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.
One user's very coherent argument in favor of the nMP
- Thread starter zephonic
- Start date
- Sort by reaction score
Running my broadcast internet show I have between 10-14 cables coming out of my 2009 Mac Pro.
[doublepost=1455464937][/doublepost]
This. I think the majority of issues from the current Mac Pro users was a shift in workflow. New and upcoming pro's will simply choose whats currently available in the market.
Last edited:
And they would not have had the PCIe lanes to do so. To get the amount of thunderbolt, dual video cards and SSD speed with current Xeon chipsets, the PCIe slots had to go. There are not enough lanes left on the CPU/Chipset.
edit:
Sounds like @MacVidCards should, you know... develop some mac pro video cards. They're on slots (or hey, build/support/sell an eGPU via thunderbolt. they work). Or, i dunno, keep whining about the new Mac Pro making his business irrelevant. That's a sure fire business model idea!
[doublepost=1455465663][/doublepost]
Pot. Meet Kettle.
Last edited:
Excellent point. And so obvious it makes me wonder why I never thought of that!
Same here. I have seven cables permanently connected to my '09 MacPro. Eight before I bought the WiFi card. Moving the internal HDD's out of the box would only add one more Thunderbolt cable.
Last edited:
AS it is, from memory the current nMP doesn't quite have enough lanes to run all the thunderbolt slots at full speed at the same time.
Honestly, I don't understand what your post is about. Having dealt with both hardware (desktop) and OS migrations for multi-state companies, there is nothing in your post that reflects value of the Mac Mini Pro.
1) A C type language, microcode then hardware - says nothing with respect to the Mac Mini Pro
2) The Mac Mini pro "going that direction" - what direction?
3) Quantum leap in that area - what area? What are you talking about?
4) Large scale customer support standpoint - any enterprise that has only one model of desktop makes support easier.
5) Executing migrations correctly - again has nothing to do with Mac Mini Pro. PM work remains the same.
6) You lose short term business betting on the long term - what are you talking about?
7) Referencing videos such as police videos etc. again has nothing to do with advantages of the Mac Mini Pro. Nothing.
The reality is that it is a somewhat closed system like Micro-Channel was for IBM. Does it work, sure..can it do certain things well yes. Is it more like an appliance in that little change can be done to it on its internals, yes. However, there are so many counterparts that do the same thing and are more flexible systems. You have presented not one tangible point that supports the Mac Mini Pro as a superior contender.
Normally, I don't reference my work experience but you brought up yours and now you have someone who is contrary to your offering and hope you will be kind enough to explain your thoughts here.
Yep, Anand covered that when he first reviewed it: http://www.anandtech.com/show/7603/mac-pro-review-late-2013/8
Yeah, it was pretty astute observation. The dual graphic cards already take up 32 lanes only leaving 8 left for everything else.
[doublepost=1455468285][/doublepost]
Just got some quick info on anandTech article on that.
[doublepost=1455468712][/doublepost]
Perhaps that is why they used DDR 3 memory rather than DDR 4 and/or 4 memory slots?
Last edited:
Don't think the Xeon chipset from 2013 supported DDR4?
The workstation/server chipsets from Intel tend to lag the consumer stuff in terms of feature set for at least 12-18 months.
edit:
according to wikipedia, DDR4 wasn't available until september 2014, so it was too late for the nMP.
Gals - this makes no sense.
An upgraded cMP with E5-2600 CPUs would have almost twice as many PCIe lanes as the cylinder. With an E5-1600 it would have the same number of lanes.
Do you think that there's something magical about a cylinder - that it can create more PCIe out of thin air?
A cMP with E5-x6xx chips would simply have the GPUs in standard form factor upgradeable cards, as well as the other PCIe consumers would either be embedded or in cards.
Also, the comment "graphic cards already take up 32 lanes only leaving 8 left for everything else" is wrong.
The MP6,1 has 48 PCIe lanes, so 16 are available. (And it practical terms, Apple could run GPU A on 8 lanes without a noticeable effect on performance for the majority of apps.)
Last edited:
Because there is significant demand for a desktop GPU connectivity now that the technology is in place to allow it. People have realised they can have ultra slim laptops with powerful CPUs and integrated graphics when moving around and then dock it with a GPU and large storage. These companies know what the demand is and are now offering the solutions. They don't care if someone on a forum is a naysayer.
I'm a dude like my user name implies.
If you think its wrong you might want to take it up with Anand on his AandTech website:
http://www.anandtech.com/show/7603/mac-pro-review-late-2013/8
With a quick check, the current Xeon processors in the nMP is still showing 40 PCI lanes and not 48.
They made the decision to gain full unrestricted access to the GPU rather than less to make full use of openGL in applications.
Last edited:
Could be use of the bridging/switching of PCI lanes as shown by the AnandTech article I already linked. Allowing more PCI lanes with using only the real 8 PCI lanes left.
Last edited:
The 4th diagram shows the switch connected to the CPU:
The third diagrams shows how the 8 PCIe lanes on the C600 are used:
40 + 8 = 48
There are even four unused PCIe 2.0 lanes still available.
The description under that diagram says "Here the CPU has a total of 40 PCIe 3.0 lanes."
I was asking a genuine question. I couldn't think why you would want to hot plug/add a GPU. Your example provides an answer. It still seems a lot of effort to go to to save a few seconds to hot add rather than connect & reboot.
I used to work with proper grown up servers that had all manner of hot swap capabilities supported by hardware & the OS but people still preferred to power down the machine to change components like RAID or Ethernet cards.
I don't think your going to squeeze 48 lanes out of a 40 lane Xeon chip as much as you can try.
Any lanes past 40 are going to be shared by other components, meaning they are going to be lower bandwidth depending on the conditions.
Apple's squeezing 56 lanes out of the system....
I think someone's moved the goalposts here....
And look what Apple's put on the 8 extra lanes:
- two GbE ports (low bandwidth)
- wifi controller (low bandwidth)
- PCIe SSD (low bandwidth)
- USB 3.0 (enough bandwidth for one port to run at full speed, but over-subscribed with four ports)
Cool. Basically they target college and uni kids with this stuff. They can keep a GPU in the dorm and then walk around with an ultra slim notebook to class. It's better than buying a console. Video editors can also take advantage by capturing video on the field and then having fun compute power on the desktop. It's been a long time people have wanted to do this.
Alpine Ridge had two firmware updates in the last month for the Thunderbolt 3 features and Nvidia has introduced beta support for hot plugging and unplugging GPUs. So sometime this year the problems will be ironed out and people won't experience system freezes.
I don't think anyone is moving goalposts when your on the wrong football field. What I said was correct in the context as I gave it. Of the 40 PCI lanes coming from the Xeon CPU, the 8 lanes left are used for everything else INCLUDING..... the bridging chips allowing for the extra PCI lanes.
Please take a closer look at the diagrams from the page that you referenced.
There are 8 free PCIe 3.0 lanes from the CPU, and 8 free PCIe 2.0 lanes from the C600 PCH.
16 lanes "for everything else", not 8. Apple uses a PLX switch to turn the 8 PCIe 3.0 lanes into 16 PCIe 2.0 lanes - and then gives 12 of those lanes to the T-Bolt 2 controllers, leaving 4 PCIe 2.0 lanes unused.
Also note that the "context" was to a branch from a comment from throAU that "There are not enough lanes left on the CPU/Chipset." The context was system lanes, not CPU lanes.