|
|
New Mac Pros available tomorrow
|
|
|
|
Moderator
Join Date: Aug 2001
Location: Nobletucky
Status:
Offline
|
|
It's official.
The new Mac Pros will be available for order starting tomorrow.
|
|
|
|
|
|
|
|
|
Mac Elite
Join Date: Apr 2003
Location: Hong Kong
Status:
Offline
|
|
Store is down.
|
|
|
|
|
|
|
|
|
Moderator
Join Date: Aug 2001
Location: Nobletucky
Status:
Offline
|
|
For chuckles, I maxed-out the options on just a MacPro (no monitor, keyboard, etc. Just the Pro)
• 2.7GHz 12-core with 30MB of L3 cache
• 64GB (4x16GB) of 1866MHz DDR3 ECC RAM
• 1TB PCIe-based flash storage
• Dual AMD FirePro D700 GPUs with 6GB of GDDR5 VRAM each
$9,599.00
Add a 27" Thunderbolt display, and the total comes to $10,598.00. Opt for the 32" Sharp 4K display, and you're talking $14,193.00. Free shipping, though.
|
|
|
|
|
|
|
|
|
Posting Junkie
Join Date: Oct 2005
Location: Houston, TX
Status:
Offline
|
|
The push for dual GPU and single CPU is so weird given how little software is fully GPU accelerated.
|
|
|
|
|
|
|
|
|
Clinically Insane
Join Date: Nov 1999
Location: 888500128, C3, 2nd soft.
Status:
Offline
|
|
Chicken/egg.
It will be, soon.
|
|
|
|
|
|
|
|
|
Moderator
Join Date: Apr 2000
Location: Gothenburg, Sweden
Status:
Offline
|
|
Sure, but it would be a nice touch to offer a cheaper version with "only" a single GPU. They can just leave that side of the triangle empty.
|
The new Mac Pro has up to 30 MB of cache inside the processor itself. That's more than the HD in my first Mac. Somehow I'm still running out of space.
|
|
|
|
|
|
|
|
Clinically Insane
Join Date: Nov 1999
Location: 888500128, C3, 2nd soft.
Status:
Offline
|
|
For what market? The segment that wants more cores than an iMac but uses tools that won't ever be updated to utilize the GPUs?
That must surely be a tiny sliver of an already small market.
|
|
|
|
|
|
|
|
|
Moderator
Join Date: Apr 2000
Location: Gothenburg, Sweden
Status:
Offline
|
|
Not sure how small that market slice is, and it's easy enough to make such a variant.
|
The new Mac Pro has up to 30 MB of cache inside the processor itself. That's more than the HD in my first Mac. Somehow I'm still running out of space.
|
|
|
|
|
|
|
|
Posting Junkie
Join Date: Oct 2005
Location: Houston, TX
Status:
Offline
|
|
I haven't seen anything conclusive on the GPUs... they're some Tahiti part (7970?) that AMD is letting Apple use non-standard FirePro markings on without any of the usual FirePro goodies (ECC RAM, software certification, SDI, etc) in bridgeless CrossFire?
|
|
|
|
|
|
|
|
|
Moderator
Join Date: Aug 2001
Location: Nobletucky
Status:
Offline
|
|
|
|
|
|
|
|
|
|
|
Administrator
Join Date: Jun 2000
Location: California
Status:
Offline
|
|
Anyone here received one yet? Especially with the 3x 4K displays. Your thoughts and pix about the system would be interesting.
|
|
|
|
|
|
|
|
|
Administrator
Join Date: Mar 2000
Location: Land of the Easily Amused
Status:
Offline
|
|
I should be getting one soon. Next week hopefully.
|
|
|
|
|
|
|
|
|
Junior Member
Join Date: Nov 2011
Status:
Offline
|
|
Originally Posted by mduell
The push for dual GPU and single CPU is so weird given how little software is fully GPU accelerated.
Why are 12 physical cores on two CPUs better than 12 on one CPU? It has to be cheaper to produce the single CPU (pricing may be another issue, particularly early in the run). Do you have an application that would benefit from 24 physical cores?
The more likely answer is that Apple spent a lot of time studying, through online surveys (and I wouldn't be surprised by if they also interviewed folks in the content production industry face to face), what people actually did with their Mac Pros. I participated in one of the surveys back in (iirc) early 2010. I suspect that, like me, there were plenty of people who drove more than one (or two) display(s) with their Mac Pros.... and that they outnumbered the users who had all four disk slots filled, or who needed more than 12 physical cores to get their production jobs down in a reasonable amount of time.
For what it's worth, I currently run one 2010 Mac Pro with six (6) LED Cinema HD displays (two on Dual Link DVI-to-mDP adapters from Atlona) driven by two (2) 5770s and one 2009 model with the same cards, each driving a single, 30-inch Cinema Display over Dual-Link DVI, in addition to a couple with more normal desktop setups (single displays).
|
|
|
|
|
|
|
|
|
Moderator
Join Date: Aug 2001
Location: Nobletucky
Status:
Offline
|
|
|
|
|
|
|
|
|
|
|
Administrator
Join Date: Jun 2000
Location: California
Status:
Offline
|
|
Originally Posted by sunman42
Why are 12 physical cores on two CPUs better than 12 on one CPU? It has to be cheaper to produce the single CPU (pricing may be another issue, particularly early in the run).
Each CPU socket will have its own memory channel(s). So splitting the cores across 2 sockets doubles system memory bandwidth. Likewise each socket has its own connection to the rest of the system. So one socket could be talking to a PCIe card, while a core on the 2nd CPU handles RAID I/O to the drive array. Each CPU has full independent I/O bandwidth.
|
|
|
|
|
|
|
|
|
Moderator
Join Date: Aug 2001
Location: Nobletucky
Status:
Offline
|
|
I had no idea just how small the new Pro is until now. Here's a pic of it next to the previous Pro...
|
|
|
|
|
|
|
|
|
Moderator
Join Date: May 2001
Location: Hilbert space
Status:
Offline
|
|
Is that your new machine?
|
I don't suffer from insanity, I enjoy every minute of it.
|
|
|
|
|
|
|
|
Moderator
Join Date: May 2001
Location: Hilbert space
Status:
Offline
|
|
Originally Posted by mduell
The push for dual GPU and single CPU is so weird given how little software is fully GPU accelerated.
If you look at the scaling of cpu vs. gpu performance in Anandtech's review, you'll see that the only way to substantially increase performance is to tap the gpus for compute. If I understand Apple's move correctly, this is the »next MacBook Air«, i. e. they're showing off where computers are going (more emphasis on gpu computing, PCIe SSDs, etc.).
|
I don't suffer from insanity, I enjoy every minute of it.
|
|
|
|
|
|
|
|
Addicted to MacNN
Join Date: Oct 2001
Location: Automatic
Status:
Offline
|
|
I find this picture from OreoCookie's link to Anandtech's review to be even more eye-opening. This machine is more and more covetable by the minute.
|
|
|
|
|
|
|
|
|
Moderator
Join Date: Aug 2001
Location: Nobletucky
Status:
Offline
|
|
Originally Posted by OreoCookie
Is that your new machine?
I wish.
There's no way I could justify a new Mac Pro. My work doesn't call for that kind of muscle, and my wallet can't handle the cost.
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: May 2008
Status:
Offline
|
|
Originally Posted by angelmb
I find this picture from OreoCookie's link to Anandtech's review to be even more eye-opening. This machine is more and more covetable by the minute.
Looks just like a small chrome trash can!
|
|
|
|
|
|
|
|
|
Professional Poster
Join Date: Feb 2000
Location: Nashua NH, USA
Status:
Offline
|
|
Originally Posted by auto_immune
Looks just like a small chrome trash can!
To kickstarter!
|
|
|
|
|
|
|
|
|
Moderator
Join Date: Apr 2000
Location: Gothenburg, Sweden
Status:
Offline
|
|
Originally Posted by reader50
Each CPU socket will have its own memory channel(s). So splitting the cores across 2 sockets doubles system memory bandwidth. Likewise each socket has its own connection to the rest of the system. So one socket could be talking to a PCIe card, while a core on the 2nd CPU handles RAID I/O to the drive array. Each CPU has full independent I/O bandwidth.
All of this is correct, but Intel is not stupid. Memory bandwidth has gone up from 3 channels at DDR3-1333 to 4 channels at DDR3-1600 - or from 32 GB/s per socket (so 64 GB/s total) to 51.2 GB/s in total. Moreover, that meant that half the system memory was always one jump away - you had to use a QPI channel to talk to the other memory controller, which added some latency and restricted bandwidth slightly (25.6 GB/s max for the QPI link for the old Westmere models). For PCIe channels, the situation is even more improved: The old Westmere Xeons had 16 PCIe 2.0 channels per socket, and the new Ivy Bridge Xeons have 40 PCIe 3.0 channels in total. Since PCIe 3.0 doubles the bandwidth from PCIe 2.0, that means the total system bandwidth has more than doubled anyway. In fact, the new MP is insanely overprovisioned in the bandwidth department.
|
The new Mac Pro has up to 30 MB of cache inside the processor itself. That's more than the HD in my first Mac. Somehow I'm still running out of space.
|
|
|
|
|
|
|
|
Moderator
Join Date: Apr 2000
Location: Gothenburg, Sweden
Status:
Offline
|
|
Originally Posted by mduell
I haven't seen anything conclusive on the GPUs... they're some Tahiti part (7970?) that AMD is letting Apple use non-standard FirePro markings on without any of the usual FirePro goodies (ECC RAM, software certification, SDI, etc) in bridgeless CrossFire?
Anand confirms that the D700 is an underclocked Tahiti, and that the others are Tahiti LE and Pitcairn with similar underclocks. The D500 is a "7930" in that it has the full memory bandwidth but even more clusters are disabled than in the 7950. The D700 has 6 gigs of RAM, which the 7970 doesn't normally get. Doesn't seem to be any differences in DP performance, and there is no ECC, so… this is the basic Radeons.
|
The new Mac Pro has up to 30 MB of cache inside the processor itself. That's more than the HD in my first Mac. Somehow I'm still running out of space.
|
|
|
|
|
|
|
|
Moderator
Join Date: May 2001
Location: Hilbert space
Status:
Offline
|
|
AFAIK the AMD Tahiti GPU uses a virtual ECC scheme (as far as I understand it's a RAID5 for RAM), so they don't actually need ECC RAM modules. If I understand the situation correctly, Apple could reconfigure their cards so that you only have 5.25 GB »video« RAM per card at your disposal.
|
I don't suffer from insanity, I enjoy every minute of it.
|
|
|
|
|
|
|
|
Administrator
Join Date: Mar 2000
Location: Land of the Easily Amused
Status:
Offline
|
|
|
|
|
|
|
|
|
|
|
Administrator
Join Date: Jun 2000
Location: California
Status:
Offline
|
|
You didn't simulate maxed-out GPUs also. Might I suggest smcFanControl, manually forcing maximum fan?
It would be most interesting if a fan control could reverse the fan direction. The MacPro is doubtless too heavy to hover, but one could hope.
|
|
|
|
|
|
|
|
|
Administrator
Join Date: Mar 2000
Location: Land of the Easily Amused
Status:
Offline
|
|
I also simulated with maxed-out GPUs, but there was no change. Using the fan control apps, it's still insufficient for lift. It does, however, swirl the ball around in the base endlessly. So that's fun.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Forum Rules
|
|
|
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
|
HTML code is Off
|
|
|
|
|
|
|
|
|
|
|
|