Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Even if it were possible for the web drivers to include a kext for GP100 soon, at the moment Maxwell GPUs are still in beta since almost a year with no official released products and the performance in OSX is downright embarrassing compared to Windows. Not to mention the OpenCL bugs that need to be fixed.

http://barefeats.com/razer_core.html

It is going to be a loooong time before Pascal can show its true muscles in OSX, if ever. So have a realistic view of what you could expect from a Pascal under OSX. Good performance and accurate compute capabilities require a lot more than a kext.

The first review of the 1070 is out. Roughly the same as a Ti/Titan X. I'm going to SLI a pair of these in my Skylake build.

http://www.clubic.com/carte-graphiq...st-nvidia-geforce-gtx-1070-pascal-gamers.html
(Taken down for breaking NDA)

Why would there be an official Mac Edition version of a Maxwell card? The cMP market is dead. Not sure what you expect NVIDIA or EVGA to do?

I don't understand why BareFeats ran everything at 1280x720, as that resolution with a 980 Ti is the very definition of CPU limited. I guess he wanted to compare the 980 Ti against the Intel GPU on the internal screen? Surprise surprise, a high-end discrete GPU is way faster than an Intel integrated GPU. As I've tried to explain many times before, the problem in CPU-limited cases is not the NVIDIA driver's fault -- it's the old and slow OpenGL framework from Apple. Maybe things would be better if those apps used Metal, but Metal is missing key features compared with DX11 so the adoption rate has been fairly slow. NVIDIA tuned the CPU performance of their driver with the new web driver from last year, so it's really not clear what else you're expecting them to do here. They can't just wave a magic wand and make the huge Apple software component that sits between the application and their driver disappear.

There have been bugs in the web drivers. Those bugs get fixed. Yes, there are still a couple of lingering issues, which will hopefully be fixed in an upcoming release. Would you rather that they stop enabling these PC cards to work since they don't meet your high quality standards? You're always free to just use a 680 Mac Edition with the stock Apple drivers, and live on GPU technology from 4-5 years ago.
 
Why would there be an official Mac Edition version of a Maxwell card? The cMP market is dead. Not sure what you expect NVIDIA or EVGA to do?

One side of me says there are still enough cMPs being used to warrant at least one more generation of official Mac Edition cards, but I know that Nvidia won't release such a thing until they can iron out all the OpenCL bugs and improve that lacklustre performance we see on various GL benchmarks. It would be irresponsible of them to release Maxwell cards when the drivers are still in beta state.

But I personally don't expect much now, although if you asked me in 2014 when I stuck a 980 in my cMP I had higher hopes at the time. You can track my history of posts from 'hopeful' over a year ago and nowadays more realistic, critical and sometimes a little too pessimistic for some people's tastes ;) that's down to there being next to zero progress on the web driver front (and even steps backwards with these OpenCL bugs that are still not completely fixed) and Apple definitely giving up on Nvidia. But these points have been debated to death by all in a dozen threads about the future of the MP.

So these days all I can do is give people a realistic expectation of what they will get when they use a Maxwell in their cMP or even when they use the web driver on any Mac with an Nvidia chip. My only duty left is to report performance and bugs so that Nvidia will fix these issues.

There are some people who think I'm trying to hurt their business or that I'm some secret agent of AMD. That's plain silly because on a personal level once I have sold my cMP then I'm out of the Mac game but I'll always be using Nvidia in my PC as long as they keep knocking out the better all round product.
 
One side of me says there are still enough cMPs being used to warrant at least one more generation of official Mac Edition cards, but I know that Nvidia won't release such a thing until they can iron out all the OpenCL bugs and improve that lacklustre performance we see on various GL benchmarks. It would be irresponsible of them to release Maxwell cards when the drivers are still in beta state.

But I personally don't expect much now, although if you asked me in 2014 when I stuck a 980 in my cMP I had higher hopes at the time. You can track my history of posts from 'hopeful' over a year ago and nowadays more realistic, critical and sometimes a little too pessimistic for some people's tastes ;) that's down to there being next to zero progress on the web driver front (and even steps backwards with these OpenCL bugs that are still not completely fixed) and Apple definitely giving up on Nvidia. But these points have been debated to death by all in a dozen threads about the future of the MP.

So these days all I can do is give people a realistic expectation of what they will get when they use a Maxwell in their cMP or even when they use the web driver on any Mac with an Nvidia chip. My only duty left is to report performance and bugs so that Nvidia will fix these issues.

There are some people who think I'm trying to hurt their business or that I'm some secret agent of AMD. That's plain silly because on a personal level once I have sold my cMP then I'm out of the Mac game but I'll always be using Nvidia in my PC as long as they keep knocking out the better all round product.

Right, so you effectively get to chose between:

1) No Maxwell or Pascal support, period.

2) Beta support, with some bugs in a small set of applications (that appear to affect you more than others).

I don't use the Adobe apps, so those issues don't affect me. If you're using Adobe apps, my guess is that it's in a professional capacity, right? So, you'd be better off with an officially supported product like the Quadro K5000 with officially supported drivers. You shouldn't be complaining too much if you're trying to do professional work with a beta/unsupported configuration, in my opinion at least. You've already indicated that many/most of the bugs have already been fixed in the latest release, and hopefully the remaining issue will be fixed soon. Right?

Please explain to me what you'd like NVIDIA to do about those "lackluster" OpenGL benchmarks, when they have no control over what version of OpenGL gets exposed to applications or the huge Apple software component (the framework) that sits between their driver and the application. You can easily verify that 99.9% of games are CPU limited on a high-end GPU like the 980 Ti by using the OpenGL Driver Monitor tool and looking at GPU Core Utilization.
 
Right, so you effectively get to chose between:

1) No Maxwell or Pascal support, period.

2) Beta support, with some bugs in a small set of applications (that appear to affect you more than others).

This is a very dishonest and distorted reply, Asgorath. Maybe you did that on purpose to win an internet argument or maybe you genuinely have a short term memory.

My position, which is easily searchable, has been:

1. Prior to the the July/August 2015 web driver update I believed, correctly, that Maxwell cards were being allowed to operate using mostly UDA compatibility because there had been no announcement of Maxwell support at all in any form. Until I mentioned UDA nobody here even new what it was. I don't blame them. Most Mac users know very little compared to the tech news that enthusiast PC users are exposed to.

2. In that July/August 2015 web driver update Nvidia announced the arrival of beta support for Maxwell and certain mobile Kepler chips. I have a Tomb Raider benchmark thread to measure it. Did anyone else here go into detail about the performance improvements???? No. This beta stage has continued until today with no further announcements, as I have mentioned to the dismay of some people.

3. I toyed in the last week with the idea of testing a Pascal in my cMP just to see what happens. That's the limit of my position on Pascal.

4. To say that "some" bugs in a "small" set of applications affects me more than others is highly deceptive, ignorant and rude to many people. Let's reword it to something that reflects the real world:

'Reproducible and critical OpenCL bugs in the web drivers that affect common operations in the world's most widely used professional applications that have been reported on more than a dozen forums.'

And that's a fact. I'm not even going to speak to you any longer if you can live in such denial and twist that situation around into a non-issue. You said yourself you aren't much of a pro app user. Well, it's better if you don't tell us about what we experience. We don't need your voice on this. If we want to know how well Warcraft runs on the Mac we'll ask you for that.
 
Someone REALLY wants to pick a fight about Nvidia. Like a tike who didn't get his Star Wars Lego Deathstar throwing a fit on Christmas Day.

Plop your 680 in and be happy. That's what Apple wants. Run their 5 year old parts with 5 year old drivers and like it.

Or buy a nnMP and throw out old one.

Nvidia isn't getting help from Cupertino. Apple has left drivers in disrepair. The OS is becoming a joke. A BIGGER joke.

But let's tear apart the last vendor who is still trying to help us. Great idea.
 
This is a very dishonest and distorted reply, Asgorath. Maybe you did that on purpose to win an internet argument or maybe you genuinely have a short term memory.

My position, which is easily searchable, has been:

1. Prior to the the July/August 2015 web driver update I believed, correctly, that Maxwell cards were being allowed to operate using mostly UDA compatibility because there had been no announcement of Maxwell support at all in any form. Until I mentioned UDA nobody here even new what it was. I don't blame them. Most Mac users know very little compared to the tech news that enthusiast PC users are exposed to.

2. In that July/August 2015 web driver update Nvidia announced the arrival of beta support for Maxwell and certain mobile Kepler chips. I have a Tomb Raider benchmark thread to measure it. Did anyone else here go into detail about the performance improvements???? No. This beta stage has continued until today with no further announcements, as I have mentioned to the dismay of some people.

3. I toyed in the last week with the idea of testing a Pascal in my cMP just to see what happens. That's the limit of my position on Pascal.

4. To say that "some" bugs in a "small" set of applications affects me more than others is highly deceptive, ignorant and rude to many people. Let's reword it to something that reflects the real world:

'Reproducible and critical OpenCL bugs in the web drivers that affect common operations in the world's most widely used professional applications that have been reported on more than a dozen forums.'

And that's a fact. I'm not even going to speak to you any longer if you can live in such denial and twist that situation around into a non-issue. You said yourself you aren't much of a pro app user. Well, it's better if you don't tell us about what we experience. We don't need your voice on this. If we want to know how well Warcraft runs on the Mac we'll ask you for that.

I know what a UDA is, and I believe you're incorrect about Maxwell operating in backwards compatibility mode. The UDA means with a single set of binaries, NVIDIA can support multiple difference GPU architectures (Fermi, Kepler, Maxwell, hopefully Pascal). This absolutely does not mean that the Fermi driver bits make Maxwell or Pascal GPUs work. If you have information that disproves this, please share it with us. All the other vendors (Intel, AMD) have separate sets of driver binaries for each GPU architecture, for example.

If you're using professional applications, then don't use an unsupported product. NVIDIA has a line of graphics cards that are specifically targeted at professional applications like all the Adobe ones. If you want full support for your professional applications, go and buy one of those cards. Don't use a gaming card with beta drivers and expect the same level of support. NVIDIA has been very clear about the level of support for Maxwell cards on OS X. For the first year or more, they said nothing at all, and simply enabled the cards via their web drivers. Once they released the new drivers with CPU performance optimizations, they mentioned beta level support. This was the first official recognition that they've been enabling PC cards via their web drivers, which they had been doing for years already.

There is only so much NVIDIA can do. My point was this: would you rather they completely abandon the Mac platform and not release drivers for PC cards? That's all I'm saying. I don't believe there will ever be a new official Mac edition card, because the cMP market is just too small and old to be worth the huge investment that it takes to get an official product to market. So, again, my point is this: NVIDIA could walk away from the Mac market entirely, or give you beta-level support for new GPU architectures as they come out. Which would you prefer? I simply don't believe there are any other realistic options here.

I'm quite happy with my TITAN X. It runs the applications I use very well. You've been linking 6+ month old forum threads discussing bugs that are already fixed in the latest drivers. Would it be nice if the remaining issues you are seeing get fixed? Of course. However, you make it sound like the Maxwell drivers are terrible and riddled with bugs so that nothing works correctly. This does not match my experience at all.

TL;DR - if you want official levels of support, buy an officially supported Mac Edition card. If they even sell them anymore.
 
  • Like
Reactions: Squuiid
Has anyone been able to test the 1080 in a pre 2013 model Mac Pro? Would there be any bottlenecks from installing this card in a PCIe 2.0 16x slot? Obviously it would only work under Windows currently, but very interested to find out :)
 
Has anyone been able to test the 1080 in a pre 2013 model Mac Pro? Would there be any bottlenecks from installing this card in a PCIe 2.0 16x slot? Obviously it would only work under Windows currently, but very interested to find out :)

Without proper EFI, I think the card can only negotiate at PCIe 1.1 x16. But that's just my guess, still waiting for someone's confirmation.
 
only on cmp 4,1/5,1, and only osx.

for 3,1 or windows one needs an efi to do the job.
 
I was checking out the offerings from MacVidCards and read that they ship used cards. Is this because they buy new cards and once flashed they can no longer be sold as new?
 
I was checking out the offerings from MacVidCards and read that they ship used cards. Is this because they buy new cards and once flashed they can no longer be sold as new?
Since moders suspended mvc again you would need to ask him about it at email/twitter/myforum
 
  • Like
Reactions: linuxcooldude
Noticed the EVGA 1080 FTW has a TDP of 215, would this possibly be able to run in the cMP with no modification, pending driver support?

http://www.evga.com/Products/Product.aspx?pn=08G-P4-6286-KR

Except the stock EVGA AIB 1080 card, they all come with 8+8 pins. That's not really safe, as it could draw more than 75w from either 8-pin, and we can only feed it 2x6-pin.

Is it safe to say no gtx for Mac Pros?

I assume you mean no GTX 1080? If you're only running Windows, sure. If you want to run OS X or both, hold off a bit. OS X web drivers by all accounts don't currently support the 1080, and when it does we'll likely be limited to PCIe 1.1 speeds due to a lack of EFI support.

Hopefully MVC figures out how to flash the 1000 series and to mod for PCIe 2.0 support :)
 
Except the stock EVGA AIB 1080 card, they all come with 8+8 pins. That's not really safe, as it could draw more than 75w from either 8-pin, and we can only feed it 2x6-pin.



I assume you mean no GTX 1080? If you're only running Windows, sure. If you want to run OS X or both, hold off a bit. OS X web drivers by all accounts don't currently support the 1080, and when it does we'll likely be limited to PCIe 1.1 speeds due to a lack of EFI support.

Hopefully MVC figures out how to flash the 1000 series and to mod for PCIe 2.0 support :)
lol. Yeah. i mean 1080. lack EFI support...lol. that's true.
To be honest, i'm looking forward to nMP if they do release it (this june 13 is killing me)....any good GPU will do either from AMD or Nvidia (i'm neutral in GPU lol).

MVC is still gone lol.
[doublepost=1465423923][/doublepost]
You mean "no P300, P500, P700" ? ;)
lol. I'm neutral in GPU..so whatever comes first would be an improvement. lol.
 
Except the stock EVGA AIB 1080 card, they all come with 8+8 pins. That's not really safe, as it could draw more than 75w from either 8-pin, and we can only feed it 2x6-pin.



I assume you mean no GTX 1080? If you're only running Windows, sure. If you want to run OS X or both, hold off a bit. OS X web drivers by all accounts don't currently support the 1080, and when it does we'll likely be limited to PCIe 1.1 speeds due to a lack of EFI support.

Hopefully MVC figures out how to flash the 1000 series and to mod for PCIe 2.0 support :)

Curious... I thought the web drivers enabled PCIe 2.0 support on stock PC cards. I thought I confirmed it ran at 2.0 speeds on the Maxwell cards a while back.
 
Curious... I thought the web drivers enabled PCIe 2.0 support on stock PC cards. I thought I confirmed it ran at 2.0 speeds on the Maxwell cards a while back.

They do, under OS X. Under Windows, so I'm lead to believe, a stock PC Card is stuck at PCIe 1.1 without the resistor mod MVC does.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.