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.
GTX Titan and GTX 780 flashed w/ EFI ROMs
- Thread starter 666sheep
- Start date
- Sort by reaction score
- Status
Not hallucinating, but "misunderstanding." Currently, the Ray-Traced 3d Renderer in AE (OS X) does not use the GTX Titan for any ray-traced 3d rendering acceleration.
While the GPU Information reports the GTX Titan, AE (OS X) will not use the GTX Titan for any ray-traced 3d acceleration.
See post here:
http://forums.adobe.com/message/5375132
June 2nd is awhile ago.
I got no error messages, it runs fine.
Note that CUDA has moved from the 5.0 in screen shot to 5.5
You stand corrected.
CS6, as noted by an Adobe rep, the issue with GK110 was with an Nvidia library, not Adobe. The Nvidia libraries have been updated, apparently fixing the issue.
Why would someone compare the 780 to the 570. Wouldn't 580 and 780 be the appropriate comparison? Can't the 580 already "run circles around" the 570 and other cards? Seems a bit odd to compare those two models to me.
I haven't looked but if the past is any indication there's probably not all that much difference between the 580, 680 and 780 on a MacPro. A minor increase one would need a very accurate stopwatch to detect.
I haven't looked but if the past is any indication there's probably not all that much difference between the 580, 680 and 780 on a MacPro. A minor increase one would need a very accurate stopwatch to detect.
The 780 cuts the AE benchmark render time down by about 35% compared to GTX680 or 580.
A 7 minute render done instead in 4.5 minutes.
People in no hurry wouldn't care much, but busy Professionals, they would care.
Chart courtesy DJenkins from this board.
A 7 minute render done instead in 4.5 minutes.
People in no hurry wouldn't care much, but busy Professionals, they would care.
Chart courtesy DJenkins from this board.
Attachments
Yeah, it looks like my guesses were right and history is indeed a good indication of the present. There's just about no difference except in a very few benchmarks.
The 780 is 3 seconds SLOWER than the 580 in this test:
The 780 is 8 seconds SLOWER than the 580 in this test:
So basically if you're running After Effects or another CUDA based video app then it might be worth it if shaving a minute off every 4-minute render is important to you. But just about nothing else is much different. And that's about what I thought too. Certainly, there is no circle-running occurring here.
The 780 is 3 seconds SLOWER than the 580 in this test:
The 780 is 8 seconds SLOWER than the 580 in this test:
So basically if you're running After Effects or another CUDA based video app then it might be worth it if shaving a minute off every 4-minute render is important to you. But just about nothing else is much different. And that's about what I thought too. Certainly, there is no circle-running occurring here.
Last edited:
Sorry, I had to pick myself up off the floor, I was having a good hearty laugh.
While some apologists for the nMP might cherry pick a few odd benchmarks where first round drivers don't do GK110 justice, there is no denying simple, obvious facts.
I have included a few more graphs from the Tom's hardware review. The 780 KILLS the 580 & 680 in all of them.
Isn't it fun how picking just a few specific benchmarks you can prove any point you want?
Picking a few specific benchmarks to prove a point you know to actually be false is what I consider to be "intellectual dishonesty".
Instead of a few, hand picked graph's to sell a particular poster's bias, why not read the WHOLE review and get a WHOLE picture?
For a good, thorough, non-cherry picked review, try reading the ENTIRE Anandtech review.
http://www.anandtech.com/show/6973/nvidia-geforce-gtx-780-review/20
Here is some of my cherry picking from last page:
"The full generational performance improvement is absolutely there, as the GTX 780 beats the last-generation GTX 580 by an average of 80%"
(Pretty sure 80% qualifies as "running circles around")
And if you want a comparison to the Top Tier FirePro 9000 in nMP (downclocked 7970), here is another good snippet:
"The 22% performance advantage that the GTX 780 enjoys over the Radeon HD 7970GHz Edition cements NVIDIA’s performance lead"
And by the way, we are only talking about GTX780 here as it can be installed in current Mac Pro with no external power. With a little more trouble you can get a Titan in there. No denying that a Titan is the FASTEST and MOST POWERFUL Single GPU on the planet right now. No "ifs" "ands" or "buts". (or cherry picked graphs)
Yep, right now, TODAY, you can put a GPU in your 2008-2012 Mac Pro that is SUPERIOR to what Apple is going to call their new "High End" option.
You can get 2 (TWO) GTX Titans AND a power supply that fits in optical bay for LESS than the retail cost of a single (1) (UNO) Fire-Pro 9000 today. Maybe Apple will cut you a deal on this lower performance Fire-Pro card, but enough to make them worth this trade off?
No waiting, and no paying Apple's "Shiny Cylinder" tax.
TODAY.
While some apologists for the nMP might cherry pick a few odd benchmarks where first round drivers don't do GK110 justice, there is no denying simple, obvious facts.
I have included a few more graphs from the Tom's hardware review. The 780 KILLS the 580 & 680 in all of them.
Isn't it fun how picking just a few specific benchmarks you can prove any point you want?
Picking a few specific benchmarks to prove a point you know to actually be false is what I consider to be "intellectual dishonesty".
Instead of a few, hand picked graph's to sell a particular poster's bias, why not read the WHOLE review and get a WHOLE picture?
For a good, thorough, non-cherry picked review, try reading the ENTIRE Anandtech review.
http://www.anandtech.com/show/6973/nvidia-geforce-gtx-780-review/20
Here is some of my cherry picking from last page:
"The full generational performance improvement is absolutely there, as the GTX 780 beats the last-generation GTX 580 by an average of 80%"
(Pretty sure 80% qualifies as "running circles around")
And if you want a comparison to the Top Tier FirePro 9000 in nMP (downclocked 7970), here is another good snippet:
"The 22% performance advantage that the GTX 780 enjoys over the Radeon HD 7970GHz Edition cements NVIDIA’s performance lead"
And by the way, we are only talking about GTX780 here as it can be installed in current Mac Pro with no external power. With a little more trouble you can get a Titan in there. No denying that a Titan is the FASTEST and MOST POWERFUL Single GPU on the planet right now. No "ifs" "ands" or "buts". (or cherry picked graphs)
Yep, right now, TODAY, you can put a GPU in your 2008-2012 Mac Pro that is SUPERIOR to what Apple is going to call their new "High End" option.
You can get 2 (TWO) GTX Titans AND a power supply that fits in optical bay for LESS than the retail cost of a single (1) (UNO) Fire-Pro 9000 today. Maybe Apple will cut you a deal on this lower performance Fire-Pro card, but enough to make them worth this trade off?
No waiting, and no paying Apple's "Shiny Cylinder" tax.
TODAY.
Attachments
Last edited:
Where's this "80%" coming from? Not even your own finding, the most extreme of them all show 80%. As you said yourself it's about 35%. And that's about right in the most extreme case. All other typical cases is between 0% and about 10%.
80% would be almost twice as fast and it's not even close to that. 35% in the most extreme case (slow old CPU benchmarking CUDA video rendering) is at least believable. Notice in the AE benchmark thread that people with newer CPUs show a smaller difference between the 580 and the 780. MacVideoCards is using a MP3,1 2.8GHz (if I gleaned that correctly) for his tests.
I'm sure the marketing department at Nvidia and resellers of Nvidia cards would have you believe otherwise but I think even guys with a GTX 570 will be disappointed in the performance increase if upgrading to the 780 and 580 owners won't even notice they have a new card in place - for the overwhelming majority of GPU intensive tasks.
It's historically been like this for a very long time too. You have to wait longer than two numerical product versions (580 --> 680 --> 780) to see a very noticeable speed increase. Usually it's about five and this case seems no different at all. So round about the time the GTX 980 ships is when GTX 580 owners should be thinking about upgrading - for then it will actually make a difference.
80% would be almost twice as fast and it's not even close to that. 35% in the most extreme case (slow old CPU benchmarking CUDA video rendering) is at least believable. Notice in the AE benchmark thread that people with newer CPUs show a smaller difference between the 580 and the 780. MacVideoCards is using a MP3,1 2.8GHz (if I gleaned that correctly) for his tests.
I'm sure the marketing department at Nvidia and resellers of Nvidia cards would have you believe otherwise but I think even guys with a GTX 570 will be disappointed in the performance increase if upgrading to the 780 and 580 owners won't even notice they have a new card in place - for the overwhelming majority of GPU intensive tasks.
It's historically been like this for a very long time too. You have to wait longer than two numerical product versions (580 --> 680 --> 780) to see a very noticeable speed increase. Usually it's about five and this case seems no different at all. So round about the time the GTX 980 ships is when GTX 580 owners should be thinking about upgrading - for then it will actually make a difference.
Not to add fuel to the fire - but the latest graphs only show DirectX performance or Windows-centric results. Aren't there any Mac-specific charts (aside from the AE CS6 Raytracing tests) we could use to really compare their performance in a native setting? I don't disagree that these newer GK110 chips are better performing under Windows than the previous generations but IMHO aren't we comparing apples and oranges? (pun intended)
It's a website called "Anandtech"
I didn't write the article, they did.
Why don't you go explain your opinions to them? I'm sure they would find your analysis fascinating. I'm sorry that the facts don't back up your opinions.
And for those wanting MORE proof beyond the CS6 graphs and Anandtech (bunch of amateurs, yes?) tests, watch Barefeats.com. They'll have the EFI Mac versions of these cards in the next week.
The apologists will have to find a new avenue of excuses then.
Meanwhile, rejoice if you have a 2008-2012 Mac Pro. You can use better video cards than the Mac-Ina-Can, TODAY.
Last edited:
I don't need proof, I just want to know how faster they are if I plan to get them for Resolve Mac or some such. I'm not really sure if AE CS6 CUDA performance improvements scale similarly with Win/Mac drivers on software such as AE/Resolve/Maya/etc. It's always nice to have numbers to compare hardware with 
Will wait for the Barefeats review!
And oh, is it a good idea to put these cards in a PCIe/TB expander? Will these cards saturate the TB port completely?
Will wait for the Barefeats review!
And oh, is it a good idea to put these cards in a PCIe/TB expander? Will these cards saturate the TB port completely?
Fastest GPUs ever made.
Why would you want to throttle their power by putting them in a 4 lane PCIE enclosure and route a rat's nest of cables over your desk just to see 2 or 3 cards strangled down to 1/4 of the speed possible in a 2008 Mac Pro?
TB was never meant for this.
The whole point is anyone with a 2008 or later MP can use cards SUPERIOR to anything available in the Mac-Ina-Can RIGHT NOW. By imposing artificial boundaries on their new design, Apple has made it obsolete before launch. It will only get worse as technology continues it's march past it.
It'll kinda be like putting a SATA3 SSD on a SATA2 connection. Some stuff will be throttled but most won't. And putting this card on any MacPro will also throttle it. It's a PCIe v3 x16 card and the nearest MacPro is half that.
You're also correct in thinking that the MacPro benchmarks will be different and less impressive than the Wintel scores and times. The PCs these cards are being tested on have PCIe v3 and all of the support environment (software) is more mature, robust, and up to data than what exists in OS X.
Using this card specifically for CUDA and OpenCL in OS X it will bench faster than it's predecessors of course but not by very much in most cases. It will also not be throttled at all over TB2 when used that way. In fact you should be able to put two or three of them on one TB2 connection with no throttling for CUDA or OpenCL operations. About a 3% reduction per card is all that's seen and that's mostly because of latency.
Also for the money in GPGPU performance you're probably better off hooking up two GTX 570 cards which are down around $90 to $100 now. And in TB2 over the nMP you could connect six or seven GTX570 cards for the price of one GTX 780 and still see just about no throttling in GPGPU performance.
Last edited:
I'm very interested in possibly getting an Asus GTX 780 for my Mac Pro... off-eBay to save some bucks. What would be the best way to get this started with MacVidCards?
EDIT: Seems you can PM on Netkas; I'll (reluctantly!) sign up for [another] forum if necessary.
Unless someone is willing to PM on my behalf!
Did you get a reply? pm'd but heard nada!?
A short example of "intellectual dishonesty"
In this thread you claim "And putting this card on any MacPro will also throttle it. It's a PCIe v3 x16 card and the nearest MacPro is half that."
While 2 weeks ago in this post:
https://forums.macrumors.com/posts/17745236/
you claimed :
"I'm pretty sure the fastest GPUs going never to almost never utilize x16 v3 (~16GB/s). I'm going to go ahead and say ya, Never! Further I'm going to say the same in regards to x16 v2 (8GB/s) - ya, never. We might wanna think so tho. "
When it fit your argument you were on one side of the fence, and when you started looking incorrect in this thread you claimed the other side. Which is it? You can't claim both things to be true and pretend to have anything resembling credibility.
Guess you cashed that check.
Most likely not, unless you use specific filters or plug ins that require OpenGl or OpenCl or CUDA.
Not sure what you mean, you can PM me here, but I thought I'd answered all of the Netkas PMs.
In this thread you claim "And putting this card on any MacPro will also throttle it. It's a PCIe v3 x16 card and the nearest MacPro is half that."
While 2 weeks ago in this post:
https://forums.macrumors.com/posts/17745236/
you claimed :
"I'm pretty sure the fastest GPUs going never to almost never utilize x16 v3 (~16GB/s). I'm going to go ahead and say ya, Never! Further I'm going to say the same in regards to x16 v2 (8GB/s) - ya, never. We might wanna think so tho. "
When it fit your argument you were on one side of the fence, and when you started looking incorrect in this thread you claimed the other side. Which is it? You can't claim both things to be true and pretend to have anything resembling credibility.
Guess you cashed that check.
Most likely not, unless you use specific filters or plug ins that require OpenGl or OpenCl or CUDA.
Not sure what you mean, you can PM me here, but I thought I'd answered all of the Netkas PMs.
Last edited:
Only if you purposely take things out of context.
Notice where I say it'll never use the full amount the conversation was about using the cards as compute devices via CUDA and/or OpenCL. And here what I said they will be throttled sometimes the context is general use - games, video, whatever...
I wouldn't expect you to be honest and quote the whole thing tho... Or maybe even just the part that says: "OpenCL and CUDA are very very low bandwidth!" ? Man, you really had to go out of your way for this one. Not really even a good try. I'm sure if you hunt harder you can find a more blatant inconsistency - with me being human and all. There probably is one somewhere - good luck.
Last edited:
I'd argue you're far more consistent than MVC is giving you credit for.
You have, in my opinion, consistently gone into existing threads and condescendingly moved the discussion from being on-topic, to an attempt to "Win The Thread" by whatever means possible for motivations unknown. I have yet to see an informative thread started by you.
I, for one, was and am more interested in reading more about the continued results of MVC/Netkas's work rather than face-palming as you attempt to "win" yet another thread by undermining their efforts through critiquing the new video card OPTION they have provided and steering the discussion into the weeds for what I can only imagine is your own self-satisfaction.
If you want a real debate on overall card value/performance vs. bandwidth constraints by application, please start another thread so this one can be kept clean.
There are those of us who are legitimately interested in knowing more about what MVC and Netkas have spent time working on and testing, and want to encourage them to continue to bring all of us more video card options for our PCIe-upgradeable Mac Pros.
I think if you read all of this thread (instead of just the last few posts) you'll find your conclusions are absurd. But here's some information threads started by tesselator, which you seem to be looking for:
https://forums.macrumors.com/threads/1594991/
https://forums.macrumors.com/threads/1626858/
https://forums.macrumors.com/threads/1598539/
https://forums.macrumors.com/threads/829222/
etc.
.
Last edited:
I'm guessing that you got your threads mixed up.
All I see here is you spouting opinions that fly in the face of facts.
You claim to know the performance of a GTX Titan despite not having one. You claim to know them better than not only myself, but Anandtech, both of whom HAVE had one. But somehow you know better than even Anandtech and deny their findings. Hubris much?
Anyhow, thanks for that link. I've got some really interesting facts to add to your icylinder fact thread.
Just some more facts that don't fit with your opinions. I'm afraid that your whole "CUDA/OpenCl don't need much bandwidth and won't be affected by being constricted through TB" theory is about to get Debunked.
This is a prime example of an off-topic post and a poor attempt to "win the thread" yet again.
Mods, could you please move Tess's post into a new thread, titled "Tesselator contributes valuable and completely unbiased information to the forums?" If not, could we have an explanation as to how this post by Tesselator is contributing to the value of this thread, or even on-topic? This could've easily been PM'd to me, however it's evident that Tess gets a sort of personal satisfaction from making a spectacle of himself.
I again, would like to get on with reading more about the 780GTX/Titan work that's been done and test results that are relevant to this thread, as well as any experiences from the OP's early customers.
I am likely not in the minority.
Much obliged.
Last edited:
- Status