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.
The 2009 (Gainestown) Mac Pro: Everything We Know
- Thread starter Tallest Skil
- Start date
- Sort by reaction score
- Status
if you actually take the time to read the article, it states a news article and NOT a rumor...
Rendering, rendering, rendering. it makes a HUGE difference.
I edit just fine on my MBP, but transcoding, compressing, rendering etc takes a lot longer. With the Mac Pro - you could continue working WHILE you are rendering (If you don't have a render farm). With QMaster - you can direct it to give Compressor 6 cores while you save 2 for the OS and an app.
http://en.wikipedia.org/wiki/Intel_Core_i7
It says it doesn't support ECC at all. Am I missing something?
It says it doesn't support ECC at all. Am I missing something?
Apple has single CPU Mac Pro's now. It leaves 1 socket open and it costs $500.00USD less than the "standard". So one question is: Will they continue with a 2 socket MP with one filled and one vacant? (Um, I think I read somewhere that's how they do it - correct me if I'm wrong - which I frequently am)http://store.apple.com/us_smb_78313/configure/MA970LL/A?mco=NzQ3Njkz I could get that, and be a 4 core whore for awhile - then toss in the other CPU when the price and my wallet converge.
Yes, currently, they have only a single dual-socket motherboard, and offer a variety of single or dual CPU's. The problem with this approach is that the entry level cost of a single processor system is unnecessarily bloated by the unused dual-socket motherboard cost. A dual-socket motherboard is more expensive to make than a single socket system... perhaps not twice, but close to it. Dual socket server boards are typically 10-12 layers vs. 6 for a single socket desktop or workstation board. So instead of a single CPU system costing $500 less, it could potentially cost $800 less.
They could offer a Mac Pro at a lower price point with the same level of expandability if they came out with a single-socket motherboard. It is less elegant from a manufacturing and SKU management perspective, but it would allow the MAC pro to start to steal more enthusiast desktop market share from Windows users. I know I'm not the only one who wants a Mac Pro that I can use to dual-boot to play Windows games on.
The Bloomfield chips with the consumer Core i7 branding have ECC support disabled. The Xeon 3500 Bloomfields and the Gainestown Xeon 5500 have full ECC support.
The parts are supposedly being released at the end of March, then it's up to Apple to ship it whenever they feel like.
The point isn't really for people to read the thread; the point is to read the first post with all the information from the thread there.
I'm very worried about this non-riser business.
Of course, it forces Apple to either make the case LONGER to support the DIMM slots, or it forces them to simply offer only... like... six slots.
Which would they do? I think we know.
Might want to update the first post with the non-riser info and speculations that come from it.
Which might not be a bad idea, if it gives added expandability. Could also help cooling with whatever ≥150 W "special" CPU Intel might decide to release for Apple (150 W CPUs already work in the current Mac Pro though).
And they'll still say "supports more RAM than ever before, up to 48 GB."
wait where did we hear they were ridding the mac pro of riser cards
Tylersburg (and DDR3) apparently can't take RAM on risers.
Therefore, no risers.
It's speculation at this point.wait where did we hear they were ridding the mac pro of riser cards
But what about that post of... whoever it was. The latest one. Last page. With the Tylersburg pictures.
What's your take on what he said about the riser business?
And 1400+ pin cards don't sound very sturdy to me, but what do I know...
Virtual Rain posted the pic of the server board and the ICH10 diagram.
I still think it's possible.
Big hint: LGA1366 = 1366 pins, and not all of them are even used.
(Reserved for future features). So how do you connect 1440 pins just to the memory? It would be 720 pins, not 1440. Each channel only gets 240 pins. It switches between the additional DIMM's on that channel. (Which is why more than 1 DIMM per channel actually will slow it down slightly, due to the added latency of switching).
As far as placement, take another look at the board pic. The riser can be positioned similarly. Imagine lifting the bank of DIMM slots perpendicular to the CPU, and the attachments (custom connector needed) to take place at a similar physical location of the DIMM nearest the CPU socket. This would still keep traces short.
Just found a new article on the performance of the X5570
http://www.anandtech.com/weblog/showpost.aspx?i=554
http://www.anandtech.com/weblog/showpost.aspx?i=554
Virtual Rain posted the pic of the server board and the ICH10 diagram.
I still think it's possible.
Big hint: LGA1366 = 1366 pins, and not all of them are even used.
It would be 720 pins, not 1440. Each channel only gets 240 pins. It switches between the additional DIMM's on that channel. (Which is why more than 1 DIMM per channel actually will slow it down slightly, due to the added latency of switching).
As far as placement, take another look at the board pic. The riser can be positioned similarly. Imagine lifting the bank of DIMM slots perpendicular to the CPU, and the attachments (custom connector needed) to take place at a similar physical location of the DIMM nearest the CPU socket. This would still keep traces short .
I'm certainly no expert so I could be proven wrong on this. FBDIMM memory is much easier to locate how ever you like given it's architecture. DDR3 is a much simpler architecture and has less latency but comes with certain constraints as a result.
From a wiring POV, FB-DIMM is much easier. Though it also uses 240 pin DIMM PCB'S, only 69 pins were actually used,
as it's a serial connection. It was just easier, and cheaper, to use parts already in the supply chain.
Any component on the board has constraints. Trace routing is just one of them, though it can be critical. DDR3 has a little give. Just take a look at the SuperMicro board I linked previously.
It has longer traces than the prototype pic you posted. 
Really, nobody makes 3.5" 'mainstream' SSDs any more. All of the high performance ones are 2.5". Mostly because the amount of flash that fits in a 2.5" case is insanely expensive already; that if *THAT* isn't enough, the amount to fill a 3.5" case would be *WAAAAAY* too much to even consider. (There are exceptions, but a 'mainstream' workstation computer is not one of them.)
*cough*... It's going to be hard for me not to violate an NDA here, but I'll stick to what has already been mentioned in this thread, and make pure Apple-related guesstimates, not directly commenting on the Intel rumors. (read my disclaimer: I have *ZERO* inside knowledge of Apple.)
Assuming that the Tylersburg chipset that would be in this Mac Pro does indeed include the rumored four x16 slots, (how's that for wishy-washy, non-NDA-breaking?) I can see no logical reason whatsoever for Apple to avoid having four full x16 slots.
As for SLI? While Tylersburg may theoretically be SLI-capable, the motherboard maker still has to license it from nVidia. It was only today that Intel licensed SLI for their own X58 board. And this board has been out for about three months. Apple would have to license SLI.
As a companion, ATI does not charge a dime for CrossFire compatibility, and the current Mac Pro's 5400X chipset is already CrossFire compatible; yet we don't see CrossFire compatibility in OS X.
I would speculate that while it is *POSSIBLE* that we will eventually see SLI/CrossFire compatibility in OS X, it likely won't arrive until Snow Leopard; and if the capability is in the board, it will be hidden until then. (Like with the first Mac Pro AirPort cards supporting 802.11n, yet having it disabled until Apple officially released their 802.11n base station.)
I fully agree. There is *NO* reason to stick with PATA for optical drives. All of the latest optical drives have SATA equivalents for the same price as their PATA kin.
Plus, the desktop X58 chipset doesn't even support PATA, so a third-party controller chips would be necessary to add it. Why waste the money and motherboard space, when there is no reason left to stick with PATA on a new system? (But I like my PATA Zip drive and PATA Imation SuperDisk drive!
)Alright, I think I stayed within the bounds of my NDA.
Note: "Tylersburg" is a codename that refers specifically to the X58 desktop chipset. According to public press releases from today, the dual-socket version is codenamed "Tylersburg-EP". Therefore, any dual-socket system, such as the Mac Pro, would almost guaranteed use the "Tylersburg-EP" chipset, not the "Tylersburg". Make from this correction what you will, I just like to be accurate.
Really, nobody makes 3.5" 'mainstream' SSDs any more. All of the high performance ones are 2.5". Mostly because the amount of flash that fits in a 2.5" case is insanely expensive already; that if *THAT* isn't enough, the amount to fill a 3.5" case would be *WAAAAAY* too much to even consider. (There are exceptions, but a 'mainstream' workstation computer is not one of them.)
*cough*... It's going to be hard for me not to violate an NDA here, but I'll stick to what has already been mentioned in this thread, and make pure Apple-related guesstimates, not directly commenting on the Intel rumors. (read my disclaimer: I have *ZERO* inside knowledge of Apple.)
Assuming that the Tylersburg chipset that would be in this Mac Pro does indeed include the rumored four x16 slots, (how's that for wishy-washy, non-NDA-breaking?) I can see no logical reason whatsoever for Apple to avoid having four full x16 slots.
As for SLI? While Tylersburg may theoretically be SLI-capable, the motherboard maker still has to license it from nVidia. It was only today that Intel licensed SLI for their own X58 board. And this board has been out for about three months. Apple would have to license SLI.
As a companion, ATI does not charge a dime for CrossFire compatibility, and the current Mac Pro's 5400X chipset is already CrossFire compatible; yet we don't see CrossFire compatibility in OS X.
I would speculate that while it is *POSSIBLE* that we will eventually see SLI/CrossFire compatibility in OS X, it likely won't arrive until Snow Leopard; and if the capability is in the board, it will be hidden until then. (Like with the first Mac Pro AirPort cards supporting 802.11n, yet having it disabled until Apple officially released their 802.11n base station.)
I fully agree. There is *NO* reason to stick with PATA for optical drives. All of the latest optical drives have SATA equivalents for the same price as their PATA kin.
Plus, the desktop X58 chipset doesn't even support PATA, so a third-party controller chips would be necessary to add it. Why waste the money and motherboard space, when there is no reason left to stick with PATA on a new system? (But I like my PATA Zip drive and PATA Imation SuperDisk drive!
Alright, I think I stayed within the bounds of my NDA.
Note: "Tylersburg" is a codename that refers specifically to the X58 desktop chipset. According to public press releases from today, the dual-socket version is codenamed "Tylersburg-EP". Therefore, any dual-socket system, such as the Mac Pro, would almost guaranteed use the "Tylersburg-EP" chipset, not the "Tylersburg". Make from this correction what you will, I just like to be accurate.
Wonderful points, I'll add them to the first post.
Wait, Tylersburg-EP or DP? I thought we'd see DP/36.
If you are telling us the truth and TRULY being wishy-washy, then I love you, and that's all there is to it.
So, in being wishy-washy, you're saying that since there is a Tylersburg board that supports 4 16x slots, there is no reason that Apple would shoot for something less than this, right?
Also, due to circumstances beyond my control, the funds I had dedicated to my Mac Pro will not be going toward a Mac Pro. Therefore, I will be stuck with this Penryn MacBook Pro until such time as I can get the money to afford a Mac Pro again.
I'm guessing that the first that this will be is Haswell or later.
I'll still update this thread, of course. I didn't make it for me; this is for everyone.
Funny, my B&W G3 cost well North of $2300 when brand new. Only one B&W was available at less than $2000 at release; the high end model started at $3000. For a while, Apple had one single sub-$2000 Power Mac G5, too.
Ironically, it was exactly what people are complaining about not having now, a single-socket tower. It would be nice to see Apple release a true single-socket X58/i7 "Mac Pro" as the low-end offering along with the dual-socket offering; like they did with the single-socket PM-G5. But, I doubt it will happen. (Which would be too bad. Because I'm not buying another Mac until they have a single-socket Nehalem system. The Mac Pro is beyond my budget (and honestly beyond my needs,) but Core 2 is now obsolete. At this point, I'll probably wait for a notebook.)
Heh. So my NDA now proves worthless. I have found Google search results referencing a "Tylersburg WS" for workstations, that includes other things, like "Tylersburg-D36".
I can honestly admit that I had not heard about those. Nor would I have had a reason to hear about them. (So my lack of knowledge of them should not be taken as confirmation of lack of existence.) So while I can't confirm it, I can't deny it either. And it sounds perfectly logical. I can say that the only search results I see for "Tylersburg-DP" are this thread, though. It looks like it's just "Tylersburg-D##"
Of course, this means I will have to go look this all up when I get in to work tomorrow; which means I won't be able to post in this thread again.
- Status