Pardon was thinking of 5770 prices, yeh Apple charges a premium over the PC version, but its not that significant.
So yeh if for 400 pounds more (over the GPU upgrade) you can have a newer machine, might be something to go for then.
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.
Sandy Bridge next year? really?
- Thread starter Bokes
- Start date
- Sort by reaction score
Pardon was thinking of 5770 prices, yeh Apple charges a premium over the PC version, but its not that significant.
So yeh if for 400 pounds more (over the GPU upgrade) you can have a newer machine, might be something to go for then.
Chupa Chupa
macrumors G5
It's really the nature of computers. Everyone is always looking ahead to the next iteration. There is always something faster coming along, and with that hope there will be other major changes too. Consider the "low end" '09 8 Core is as fast as the top of the line '08 8 Core. But MPs are not "sexy" machines. They are work horses. No one ever got excited about a work horse.
Bottom line: buy when you need to. Obviously if a model year is about to get replaced wait a month or so, but its hard to pin down when Apple will update the MPs next. This is a great time to buy b/c you have your choice of discounted '09s or a brand new model year that will surely last 18 mos. at a min.
They don't have to add 5 months to the Sandy Bridge update release date since the projections so far added 4-5 months to the Xeon update date (used last 2 years) anyway. Intel is moving there release of Xeons from March to June-September timeframe. So at least as likely, given current information, that Apple will be back on the release +/- 1 month window that the other Xeon system vendors will be on in 2011 as they would be on some + 5 month schedule.
Decently likelihood that what happened this Summer to folks waiting for Mac Pro update will happen next Summer for those waiting on HP/Dell/etc. updates.
Since there was as an extremely minor design/engineering update, Apple can have spent lots of time this spring/summer working on prep for the next generation design issues. Unless they cannibalized the folks assigned to Mac Pro to other stuff, there really shouldn't be much of delay since it takes over a year to get completed systems out of the door. There is nothing in shipping this version late that should significantly impede work on the getting the next version out.
The current info is that the next generation Xeons have 4x memory channel/controller capabilities so the 4 slot design of the current Mac Pro doesn't have to be tweaked much. Just some spliting of the line routing to the CPU for the 4th DIMM socket (not shared anymore).
Probably will be some choices to be made about PCI-e v3 , USB 3.0 , and other additional I/O going to add to box that could slow it past the Xeon update window that the other vendors will use, but those shouldn't be show stoppers for Apple.
Sandy Bridge is going to be incrementally faster than the current Westmere CPUs. Mainly:
1. A bit better Memory I/O bandwidth when using 4 DIMMs. (so perhaps the 3 vs. 4 threads will die off some time in late 2011- early 2012. 😉 )
2. Better internal network between the cores. ( better multi-core application performance )
3. Better vector computing comparabilities. ( if have apps that heavily leverage SSE going to see significant jump in performance. )
4. Boosts in GHz. Sandy Bridge isn't a process shrink (transistors didn't get smaller) so the boost comes in "better design" in the circuits ( not smaller closer together ones). The "I must have bigger Hz" crowd will be happier. Also the next entry level dual Quad (8-core) model will most likely beat the current 3.33 Hex in more contexts (and still be cheaper. )
Will probably see speed ups from Westmere that range between 10-30% over a variety of apps. They'll be faster, but Westmere won't instantly be "stone age" antiques either.
I think there is also a misconception that the next Mac Pro will come quicker than the current one did because this one took so long. That is bogus. In part this is because the general Sandy Bridge line up is coming out for desktop systems later this Fall. The misconception is flawed because think the Sandy Bridge flavor of the Xeons is coming in March. It is not. It is going to come later. Much later.
The other secondary misconception is that Apple is going to dump Xeons an go desktop extreme version ... again... not very likely. Doesn't really buy anything in terms of cost reduction. Nevermind that Apple isn't slashing the prices of the iMac to open up room for the Mac Pro to move down into.
I think the bigger buzz was not so much the Sandy Bridge Xeon update itself but the other stuff folks speculate will be in the box on the I/O front. Namely USB 3.0 (or not ) , Light Peak ( or not ) , and better I/O (faster PCI-e slots if PCI-e v3 makes cut and the already mentioned better DIMM I/O in max configuration profile. ).
All of those makes dealing with just 4 DIMM (per CPU package) and 4 PCi-e slots easier to manage for larger workloads. The Mac Pro is more viable to the set of users to pack it to the gills with stuff.
For example you could do high end video capture off the USB 3.0 socket. (bingo don't need PCI-e slot for that. ). Likewise some "cures world hunger" capability that the Light Peak socket will bring.
P.S. With the AMD/ATI update for HD6000 series coming later this Fall, a 12 month window for Mac Pro would make it likely that will get HD6000 cards in the next gen Mac Pro given the standard "measured in months" delay between new GPU cards and Mac versions. In other words putting the Mac Pro release in March made that worse. (it was a factor contributing to older video cards in the released configs. )
That 45 W includes the integrated graphics but on the other hand some of the lower power would be from the process shrink.
8 cores of what instruction set? Once merge CPU and GPU onto a single die there are a heterogeneous set of cores on the die. So that space folks think is going to get devoted to classic x86 cores will more likely be pushed to GPGPU ones. Intel's graphics are behind the curve on performance. I can't see them staring the GPU performance to push a generic mainstream user core mix up over 4 cores.
Bringing it back to a Mac Pro and Xeon context... Yeah probably would see 8 cores on midrange Xeon models since integrating a GPU into a workstation CPU package doesn't present much sense. There may be a Xeon at that point with an integrated GPU, but it will be aimed at 1U and Blade servers where board space and costs are a premium and the graphics requirements are minimal (if there is any monitor hooked to the server at all). Any product line up that was aim at 2 package capabilities wouldn't have one.
Just depends on how much rendering you do and how truly multi threaded the rendering app that you use is. Also factor in whether faster rendering speed equates to more productivity or just that your new machine idles between 6am and 9 am when your current machine might still be working away.....And whether this is relevant or not because you're still in bed sleeping anyway.
The problem with more cores is that if you're not careful it becomes some kind of illusionary productivity boost that never actually materializes. If you really need as many cores as you can for rendering and you're not using Mac exclusive software to do it, you're better off building or buying a PC to crunch the numbers. Mac Pro's are not good value for this at the moment.....Apple and Intel are both adding a premium for the latest processors that don't stack up under most fiscal analysis. This is particularly true if you don't use a lot of multi threaded software or you don't render much of the time.
As an 08 owner, I'd recommend sticking a 5870 in there and getting another couple of years out of your current machine. Until more software catches up with all the multi core hardware tech, it's mostly just a p*ssing contest.
GroundLoop
macrumors 68000
This is my thought exactly. I am going to get at least another 3-4 years out of my 2008 (it is not a business machine, just a personal video editing machine/server).
GL
Hellhammer
Moderator emeritus
With your logic, the 12-core Mac Pro with 5870 should be called 1612 core because 5870 has 1600 stream cores. Your point sounds quite far fetched to me, when we talk about CPUs, it's always the CPU cores, not GPU.
My point is that it doesn't just "emulate" a second core (which was the old way of doing it and provided very minimal benefit), it actually has additional hardware, which does boost performance almost as much as having additional full cores in some cases.
I guess you are right. Thanks for that.
Lets hope the card will be available soon and will work in 08 without any need of flashing or dumping 🙂
Most likely x86.
How about…
12–1600 core Mac Pro? 12/1600? 12C/320[4+1]SP?
😀
Seriously though, as the CPU becomes more heterogeneous I think more numbers are needed to quantify things. First it was clock speed. Now it's clock speed and core count. Soon it will be CPU clock speed, CPU cores, GPU clock speed, GPU cores, and whatever specialized stuff also ends up on the chip.
That is not my logic at all. Not only are those cores on different dies but in completely separate packages with no direct interconnect.
When the x86 and graphics gpcpu cores are on the same die and leverage the same memory controller(s) you just cannot add cores of one type without impacting the performance of cores of the other type. You have to balance the allocation of cores.
So yes if intel wants the intgrated graphics to go faster they will have to cap the number of x86 cores significantly under 8. There is not that much memory bandwidth to go around.
I was addressing the point of 8 being a default. I don't think the overall archietcure familys going forward are going to have some fixed default. There will be variants within a family where the cores are mixed and matched in different balances.
There is a fixed transistor and bandwidth budget. You can't willy nilly just replicate cores on a die without balancing the system. And it isn't far fetched since CPU + GPU packages will be shipping in volume in about 4 months which have to make exactly these kinds of tradeoffs.
What is more far fetched is 8 cores sharing memory bandwidth some moderly fast GPU cores. Strip off the GPU and it gets more believable. So it does come down to how many of what kind.
VirtualRain
macrumors 603
I'll be curious to see how they configure things since there's also a TDP budget constraint as well. I really don't see much value outside the low-end of the market in having a GPU (that's relatively poor performing by Intel's track record) stealing transistors, bandwidth and thermal budget from the CPU.