or Connect
AppleInsider › Forums › Mac Hardware › Future Apple Hardware › Ivy Bridge E due in Q3 2013 - What does this mean for the Mac Pro?
New Posts  All Forums:Forum Nav:

Ivy Bridge E due in Q3 2013 - What does this mean for the Mac Pro?

post #1 of 33
Thread Starter 
How is progress on Thunderbolt integration with Sandy Bridge E? Will we see a new Mac Pro Q1 2013?
post #2 of 33

Doesn't matter. The next Mac Pro is Sandy Bridge Xeon, which are already out. We should (better) absolutely see a Mac Pro next year.

Originally posted by Relic

...those little naked weirdos are going to get me investigated.
Reply

Originally posted by Relic

...those little naked weirdos are going to get me investigated.
Reply
post #3 of 33
Thread Starter 
Is Intel the only option at the moment for the Mac Pro if it stays long-term? Just curious.
post #4 of 33

Yep. 

Originally posted by Relic

...those little naked weirdos are going to get me investigated.
Reply

Originally posted by Relic

...those little naked weirdos are going to get me investigated.
Reply
post #5 of 33
I had a long answer for you then I hit cancel accidentally.
Quote:
Originally Posted by Winter View Post

How is progress on Thunderbolt integration with Sandy Bridge E?
Don't know don't care. TB can be implemented externally no problem.
Quote:

Will we see a new Mac Pro Q1 2013?

Well if we don't, Apples leadership team will look like idiots. These people don't like to look like idiots!

As to what the machine will look like, I have a strong feeling that Intels MIC architecture will play a role in that new machine. The same old Intel XEON architecture won't cut it for future Mac Pros.
post #6 of 33
Quote:
Originally Posted by Winter View Post

Is Intel the only option at the moment for the Mac Pro if it stays long-term? Just curious.

It depends upon what you mean. They could go Intel MIC architecture. The only other option is AMD and that would only work if they could beef up their high performance chips.
post #7 of 33
Quote:
Originally Posted by Tallest Skil View Post

Doesn't matter. The next Mac Pro is Sandy Bridge Xeon, which are already out. We should (better) absolutely see a Mac Pro next year.

I figured they'd wait for ivy at this point if it has a relatively complete lineup (unlike westmere). It could be a mix. It's unlikely that they're going to update this on short cycles even given the relatively minimal testing required with pin compliant updates.

Quote:
Originally Posted by wizard69 View Post

I had a long answer for you then I hit cancel accidentally.
Don't know don't care. TB can be implemented externally no problem.
Well if we don't, Apples leadership team will look like idiots. These people don't like to look like idiots!
As to what the machine will look like, I have a strong feeling that Intels MIC architecture will play a role in that new machine. The same old Intel XEON architecture won't cut it for future Mac Pros.

 

Thunderbolt currently requires some kind of displayport connection on the logic board. I don't think it's as important as people suggest on such a machine. You gain some compliance with recent Apple peripherals, but there are better existing solutions available for mac pro like hardware. The thunderbolt thing seems to be more of an issue of whether people feel Apple cares here. They may have to change board designs somewhat. The daughter board + backplane design may have issues depending on how the thunderbolt chip must be situated relative to the cpus as the pci lanes themselves migrated onto the cpu package. I'm not sure what the placement constraints are on the thunderbolt chip itself. If they do go this route, they may end up locked into embedded graphics options. Again I'm not sure of good options here. I can't think of any extremely powerful options available in such a form factor. They could implement something cheap in that regard with an option for a second card via traditional PCI cards, but you may run into bandwidth constraints on the single socket models assuming they don't go back to over-subscribing the lanes as they've done with prior generations. I know I'm leaving out some detail, but I couldn't find any definitive references on some of it.

 

I don't entirely agree with Marvin that everything will go to thunderbolt. Given the trend toward further integration, I can't see how it would be advantageous to add further distance even at his claimed latency specs. It also ignores that 200W+ graphics cards are still produced today. It will be interesting to see where thermal profiles and power draw go over the next few years. They're not a complete non issue in any market segment. Even in large server farms density and power consumption are always factors that would be addressed. If Apple is going to retain the mac pro they may need to retain some high performance gpus options. Apple lacks cards with workstation drivers and large amounts of vram, so the mid to high end gaming card designs fill this niche under OSX. Companies that would test Quadro/Firepro drivers under Windows test these on OSX because they represent the majority of the user base. When you drop to the imac realm of embedded graphics, it's a pretty significant hit at equivalent price levels. The 6970m probably cost as much as the 5870 if examining time of release on both cards. Something like a 7970 would provide some amazing benefits for the mac pro, but I think they'll delay another generation. If updates were only months away, I don't think we would have seen the price adjustments on the mac pro line. That was the first time they did such a thing since the PowerPC era.

post #8 of 33
Originally Posted by hmm View Post
I figured they'd wait for ivy at this point if it has a relatively complete lineup (unlike westmere). It could be a mix. It's unlikely that they're going to update this on short cycles even given the relatively minimal testing required with pin compliant updates.

 

Wait ANOTHER ENTIRE YEAR for a new product when it should have come out months ago? I don't think so. If they really wanted to kill it (and that would kill it), they would have discontinued it at WWDC.

Originally posted by Relic

...those little naked weirdos are going to get me investigated.
Reply

Originally posted by Relic

...those little naked weirdos are going to get me investigated.
Reply
post #9 of 33
Quote:
Originally Posted by Tallest Skil View Post

 

Wait ANOTHER ENTIRE YEAR for a new product when it should have come out months ago? I don't think so. If they really wanted to kill it (and that would kill it), they would have discontinued it at WWDC.


I agree that could kill it, but I remain skeptical that they would make such adjustments (adding in one more cpu and shuffling pricing slightly) for a 6 month extension. Other oems weren't shipping until around June, and Apple has fallen behind them in the past. I would also note that Ivy Bridge E is still scheduled for next year. They haven't given any signs that their plans have changed there. The x86 performance gains aren't that impressive at the lower end compared to what we have currently. Sandy Bridge E or the Ivy equivalents would bring usb3, PCIe 3.0, updated gpus, etc.

 

Anyway my expectations are set pretty low regarding that line at the moment. Most of my own computing requirements have been relatively flat over the past couple years. I can get by with 16GB of ram and almost whatever gpu under OSX. If I ran primarily Windows they'd be a bit more specific, yet I couple pick up some stuff that is unavailable under OSX. Anyway my concerns with the future of the line are somewhat diminished from what they were. I guess if Ivy is sliding that far back you could see a new Mac Pro early in the year with Ivy at the beginning of 2014 if it really slid as far as Q3. I kind of wonder what intel will do with consumer Haswell at this point. Do they want Haswell slipping out prior to Ivy E? It could kill enthusiast class hardware sales, and i7 versions usually come out prior to Xeons. Overall I'm not sure where all of this will end. I primarily disagree that outside the case modularity is the future when the primary long term trend is toward tighter integration. I don't foresee people ever running 200W+ gpus over thunderbolt even though macrumors is plagued by these suggestions.

post #10 of 33
Quote:
Originally Posted by hmm View Post

I figured they'd wait for ivy at this point if it has a relatively complete lineup (unlike westmere). It could be a mix. It's unlikely that they're going to update this on short cycles even given the relatively minimal testing required with pin compliant updates.

Thunderbolt currently requires some kind of displayport connection on the logic board.
Interesting. I'm not sure it is required, after all a Mini can run to an HDMI display with no Display Port data going over TB. I would suspect that the "requirement" is really one of Apples.
Quote:
I don't think it's as important as people suggest on such a machine. You gain some compliance with recent Apple peripherals, but there are better existing solutions available for mac pro like hardware. The thunderbolt thing seems to be more of an issue of whether people feel Apple cares here.
I disagree here, TB could be very important to a Mac Pro replacement machine.
Quote:
They may have to change board designs somewhat. The daughter board + backplane design may have issues depending on how the thunderbolt chip must be situated relative to the cpus as the pci lanes themselves migrated onto the cpu package. I'm not sure what the placement constraints are on the thunderbolt chip itself. If they do go this route, they may end up locked into embedded graphics options.
Actually this is what I expected in this years Mac Pro update. Sadly we got screwed there.
Quote:
Again I'm not sure of good options here. I can't think of any extremely powerful options available in such a form factor. They could implement something cheap in that regard with an option for a second card via traditional PCI cards, but you may run into bandwidth constraints on the single socket models assuming they don't go back to over-subscribing the lanes as they've done with prior generations. I know I'm leaving out some detail, but I couldn't find any definitive references on some of it.
PCI Express 3 takes care of many of these issues. As for an on board GPU it is easier than ever to put a high performance chip on the motherboard.
Quote:

I don't entirely agree with Marvin that everything will go to thunderbolt.
Yes that is nonsense.
Quote:
Given the trend toward further integration, I can't see how it would be advantageous to add further distance even at his claimed latency specs. It also ignores that 200W+ graphics cards are still produced today. It will be interesting to see where thermal profiles and power draw go over the next few years. They're not a complete non issue in any market segment.
The trend is clear, it is towards denser motherboards, motherboards that support far more functionality. As to a GPU it is pretty easy to put one with respectable performance on a motherboard these days. As to total power there isn't much difference between the power being used on the motherboard or on a plug in card.
Quote:
Even in large server farms density and power consumption are always factors that would be addressed. If Apple is going to retain the mac pro they may need to retain some high performance gpus options. Apple lacks cards with workstation drivers and large amounts of vram, so the mid to high end gaming card designs fill this niche under OSX. Companies that would test Quadro/Firepro drivers under Windows test these on OSX because they represent the majority of the user base. When you drop to the imac realm of embedded graphics, it's a pretty significant hit at equivalent price levels. The 6970m probably cost as much as the 5870 if examining time of release on both cards. Something like a 7970 would provide some amazing benefits for the mac pro, but I think they'll delay another generation.
2013? By that time the 7970 will be old.
Quote:
If updates were only months away, I don't think we would have seen the price adjustments on the mac pro line. That was the first time they did such a thing since the PowerPC era.
Well even with the price adjustments they still screwed up what should have been an update. The Mac Pro update was handled so poorly that Apple will have to make good on its 2013 promises. The current Mac Pro is so antiquated that the 2013 machine will have to be rather impressive to recapture lost momentum.
post #11 of 33
Quote:
Originally Posted by wizard69 View Post


Interesting. I'm not sure it is required, after all a Mini can run to an HDMI display with no Display Port data going over TB. I would suspect that the "requirement" is really one of Apples.

Displayport 1.2 to hdmi adapters do exist. I'm not sure whether mini displayport to hdmi (what you'd find here) would offer audio output. Displayport 1.2 added audio and a few other things, where thunderbolt seems to  use the older standard which has fewer features.

Quote:
Actually this is what I expected in this years Mac Pro update. Sadly we got screwed there.

The daughterboard design is likely quite cost effective. They've got a single backplane, yet they're not stuck absorbing the cost of using a full dual package board with one socket vacant, and they're able to use the cheaper cpus in the single package variant this way. I'm just not sure if thunderbolt presents any problems with such a configuration. It's also possible they may run into lane restrictions with the single package model if implementing thunderbolt + their typical configuration. They have previously over-subscribed the PCI bus, so it's always a possibility. The 1,1 allowed them to be allocated in several configurations.

 

 

Quote:
PCI Express 3 takes care of many of these issues. As for an on board GPU it is easier than ever to put a high performance chip on the motherboard.

It does, but newer cards are still carrying x16 PCI 3 specifications even if they don't fully saturate such a connection. It also makes no difference when it comes to single lane connections, so you can still run into issues. I am not as familiar with embedded graphics. Some of the rather large form factors allow for reasonably quiet cooling of cards in the 200W+ realm. Some desktop and workstation class gpus draw a ton of power, although gaming gpus are typically clocked higher than workstation variants for whatever reason aside from the differences in drivers.

 

 

Quote:
Yes that is nonsense.

I think he gets carried away with stuff sometimes, but I don't see a trend toward modular computing as a way to continue to accommodate specialty devices. The topic of external gpus comes up frequently when there isn't a chassis out there that can take anything past a mid range desktop gpu. Even a 150W external PCI chassis is quite expensive, and it's not likely to ever become a mainstream item. It quickly reaches the point of being truly pointless as there are often better ways to approach the problem for less money.

 

 

Quote:

 

The trend is clear, it is towards denser motherboards, motherboards that support far more functionality. As to a GPU it is pretty easy to put one with respectable performance on a motherboard these days. As to total power there isn't much difference between the power being used on the motherboard or on a plug in card.
2013? By that time the 7970 will be old.

 

Intel has been moving more things on to the cpu package. I wonder if they'll have more problems like last year with the sandy bridge recall. The issue of power was more in its dissipation. Some gaming cards can hit 300W of power draw. Anandtech has tested some higher than that (hehe...that was an intentional reference >=)> ). On the 79700, a newer card would be dependent on when such an update is released.

 

 

Quote:
Well even with the price adjustments they still screwed up what should have been an update. The Mac Pro update was handled so poorly that Apple will have to make good on its 2013 promises. The current Mac Pro is so antiquated that the 2013 machine will have to be rather impressive to recapture lost momentum.

 

It's shocking to me that they released that. I took it to signify that a real update was pretty far out. In 2010 Apple updated months after some of the other oems. Even at that time Sandy Bridge E was already behind. It was scheduled for the second half of 2011 around September and slipped further from there. In this case the other oems had just released their updates. I don't think you would have seen the price adjustments + additional cpu option if they weren't quite far behind.


Edited by hmm - 8/3/12 at 8:51am
post #12 of 33
Quote:
Originally Posted by hmm View Post

Displayport 1.2 to hdmi adapters do exist. I'm not sure whether mini displayport to hdmi (what you'd find here) would offer audio output. Displayport 1.2 added audio and a few other things, where thunderbolt seems to  use the older standard which has fewer features.
I probably was tired when writing that last response. What I meant it that the Mini supports both TB and an HDMI port, so it is likely that TB ports could function with no internal Display Port connections at all. in otherwords a data only TB port on a Mac Pro seems to be possible. I do see it is unlikely though because Apple wants to be consistent with its products.
Quote:

The daughterboard design is likely quite cost effective. They've got a single backplane, yet they're not stuck absorbing the cost of using a full dual package board with one socket vacant, and they're able to use the cheaper cpus in the single package variant this way. I'm just not sure if thunderbolt presents any problems with such a configuration. It's also possible they may run into lane restrictions with the single package model if implementing thunderbolt + their typical configuration. They have previously over-subscribed the PCI bus, so it's always a possibility. The 1,1 allowed them to be allocated in several configurations.
The problem becomes even worst if they do decide to support Intels MIC hardware which also interfaces over PCI Express. MIC makes me wonder if Intel has plans for a new XEON to support it in a workstation design.

Daughter cards for GPUs and even SSDs do give Apple a chance to address one of the big issues with desktop PC design which is card cooling. PC expansion card architecture is rather pathetic in this regard.
Quote:

It does, but newer cards are still carrying x16 PCI 3 specifications even if they don't fully saturate such a connection. It also makes no difference when it comes to single lane connections, so you can still run into issues. I am not as familiar with embedded graphics. Some of the rather large form factors allow for reasonably quiet cooling of cards in the 200W+ realm. Some desktop and workstation class gpus draw a ton of power, although gaming gpus are typically clocked higher than workstation variants for whatever reason aside from the differences in drivers.
As to embedded graphics my thought there is embedded as in soldered to the motherboard. With the current GPU hardware on the market a decent sub 200 watt GPU can be had right on the motherboard solving the TB routing problem. Yes a daughter card would also work but would also be expensive and proprietary.

I think the thing to remember here is that this will be a 2013 machine. As such it will have access to, hopefully stable, 28nm and smaller GPU designs.
Quote:

I think he gets carried away with stuff sometimes, but I don't see a trend toward modular computing as a way to continue to accommodate specialty devices.
It isn't. For all of its faults PC slots provide for a well accepted standard. The economics of specialty devices does not allow for random external implementations that are specific to one brand of interface. Believe me TB is a branded interface that many users will avoid because of the lack of wide spread support.
Quote:
The topic of external gpus comes up frequently when there isn't a chassis out there that can take anything past a mid range desktop gpu. Even a 150W external PCI chassis is quite expensive, and it's not likely to ever become a mainstream item. It quickly reaches the point of being truly pointless as there are often better ways to approach the problem for less money.
I've never understood this either, external GPUs will never get large scale traction in the marketplace simply because the demand isn't strong enough to make them economical and right now TB is no where near fast enough.
Quote:
Intel has been moving more things on to the cpu package. I wonder if they'll have more problems like last year with the sandy bridge recall. The issue of power was more in its dissipation. Some gaming cards can hit 300W of power draw. Anandtech has tested some higher than that (hehe...that was an intentional reference >=)> ). On the 79700, a newer card would be dependent on when such an update is released.
What Sandy Bridge recall? Honestly I must have been sleeping when that happened as I never heard of a thermal related SB recall. I did hear of a design issue with a transistor but really that is nothing new and is why companies come out with new steppings from time to time.

As to power draw engineers today can basically engineer a card for any thermal point. It is a matter of how much performance you can accept in a given thermal envelope that determines if it is practical to build the GPU into the motherboard.
Quote:
It's shocking to me that they released that. I took it to signify that a real update was pretty far out. In 2010 Apple updated months after some of the other oems. Even at that time Sandy Bridge E was already behind. It was scheduled for the second half of 2011 around September and slipped further from there. In this case the other oems had just released their updates. I don't think you would have seen the price adjustments + additional cpu option if they weren't quite far behind.

I took it to signify that the current Mac Pro architecture is dead. If Apple comes out with a new machine targeting the pro market next year I expect it to be radically different than today's Mac Pro. It is really the only justification I can see for the approach they are taking here. Deader than a door nail is probably a phrase from my childhood that applies to the Mac Pro.

Such a machine could be a good thing or a bad thing depending upon how forward looking Apple is. I just see Intel making lots of noise about new technologies that could go into such a machine. Combine that with the efforts of other chip makers and I see a potentially ground breaking Mac Pro in the future. That is if Apple has the stones to move forward here as it would be a gamble.
post #13 of 33
Thread Starter 
Quote:
Originally Posted by wizard69 View Post

I just see Intel making lots of noise about new technologies that could go into such a machine. Combine that with the efforts of other chip makers and I see a potentially ground breaking Mac Pro in the future. That is if Apple has the stones to move forward here as it would be a gamble.

At some point, the consumer buzz with Apple is probably going wane, no? By then, will the Pro market still be around to grab hold.
post #14 of 33
Quote:
Originally Posted by Winter View Post

At some point, the consumer buzz with Apple is probably going wane, no? By then, will the Pro market still be around to grab hold.

Will pro users still be around? Yes! Not every pro user is emotionally short changed like some of the video pros that go around demanding that Apple build the machine they want. On this forum people seem to equate Pro users with graphics / video professionals, but there are far more pro users than that niche. Those users should actually be expanding as more and more software houses port their software to the platform.

Realize also that many pro users have moved to laptops. If you are an engineer it is far more desirable to move your hardware to the job than it is to move the job to your hardware. It is only recently that the trade offs between a desktop and a laptop became favorable to the laptop. Even then many engineers just love a performance machine on the desk. They are also smart enough to wait for the idea machine.

So yeah Apple has screwed up the handling of the Mac Pro significantly. This will and likely has caused people to leave the platform. Personally the farther away they are the better. However a significant advancement in the Mac Pros capability could bring people back. Especially if Apple has signed up intel for exclusivity or special products. Look at it this way, if Apple can deliver a 24 core machine in the same price range as today's Mac Pro people will take interest. More so if they can demonstrate could software utilization of those cores. 24 is only a target here, intel has already indicated far more cores will be supported in the MIC family.

So the possibility is there for Apple to recover from this blunder. But it won't be a freebie they will have to work hard to recapture market share and demostrate value in the new machine.
post #15 of 33
Quote:
Originally Posted by wizard69 View Post


Will pro users still be around? Yes! Not every pro user is emotionally short changed like some of the video pros that go around demanding that Apple build the machine they want. On this forum people seem to equate Pro users with graphics / video professionals, but there are far more pro users than that niche. Those users should actually be expanding as more and more software houses port their software to the platform.
Realize also that many pro users have moved to laptops. If you are an engineer it is far more desirable to move your hardware to the job than it is to move the job to your hardware. It is only recently that the trade offs between a desktop and a laptop became favorable to the laptop. Even then many engineers just love a performance machine on the desk. They are also smart enough to wait for the idea machine.
So yeah Apple has screwed up the handling of the Mac Pro significantly. This will and likely has caused people to leave the platform. Personally the farther away they are the better. However a significant advancement in the Mac Pros capability could bring people back. Especially if Apple has signed up intel for exclusivity or special products. Look at it this way, if Apple can deliver a 24 core machine in the same price range as today's Mac Pro people will take interest. More so if they can demonstrate could software utilization of those cores. 24 is only a target here, intel has already indicated far more cores will be supported in the MIC family.
So the possibility is there for Apple to recover from this blunder. But it won't be a freebie they will have to work hard to recapture market share and demostrate value in the new machine.

I don't entirely agree with you here. There was a lot of tension around Apple's direction for a long time. FCP 7 was still 32 bit which was somewhat limiting for an application that addresses large sets of data. One of the main sticking complaints seemed to be lack of multi-cam support. The reason people equate professionals with such trades on Mac forums is pretty much an issue of industry association. Final Cut Pro and possibly Shake brought a lot of that to early OSX. Desktop publishing has an extremely long history on the Mac. I would imagine that the need to upgrade now group probably purchased their last machines in more like 2008-2009. If they're on a 6-12 core mac pro, it's unlikely that they already need to upgrade. Many of these workloads don't gain weight very quickly, and some of them aren't even totally dependent on the number of X86 cores. With the desktop you have some ability to customize a solution if the available stock configurations are not a good fit. The late rollout with Sandy Bridge E, rising cpu prices at the high end, and slow growth at equivalent price levels may have significantly offset the volume in this segment as budgets aren't totally elastic, and for such users stability is often a prime factor. It's not always necessary to be on the latest thing to accomplish work unless certain things aren't supported on the old one. Speaking of that, Adobe made some improvements with Photoshop CS6 that finally brought a couple portions of that silly program up to par with Painter and Manga Studio. It's still weak for linear compositing. Some of these iconic applications become incredibly stagnant over time. People still use them and maintain existing workflows. No matter how many bugs, users find something that enables them to finish work and they go with it. In the case of FCPX, some of the features that were not included likely produced a kneejerk effect. Overall I wouldn't take the forum complaints as an unbiased sampling. Creativecow had a lot of discussion on the issue too, yet it maintained a different tone as the primary focus of the site isn't Apple.

 

Regarding a 24 core mac pro, I'm not sure how they'd really implement one of Intel's solutions for the HPC market. I don't see that as likely. Looking at the hardware family currently in use, it will go up to 20 cores next year with Ivy Bridge E, but that might be a year away. I kind of wonder regarding Intel's long term refresh cycle goals are at this point. 

Quote:
Originally Posted by wizard69 View Post


I probably was tired when writing that last response. What I meant it that the Mini supports both TB and an HDMI port, so it is likely that TB ports could function with no internal Display Port connections at all. in otherwords a data only TB port on a Mac Pro seems to be possible. I do see it is unlikely though because Apple wants to be consistent with its products.
 

Intel doesn't have a currently certified data only thunderbolt spec. Light ridge supposedly supports 4 channels x 10Gb/s which makes sense given the breakdown. That can feed two ports and display data consumes half the bandwidth. That bandwidth can't be used for raw data. Displayport 1.2 is a displayport only setup using 4 lanes with an effective throughput of around 16.5Gb/s. It has a raw bandwidth of more like 20 over 4 channels, but it loses some of that in coding overhead. While it may not be fully saturated, to run displayport 1.2 devices over thunderbolt properly, I believe it still takes consumes all channels on a port due to the extra bits that are transported. It allows for audio output and 10 bit displayport connections, all of which will eventually become more mainstream. The more wider gamut displays become popular (already popular at a prosumer level on new purchases), the 10 bit displayport will become a desirable feature to alleviate some of the dithering concerns. This is also valuable with the increasing contrast ratios to avoid stair stepping. While dithering works, it can cause problems if it isn't implemented well, and not all of these displays are sold at incredible margins. Okay I'm starting to ramble.

 

 

 

Quote:

 

The problem becomes even worst if they do decide to support Intels MIC hardware which also interfaces over PCI Express. MIC makes me wonder if Intel has plans for a new XEON to support it in a workstation design.
Daughter cards for GPUs and even SSDs do give Apple a chance to address one of the big issues with desktop PC design which is card cooling. PC expansion card architecture is rather pathetic in this regard.
As to embedded graphics my thought there is embedded as in soldered to the motherboard. With the current GPU hardware on the market a decent sub 200 watt GPU can be had right on the motherboard solving the TB routing problem. Yes a daughter card would also work but would also be expensive and proprietary.

 

I don't think the imac embedded graphics are soldered in. The 6970m is around 100W. The 5870 option in the mac pro should be around 300-400W under load according to some of the articles on the web. This isn't that atypical. Most workstations use workstation cards by default. These are typically in the 100-200W range. NVidia lists the Quadro 4000 as 162W maximum. I don't think it would really hinder anyone if they went this route. I doubt many people buy a mac pro for gaming.

 

Quote:

 

I think the thing to remember here is that this will be a 2013 machine. As such it will have access to, hopefully stable, 28nm and smaller GPU designs.
It isn't. For all of its faults PC slots provide for a well accepted standard. The economics of specialty devices does not allow for random external implementations that are specific to one brand of interface. Believe me TB is a branded interface that many users will avoid because of the lack of wide spread support.

 

I'm somewhat concerned about it becoming much like firewire. Thunderbolt isn't fast enough for some super high bandwidth requirements. Mini SAS is a better connection choice for truly high performance storage solutions. USB3 is much cheaper. The thing that's nice about TB is that it functions quite well in some of Apple's reference configurations. You can hook up a couple thunderbolt displays and the Pegasus raid over a port requiring a single cable. That is kind of cool. Future revisions may make it more attractive if it outpaces other technologies, but it's not that efficient where it stands today, and it's not scheduled to go to PCIe 3.0 standards until 2014. It remains to be seen how expensive it will be at that time. Intel has warned it could be costly to implement such bandwidth.

 

 

Quote:
I've never understood this either, external GPUs will never get large scale traction in the marketplace simply because the demand isn't strong enough to make them economical and right now TB is no where near fast enough.

 

I guess a few people would like to run the latest games at the highest settings on their macbook air machines, although it's still possible to hit a point where the cpu becomes the bottleneck. Beyond that right now to get a 150W power source external PCI chassis, you can spend $500+ for the chassis alone. At some point it becomes more practical just to own a desktop solution for such a thing. Given that acceptable integrated graphics solutions are starting to pop up for many tasks, I don't see a bright future for a cost effective external solution. Integrated graphics even ship on some desktop hardware today. While they aren't currently capable of displacing the $100+ discrete gpu market, they're getting better. In intel's case they're no longer an after thought, and that in itself is an improvement.

 

 

Quote:
What Sandy Bridge recall? Honestly I must have been sleeping when that happened as I never heard of a thermal related SB recall. I did hear of a design issue with a transistor but really that is nothing new and is why companies come out with new steppings from time to time.
As to power draw engineers today can basically engineer a card for any thermal point. It is a matter of how much performance you can accept in a given thermal envelope that determines if it is practical to build the GPU into the motherboard.
I took it to signify that the current Mac Pro architecture is dead. If Apple comes out with a new machine targeting the pro market next year I expect it to be radically different than today's Mac Pro. It is really the only justification I can see for the approach they are taking here. Deader than a door nail is probably a phrase from my childhood that applies to the Mac Pro.

 

Sandy Bridge had problems with its SATA implementation. It cost intel a huge amount of money, and this probably influenced them to to be more cautious on Sandy Bridge E. Here's a description from your favorite site. Tom's hardware posted a recap article as well.

post #16 of 33
Quote:
Originally Posted by wizard69 View Post


Will pro users still be around? Yes! Not every pro user is emotionally short changed like some of the video pros that go around demanding that Apple build the machine they want. On this forum people seem to equate Pro users with graphics / video professionals, but there are far more pro users than that niche. 

Researchers doing lab work are too busy to frequent apple fan boy boards, but anyone involved in research is not depending on any type of desktop machine to do the actual number crunching anyway.  They're using distributed computing and clusters of servers.  And those people don't give Apple a second thought.

post #17 of 33
You kinda prove my point, this board is populated by sxtremist views on what it is to be a professional. The fact is many professionals have switched to laptop use for the advantageous they provide. Yes even professionals in research labs. I know of a few of these organizations and frankly your image of an R & D facility with clusters of servers at their disposal is not the norm.

Professional use of Mac hardware encompasses a far wider array of users than many seem to want to admit to here. Few out there work on projects for NASA or even the automobile companies for that matter. The same thing goes for people doing movie editing, not many are making feature length movies.
Quote:
Originally Posted by Conrail View Post

Researchers doing lab work are too busy to frequent apple fan boy boards, but anyone involved in research is not depending on any type of desktop machine to do the actual number crunching anyway.  They're using distributed computing and clusters of servers.  And those people don't give Apple a second thought.
post #18 of 33
Quote:
Originally Posted by wizard69 View Post

You kinda prove my point, this board is populated by sxtremist views on what it is to be a professional. The fact is many professionals have switched to laptop use for the advantageous they provide. Yes even professionals in research labs. I know of a few of these organizations and frankly your image of an R & D facility with clusters of servers at their disposal is not the norm.
Professional use of Mac hardware encompasses a far wider array of users than many seem to want to admit to here. Few out there work on projects for NASA or even the automobile companies for that matter. The same thing goes for people doing movie editing, not many are making feature length movies.

Almost every major institution or company uses their own flavor of linux and uses extra server capacity in their data centers or something like AWS or Azure to provision on the fly. My notice of Mac Pro's has been more for the freelancers, small business folk who cannot afford large scale efforts or hire specialist who can tailor provisioning services for their needs etc. Mac Pro borders a niche market but never the less a large enough population still need them. 

post #19 of 33
Quote:
Originally Posted by AandcMedia View Post

Almost every major institution or company uses their own flavor of linux and uses extra server capacity in their data centers or something like AWS or Azure to provision on the fly. My notice of Mac Pro's has been more for the freelancers, small business folk who cannot afford large scale efforts or hire specialist who can tailor provisioning services for their needs etc. Mac Pro borders a niche market but never the less a large enough population still need them. 

That's because by being open source, Linux is extremely flexible.

Quote:
Originally Posted by wizard69 View Post

You kinda prove my point, this board is populated by sxtremist views on what it is to be a professional. The fact is many professionals have switched to laptop use for the advantageous they provide. Yes even professionals in research labs. I know of a few of these organizations and frankly your image of an R & D facility with clusters of servers at their disposal is not the norm.
Professional use of Mac hardware encompasses a far wider array of users than many seem to want to admit to here. Few out there work on projects for NASA or even the automobile companies for that matter. The same thing goes for people doing movie editing, not many are making feature length movies.

I kind of wonder how many of them are potential mac users. I was suggesting that these notions were due to software availability. Shake and Final Cut Pro were for all practical purposes Mac only applications. I know Shake had Irix and Windows NT support, but Apple bought them out. Quark Express started on the Mac. Photoshop started on the Mac. There are a few other desktop publishing programs that started there establishing a major presence. A few turnkey solutions existed prior to applications like those, but they were extremely costly. Autodesk also ported Smoke to OSX from their Linux turnkey solution. I'm just saying these perceptions are from a prior era as Apple actively catered to these markets. Regarding notebook hardware, its improvements in terms of raw percentages have outpaced some of the gains at the workstation level in recent years, especially with an inelastic budget. A lot of software lacks n-core scaling, even in some of the markets commonly addressed on here. A lot of it hits a wall past 4 cores and still holds onto single threaded processes in some areas. This is a an overall problem for the general health of the workstation market. Even if you can benefit from greater speeds, it doesn't mean you necessarily benefit from ever increasing core counts. Sandy Bridge E is a little bettter in that some of the core count/clock speed tradeoffs aren't as severe at the 8-12 core level, but damn those cpus run hot.

post #20 of 33
Thread Starter 
I am going to throw this curveball into the equation (and I'm sure Intel is working on it) but how long will it take for a Pro level desktop to be developed that runs fairly cool?
post #21 of 33
Originally Posted by Winter View Post
I am going to throw this curveball into the equation (and I'm sure Intel is working on it) but how long will it take for a Pro level desktop to be developed that runs fairly cool?


Define 'cool'. My northbridge idles at 149(THANKS HUDDLER I CAN'T TYPE THE DEGREE SYMBOL)F, and that's the hottest part when idle.

Originally posted by Relic

...those little naked weirdos are going to get me investigated.
Reply

Originally posted by Relic

...those little naked weirdos are going to get me investigated.
Reply
post #22 of 33
Thread Starter 
70-80 degrees Fahrenheit idle, 100 degrees under heavy load.
post #23 of 33
Originally Posted by Winter View Post
70-80 degrees Fahrenheit idle, 100 degrees under heavy load.

 

That's a pretty small range there… Does the Mac Mini even idle at 100?

 

The first-gen (1GHz Intel Atom-based) Apple TV even ran at 104! lol.gif

Originally posted by Relic

...those little naked weirdos are going to get me investigated.
Reply

Originally posted by Relic

...those little naked weirdos are going to get me investigated.
Reply
post #24 of 33
Thread Starter 
Yeah it's probably going to take 5-10 years but I'd love to see it.
post #25 of 33
Interesting points.
Quote:
Originally Posted by hmm View Post

That's because by being open source, Linux is extremely flexible.
Linux is flexible but his post was BS. Many companies won't even consider a Linux based machine in their shops. Again this comes back to my point that the board here is filled with extremist view points. I use Linux at home but it isn't even worth suggesting it at work nor would many other companies near by bother with Linux.
Quote:
I kind of wonder how many of them are potential mac users. I was suggesting that these notions were due to software availability. Shake and Final Cut Pro were for all practical purposes Mac only applications. I know Shake had Irix and Windows NT support, but Apple bought them out. Quark Express started on the Mac. Photoshop started on the Mac. There are a few other desktop publishing programs that started there establishing a major presence. A few turnkey solutions existed prior to applications like those, but they were extremely costly. Autodesk also ported Smoke to OSX from their Linux turnkey solution. I'm just saying these perceptions are from a prior era as Apple actively catered to these markets.
Apple still does! However third party vendors would be foolish to remain Mac only for many of their products.
Quote:
Regarding notebook hardware, its improvements in terms of raw percentages have outpaced some of the gains at the workstation level in recent years, especially with an inelastic budget. A lot of software lacks n-core scaling, even in some of the markets commonly addressed on here.
Some software does suffer in that regard. Like wise some doesn't but that is hardly the point these days. Multi processors allow for many apps to run at the same time without suffering. That can be huge for many users.
Quote:
A lot of it hits a wall past 4 cores and still holds onto single threaded processes in some areas. This is a an overall problem for the general health of the workstation market. Even if you can benefit from greater speeds, it doesn't mean you necessarily benefit from ever increasing core counts.
It really isn't as huge a problem as people make it out to be.
Quote:
Sandy Bridge E is a little bettter in that some of the core count/clock speed tradeoffs aren't as severe at the 8-12 core level, but damn those cpus run hot.
The trade off of clock speed and heat will go on forever. What is interesting is that we are nearing 4Ghz clocks in desktop machines these days. That is almost a full GHz over what was possible a couple of years ago. This means apps bound to a single thread are still seeing speed ups which is why I say it isn't an issue now.

In any event apps like those are yesterday's issues. The interesting apps of tomorrow will be highly threaded. The question is what will grab the community and become successful. Imagine the power of Siri right on your desktop. Or modeling and designing the machines of the future in real time. Or how about a genetics work station app. Lots of cores will makes such things possible and economical.
post #26 of 33
Never!

Think about it Pro Level hardware is always at the bleeding edge. Even if the average desktop was operating fabless and cool to the touch pro users will be demanding far better performance. With performance comes heat.

Quote:
Originally Posted by Winter View Post

I am going to throw this curveball into the equation (and I'm sure Intel is working on it) but how long will it take for a Pro level desktop to be developed that runs fairly cool?
post #27 of 33
Thread Starter 
Quote:
Originally Posted by wizard69 View Post

Never!
Think about it Pro Level hardware is always at the bleeding edge. Even if the average desktop was operating fabless and cool to the touch pro users will be demanding far better performance. With performance comes heat.

Thankfully there's smcfancontrol, right? *laughs*
post #28 of 33
Quote:
Originally Posted by wizard69 View Post

Never!
Think about it Pro Level hardware is always at the bleeding edge. Even if the average desktop was operating fabless and cool to the touch pro users will be demanding far better performance. With performance comes heat.

This isn't always true. Xeons tend to be tested for 24/7 use, and operating temperatures that are considered within spec for constant use are significantly lower than those that cause thermal shutdown. It's normal to have a beefy cooling system in place. If you're examining workstation gpus or those like the tesla processors that are optimized for computation, they're often clocked lower. They also run at lower levels of power consumption and much cooler. Gaming gpus might die if put through those kinds of workloads. I mean yes they can run hot if you're examining things tailored for bleeding edge workloads, yet temperatures are a relative thing as you do need lower ones for reliable service if they are to be subjected to constant use.

 

Quote:
Originally Posted by Winter View Post


Thankfully there's smcfancontrol, right? *laughs*


I used to use that, yet it's hard to get it just right, and I found it difficult to test whether it messed with the normal settings post installation even if uninstalled. People claimed this, and I meant to test it but didn't get around to it. I will say that there's always after market thermal paste if you're trying to bring things down a few degrees. Considering doing that on a couple older machines. You just need to have all of the tools handy and a suitable method of keeping screws and things organized. Otherwise it could be bad.

post #29 of 33

The next Mac Pro is Sandy Bridge Xeon400

post #30 of 33
Quote:
Originally Posted by hmm View Post

This isn't always true. Xeons tend to be tested for 24/7 use, and operating temperatures that are considered within spec for constant use are significantly lower than those that cause thermal shutdown.
I think you missed the point machines built for pro usage run hotter than those of lower performance given the same generation of chips. When someone talks workstation it is rational to assume it will run hotter than the run of the mill office PC.
Quote:
It's normal to have a beefy cooling system in place.
Of course it is because they run hotter.
Quote:
If you're examining workstation gpus or those like the tesla processors that are optimized for computation, they're often clocked lower. They also run at lower levels of power consumption and much cooler. Gaming gpus might die if put through those kinds of workloads. I mean yes they can run hot if you're examining things tailored for bleeding edge workloads, yet temperatures are a relative thing as you do need lower ones for reliable service if they are to be subjected to constant use.
I have no doubt that running a chip cooler makes it more reliable. However those big heat sinks on those Tesla cards are not there for good looks.
Quote:
I used to use that, yet it's hard to get it just right, and I found it difficult to test whether it messed with the normal settings post installation even if uninstalled. People claimed this, and I meant to test it but didn't get around to it. I will say that there's always after market thermal paste if you're trying to bring things down a few degrees. Considering doing that on a couple older machines. You just need to have all of the tools handy and a suitable method of keeping screws and things organized. Otherwise it could be bad.
After market thermal paste just improves thermal transfer, power into the chip remains the same or increases. Why would it increase? Better thermal transfer means further opportunities to speed step thus putting more power into the chip. It is certainly good to see a chip run cooler but you need to remember the same heat is being put into the room. If you are concerned about power usage the route to take is lower power chips.
post #31 of 33
Quote:
Originally Posted by KremerOtha View Post

The next Mac Pro is Sandy Bridge Xeon

If that was the case why the wait till 2013?
post #32 of 33
Originally Posted by wizard69 View Post
If that was the case why the wait till 2013?

 

Gotta hold onto the hope that it'll even exist… 

Originally posted by Relic

...those little naked weirdos are going to get me investigated.
Reply

Originally posted by Relic

...those little naked weirdos are going to get me investigated.
Reply
post #33 of 33
Quote:
Originally Posted by wizard69 View Post


I think you missed the point machines built for pro usage run hotter than those of lower performance given the same generation of chips. When someone talks workstation it is rational to assume it will run hotter than the run of the mill office PC.
Of course it is because they run hotter.
I have no doubt that running a chip cooler makes it more reliable. However those big heat sinks on those Tesla cards are not there for good looks.
After market thermal paste just improves thermal transfer, power into the chip remains the same or increases. Why would it increase? Better thermal transfer means further opportunities to speed step thus putting more power into the chip. It is certainly good to see a chip run cooler but you need to remember the same heat is being put into the room. If you are concerned about power usage the route to take is lower power chips.

Tesla gpus still don't have quite the max power consumption of some of the gaming cards. It's not just the heat sinks. It is important to remember that these are set up to be run hard for hours or days at a time. Similar use of a gaming card could literally destroy the thing as they can reach some insane temperatures with their stock cooling solutions. As for after market paste, I'm aware of what it does. I considered it to see if it would improve efficiency. Right now the readings for the cpu and its heatsink are quite far apart even though the distance between sensors is relatively minor. I expect that these machines have the classic bad oem thermal paste application. Perhaps I should have worded my statement differently though. Performance parts tend to generate more heat, but it's typical to run them at lower absolute temperatures than might be considered acceptable in something like a gaming machine /.

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Future Apple Hardware
AppleInsider › Forums › Mac Hardware › Future Apple Hardware › Ivy Bridge E due in Q3 2013 - What does this mean for the Mac Pro?