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.
2016 nMP
- Thread starter ManuelGomes
- Start date
-
- Tags
- nmp
- Sort by reaction score
- Status
it all depends on what exactly you're wanting to do.. if it's renderings of CAD/3D Models/Animations then i think it's Autodesk who is doing it best right now:
it's built in to the applications themselves so there's no exporting/manually uploading etc.. it's taken care of for you in the background.. literally one button click to move from local render to cloud render
...
there are other services available such as Rebus :
https://us.rebusfarm.net/en/
..who i've used in the past but i've been won over by autodesk. (not necessarily because the ease of rendering on the cloud.. more like the ease of setting up my renders in fusion360.. texturing is sweet in that app (things like 3D textures which are incredible time savers for the type of stuff i generally do)
---
edit- i think that's not a real recent video i posted.. there's more app integration than shown there nowadays.
also, you can try this stuff out for free.. for a year.. and if you're not doing commercial work with the software, you can continue to use it for free after the trial is up.. you'll be paying a dollar or two for full sized/final quality commercial renders.
How will it be a tough spot? Dual Polaris 10 chips should result in around 10-12 TFLOPS of performance. Compare this to 7 in the current mac pro, 8 TFLOPS on the Nvidia GTX 1080 or 10 TFLOPS for Nvidia's P100. Doesn't sound like a tough spot to me.
Because if you run an app that only uses one GPU that's cut down to 5-6 TFLOPs. I could see Apple making the 10-12 TFLOPS claim, but that's harder to take seriously.
It's not impossible, but they'd have to possibly use a larger power supply. I think given the complaints so far a larger power supply isn't out of the question. They could just delay Vega Mac Pro shipments until October.
If they do implement vega in October, I'm sure people who needs upgrade could wait couple more months.
Seems you are back pedaling a bit now. How convenient.
It always makes me chuckle when someones tries to discredit a product by saying that its only good enough for making Youtube videos. Considering some Youtube channels have more viewership than most cable channels. It also propels the channel operators into TV commercials, TV Series, Award shows and the like. Often big production thats centered around these channels as well.
What apps out there 1. Have a wide install base on the mac and 2. are heavily GPU compute dependent that don't take advantage of openCL and the mac pro's dual GPUs? This is an honest question as barefeats has lots of examples of apps like Final Cut Pro, Adobe CS, Da Vinci Resolve and Luxmark that all take advantage of both GPUs.
That's not typically how applications use the dual GPUs on a Mac Pro. One GPU is doing OpenGL and one is doing OpenCL. You're still going to be tapping some of the GPU power, but it's very hard to fully tap both GPUs in that configuration. You won't be getting the full teraflops from the cards.
Plus 18 teraflops is still > 12 teraflops.
1) Those are OpenCL only benchmarks, as a mentioned, stuff like FCPX does not run both GPUs in OpenCL. One does OpenGL, one does OpenCL.
2) Those are also synthetic benchmarks. Much harder to scale things in real world performance.
It's really easy to scale OpenCL across dual GPUs, especially in synthetic benchmarks. Trying to scale with one GPU doing OpenGL and one doing OpenCL is harder.
There is some improvement with the addition of a second GPU. Since these tests are a combination of CPU+GPU my guess is even a more powerful single GPU would be limited by the CPU in these cases.
Unfortunately these "synthetic" benchmarks are the best way we have of measuring performance. Results using Final Cut and Adobe CS I imagine are fairly representative of real world use cases of these apps. If you have better tests or can show configurations with something like the AMD R9 290X doing better than a pair of D700s than I would be interested in seeing that.
I am not convinced by your argument that apps are splitting openGL and openCL across the GPUs. If apps are correctly taking advantage of openCL than it should use all the GPU resources available. All the OpenCL tests including final cut, adobe CS, resolve and luxmark seem to do that with the biggest performance limiting factor being the CPU.
http://videocardz.com/60232/amd-radeon-r9-480-based-on-polaris-10
In other news, bits&chips says that GP102 will be very different from GP100. 3840 CUDA cores, with 1.6 GHz, 384 Bit memory bus of GDDR5X memory, 96 ROPs, and 24 GB of VRAM and 250W of TDP. Most importantly: based on GP104 SM's layout, not GP100. GP100 will be only non-consumer GPU.
Even I will be staggered if this turns out to be true.
Remember what we discussed on previous pages about connecting multiple Macs into one cloud?
http://www.ccixconsortium.com
Im wondering how this can be helpful with this idea...
In other news, bits&chips says that GP102 will be very different from GP100. 3840 CUDA cores, with 1.6 GHz, 384 Bit memory bus of GDDR5X memory, 96 ROPs, and 24 GB of VRAM and 250W of TDP. Most importantly: based on GP104 SM's layout, not GP100. GP100 will be only non-consumer GPU.
Even I will be staggered if this turns out to be true.
Remember what we discussed on previous pages about connecting multiple Macs into one cloud?
http://www.ccixconsortium.com
Im wondering how this can be helpful with this idea...
Last edited:
http://videocardz.com/60253/amd-radeon-r9-480-3dmark11-benchmarks
Not bad for a Mainstream offering.
Edit, I got excited. It looks like it is the 2304 GCN core 1266 MHz engineering sample.
Not bad for a Mainstream offering
.Edit, I got excited. It looks like it is the 2304 GCN core 1266 MHz engineering sample.
http://videocardz.com/60232/amd-radeon-r9-480-based-on-polaris-10
In other news, bits&chips says that GP102 will be very different from GP100. 3840 CUDA cores, with 1.6 GHz, 384 Bit memory bus of GDDR5X memory, 96 ROPs, and 24 GB of VRAM and 250W of TDP. Most importantly: based on GP104 SM's layout, not GP100. GP100 will be only non-consumer GPU.
Even I will be staggered if this turns out to be true.
Remember what we discussed on previous pages about connecting multiple Macs into one cloud?
http://www.ccixconsortium.com
Im wondering how this can be helpful with this idea...
What will really make things interesting is if the Quadro variants are full GP100.
The Tesla variants should be GP100 - that was announced. Quadro is anyone's guess - not sure how important 1:2 FP64:FP32 is for graphics.
If macOS is very different, and requires 2012 and later computer, they might update 10.11 to the end of 2017.
I have been wondering when we'll see the OS requirements get a boost. We've been at the same requirements for four years now. Realistically though I don't think there's any reason they *have* to boost them. The biggest difference between old computers now is SSDs and since they haven't made fusion drives standard someone who buys a cheap Mac now is going to have just as mediocre an experience as someone from four years ago.
It's going to go Xeon, and word is touch screen...or Apple Pencil functionality. And availability will be day of or a few days after.
Wait, did you mix up Macbook Pro and Mac Pro?sorry was editing as info was coming in
YEs! lol whoops...macbook Pro folks...macbook proWait, did you mix up Macbook Pro and Mac Pro?
- Status