon before numbers even come into play).

As for the methodology, how can you say it's accurate when there's no indication whatsoever of what methodology that would be?
No indication of drivers, number of iterations and type of weight (average vs mean vs weighted) and so on.

Quote:
If you can make gaming cards work for you and your pipeline, great! But your experience doesn't/can't de-legitimize the market for pro-class graphics hardware.

Huh? Man, do you have an axe to grind or something? I didn't say that anywhere either. You seem to be having a go at me by putting a slew of words I didn't say in my mouth.

Quote:
Your assertion that AMD FP cards can't competently provide stereo, 10-bit color, etc is also not well-supported by the facts. There are many high-end visual simulation and digital media solutions leveraging these features on FirePro cards with perfectly acceptable results.

OK, this is going completely in the absurd.
I said the GTX cards can't do that compared to quadros, not that AMD cards can't.
I mostly mention nVIDIA cards in my post because that's what I have applicable experience with, while I have little with AMD therefore I held comments on that front.

Quote:
Come on, man. Why not give credit where credit is due? AMD seem to have been doing their homework. Looking at straight up performance and price-performance it looks to me like FirePro has the strongest pro gfx offerings for ADSK 3D and a few other toolsets.

But Individual mileage may vary, right?

You completely mis-interpreted (I assume unintentionally) my post and have twisted everything I said entirely out of context and meaning.
I not ONE sentence I drew ANY comparison of any sorts between AMD and nVIDIA, and re-reading my post there is nothing that might even remotely suggest it unless you are somehow projecting a bias I don't have onto it. As I said if I mention nVIDIA more prominently it's because I have considerable experience with their brand and cards, while nowhere nears as much with AMD's therefore I didn't feel qualified to talk about objectivity and factual accuracy.

Again you sound like you have a vested interest. Can I ask what your current employer is? I seem to sense a bias.
You are very close to an ad hominem in your post.
If I had to be as assuming and aggressive as you are I would probably imagine you work for AMD or for some associate of it and are particularly defensive of this article because it paints them in a somewhat favourable light. Would that be an accurate assumption?

Edit:
Ops, apparently it would be accurate:
2013-10-1
AMD Selects SAPPHIRE as Exclusive Global Distribution Partner for AMD FirePro™
Aren't you an evangelist for them?
__________________
"As an online CG discussion grows longer, the probability of the topic being shifted to subsidies approaches 1"

Free Maya Nodes

Last edited by ThE_JacO : 11-11-2013 at 12:56 AM.
 
Old 11-11-2013, 12:57 AM   #7
aglick
sushi
 
aglick's Avatar
CGSociety Member
Adam Glick
Sapphire
Austin, USA
 
Join Date: Nov 2006
Posts: 1,281
ok fair enough. Maybe I was interpreting your words with a denfensive ear.

I work for Sapphire. And I am biased. I'm an AMD fan. I also have a kneejerk reaction against kneejerk reactions.

It sounded to me like you were discrediting the article, the author and the veracity of his conclusions. If you were not, then I apologize for being overly sensitive.

Much Respect,

-Adam
__________________
http://adamglick.net

Last edited by aglick : 11-11-2013 at 01:03 AM.
 
Old 11-11-2013, 01:04 AM   #8
ThE_JacO
MOBerator-X
 
ThE_JacO's Avatar
CGSociety Member
portfolio
Raffaele Fragapane
That Creature Dude
Animal Logic
Sydney, Australia
 
Join Date: Jul 2002
Posts: 10,954
I appreciate the mature reply.
No offense taken, none intended, to you or to the author. I hope the article will be corrected and completed because hardware info like that is surprisingly rare and it's a shame to see a solid effort framed in a somewhat poor context (purely in terms of source checking and accessory info), that's about it.

I'm not questioning or discrediting the numbers (I would have no way to do so not having access to 80% of those cards, if I had the inclination to begin with) and, believe it or not, have no bias and more than a little hope that OCL will mature and AMD will do well, more so on the CPU and next gen hybrid cards than on GPUs, for the sake of competition and my interests as a consumer.
__________________
"As an online CG discussion grows longer, the probability of the topic being shifted to subsidies approaches 1"

Free Maya Nodes
 
Old 11-11-2013, 07:51 AM   #9
sentry66
Expert
 
sentry66's Avatar
portfolio
node crazy
USA
 
Join Date: May 2008
Posts: 2,076
Websites always seem to benchmark pro cards against other pro cards or consumer gamer cards against consumer gamer cards. I think it's a gross oversight to not include at least one or two gaming cards among the group of professional cards so people can understand the pros and cons of what's out there.

For a lot of people, it's getting harder to justify a pro card. Especially when 3d software and rendering companies seem to be going out of their way to make their software run well on gaming cards. Also not to mention gaming cards do have the bleeding edge fastest hardware which run better on other apps - like adobe software or game engines.

Features like 10 bpc color are great in theory, except the part where hardly any software takes advantage of those display color depths.....and for a lot of people, the final destination is going to be 8bpc anyway. IMO it's better to see banding issues up front than to discover them after the "final" was delivered because you didn't see the banding on your 10bpc color card.

Regardless what card a person chooses, there's going to be pros and cons. It's just always a shame when articles don't bother to include the full range of what's on the market that many pros actually are considering. Tom's Hardware has had some nice video card articles lately with a good mixture of cards.

Last edited by sentry66 : 11-11-2013 at 07:55 AM.
 
Old 11-11-2013, 04:07 PM   #10
BoostAbuse
slerping quaternions
 
BoostAbuse's Avatar
John Appleseed
Zimbabwe
 
Join Date: Dec 2006
Posts: 1,283
I'll side with Raff on this one as well that the article needs a bit more information and probably in good form you might want to provide a full disclosure statement that you represent AMD Adam Would be good to provide a rundown of the drivers used and in-depth settings much like you see the major review sites run in order to assure the right information is being sent across to the viewers.

sentry66 raises a good point as well that the days of dedicated pro level cards being the only choice are pretty much gone and many review sites are benching the higher end consumer level cards in their tests now as well. Sure there are feature defining benefits to the pro level cards that will be a requirement for some but for the other vast majority I'm sure they'd be more eager to see how an R9 290X stands up against a GeForce 780, 780 TI or Titan in the grand scheme of things.

I'll be waiting for the 4K consumer card tests

-s
__________________
My opinions do not represent those of my employer.
 
Old 11-11-2013, 04:54 PM   #11
aglick
sushi
 
aglick's Avatar
CGSociety Member
Adam Glick
Sapphire
Austin, USA
 
Join Date: Nov 2006
Posts: 1,281
I agree that it will be interesting to see a few consumer cards tested, and I think the author said he's working on this.

My guess is that for most apps (esp. the ones that use OpenGL), gaming cards will be slower in general than their pro graphics counterparts due to the application-specific performance tuning that happens in pro gfx drivers. Of course, looking at price-performance, gaming cards will often score really well.

One reason to be wary of making direct comparisons is that there is a misconception/misperception that performance (or price-performance) is the most important consideration for most pro graphics users. It's not.

Productivity is. -And *reliability* seems to be a key for delivering that. On-going and exhaustive testing, driver tuning, bug fixes and certification efforts carried out between the ISVs and the hardware makers is expensive. It's valuable. It's the primary reason pro cards cost more.

There seems to be less perceived importance on these issues here in the M&E market -probably for a lot of different reasons. Nobody's saying that gaming cards can't work for some 3D artists/pipelines. They absolutely might. In fact, some really will. -Depends on the specific app and 3D card and driver version and proper settings, etc etc.

In the CAD/CAM/CAE world (i.e. 80% of the pro graphics market), certification and support (i.e. quick bug fixes, performance optimization, ISV support) seem to continue to be much more of a concern for users.

In any case, forums like this are a great resource for folks looking to research a bit about their apps and needs so they can make smarter buying decisions for themselves.

*(edited)*
__________________
http://adamglick.net

Last edited by aglick : 11-11-2013 at 07:31 PM.
 
Old 11-11-2013, 05:38 PM   #12
Tuxon86
PRO
portfolio
Alain Ménard
Beauharnois, Canada
 
Join Date: Apr 2013
Posts: 169
Quote:
Originally Posted by cgbeige
ya, I saw that. Good stuff but wondering why the FirePro W9000 isn't there. I'm considering it because it's a BTO option in the new Mac Pro.


Isn't the w9000 starting to get old. I mean, isn't it last year tech?
 
Old 11-11-2013, 05:56 PM   #13
aglick
sushi
 
aglick's Avatar
CGSociety Member
Adam Glick
Sapphire
Austin, USA
 
Join Date: Nov 2006
Posts: 1,281
Quote:
Originally Posted by Tuxon86
Isn't the w9000 starting to get old. I mean, isn't it last year tech?


Apple are going for OpenCL performance. W9000 is still the king of OpenCL/compute in the pro gfx space.

https://compubench.com/result.jsp?test=CLB10101
__________________
http://adamglick.net
 
Old 11-11-2013, 06:07 PM   #14
sentry66
Expert
 
sentry66's Avatar
portfolio
node crazy
USA
 
Join Date: May 2008
Posts: 2,076
Most reviews always use benchmarks as the primary core of the review, where the commentary is along the lines of "yup looks like the quadro did well in this test". The rest of the review is usually typical fluff like how many video ports it has or how the card's fit and finish is. IMO that's seriously lazy. Any computer geek off the street can click the 'run benchmark' button and record the time.

What people want is an actual professional who knows enough of the basics of the different apps to sit down and actually try to work on a real-world heavy scene file. Take 10 minutes to load up a 3ds max scene and start selecting objects, moving vertices/curve points, change some materials around, try out the sculpt tools, get some particles going. Take note of if certain cards take a long time to select a heavy poly model or don't draw something on the screen correctly. Then do the same with maya, cinema 4d, etc. People are interested in what cards struggle with these basic functions.

Talking about pro features, how is the stereo 3D on the pro cards? How much faster do quadros switch to stereo mode than the geforces or AMD vs Nvidia on the matter? How seemless is the experience? I can tell you it kinda sucks on geforces, where your screens will flicker and blank out for a few seconds before coming back on in stereo mode. Then it happens all over again when you go back to 2D mode or switch from windowed to fullscreen. Stuff like that, IMO are things pros who might pay extra for a pro card want to know.

It bothers me when a review raves about a certain card that performed benchmarks well, you buy it, and within 3 minutes you know it's not going to work for you. Aside from fiddling with control panel settings, your only option is to try older or certified drivers, or upgrade your 3D software if you're using an older version.

Last edited by sentry66 : 11-11-2013 at 06:57 PM.
 
Old 11-11-2013, 06:19 PM   #15
Tuxon86
PRO
portfolio
Alain Ménard
Beauharnois, Canada
 
Join Date: Apr 2013
Posts: 169
Quote:
Originally Posted by aglick
Apple are going for OpenCL performance. W9000 is still the king of OpenCL/compute in the pro gfx space.

https://compubench.com/result.jsp?test=CLB10101


I know.... But it doesn't change the fact that the w9000 is "old news" by now tech wise. As for OpenCL performance, this will be great when my application will be updated to use it
 
Thread Closed share thread


Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
CGSociety
Society of Digital Artists
www.cgsociety.org

Powered by vBulletin
Copyright ©2000 - 2006,
Jelsoft Enterprises Ltd.
Minimize Ads
Forum Jump
Miscellaneous

All times are GMT. The time now is 06:12 AM.


Powered by vBulletin
Copyright ©2000 - 2016, Jelsoft Enterprises Ltd.