ie, "i've heard some really cool vaporware may solve some of these problems someday"
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.
New Mac Pro Misconceptions
- Thread starter Radiating
- Start date
- Sort by reaction score
ie, "i've heard some really cool vaporware may solve some of these problems someday"
Even just the small one the OP posted works with GPU's and has been proven to do so. It won't work with macs yet due to the lack of driver support, but many of them work with PC's for lower end videocards with low power requirements.
The fact that this workstation has 6 thunderbolt 2 ports gives a good indication that Apple plans on allowing driver support of external video cards.
So, by your own admission, currently there are ZERO (0) TB enclosures that work with OSX and GPUs.
Why exactly would you ask the above question unless you were a paid shill from Apple here to try to blow some of the stink off this pile of xxxx?
Come back when a single enclosure actually works. Until then, as Tesselator noted, you are arguing a complete fallacy.
Well, I asked about this on another MacPro thread, but it looks like it might belong here.
Has anyone heard more details about this TB GPU chassis?
http://www.anandtech.com/show/7040/computex-2013-thunderbolt-graphics-from-silverstone
They specifically mention using with a MacBook Air to run an external GPU.
http://www.bit-tech.net/news/hardware/2013/06/06/silverstone-external-graphics-card-case-deb/1
Has anyone heard more details about this TB GPU chassis?
http://www.anandtech.com/show/7040/computex-2013-thunderbolt-graphics-from-silverstone
They specifically mention using with a MacBook Air to run an external GPU.
http://www.bit-tech.net/news/hardware/2013/06/06/silverstone-external-graphics-card-case-deb/1
The only promising entry.
Note that 5 days ago when story was written "Silverstone told me there is no release date as such as they are still in the alpha phase of development"
So, still somewhere off in a rosy future where unicorns jump over rainbows.
The GTX480 inside appears to be sitting at an uncomfortable angle, not how finished products will likely look.
This is the best and most likely option, though a cobbled together prototype like that isn't going to bring much peace to the issue at hand.
There are USB 3.0 graphics that work with Macs (I was told). You of course need a USB 3.0 card if you're on a MacPro tho.
.
Last edited:
I know we have discussed this article in the past, but it seems like we have all forgotten it. Even on medium range graphics card the difference between going over TB and PCIe is easily demonstrable. I am wary of people claiming to be using a Titan over TB 1 and claiming a 5% reduction. I have to question their testing techniques.... Possibly they were using their eyes?
http://www.tomshardware.co.uk/pci-express-graphics-thunderbolt,review-32525.html
We don't have any such benchmarks for TB2 since we have no TB2 devices, but graphics cards are not going to slow down over the next couple of years. It is safe to assume that trying to upgrade to something beefier that the new Mac Pro will come out with will see you hitting bandwidth issues with TB2.
http://www.tomshardware.co.uk/pci-express-graphics-thunderbolt,review-32525.html
We don't have any such benchmarks for TB2 since we have no TB2 devices, but graphics cards are not going to slow down over the next couple of years. It is safe to assume that trying to upgrade to something beefier that the new Mac Pro will come out with will see you hitting bandwidth issues with TB2.
The bottom line is that the success of the MacPro is dependent on software developers. If the developer of DaVinci Resolve 10 which has primarily been based on CUDA support says this machine kicks ass, I am inclined to believe him.
People can bitch and moan about CUDA support all they want but CUDA support doesn't mean jack **** when people have no ideas what software developers have been up to with Apple.
Again, DaVinci Resolve 10 is a perfect example.
People can bitch and moan about CUDA support all they want but CUDA support doesn't mean jack **** when people have no ideas what software developers have been up to with Apple.
Again, DaVinci Resolve 10 is a perfect example.
I suggest you either read that site or correct your maths.
55/91 is 60% of the performance and the gap is only likely to get wider on new gen cards.
TB2 is NOT a replacement for high speed PCI-E. period.
Won't latency be the same between TB2 and TB1... I would assume so. Anyway, WOW! Look at the small file I/O (pp.5) where it's not even bumping the speed limits yet. 0.5k ~ 32k... That thunderbolt PCIe device just smashed all the performance goodness to heck. That's MUCH worse than I had assumed it might be. Crap!
It looks like some games also take a dive...
.
Last edited:
What would be the point of running an external GPU for video output when the internal solution seems (on paper I admit) so powerful, in "pro" terms (not gaming)?
Also - this has been bothering me for a while - in GPGPU terms, why is it so outlandish to imagine an external solution? BlackMagic in their Da Vinci Mac hardware setup guide SUPPORT plugging in 3 Quadro 4000 cards + 1 RED Rocket in an expansion chassis, all 4 cards running on a single x16 lane.
There is a lot of bitching here, but I guess if I sold modded cards I'd be pissed too seeing as the new Mac Pro just vaporised my future market.
Also - this has been bothering me for a while - in GPGPU terms, why is it so outlandish to imagine an external solution? BlackMagic in their Da Vinci Mac hardware setup guide SUPPORT plugging in 3 Quadro 4000 cards + 1 RED Rocket in an expansion chassis, all 4 cards running on a single x16 lane.
There is a lot of bitching here, but I guess if I sold modded cards I'd be pissed too seeing as the new Mac Pro just vaporised my future market.
You may not have paid attention to this in the past. The thunderbolt eGPU discussion has been going since 2011. It doesn't matter as it's not shipping today, and the mac pro isn't the thing that would really motivate vendors. Notebooks greatly outsell it and are far better aligned with such a solution. It seems like you're just linking stuff to show proof of concepts when the proof of concept came long ago. In terms of general computing, pushing everything outside the box is a clunky solution that is really motivated by the popularity of mobile form factors. Otherwise performance is typically pushed by integration. Distance is a huge factor with critical components.
Also note the hacked solution that was required to make that work. It shows a proof of concept that is really aimed at notebook use. I'm sure some people would buy it if it hit good enough on price to performance ratio, but that doesn't make it a good idea to leverage the problem and solution upwards as a set. I wouldn't even consider thunderbolt gpus as an option with the future mac pro. Additional external storage would have to be one, but with gpus they're either good enough to last the life of the machine or they aren't. If they aren't it's not a good purchase.
I don't follow the exact list you do on a purchase, in that I don't give resale value too much consideration. It's not always predictable, so I tend to calculate the value of use and ROI based around 0 salvage value. Usually I'll gauge the validity of a purchase based on an assumed 3 year cycle.
Yup, agree. Sounds right.
Apple manufacturing costs. No, really, I'm serious. That's exactly what this design addresses. And if the previous MP line wasn't profitable I can't really blame them either.
Not sure on the latency, but that sounds right. To be honest in the last couple of months I have spent most of my time playing with my new tractor (long story) rather than following new tech.
Yes, exactly.
And with a new Mac Pro you would only have a single X4 lane. So 1/4 of the bandwidth. What is not to understand about this being a bad thing?
It's like buying an 8 cylinder car and trying to drive around with 6 of the spark plug wires yanked off. Progress?
Last edited:
The trolls on this site who have nothing better to do than make snide trash can remarks.
John Deere? Cool beans!
Here's mine with my brother and a load of freshly inoculated logs - on the mushroom farm in Texas:
Probably not (unless push TB 2 into a corner case). Not so much that the difference in TB protocols and configuration ( TB 2 is just remapping same aggregate bandwidth), but pragmatic fact that TB2 will be implemented on a smaller , faster process technology. Intel is going to do a shrink to get cost down and put some more hardware ( so cost won't radically drop. )
The switching latency though should go down a small amount. (going to be easier for them to design something that switches incrementally faster).
Now if throw a traffic jam onto the TB 2 network ( mega 4K video streams with concurrent high PCI-e data streams ..) then probably would go up in some configurations. In contrast though when no traffic jam the now larger TB 2 network backbone bandwidth should allow more transparent x4 PCI-e v2 data transmissions.