Oh ok, my bad then. Thanks for the correction
Ok good to know... A seriously handicapped version of the XT2.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
To those nMP owners who installed Windows
- Thread starter wheelhot
- Start date
- Sort by reaction score
Ok good to know... A seriously handicapped version of the XT2.
Makes sense... What's your take on the D500... It doesn't seem to offer any added performance on the D300 in most benchmarks that have been run... Why would Apple have even bothered offering both the D500 and D300 given their near identical performance in most benchmarks?
Better performance with 4k video in FCP perhaps? Anand said that VRAM usage in 4k editing exceeded 3GB easily. IDK precisely what OS X apps take the advantage of double precision, but if any, it also could be a good reason.
The only OSX app that I can think of that's not scientific simulation software would be Siemens NX, it's probably use by Apple themselves
Glad to help guys, the more we learn about these cards the easier it will be to decide on which to purchase.
For those asking if crossfire was enabled, it was enabled but it wasn't forced on in the driver, there is an option to just plain "enable crossfire" which I had ticked, then there is an option for "enable crossfire for programs without a application profile in the driver" which I didn't have ticked. I'm assuming that's to allow per application use of crossfire.
Anyway, I don't think many 3d creation cad apps even utilize crossfire but if you guys really want to see I can re run the bench mark with it forced on to see any difference.
I'm a freelance 3d artist, and I bought this machine as my work computer since it seemed like a good value for the 6 core which I use for rendering and the dual firepros. So far polygon performance and viewport performance I'm getting for high detail geo is a tad disappointing compared to my previous gtx 760 PC, on average it's exactly the same if not a few Fps slower (again no crossfire). Crossfire seems to make no difference for 3ds max or maya.
Overall though I'm happy with the machine and it's 6 core is awesomely fast, just a little pricy considering we now know they are ~w5000s let's hope the drivers are just poor and not getting the performance it should.
For those asking if crossfire was enabled, it was enabled but it wasn't forced on in the driver, there is an option to just plain "enable crossfire" which I had ticked, then there is an option for "enable crossfire for programs without a application profile in the driver" which I didn't have ticked. I'm assuming that's to allow per application use of crossfire.
Anyway, I don't think many 3d creation cad apps even utilize crossfire but if you guys really want to see I can re run the bench mark with it forced on to see any difference.
I'm a freelance 3d artist, and I bought this machine as my work computer since it seemed like a good value for the 6 core which I use for rendering and the dual firepros. So far polygon performance and viewport performance I'm getting for high detail geo is a tad disappointing compared to my previous gtx 760 PC, on average it's exactly the same if not a few Fps slower (again no crossfire). Crossfire seems to make no difference for 3ds max or maya.
Overall though I'm happy with the machine and it's 6 core is awesomely fast, just a little pricy considering we now know they are ~w5000s let's hope the drivers are just poor and not getting the performance it should.
I ordered the 6 core D700 and use 3DSMax too, I have a gaming system that has an EVGA GTX 760 in it too so will be able to compare as far as that is concerned. I would make sure that you have OpenGL selected in 3dsmax in the viewport configuration and you should see a noticeable improvement with hardware lighting and viewport speed on heavy scenes. Well I hope so anyway haha.
Yes, the progress on dual GPUs front has been pretty much nil in many softwares, hence not many people even bothered to buy a dual Quadro, but I noticed that some 3D rendering software uses GPUs instead of CPUs to render and my bet is those rendering softwares will notice a huge boost in their performance?
Well we still need results from a D300 and D700, before we can confirm that it's indeed the drivers that needs to be optimized cause spec wise, the D500 should perform like a W7000 and not a W5000
----------
Hmm, isn't 15fps considered as a big improvement?
Well we still need results from a D300 and D700, before we can confirm that it's indeed the drivers that needs to be optimized cause spec wise, the D500 should perform like a W7000 and not a W5000
----------
just tried opengl in max 2014, really really slow
opengl in max has always been kinda odd, but side note, I DID notice at least 15fps improvment in maya over the gtx 760
Hmm, isn't 15fps considered as a big improvement?
Last edited:
I wonder if the slower performance of the nMP FirePros compared to AMD FirePros is cause of the lower clock speed on the nMP FirePros?
Hmm, if I'm not mistaken, a few years ago, Autodesk unified all the 3D programs (Inventor, Revit, Maya) into using a single unified 3D engine and instead of using OpenGL, it uses DirectX instead, so theoretically, if you use Autodesk 3D softwares, you wont see much difference (would probably see a performance decrease) with a workstation GPU.
Hmm, if I'm not mistaken, a few years ago, Autodesk unified all the 3D programs (Inventor, Revit, Maya) into using a single unified 3D engine and instead of using OpenGL, it uses DirectX instead, so theoretically, if you use Autodesk 3D softwares, you wont see much difference (would probably see a performance decrease) with a workstation GPU.
oh possibly, im not sure actually. I do know in the settings in 3ds max its still showing direct x 11 with custom nitrous view, and im getting gains with the fire pro in maya vs an nvidia gaming card
Yeah they used to have a Maxstream driver that really helped back in the day when I used Quadro's.
But seems you are correct based on this post:
http://forums.autodesk.com/t5/3ds-M...t-3d-or-OpenGL-as-display-driver/td-p/4287853
Found the old drivers
http://www.nvidia.com/object/maxtreme-archive.html
Moving to DirectX doesn't make much sense at all for multiplatform applications. The only reason to pick DirectX is that have zero plans to move the app to another platform.
http://usa.autodesk.com/products/mac-compatible-products
[ right column is the Windows, but runs in VM/BootCamp products. ]
Revit is Windows only but AutoCAD and Maya are multiplatform. Revit , 3ds , and Inventor all being on a veneer over DirectX would make alot more sense.
Perhaps they are doing entirely different frameworks for different platforms but that is alot of work (although some Cocoa / Win32 differences drive that anyway). And really isn't going to help with movement to the other OpenGL platforms like iOS / Android.
----------
Unlike the W7000 (with 4GB VRAM), the D300 is impaired for 3GB nominal workloads. To get closer to a W7000 rough eqivalent they had move to something higher in the GPU solution mix to offset the lower clock.
The D300 is more so for the folks who aren't hard core FCPX and just need a nominal card. Probably why it has incrementally better gaming and non-VRAM heavy 3D or computational work at a lower price.
The VRAM gap makes it not just double precision. Some of these Viewperf benchmarks load in a pretty substantive amount of data into VRAM. The D300 is going to trail a bit on those too but probably pretty close when VRAM isn't as large of a factor ( so the scores are likely similar in some subtests. )
Coupled with the nuked ECC these results do smell alot like mainstream configured GPUs being dressed up with FirePro drivers but the "FirePro" differentiators being turned off. ( Hence Apple and AMD slapping large mark-up on them. ) The GPU in a W5000 isn't that much better than the iMac max BTO option.
Last edited:
Odd that they would be using a die with 25% of the cores flipped off. That is alot of stuff had to fab that isn't doing anything. Although even more odd that it is 25% + 10 cores unit off. (1526 instead of 1536 ). Every 4 dies on that wafer is a whole die worth of "dead weight".
At the mark-up AMD is perhaps charging it probably isn't hurting them all that much. It may get Apple better pricing on the full XT2 models they getting too if all count to same volume discount.
Ok, I've just rechecked and they unified their material library (link), so it means that Revit, Inventor, 3Ds (not sure about Maya), must've shared the same 3D engine cause if they don't, I'm not sure how Autodesk manages interoperability between those softwares as they are very different in application. AutoCAD is primarily a 2D Based software (I know it can do 3D, but not like Revit) so porting it to Mac is possible, but even then AutoCAD Mac is missing some features which Autodesk themselves admit might not be ported ever as its complexity is off the chart.
I tried searching about the project Autodesk had in unifying their 3D engine but can't find it, what I did find is they are using Direct3D which is Windows only, basically it allows the user to use gaming GPU without having to use workstation GPU.
Check out this result and notice the difference between workstation and gaming GPU when it comes DirectX/Direct3D and OpenGL (Notice that both is Autodesk softwares)
Inventor DirectX/Direct3D test
Maya OpenGL test
Hmm, wouldn't ECC enable will make the result slower? Also, isn't a workstation GPU similar to a gaming GPU except the biggest difference being the driver?
you need to think about binning
You are looking at "flipped off" from the wrong angle.
For a long time silicon vendors have been faced with the problem of what to do with chips that "mostly work" or "work at less than the expected clock".
If you make a 12-core chip, but a particular one only has 9 working cores - to you throw it away, or do you disable the 9th core and sell it as an 8-core? "Sell it as an 8-core" is the answer. (No difference if a GPU should have 1024 cores, but only 803 work - sell it as 768 core.)
Same thing happens with cache. If you try to make a chip with 30 MiB cache, but only 20 MiB is good - do you throw it away or sell it as an SKU for 20 MiB? "Sell it as a 20 MiB" is the answer.
Apple is probably getting a great price on GPUs that ATI would have thrown in the trash for being too slow or too few working cores.
You are looking at "flipped off" from the wrong angle.
For a long time silicon vendors have been faced with the problem of what to do with chips that "mostly work" or "work at less than the expected clock".
If you make a 12-core chip, but a particular one only has 9 working cores - to you throw it away, or do you disable the 9th core and sell it as an 8-core? "Sell it as an 8-core" is the answer. (No difference if a GPU should have 1024 cores, but only 803 work - sell it as 768 core.)
Same thing happens with cache. If you try to make a chip with 30 MiB cache, but only 20 MiB is good - do you throw it away or sell it as an SKU for 20 MiB? "Sell it as a 20 MiB" is the answer.
Apple is probably getting a great price on GPUs that ATI would have thrown in the trash for being too slow or too few working cores.
As for CUDA and OpenCL, considering that Apple is pushing OpenCL to be adopted more into the market (and probably want it to be the dominant standard as it's open), sticking with AMD card when it comes to OpenCL is better.
Check out the results below, I choose these 2 as they highlight Apple primary target market for MacPro users
BaseMarkCL Image Processing
BaseMarkCL Video Processing
Check out the results below, I choose these 2 as they highlight Apple primary target market for MacPro users
BaseMarkCL Image Processing
BaseMarkCL Video Processing
Last edited:
Yeah, it's mainly CUDA but on the positive note, if OpenCL becomes more widely adopted, chances are the market will respond and it'll break Nvidia monopoly in that market. I believe AMD will be more then willing to help Apple accomplish it, as it'll greatly benefit their FirePro sales.
----------
Yeah, if my memory serves me right, this was exactly the case in Intel Core2Duo and Intel CoreDuo. Core Duo is basically the rejected (ok, that sounds a lil harsh, let's call it, didn't meet the required spec
) version of Core 2 DuoThat's how it usually works AFAIK. Next thing is, that XT2 has been replaced by XTL in all currently offered cards. AMD got rid of GPUs that are no longer on the market, Apple got a decent discount. Great deal for both.
Are you suggesting that companies like AMD keep chips in massive warehouses hoping to sell them one day?