2016 nMP

[goMac]

Yes, MfM stuff is supported, but not the other way around. A R9 290X device ID in the drivers doesn't automatically make it "supported" or "Made for Mac". Apple doesn't ship one single device with a GCN 1.1 card in it. All recent AMD based Macs have one of those: Cape Verde, Pitcairn, Tahiti, Tonga. That's it, no Bonaire, no Hawaii, no Fiji.

It doesn't make it MfM, but it does mean you can actually file bugs with Apple on the device. They may remove it in response to filing bugs, but if it's there it's fair game. If Apple shipped a generic USB 3 driver, and it worked with a card but corrupted your data every week, that would be fair game for a bug too, even if the card wasn't MfM.

Hard to say without looking at the source code, but I'm quite sure that Apple is using a branch based on an older Nvidia driver where Metal supported has been merged in. The performance differences between those two (in earlier OS X versions, maybe still today, has to be checked) is an evidence for that.

It could be. MacVidCards claim that the Nvidia driver hasn't been updated at least just doesn't clearly hold water. If you look at the right version number, it's pretty clearly moving ahead each OS release.

Apple would require Nvidia provide them the source to the driver so Apple could actually compile it themselves, so it makes sense that the Apple version numbers would be moving, and the Nvidia one, which would probably be set by Nvidia's compile process would not be. Apple isn't taking the built versions and just bundling them.

The other reason I don't tend to think Apple forked it is because Apple compiles the Nvidia drivers themselves, but they don't maintain them. Nvidia would have contributed the Metal driver. It would be also really weird if Nvidia actually wrote features like OpenGL 4.1 in a fork.

The built in drivers will lag behind the web drivers though just due to testing schedules, and Apple adding another testing process on top. Code freezes can happen weeks or months in advance of a release, and if Nvidia misses a code freeze, their changes won't get in.
[doublepost=1455220809][/doublepost]

None of them have support for intel quicksync (FCPX fast encoding).
The low frequencies will make them slower than the iMac 5k in most creative apps. (Adobe CC, Logic ...)

These machines have built in GPUs. I don't see why QuickSync would be a requirement at all. That's why the Xeons don't ship with QuickSync. QuickSync is only needed if you're using integrated graphics, and those Xeons don't even support integrated graphics.

 

[MacVidCards]

Is it press release, or Interview? Reading comprehension is THAT bad in your case?

An "interview" with an AMD guy about all their great upcoming products. It's a PR link, like most of your posts. I'm sorry that you don't understand that asking an AMD guy about AMD products isn't really an "interview". I suspect Google Translate has failed you.

 

[pigsyn]

Pretty sure the Xeon E5 v2 line up does not have quick sync either and Apple was fine with building a computer around that architecture. Quick sync is only there with integrated graphics, which these E5 Xeons never have... so not sure I follow your logic there that Apple would wait for the v5's or something.

As for clock speeds, we only know what's been leaked so far, and they're pretty much inline with the v3 chips. Even though the base clocks are a little lower than the v2 series, there will be a massive leap (~15%) in instructions per clock going from v2 to v4, so even at lower clock speeds these new chips will perform much better than Ivy Bridge-EP.

So is your thought apple would wait for the Xeon E5 v5's then? What makes you think Intel would have those chips running at much higher clock speeds than the v4's?

You're absolutely right.
I was just saying that the Xeon E5 v4 lineup does not look very interesting for a single cpu workstation.
What's the point for Apple to release a new MacPro with E5 v4 cpu's, when they know a lot of mac users have high expectation for the next MacPro ?

Who wants a 12 cores 2.2 GHz MacPro ? At best it will be on-par with my current 5,1 12 cores on most apps that I use.

Since Apple love AMD so much, maybe they're waiting for a Zen Opteron.

 

[goMac]

Clock speeds in general are going down. It's just the way the industry is going. My current top end Macbook Pro has a lower clock speed than my old top end Macbook Pro. Reducing clock and upping cores is more efficient. As we get closer to pushing fabrication processes to their limits this will just get worse and worse.

Just expect year over year chips are going to have less clock and more cores.

 

[tuxon86]

Clock speeds in general are going down. It's just the way the industry is going. My current top end Macbook Pro has a lower clock speed than my old top end Macbook Pro. Reducing clock and upping cores is more efficient. As we get closer to pushing fabrication processes to their limits this will just get worse and worse.

Just expect year over year chips are going to have less clock and more cores.

Not when the vast majority of application are still single thread.

 

[Xteec]

Not when the vast majority of application are still single thread.

Those applications are designed for people who don't need workstations.

 

[jonisign]

You're absolutely right.
I was just saying that the Xeon E5 v4 lineup does not look very interesting for a single cpu workstation.
What's the point for Apple to release a new MacPro with E5 v4 cpu's, when they know a lot of mac users have high expectation for the next MacPro ?

Who wants a 12 cores 2.2 GHz MacPro ? At best it will be on-par with my current 5,1 12 cores on most apps that I use.

Since Apple love AMD so much, maybe they're waiting for a Zen Opteron.

I think it's safe to say we don't know the whole Xeon E5 v4 lineup yet, as lately Intel has had a large range of power targets for every core count option. If you look at the E5-16xx v4 options on wikipedia there is only 1 chip listed-- An 8 core chip which is clocked higher than Apple's current 8 core offering! Also, take a look at the top of the line Xeon E5-2699 v4 22 cores with a turbo up to 3.6 GHz! This chip is turbos higher than Apple's current 12 core offering and will surly trounce the singled threaded operation of the 12 core Xeon E5-2697 v2 used in the current Mac Pro which turbos up to a max of 3.5 GHz.

Yes, if you look at wikipedia there is a 2.2 GHz 12 core chip @ 105 W planned, but Apple seems to be targeting the ~130W chips for the Mac Pro. While a 2.2 GHz 12 core V4 @ 105W may not beat out the current Ivy Bridge 2.7 GHz 12 core in single threaded operation, I'd bet you whatever 12 core V4 Intel releases @ 130W will surely beat it in single threaded operation.

Also, what is the point of getting a workstation like this unless you're taking advantage of multi-threading? You're always going to get better single threaded performance from the iMac line because Intel's latest gen desktop chips are usually an architecture jump or process node ahead of Intel's server line chips.

 

[tuxon86]

Those applications are designed for people who don't need workstations.

Wrong...

Plenty of scientific and simulation softwares are single thread.

 

[Stacc]

Clock speeds in general are going down. It's just the way the industry is going. My current top end Macbook Pro has a lower clock speed than my old top end Macbook Pro. Reducing clock and upping cores is more efficient. As we get closer to pushing fabrication processes to their limits this will just get worse and worse.

Just expect year over year chips are going to have less clock and more cores.

Remember, just because clock speeds have stayed the same doesn't mean that performance has. Intel has been increasing IPC (instructions per clock, i.e. performance per clock) with each generation. Also, frequencies have been moving slowly up since sandy bridge. A dual core 2.66 Ghz processor in a macbook pro from 2010 will be much slower than a quad core 2.2 Ghz Macbook Pro from 2015, even in single threaded tasks.

Those applications are designed for people who don't need workstations.

Not true. Many tasks are still inherently single threaded. For instance many physics simulations are single threaded because you can't compute the next timestep until you have computed the current one. You can't divide the workload up like you can when you are encoding video. Obviously multithreaded performance is important but saying that single threaded performance doesn't matter for workstations is silly.

 

[tuxon86]

Remember, just because clock speeds have stayed the same doesn't mean that performance has. Intel has been increasing IPC (instructions per clock, i.e. performance per clock) with each generation. Also, frequencies have been moving slowly up since sandy bridge. A dual core 2.66 Ghz processor in a macbook pro from 2010 will be much slower than a quad core 2.2 Ghz Macbook Pro from 2015, even in single threaded tasks.



Not true. Many tasks are still inherently single threaded. For instance many physics simulations are single threaded because you can't compute the next timestep until you have computed the current one. You can't divide the workload up like you can when you are encoding video. Obviously multithreaded performance is important but saying that single threaded performance doesn't matter for workstations is silly.

Exactly.

 

[koyoot]

Im sure Aiden will be interested in this: http://www.bitsandchips.it/9-hardware/6621-rumor-scaletta-di-presentazione-delle-gpu-pascal

In April only Tesla GPU based on Pascal arch. It will be only presented. Late Q3 for consumer GPUs. It is obvious why. GDDR5X mass production starts in summer, and HBM2 will make the GPUs appear in late Q3 at best. That is unfortunate. Also we cannot forget that those GPUs are made on 16 nm FF+ from TSMC. Which is supposed to be massively bought by Apple and their SoC's. Which also slows down production on the rest of the GPUs.

So it really looks like AMD is few months faster in terms of releasing the silicon to market.

 

[tuxon86]

Im sure Aiden will be interested in this: http://www.bitsandchips.it/9-hardware/6621-rumor-scaletta-di-presentazione-delle-gpu-pascal

In April only Tesla GPU based on Pascal arch. It will be only presented. Late Q3 for consumer GPUs. It is obvious why. GDDR5X mass production starts in summer, and HBM2 will make the GPUs appear in late Q3 at best. That is unfortunate. Also we cannot forget that those GPUs are made on 16 nm FF+ from TSMC. Which is supposed to be massively bought by Apple and their SoC's. Which also slows down production on the rest of the GPUs.

So it really looks like AMD is few months faster in terms of releasing the silicon to market.

You had to search and find an italian site to support your claim...

 

[Xteec]

Not true. Many tasks are still inherently single threaded. For instance many physics simulations are single threaded because you can't compute the next timestep until you have computed the current one. You can't divide the workload up like you can when you are encoding video. Obviously multithreaded performance is important but saying that single threaded performance doesn't matter for workstations is silly.

I'm not going to pretend I knew this but that is an interesting thing I will file away in my nice to know box so thanks for the info.

 

[ManuelGomes]

Anything against Italians? I'm not Italian mind you, but I can't possibly see why they should be any worse or better than anyone else. Care to explain? Or are now only the US based sites the ones to go by?
Or is it that maybe the news in question aren't of your liking?
There's always Google Translate

 

[tuxon86]

Anything against Italians? I'm not Italian mind you, but I can't possibly see why they should be any worse or better than anyone else. Care to explain? Or are now only the US based sites the ones to go by?
Or is it that maybe the news in question aren't of your liking?
There's always Google Translate

If you search long enough on google you'll find a site somewhere that validate your own personnal point of view. The fact that Koyoot had to go to an obscure italian hardware site to find an article to validate his view instead of one of the major one is quite telling. This isn't an attack against Italian, it's against a deceitful tactic to push an agenda.

 

[koyoot]

If you search long enough on google you'll find a site somewhere that validate your own personnal point of view. The fact that Koyoot had to go to an obscure italian hardware site to find an article to validate his view instead of one of the major one is quite telling. This isn't an attack against Italian, it's against a deceitful tactic to push an agenda.

So far, the only one who is trying to prove anything is you. I found it on anandtech Forum, so I reposted here.

http://forums.anandtech.com/showthread.php?t=2463869 Look at time, and date, and compare it to my post.

 

[Stacc]

If you search long enough on google you'll find a site somewhere that validate your own personnal point of view. The fact that Koyoot had to go to an obscure italian hardware site to find an article to validate his view instead of one of the major one is quite telling. This isn't an attack against Italian, it's against a deceitful tactic to push an agenda.

Koyoot likes to present his statements as fact but the reality is very little is known regarding Nvidia's and AMD's GPUs that will be released this year. Any specifics regarding which GPUs and when are pure speculation at this point.

At best his information is based on rumors from various places and message boards around the internet. It doesn't mean he is wrong, it just means it should be taken with a grain of salt.

 

[tuxon86]

So far, the only one who is trying to prove anything is you. I found it on anandtech Forum, so I reposted here.

http://forums.anandtech.com/showthread.php?t=2463869 Look at time, and date, and compare it to my post.

So a forum post that link to the same website is a proof that NVidia won't deliver? Really?

 

[lowendlinux]

My gut says that AMD will beat nVidia with new production GPU's this gen. I'm also not convinced Pascal is going to be all that great.

 

[tuxon86]

My gut says that AMD will beat nVidia with new production GPU's this gen. I'm also not convinced Pascal is going to be all that great.

Maybe, maybe not, no one but AMD/NVidia really know which also happens to disqualify Koyoot magic crystal ball...

 

[koyoot]

So a forum post that link to the same website is a proof that NVidia won't deliver? Really?

So thats what is bothering you...

No. I never said they will not deliver. The fact that there are no GPUs on Zauba is worrying, for Nvidia here. If there would be Pascal GPUs already shipping on Zauba - there would not be problem. So far, there are only boards for GPUs, without GPUs. Second thing: Nvidia will use GDDR5X and HBM2. GDDR5X Production starts in the summer. HBM2 started, but GPUs available with it for consumer market are not due to late Q3.

As for AMD, first silicon on Zauba shipped as early as late november last year. 3-6 months, and it is ready for release on market.

Edit: http://semiaccurate.com/2016/02/01/news-of-nvidias-pascal-tapeout-and-silicon-is-important/ Read it.

 

[MacVidCards]

So thats what is bothering you...

I think what is bothering him is the never-ending posting of AMD PR. Link after link claiming great GRAND things coming from AMD. Links to AMAZING things coming in the "Vapourware 5000" line, due any day now.

It gets old. And when you post two foreign language links back to same rumour post, it gets into the "clutching at straws" realm.

 

[ManuelGomes]

The info on Pascal coming later in the year is pretty much everywhere.

 

[tuxon86]

So thats what is bothering you...

No. I never said they will not deliver. The fact that there are no GPUs on Zauba is worrying, for Nvidia here. If there would be Pascal GPUs already shipping on Zauba - there would not be problem. So far, there are only boards for GPUs, without GPUs. Second thing: Nvidia will use GDDR5X and HBM2. GDDR5X Production starts in the summer. HBM2 started, but GPUs available with it for consumer market are not due to late Q3.

As for AMD, first silicon on Zauba shipped as early as late november last year. 3-6 months, and it is ready for release on market.

Edit: http://semiaccurate.com/2016/02/01/news-of-nvidias-pascal-tapeout-and-silicon-is-important/ Read it.

An article based on rumors... Got anything official? I know you love posting AMD press release as gospel, maybe you could find some credible information instead of C&P links to rumors sites. Also the gddr5x/pascal is also a rumor...

 

[ManuelGomes]

Foreign to you maybe, but it all seems foreign to you anyway...
[doublepost=1455309548][/doublepost]koyoot, Intel Extreme Masters in Poland next month, going?