MacPro Wishlist
I know many will say that this is absurd, given that physical units of the new model are only just flowing, but I have to post this as do not see myself buying a 2013 MP to replace my 8-core MP. A supermicro board awaits, quite frankly, if I cannot delay. I hope I can.
1. Dual Socket option
You have two cores instead of one that can go hyper.
Some tasks may be better on one socket, but dual 6's are cheaper and can go to higher clocks than a single 12.
Scope to go to many more cores.
2. LGA2011-13
This is the new socket for the Haswell-EP and Broadwell CPUs, that are arriving this year.
Will give a couple of years life and a longer upgrade horizon.
3. 8 DIMMs, or 64GB support in 4 DIMMs. 128GB is minimum looking ahead, with 256GB far more fitting, IMHO.
I do not ask for more internals. In fact, I hoped for a device like this many years ago when I first heard of Thunderbolt, namely a machine connected to expansion like some kind of Dr Octopus, though as a mid-tier device.
Which reminds me...
4. Mid tier single-socket Haswell version. Mac Midi.
Those puppies will stretch to 8 cores.
No ECC, for sure, but that is not the point.
1. Dual Socket option
You have two cores instead of one that can go hyper.
Some tasks may be better on one socket, but dual 6's are cheaper and can go to higher clocks than a single 12.
Scope to go to many more cores.
2. LGA2011-13
This is the new socket for the Haswell-EP and Broadwell CPUs, that are arriving this year.
Will give a couple of years life and a longer upgrade horizon.
3. 8 DIMMs, or 64GB support in 4 DIMMs. 128GB is minimum looking ahead, with 256GB far more fitting, IMHO.
I do not ask for more internals. In fact, I hoped for a device like this many years ago when I first heard of Thunderbolt, namely a machine connected to expansion like some kind of Dr Octopus, though as a mid-tier device.
Which reminds me...
4. Mid tier single-socket Haswell version. Mac Midi.
Those puppies will stretch to 8 cores.
No ECC, for sure, but that is not the point.
Comments
By the way I don't want anybody to confuse the thought about "proof of concept" with the thought that the new Pro is in anyway junk. It obviously isn't. What it is is a very forward looking machine that is designed with an obvious eye toward technology is going. You can do anything you want! Just don't come here expecting a lot of support for your idea. Admittedly this is possibly the new Mac Pros biggest short coming. However the question then becomes why do you think you need all of those cores. Having trouble running Crisis? What is the point? You can't compare the Mac Pros hardware, designed a couple of years ago, with hardware that hasn't even been delivered yet. As for Broadwell based XEONs, Intel hasn't actually shipped 14 nm yet. Im hoping and probably more than a few others, that Intel has XEON based on 14nm technology by mid year 2015 but that is a ways off.
Beyond that buying a specific motherboard now in the hopes that it will optimally support future hardware is a big gamble. It is far better to wait for the processor advancement you are looking for to hit the shelves in volume. Far to many things can change between now and when Broadwell XEON actually ships. Again we have no idea if you even have a light inmate need for all that RAM. At some point it might make more sense to cluster machines rather than to try to overload them with cores and lots of RAM. Why? Current Intel technology doesn't really handle a large number of cores on each motherboard well. There isn't enough bandwidth to memory to feed the cores properly for some workloads. What is right is very much user dependent.
It should be noted that Intel is doing a lot of work on the bandwidth problem and as such the new XEON architectures are very interesting. However proof is in the pudding and as such I wouldn't buy until the pudding is served. I certainly don't get as worked up over it as I did at first. However Apple would do well to implement another SSD slot in the machine. Frankly it looks like it is already a possibility but that they just ran out of PCI Express lanes. A fast internal SSD would be a wonderful scratch disk and for some apps could make the Mac Pro less dependent upon external disks for normal work flows.
Other than that I have to agree the architecture of the new Mac Pro is surprisingly satisfying. XMac rises again!!! Thank you????????????????????????????????
Honestly there has never been a better time for a $1500 Mac that is a real desktop performance machine. They can even build it into the same Mac Pro chassis for all I care. Haswell however will likely be a thing of the past until this actually ships. An ideal machine here would support one of the new RAM standards such as DDR 4, Memory Cube or something else. Given a substantially faster memory system and Intels revised Broadwell GPUs it ought to be a fine machine that can run without the need for an external GPU for many users.
In any event I feel a sense of technology lust in your post. That is never a good thing because it can blind you to what is your real needs. No where in this post of yours did you cover what your needs where, rather it is a post glamorizing stuff that Intel hasn't even shipped yet. As such I really doubt that you have a pressing need for any new hardware. If so why not wait for the next Mac Pro revision.
[IMG]http://forums.appleinsider.com/content/type/61/id/42017/width/200/height/400[/IMG]
Is very Portable
The last of many baseless presumptions and errors in your post.
To think that it is somehow wrong or "lusting" to not wish to include a 3 year old chipset and only supporting 64GB of DDR3, but, instead, begin with a socket and chipset that is due to be available next quarter, beggars belief.
So yell at Intel. Or show us Xeon-E chips on a newer chipset.
To think that it is somehow wrong or "lusting" to not wish to include a 3 year old chipset and only supporting 64GB of DDR3, but, instead, begin with a socket and chipset that is due to be available next quarter, beggars belief.
Well Apple supports what they support, but OWC has tested it with 128GB. It will also show up elsewhere. As for the chipset, it actually started shipping mid 2012. In terms of shipping time, it's approaching the two year mark. Intel always changes Xeon EP chipsets once every two cycles. Haswell will update chipsets, and it will still be some time before that is available. I'm not going to make a guess on when version 2 will come out, but considering they just redesigned it, I don't expect it to languish for long cycles beyond hardware availability.
1. Haswell EP is coming in late this year. Which may feature an native 12 Core Die. And as we move to 14nm next year, We may even get a native 16 Core Broadwell EP as well. So no, i dont see the need for a 2 Socket version.
2. With Haswell EP you get to use DDR4, which will allow double the memory capacity. And you could even get 128GB per DIMM. So no, you dont need 8 DIMM at all.
The biggest Short coming of Mac Pro is amount of PCI-E lane. Which i dont see Intel increasing in the next 2 years roadmap. So the problem with Thunderbolt Port usage requiring special care will continue to stay until PCI-E 4.0 comes. And PCI-E SSD wont see much improvement either due to the same lanes issues.
So basically all you needs will be fulfilled in the next update of Mac Pro. For some reason I dont think Apple will jump on the Haswell-EP bandwagon later this year. One because DDR4 is expensive, two because its mass availability between first Mac Pro and the Haswell-EP will be too short. Heck the shipping time now, 4 months after it being announced is still 4 - 6 weeks. It may wait until Broadwell-EP.
And I would love a Haswell-E , Prosumer Grade Mac Pro. But knowing Apple, i dont think they will ever make one.
In any event I really wonder what the problem is with the Mac Pro and ship times. Did they underestimate demand? Are some parts hard to obtain? Poor yields? AMD having problems delivering? Intel having problems delivering? Lots of possibilities but it is odd for Appple to be this far behind so many months after starting to ship. A plain old six core Haswell desktop chip would do it for me. One processor card and one video card would make for a very nice workstation. As for making one I really see a need for them to do something to fill the now huge gap between the Mac Pros starting price and the Minis base price. Here a $1500 machine (starting price) would be ideal. It puts the machine into the line up at about twice the cost of the Mini and half the cost of the Mac Pro. Not a bad position to be in really.
Yes this is a huge problem and unfortunately Intel actually cut back on PCI Express lanes in some of its XEONs. So unless Apple has some influence here, we might not see the types of chips we want to see or the improvements in the Mac Pro we would like to see. Ideally the Mac Pro would support two PCI Express interfaced SSD modules in the future, however that requires the right Intel chip to support that port.
What do you mean? They just moved the PCI express lanes onto the cpu package. The number of lanes is merely dependent on how many sockets are populated. It's only constrained because there are a number of medium to high bandwidth things hooked up regardless of whether they are in use at any given time.
I'm not sure what you mean by that, lanes go to devices or cards, use them all and you are stuck configuration wise. In the context of the discussion I was reading the feeling was that Intel has lost touch with the market considering that PCI Express lanes are often dedicated to storage in new machine designs. Thus if you want to build a server supporting PCI Express based SSDs you end up needing to buy more processor than maybe you really need.
The other interesting thing here is that Intels new advanced XEON cores are arranged with CPU complexes of five cores. So chips will come in 5, 10 & 15 core variants if the tech migrates down to the lower end XEONs. I just find 5 to be a curious number.
To me, Apple somehow knew there were going to be hike up. Otherwise they could have launch it when they announced it. I dont see it being parts issues since most of them are off the shelves and even custom parts should have no problem raping up production. Case making are just robotics and machinery.
My guess It is more likely work force issues where it is now made in US. Compared to China Foxconn where you could have 10s of thousands workforce within days ( or more like hours ).
It is an issue on one series of low end chips. I was reading an article in one of the design forums where people where upset over one new XEONs because Intel didn't offer them enough PCI express lanes.
I'm not sure what you mean by that, lanes go to devices or cards, use them all and you are stuck configuration wise. In the context of the discussion I was reading the feeling was that Intel has lost touch with the market considering that PCI Express lanes are often dedicated to storage in new machine designs. Thus if you want to build a server supporting PCI Express based SSDs you end up needing to buy more processor than maybe you really need.
Are you referring to E3 variants? Those have 20 lanes. They are based on the same chip types as those used in the imacs. The E5s have 40 lanes per cpu package.
Quote:
The other interesting thing here is that Intels new advanced XEON cores are arranged with CPU complexes of five cores. So chips will come in 5, 10 & 15 core variants if the tech migrates down to the lower end XEONs. I just find 5 to be a curious number.
I was unaware of that. Well I did notice something of a pattern in the E7s, which now use a comparable socket type. Do you find it odd due to not being some variant of 2^n (obvious exponentiation for anything built on a square)?
Quote:
To me, Apple somehow knew there were going to be hike up. Otherwise they could have launch it when they announced it. I dont see it being parts issues since most of them are off the shelves and even custom parts should have no problem raping up production. Case making are just robotics and machinery.
My guess It is more likely work force issues where it is now made in US. Compared to China Foxconn where you could have 10s of thousands workforce within days ( or more like hours ).
It may not have made sense even then, considering that the sustained volume may be much less. The machine hadn't seen measurable changes since 2010. It was out of stock in Europe for a certain amount of time. Even if sales were poor on the old one, those factors would naturally lead to some amount of pent up demand. It's just that it makes no sense to build infrastructure around short term pent up demand. As for Foxconn, do they really hire large workforces for 2-3 month contracts? Also 10s of thousands in hours is completely unrealistic. You don't need to exaggerate to make a point.
As for the machines success I see one big problem within Apple, that seems to get repeated often, a machine isn't selling well so that is an excuse to ignore it. Of course it isn't selling well because people want Apple to take the product seriously and keep it fresh. This was the problem with the Mac Pro well before 2010 and you can see a similar pattern with the iPod Touch which Apple blissfully ignores. There are probably other examples but there does seem to be a self fulfilling pattern that Apple drops into here where something isn't selling well, usually because they aren't paying attention, so they ignore the product even more.
I mention the Touch as another example beyond the Mac Pro because Apple hasn't kept the product current with the technology seen in the iPhones as they have in the past. Now we hear that iPod sales are in decline. Duh! If you suddenly let a product get way behind the technology curve can you really expect people to continue to buy the product?
The Old Mac Pro suffered in the same way often in an inexcusable manner. For example people can understand the Intel screw ups that left us without a real upgrade option. However why did Apple have to maintain the same damn video card in the machine for just about ever? The lack of real improvements to the machine over successive years really pissed people off. So it is rational that there is some pent up demand. The problem is it has been 5 months now and that is just a little excessive when the backlog is six weeks.
I'm not sure it is such a huge exaggeration. There have been claims of companies hiring several thousand in a day. Maybe not ten thousand but it is certainly possible to raise a large work force quickly. However nothing about this product suggest to me that people are a problem.
No exaggeration. Foxconn have certainly done that in its best days. Not anymore though, where you could only get a few thousands now in a day.
I have the same thoughts, but from what i saw in the Mac Pro Production video. It was more Labor Intensive then I thought it would be. But like you said it doesn't seems to be a unsolvable problem either.
I once thought on that but I have an had time imagine Intel have problem ramping up a controller chip. The original ETA were likely on certification and Q&A testing procedure. Since they managed to ship first batch out, certainly it cant be that much more difficult to get more done.
I think it is hard for people outside of China to realize just how densely populated the country is. A single factory in China can employ more people than we have in some states. Actually I'm not that surprised but even then the automation is such that you can't throw???????????? people at it. I have no idea where the pinch points are but I wouldn't be surprised if there where more than a few. Why? Intel isn't invincible and frankly they have had significant delays in getting new processes up. Most of the world doesn't care because Intel still leads the rest of the industry. I have a hard time believing AMD has a problem with the video cards as there is little new there.
The meeting earlier in the week has me believing that the Mac Pro is doing better sales wise than Apple expected. The strong Mac Sales seem to indicate that the Mac Pro played a significant part in those strong sales.