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.
AMD Confirms RX 480 At $199 USD, Other APU & Polaris Announcements
- Thread starter lowendlinux
- Start date
- Sort by reaction score
atm, the p10 is in same situation as Fiji, only framebuffer support, no acceleration.
for sure Apple is preparing something based on P11 (baffin) but it will be rmbp only, maybe 4k imac, it's too weak for 5k imac or n^2MP.
P10 aren't used so far. I guess no n^2MP this year. Maybe next spring, with Vega/HBM2
for sure Apple is preparing something based on P11 (baffin) but it will be rmbp only, maybe 4k imac, it's too weak for 5k imac or n^2MP.
P10 aren't used so far. I guess no n^2MP this year. Maybe next spring, with Vega/HBM2
480 tested and benched. Typically draws 100W while matching an overclocked GTX980. The 150W spec is just because of the connections required from slot and 6pin.
http://wccftech.com/amd-rx-480-faster-than-nano-980/
Only $199. For the first time in a long time I'm impressed by Radeon. I have no doubt 490/490X will match the 1070/1080 at a lower cost and continue to dominate on OpenCL and double precision. Vega should amaze even further.
Apple has made the right choice.
http://wccftech.com/amd-rx-480-faster-than-nano-980/
Only $199. For the first time in a long time I'm impressed by Radeon. I have no doubt 490/490X will match the 1070/1080 at a lower cost and continue to dominate on OpenCL and double precision. Vega should amaze even further.
Apple has made the right choice.
Last edited:
SCSC, is "480X" a typo? I couldn't find anything about it in the link you provided. I only see 480 information.
Yes sorry typo. Fixed.
[doublepost=1466013190][/doublepost]
Too early to tell if or when macOS will have drivers.
If you won't flash it. R9 280X is the best card, which has exactly the same Device ID as the nMP D700, therefore, must be supported by Apple.
However, this card draw quite a lot of power, you have to seriously think about how to power a 6+8 Pin card.
For easy power management, a 2x6pin 7950 (also name as R9 280) is another good choice, this card also has licensed Mac Edition card. So, technically, Apple will keep supporting this card as well.
Both these card can be flashed by user, but also work OOTB without flash (no boot screen).
You can click the R9 280 in my signature, which link to the PC R9 280 I am now using.
Just use your GTX 670, it should be plug n play.
Drivers are fine, but some people are reporting dead ports when using the card without framebuffer (which you have to do on non-Hackintoshs, since Apple didn't provide any framebuffer personalities that match PCIE cards).
It's the other way around: Framebuffer personalities included in AMD9500 can give a hint on upcoming Macs. A nnMP is likely to get 6 DisplayPort connectors, so that's what you'll be looking for.
Oh okay. Thanks for clearing that up.Just use your GTX 670, it should be plug n play.
Drivers are fine, but some people are reporting dead ports when using the card without framebuffer (which you have to do on non-Hackintoshs, since Apple didn't provide any framebuffer personalities that match PCIE cards).
It's the other way around: Framebuffer personalities included in AMD9500 can give a hint on upcoming Macs. A nnMP is likely to get 6 DisplayPort connectors, so that's what you'll be looking for.
Just use your GTX 670, it should be plug n play.
...
YA, it is working fine, but since I only use the computer for Photoshop and FCPX, those programs don't seem to speed up as much as I was expecting. AMD cards work better in FCPX right?
[doublepost=1466174362][/doublepost]
Thanks for the TIP!
It isn't.
https://forums.macrumors.com/threads/fcpx-amd-vs-nvidia.1956128/page-2#post-22937432
BruceX test in FCPX made with iMac and GTX 980 Ti - 28 seconds in external enclosure. GPU has 6 TFLOPs of compute power.
The same test made on Trash can with Dual D500. - 8.9 seconds.
https://forums.macrumors.com/threads/fcpx-amd-vs-nvidia.1956128/page-2#post-22971122 The GPU setup has 4.4 TFLOPs of compute power. Titan X would still be beaten by Mac Pro.
Both GTX 980 TI and Titan X are the same 600mm2 die that is called GM200.
Not exactly. AMD GPUs work better with OpenCL, and because of that they work better with FCPX.
While I agree that in general that AMD is better for FCPx, I don't like this particular comparison you've made.
This is comparing a single Thunderbolt2-connected eGPU on a 4-core iMac to two video cards connected directly in PCIe 3 on a 6-core Mac Pro. Even if you throw out Thunderbolt 2 vs PCIe 3, it's still comparing a single video card to two, and 4 cores to 6 cores. Video work is massively parallelized, and in my experience makes extremely good use of all cores you can throw at it.
We also don't know which OS is running on the two vastly different computers, and we saw in the thread a doubling of performance from Yosemite to El Capitain. This is why Synchro rightly asks everyone to report the entire system including OS version.
On which hardware, and which software?
FCPX, I think, fully benefits from amount of compute power of the GPUs. Additional cores would not bring any meaningful benefit on this application. Again, look at difference in compute power of the GPUs. OpenCL exposes the compute capabilities for application. It sees only compute power. Not the amount of GPUs.
Not too hard, my dual 7950 setup only cost no more then $300 (new cards), which of course don't have 12G VRAM but only 6G, however, this setup should perform better than a single TitanX in FCPX (finish BruceX in 15s with the W3690).
May be I didn't make it clear, I was't emphasis enough on the "same cost". How much you have to pay for a TitanX? And for the same cost, what setup (not necessary single card) I can use with AMD?
A single used 7950 may cost around $100, how to get a $100 Nvidia GPU which can beat the 7950 in FCPX?
I know it's not a fair comparison to use 2 AMD card to beat a single Nvidia card. However, what I want to say is for the same initial setup cost, AMD usually works better.
Apple and AMD screw us up, make us almost can't use any better GPU then the 280X. On the other hand, Nvidia keep updating it's web driver, and allow us to use the TitanX. There is no doubt that the TitanX should be the fastest card (in general). However, that's not cheap, and obviously AMD mid level card has very good cost to performance ratio in OpenCL, also, FCPX is highly optimised for the AMD card (may be because of the nMP), this make the Nvidia card (not just the expensive TitanX) even harder to beat the AMD card in cost to performance ratio in FCPX.
Little sidenote:
you have 3 GB of VRAM.
Each card has it's own VRAM. It doesn't combine. You have to cards with 3 GB each.
In this case/rendering depends on if a large video buffer is required or not. The compute cores don't always need '12GB' to achieve faster rendering.
If he does need more VRAM to go with that 7000 series then there are 7970 with 6GB for about 110-120 bucks. Dual GPU 7990 series you can pick up for 200 bucks. You can even pick up a monster 295X 8GB for about 300 and the price drops fast. These require more power though but have very high performance in OpenCL...with none of the bugs that Nvidia has been having problems with.