Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Will there be a new Mac Pro in 2015?
- Thread starter Anto38x
- Start date
- Sort by reaction score
- Status
One more little thing.
I think that even if we will see Fiji in Mac Pro I don't believe it will be mainstream option.
Lets look at this proposition.
D310: Tonga or Tonga XT. with 2 or 3 GB of RAM.
D510: Grenada Pro with 4 GB of RAM and 512 Bit bus.
D710: Grenada XT with 8 GB of RAM and 512 Bit bus.
D910: Fiji XT with 4 GB of HBM.
Why is that? Simple answer - pricing.
Apple loves big portions of money, and very high profit margins.
Tonga XT is rumored to have 249$ price tag, right? So AMD can sell it to Apple for 199$ a piece.
Grenada Pro goes for around 330$ right? So AMD can sell it to Apple for 299$ a piece, right?
Grenada XT sells for 450$(around) right? You get what Im referring?
AMD is not willing to sell the full fiji chip for less than 600 USD. Nobody. What that means is dual Fiji setup could have even 2000$ price in Mac Pro. Could Apple justify that in any way? I don't believe so.
With the Tonga XT, Grenada Pro, and Grenada XT they are able to maintain 400 and 1000$ price tags over the base model of GPU, and have enough profit margin. And with Fiji being another part of lineup, some sort premium version, they are able to get even higher profit margin with even higher price tag.
This is only my theory, you don't have to take my word for this.
I think that even if we will see Fiji in Mac Pro I don't believe it will be mainstream option.
Lets look at this proposition.
D310: Tonga or Tonga XT. with 2 or 3 GB of RAM.
D510: Grenada Pro with 4 GB of RAM and 512 Bit bus.
D710: Grenada XT with 8 GB of RAM and 512 Bit bus.
D910: Fiji XT with 4 GB of HBM.
Why is that? Simple answer - pricing.
Apple loves big portions of money, and very high profit margins.
Tonga XT is rumored to have 249$ price tag, right? So AMD can sell it to Apple for 199$ a piece.
Grenada Pro goes for around 330$ right? So AMD can sell it to Apple for 299$ a piece, right?
Grenada XT sells for 450$(around) right? You get what Im referring?
AMD is not willing to sell the full fiji chip for less than 600 USD. Nobody. What that means is dual Fiji setup could have even 2000$ price in Mac Pro. Could Apple justify that in any way? I don't believe so.
With the Tonga XT, Grenada Pro, and Grenada XT they are able to maintain 400 and 1000$ price tags over the base model of GPU, and have enough profit margin. And with Fiji being another part of lineup, some sort premium version, they are able to get even higher profit margin with even higher price tag.
This is only my theory, you don't have to take my word for this.
That was the setup I mentioned a while back, and right now it's what makes sense really.
The Fiji model is somewhat of a mystery though. It is in fact a high end part, and a newer one, but like I said before people will look at it as having "only" 4GB which could make it look like a lower part than Grenada XT with 8GB. We all know it's a different ball game but from the marketing stand point it would be a tough call to make it worth it the extra dough.
It seems AMD is re-gaining market share, too bad the short supply with the Fiji chips, it would help here.
Pascal orders seem to have landed at TSMC. FP64 seems to be back as a must have feature again finally.
The Fiji model is somewhat of a mystery though. It is in fact a high end part, and a newer one, but like I said before people will look at it as having "only" 4GB which could make it look like a lower part than Grenada XT with 8GB. We all know it's a different ball game but from the marketing stand point it would be a tough call to make it worth it the extra dough.
It seems AMD is re-gaining market share, too bad the short supply with the Fiji chips, it would help here.
Pascal orders seem to have landed at TSMC. FP64 seems to be back as a must have feature again finally.
4 GPU options seems like a lot. One problem is Grenada Pro is likely very similar in performance to Tonga XT, especially given the limited thermal headroom of the mac pro. This should become more clear when the AMD R9 380X gets released, since we haven't seen a Tonga XT retail card yet.
An option I could see would be two top end choices for the mac pro. Say a D710 (Grenada/Hawaii XT with 8 or 16 GB VRAM) and a E710 (Fiji with 4 GB VRAM). This would be more confusing though.
I still think Apple will choose to stick with Tonga/Fiji. They share the same GCN architecture and are more efficient than Hawaii. They could always put 6 or 8 GB VRAM on the lower end Tonga (D310 equivalent) for people who need it. I am sure Apple can find a way to market why HBM is better. If its one thing AMD could use, its some marketing help from Apple.
An option I could see would be two top end choices for the mac pro. Say a D710 (Grenada/Hawaii XT with 8 or 16 GB VRAM) and a E710 (Fiji with 4 GB VRAM). This would be more confusing though.
I still think Apple will choose to stick with Tonga/Fiji. They share the same GCN architecture and are more efficient than Hawaii. They could always put 6 or 8 GB VRAM on the lower end Tonga (D310 equivalent) for people who need it. I am sure Apple can find a way to market why HBM is better. If its one thing AMD could use, its some marketing help from Apple.
I made a mistake a year ago. I rushed out and bought a late 2013 iMac 27" and soon after they made the new ones available... But that's not the problem. The problem is that I made a mistake with the Fusion Drive... I should have bought the iMac with a 'normal' SSD of the maximum capacity then available. Some of the tasks I do involve large amounts of data, and waiting hours for stupid things to get done due to the slow HDD part of the Fusion Drive is very frustrating....
I am looking to sell this iMac and invest in the Mac Pro with whatever maximum flash storage it will be available with. But of course the buyers' guide suggests to wait as new ones may be available soon....
Anyone else who really can't wait?
I am looking to sell this iMac and invest in the Mac Pro with whatever maximum flash storage it will be available with. But of course the buyers' guide suggests to wait as new ones may be available soon....
Anyone else who really can't wait?
dont trust the buyer's guide. It has become unreliable for some products, however, the new version should come soon.I made a mistake a year ago. I rushed out and bought a late 2013 iMac 27" and soon after they made the new ones available... But that's not the problem. The problem is that I made a mistake with the Fusion Drive... I should have bought the iMac with a 'normal' SSD of the maximum capacity then available. Some of the tasks I do involve large amounts of data, and waiting hours for stupid things to get done due to the slow HDD part of the Fusion Drive is very frustrating....
I am looking to sell this iMac and invest in the Mac Pro with whatever maximum flash storage it will be available with. But of course the buyers' guide suggests to wait as new ones may be available soon....
Anyone else who really can't wait?
380X seems to be 256b after all, late October launch.
Why hasn't AMD enabled the full mem controllers in the die? No need due to mem optimization or poor yields (unlikely)?
Still, looks great to me, 2 of these on the nMP would make my day, look no further...
Why hasn't AMD enabled the full mem controllers in the die? No need due to mem optimization or poor yields (unlikely)?
Still, looks great to me, 2 of these on the nMP would make my day, look no further...
FirePro W8100 and W9100 were used for rendering an Indian film. Cool...
It could have been done with nMP
It could have been done with nMP
This is nice to see
“Cook is quick to point out, however, that this doesn’t foreshadow the end of the Mac. “I think there are other people — like myself — that will continue to buy a Mac and that it will continue to be a part of the digital solution for us,” he adds. “I see the Mac being a key part of Apple for the long term and I see growth in the Mac for the long term.””
20 Minutes With Tim Cook
http://www.buzzfeed.com/johnpaczkowski/twenty-minutes-with-tim-cook
“Cook is quick to point out, however, that this doesn’t foreshadow the end of the Mac. “I think there are other people — like myself — that will continue to buy a Mac and that it will continue to be a part of the digital solution for us,” he adds. “I see the Mac being a key part of Apple for the long term and I see growth in the Mac for the long term.””
20 Minutes With Tim Cook
http://www.buzzfeed.com/johnpaczkowski/twenty-minutes-with-tim-cook
My thoughts exactly. If nMP update is about to come soon, it could be all Tonga.
2x D310 = Tonga Pro, 2GB vram
2x D510 = Tonga XT, 4GB
2x D710 = Tonga XTX, 6GB (if 384-bit bus, 8GB if 256-bit)
2x D800 = Fury XT, 4GB
And maybe Metal v2 will introduce CrossFire. And it could be introduced as a revision update of El Captain (10.11.3 or 4)
Last edited:
HBM2 might be a while still. FirePro on HBM2 even more so, with validation and all. That's way too long, but who knows.
We still don't know about Broadwell for sure, although it's almost certain that Intel won't flog it. It might be coming in Q4, before the Extreme version, but there's no official confirmation that I know of.
So, if Xeon -EP comes soon enough, TB3 is shipping without constrains and Apple did take the time to re-engineer the nMP for it, we might be seeing it soon.
If they were too busy with iDevices, iCar and whatever not... well, be prepared for a long wait, still.
Full Tonga lineup doesn't seem likely.
We still don't know about Broadwell for sure, although it's almost certain that Intel won't flog it. It might be coming in Q4, before the Extreme version, but there's no official confirmation that I know of.
So, if Xeon -EP comes soon enough, TB3 is shipping without constrains and Apple did take the time to re-engineer the nMP for it, we might be seeing it soon.
If they were too busy with iDevices, iCar and whatever not... well, be prepared for a long wait, still.
Full Tonga lineup doesn't seem likely.
The rather straightforward solution to that is to just take your large data sets and put them on another drive. There are new "higher capacity" SSD drives that are SATA based.
For very large data sets trying to target on SSD "slim stick" drive is not really a good match to the problem. Even if go with an SSD actually want more than just 4 Flash chip packages and limited board space ( for RAM ,etc. ) . Or at least a cheaper $/GB target rather than max x4 PCIe transfer speed.
For example, a 1TB 850 Pro.
http://anandtech.com/show/9451/the-2tb-samsung-850-pro-evo-ssd-review
Given Apple's "max sdd" $1K pricing you can find a decent a decent 1TB drive and a 2-4 bay TB enclosure for less than $1K. If the working set size is smaller than 1TB but dynamically varies over time then a bigger SSD drive in a fusion pair would help. Same 2-4 bay enclosure but pair a 500GB SSD with a 2-3TB HDD. If you picked the 3TB combo then probably suffering in part from an issue that the SSD isn't a high enough percentage of your working set over time. The 1TB combo's roughly 13% is much better than the 3TB combo's 3%. Caching percentages of 10-20% are typically highly effective. Low single digits no so much. If crank that up to 480GB (add some spare area to a 512 drive) for 2TB for a 24% target it will likely perform better over a wider set of workloads. Don't really need a whole new system to do that. It could be added to the 2013 27" iMac ( and later attached to a future Mac Pro )
Anyone else who really can't wait?
A upgraded Mac Pro with 2015-2016 components probably isn't going to be a major change in solving storage I/O bandwidth issues. Apple's internal, max capacity SSD prices are still likely going to be priced way above the market even after upgrade. Highly likely they are going to be selling "speed" of internal SSDs more so than their capacity. If have a bulk capacity problem then there may not be much change between what is other there now and any upgrade.
Folks who are RAIDing multiple TB drive enclosures to get "faster than a single TB v2" bandwidth will see some relief (with new external enclosures ) with TB v3, but overall bandwidth probably isn't going to go up.
No they didn't. Already covered back at post 1718 a month ago in this very same thread ....
https://forums.macrumors.com/thread...mac-pro-in-2015.1840458/page-69#post-21733384
Most likely this "coming at the end of the year" was pretty much literal. Maybe last week in December or first couple days of January. The actual end of 2015. However, even your "information source" article stated that it was Q1 2016 arrival time. Add 6 weeks to the literal end of the year ( the delay the bug is causing ) and you end up with .... ta da ... mid February. Right around where this new information places the Broadwell E after the slide.
What might have happened if no bug was that Intel started production in 2015 and shipped to some system vendor factories in late 2015 so that could release in very early Q1 2016. But releasing finished system products to end users. Not much out there to support that.
Broadwell E and Broadwell EP share dies over a subset of models. It is a matter of which features are flipped on/off. The in the shared zone, the transistor usage overlap is pretty high ( probably in the 90+% range). If the found defect is in that common subset then the delay will impact both. -EP has a longer testing acceptance cycle ( with delay slots designed to soak up minor fixes ). Maybe it takes a marginally smaller than 6 week hit but the evidence they were shooting for 2015 release is extraordinarily slim and not much of that is coming from Intel. (unless looking a 'pie in the sky' projection roadmaps from 2-3 years ago before even started trying out the 14nm process at scale. )
iDevices and iCar ... probably have very little to do with Mac Pro team being pulled off to do other Mac projects. On board discrete GPU skills are transferable to MBP and iMac products. Wifi systems ... largely the same across all macs. Thunderbolt ... on all Macs except for MacBook which cause use a TB v3 update on that 'yesterday'. Very decent chance there are experiment Mac products that does see light of day for next 1-2 years that also could consume the Mac Pr members.
Don't really have to invoke iOS or car stuff to put a cap on resource allocation.
The car thing is probably easier to ramp at Apple because Apple doesn't have any car products. They have to do lots of hiring from the outside. There is little experienced notion of how many folks are too many or too few to get something done. Management is content to throw money over the wall into a largely decoupled "skunk works project" that is largely disconnected from marketing , sales, internal design center , etc. [ Kind of how the Mac was a 'pirate' renegade operation inside of Apple early in its lifecycle. ] . If this is an actual whole car I don't think this is going to end well (let's be hipper than Tesla ... a have more money than sense idea ). Apple building a infotainment computer subsystem for a car that handles security (which is spectacularly bad in modern cars) , connectivity , GUI , personal assistant, etc. that makes far more sense.
The Mac products are all far more tightly coupled to each other both in resource assignment pool, resources, and customer base. The Mac Pro probably competes with other Mac products far more than the other stuff with their own revenue and investment streams.
Neither.
The bulk of the Mac line up is probably not going to change. So the "rest of the line up" is most likely a huge over statement. If there was alot of stuff to intro they probably would be another dog and pony show scheduled. There isn't.
iMacs are way out of date ( outside of the bumped 5K models ... which are also behind the curve CPU wise. ). Broadwell-EP isn't coming in 2015. Also at this stage AMD probably doesn't have an overstock of special bin Fijis to sell. Intel's Gen 6 (skylake) models with upper end integrated graphics are tagged for early 2016.. so most of the laptops don't have much to move to that would be a significant increase. The MacBook does. It is relatively underpowered (versus Skylake option) and could use TB v3 quite a bit.
WWDC is the 'broken clock' pick. Even a broken clock is right twice a day. Not good components motivation there. Xeon E5 v4 Broadwell-EP will be out well before June 2016. Xeon E5 v5 is being queued up for Q4 '16 - Q1 '17 so an even longer delay doesn't make sense. Perhaps a "you can't buy it for another 6-8 months but we aren't dead ... really trust us" announcement for v5? They could if actually trying to kill of the product slowly.
HBM2 isn't due until Q2 '16. Vast majority of that initial product is probably going to go into discrete cards for the general Windows PC market as Nvidia and AMD do battle first there. Apple has been a consistent "buy in the 2nd half of lifecyle " on GPUs. If they have to do extra special component binning to fit the narrow thermal profile that is probably going to continue.
The first set of hardware but between Fall '15 and WWDC '16 in timing is more likely.
So all that being said, it's fair to say even though the mac pro is long in the tooth, it's still a safe buy if you need something to last through December?
dec, I'm surely aware that the people that work on iDevices and iCar aren't probably the same on the Mac products. The car division is a new collection of people from other companies Apple has been poaching for a while.
It was more of a "criticism" towards the current investments (time and money wise), which we must accept as the right ones in fact since it's where the money is. iPhone sales are through the roof as usual.
But the Macs get no love, or not so much. Specially the Mac Pro.
It was more of a "criticism" towards the current investments (time and money wise), which we must accept as the right ones in fact since it's where the money is. iPhone sales are through the roof as usual.
But the Macs get no love, or not so much. Specially the Mac Pro.
I'd hunt for a decent used config if that's your choice, but yeah, I think that's probably a safe bet at this point.
What most if not all people here have missed about Apple interest in Radeons is not the theoretical discount prices Apple may or may not have gotten, but that Raja Koduri, who was a chief architect and director at ATI and then AMD became an employee at Apple for a short time (He's now back in charge of Radeon development). Apple's use of Radeons, the development of the Mac Pro D series cards, and any discount prices for the chips were probably all down to Koduri's work, even still to this day. Prior to Koduri, Apple had also acquired AMD's Bob Drebin and Jim Keller. All three top architects at AMD who would have decided which GPUs would end up in Macs. They may have also pushed for more OpenCL adoption and helped development of Metal.
Last edited:
Rather dubious. How about Apple hired three GPU architects to work on the Apple implementation GPUs that they need designed in implemented. Apple is going to hire 3 architects primarily just to sit around and pick which parts to buy from a company that it was selling anyway?
Certainly there was going to be no Nvidia "fanboy" , Nvidia automatcally wins syndrome going on with them at Apple. But I doubt that was present at Apple in the first place because they have switched back and forth between integrated GPU vendors on a regular basis as designs changed and new "bake offs" held.
" ...when AMD’s CTO of the Graphics Product Group, Raja Koduri, first quietly left the company for Apple. This was hot on the heels of Apple’s hiring of another AMD GPU CTO, Bob Drebin. At the time (2009) I didn’t understand why Apple would want so many smart graphics guys on staff, were they working on their own GPU? .... It turns out that Steve Jobs wanted to surround himself with the absolute best in the business. Today, the impact of the work of folks like Bob Drebin, Raja Koduri, Jim Keller and others is quite evident. Apple tends to ship some of the fastest GPU hardware in the mobile industry, and its work in bringing high-DPI displays to virtually all of its products is unparalleled. ... "
http://www.anandtech.com/show/6907/the-king-is-back-raja-koduri-leaves-apple-returns-to-amd
The vast majority of the "implement stuff with transistors" folks that Apple hires don't have much to do with the Mac products. The Mac stuff is mostly highly industry standards based and available components with some minor tweaks that you are don't need high powered digital implementation architects to map out..
Apple helped invent OpenCL. They needed to go hire 3 outside folks to push for the internal adoption of it? Just how dysfunctional do you think Apple is?
Metal first appeared which kinds of devices? And on which GPU architecture? [ answer is not Mac and AMD ]
There is relatively ( compared to iOS devices) little money in cars. Tesla isn't making any money. They might make money in a couple of years if the stars all line up. $40-60 oil points to that perhaps not happening.
Car parts ... .yeah maybe. The percentage of car bill of material spent on computers/electronics is probably going up over next 10 years. Another complete car vendor? There are too many now. World doesn't really need another and there probably are too many now.
This seems more like a "piss away the money because we can and other Silicon Valley hipsters are doing it too" thing that something that is a reasonable strategy.
Macs get love. All signs point to a Retina 21.5" iMac coming. The MacBook has reappeared (after being retired when MBA 11" came). Mac Pro got major adjustment. Apple poured about 25-40% of the engineering resources that went into creating the Thunderbolt 3 spec. Apple is investing. They aren't making $500M-1,000M kinds of bets, but there are many 10's of millions going into Mac product development.
That $100B stockpile? It largely isn't being spent on iOS devices either. It isn't money that is even remotely part of the possible allocation. That money isn't there to "make it rain" on the R&D teams to do anything they what are "cost doesn't matter" budgets.
You misread me, although I was brief because I have no interest in debating this beyond mentioning the people and connections. I had basically repeated the knowledge in the Anandtech passage you quote, but I wanted to emphasise that the guys who went to Apple helped make sure AMD offered the best OpenCL performance for Apple's needs. In that sense they continued to work for AMD within Apple. They pitched to Apple and then let AMD know what Apple needed.
Where does Metal figure? Because all three guys were partly responsible for the early development of Vulcan and Mantle and would have brought their expertise to help Apple develop Metal. The timeframe matches precisely.
- Status