WWDC wishlist

[rockyromero]

The Mac Pro assembly factory will continue to produce a 2013 Mac Pro form factor but it will now be a high end kitchen appliance. It's exact function and advanced capabilities are still very secret but this will be announced at a later date well within time for the holiday shopping season. It will also come in a limited choice of colors."

I'll celebrate with that and buy all the MPs that I can.

The form, the function & potential is perfect for semi Pro applications.

At a fraction of the price, the secondary market would go wild.

This is a machine that is built for business.

➰

 

[H2SO4]

I’m actually mostly wishing for a nMP that will be best with a VM. Should I therefore be hoping Apple announce one with more cores of the same speed, faster RAM, faster CPU but same number of cores or even bigger GPU(s)?

 

[rockyromero]

I’m actually mostly wishing for a nMP that will be best with a VM. Should I therefore be hoping Apple announce one with more cores of the same speed, faster RAM, faster CPU but same number of cores or even bigger GPU(s)?

Do you mean VM: Virtual Machine?

I use a nMP with multiple VM partitions successfully and want to improve on that.

If so, could you expand on how that is related?

➰

 

[H2SO4]

Do you mean VM: Virtual Machine?

I use a nMP with multiple VM partitions successfully and want to improve on that.

If so, could you expand on how that is related?

➰

I do mean virtual machine yes. What I’m getting at is what is the focus for hardware if you are running Parallels for example?

 

[koyoot]

I do mean virtual machine yes. What I’m getting at is what is the focus for hardware if you are running Parallels for example?

You should ask for good and reliable OS in the first place. That is where it lacks the most.

From hardware point of view: RAM> CPU> SSD> GPU in that order. With highest possible attention to first 3.

 

[rockyromero]

You should ask for good and reliable OS in the first place. That is where it lacks the most.

MacOS has been extremely reliable as compared to the W7 world.

I do have an unresolved issue of running over 32 concurrent VM and the internet connectivity drops over that for each subsequent VM.

Don't know if it's a hardware or software issue and have no way of easily troubleshooting.

Using an additional nMP may solve that limitation. I can only guess.

I'll know as I start to expand my capabilities.

➰

 

[koyoot]

Different people have different experiences 😉.

I have switched to Linux/Windows Xeon E3v5/GTX 980 setup because it was more reliable in stability than OSX. There is no OS as stable as Linux in my experience.

 

[H2SO4]

You should ask for good and reliable OS in the first place. That is where it lacks the most.

From hardware point of view: RAM> CPU> SSD> GPU in that order. With highest possible attention to first 3.

Wasn’t sure where you were going with this until post 32. A good reliable OS is great but you’ll never be able to accomplish everything you want to and more, if you stick to just one.
So then my 2016 nMP ideal BTO config becomes;
24GB 2133 DDR4 RAM.
6 Cores @ >3.5GHZ
Dual NVMe SSD.
Single D500, (or whatever they replace it with).

 

[koyoot]

Wasn’t sure where you were going with this until post 32. A good reliable OS is great but you’ll never be able to accomplish everything you want to and more, if you stick to just one.
So then my 2016 nMP ideal BTO config becomes;
24GB 2133 DDR4 RAM.
6 Cores @ >3.5GHZ
Dual NVMe SSD.
Single D500, (or whatever they replace it with).

RAM for Broadwell-EP CPUs should have 2400 MHz 😉.

 

[rockyromero]

From hardware point of view: RAM> CPU> SSD> GPU in that order. With highest possible attention to first 3.

Agreed.

My hardware now far exceeds my software limitations and improvements have minor effects.

I'm of the thought that more boxes (cylinders) are better than one powerful machine for a variety of reasons including redundancy.

➰

 

[koyoot]

Wasn’t sure where you were going with this until post 32. A good reliable OS is great but you’ll never be able to accomplish everything you want to and more, if you stick to just one.
So then my 2016 nMP ideal BTO config becomes;
24GB 2133 DDR4 RAM.
6 Cores @ >3.5GHZ
Dual NVMe SSD.
Single D500, (or whatever they replace it with).

I forgot about one thing. Software is most important. You know, this board is full of people who as a cure for software bottlenecks are suggesting buying faster GPUs. What is the point of this? Start demanding from developers faster software that can use all of capabilities from your hardware.

"A client comes to hardware vendor with a question: "- My software bottlenecks me. No new hardware is able to influence my workability on projects, what should I do? How can I contact with developers?"
Customer service tries to convince client that all he needs is brand new hardware, that is barely faster than previous version."

What is the point?

 

[Peace]

Man all I want is to be able to upgrade the video cards in my current nMP.

 

[lowendlinux]

I forgot about one thing. Software is most important. You know, this board is full of people who as a cure for software bottlenecks are suggesting buying faster GPUs. What is the point of this? Start demanding from developers faster software that can use all of capabilities from your hardware.

"A client comes to hardware vendor with a question: "- My software bottlenecks me. No new hardware is able to influence my workability on projects, what should I do? How can I contact with developers?"
Customer service tries to convince client that all he needs is brand new hardware, that is barely faster than previous version."

What is the point?

There isn't but I also wouldn't be trying to run 32 VM's on a Mac Pro nor a cluster of them unless they were "just because". I'd go IX systems for the HW and Red Hat for the SW and depending on what's being done on those machines I might not even run VM's

 

[AidenShaw]

There isn't but I also wouldn't be trying to run 32 VM's on a Mac Pro nor a cluster of them unless they were "just because"....

A small VM for me is 8 cores, 64 GiB RAM, and 2 TB disk. vCenter with SAN storage.

 

[rockyromero]

There isn't but I also wouldn't be trying to run 32 VM's on a Mac Pro nor a cluster of them unless they were "just because". I'd go IX systems for the HW and Red Hat for the SW and depending on what's being done on those machines I might not even run VM's

I'm ramping up to 100+ VM.

It's production.

Business, you know.

I don't know enough about IX and Red Hat, yet.

All in all, WWDC will bring MacOS, iOS, TVOS, WatchOS and cumulative, it will amount to the whole experience.

The hardware is incremental and would allow me to plan ahead.

Riding the Apple ecosystem is better, for now, than the alternative.

➰

 

[briloronmacrumo]

new Mac Pro:

• E5v4 at upto 135W TDP
• GPUs AMD Polaris for D300/D500 replacement, and rebranded W8100 or Fury Nano for D700 Replacement
• Propietary SSD NVMe
  
• 6 TB3 +4 USB-C (2nd GPU on x8 Bus)
• No 10Gbt Ethernet.
• 550W PSU

Thanks Mago, I could live with that 😉 but Apple's absent Mac Pro-related deliverables ( i.e. monitor, Mac Pro and upgrades to same ) don't leave me sanguine about their Pro desktops ( either at WWDC or later ). I'd like to be shown/persuaded otherwise.

 

[AidenShaw]

I'm ramping up to 100+ VM.
➰

Which OS(s) on the VMs?

 

[rockyromero]

Which OS(s) on the VMs?

I use Ubuntu OS on the VM for the most part.

I do have W7 and W10 also, but they use more memory & storage needlessly.

Unless I can solve my 32 VM limitation with dropping internet connectivity, I will continue with nMP cylinders.

To get over 100 VMs, I anticipate having four cylinders.

Hey, I could be running on all 4 cylinders at some point!

➰

 

[lowendlinux]

Docker..

You really need to get ahold of someone who does IT infrastructure. What are the VMs used for, what are the people doing on the VM, are they using PCI pass through what's the expectation with on close, what are their real storage requirements, and on and on. If you try to run 100 VMs even on a maxed out MP not one will be happy least of all you there's just not enough cores, networking or RAM to make that pleasant let alone things like redundancy.

Before you spend your bosses money give this stuff serious thought. You're plopping a round peg into a square hole and the results won't be pretty

 

[rockyromero]

You really need to get ahold of someone who does IT infrastructure.

That would be me. And that's why I'm here to learn until I find someone that knows better.


What are the VMs used for, what are the people doing on the VM, are they using PCI pass through what's the expectation with on close, what are their real storage requirements, and on and on.


The VMs each run a stand alone application for each client automatically, without the client interfacing. Local storage is negligible besides the standard 10G, since the work is cloud based. PCI pass through is non existent or non important, I believe.


If you try to run 100 VMs even on a maxed out MP not one will be happy least of all you there's just not enough cores, networking or RAM to make that pleasant let alone things like redundancy.


Each VM requires 1G of RAM and with 128G on the nMP, I have more than enough. On the Mac Mini, I struggle with 16G maxed out, slower speeds, and capacities.

Scaling seems to simply mean adding additional cylinders as the requirements increase.

Obviously, I want to expand as judiciously as I can in an area that I know less about which is the hardware aspect. The software requirements are static, and consistent.

The feedback here is helpful and I'm giving this ongoing thought.

➰

 

[Mago]

Thanks Mago, I could live with that 😉 but Apple's absent Mac Pro-related deliverables ( i.e. monitor, Mac Pro and upgrades to same ) don't leave me sanguine about their Pro desktops ( either at WWDC or later ). I'd like to be shown/persuaded otherwise.

Ok, my last minute leaks aware WWDC Annoucements should be the following:

1. New MacBook Pro 14"/16" Introduction, Available Q3 with MacOS, 2TB3 + 2 USB-C, 16" models with dGPU (AMD Polaris) also dp1.3 capable on USB-C ports.
2. New MacPro introduction maybe Available Next Month, Xeon E5v4 upto 145W TDP, AMD Polaris for D300/D500 Replacement and Fury Nano or W81000 nonECC for D700, 4 TB3 ports 6 USB-C Usb-C ports DP1.3 Capables, Dual PCIE2 based NVMe SSD upto 2TB, Upto 128GB ECC Ram, Ehternet Ports Either Single or Dual at 1Gbps, no 10GbT yet.
3. Thunderbolt 3 or Usb-C (dp1.3) 5K Retina Cinema Display, no way integrated GPU, not yet not never.
4. New Airport Extremme Timecapsule, maybe loaded with full headless MacOS Server / ARM A9 Cpu, upto Dual HDD models, and Siri on Command.

That's all. let's see how much of this I Match.

Later This Year Apple should introduce Updated iMac maybe based on AMD Zen APU, same should introduce a Mac Nano, Quad Core no integrated PSU (powered by USB-C Display or Brick) 4 USB-C Dp1.2, 2 TB3, 1 Hdmi 1 Ethernet, similar design as current mini on half volume.

 

[pat500000]

Man all I want is to be able to upgrade the video cards in my current nMP.

You could..buy a Mac Pro 2016.

 

[lowendlinux]

That would be me. And that's why I'm here to learn until I find someone that knows better.


What are the VMs used for, what are the people doing on the VM, are they using PCI pass through what's the expectation with on close, what are their real storage requirements, and on and on.


The VMs each run a stand alone application for each client automatically, without the client interfacing. Local storage is negligible besides the standard 10G, since the work is cloud based. PCI pass through is non existent or non important, I believe.


If you try to run 100 VMs even on a maxed out MP not one will be happy least of all you there's just not enough cores, networking or RAM to make that pleasant let alone things like redundancy.


Each VM requires 1G of RAM and with 128G on the nMP, I have more than enough. On the Mac Mini, I struggle with 16G maxed out, slower speeds, and capacities.

Scaling seems to simply mean adding additional cylinders as the requirements increase.

Obviously, I want to expand as judiciously as I can in an area that I know less about which is the hardware aspect. The software requirements are static, and consistent.

The feedback here is helpful and I'm giving this ongoing thought.

➰

If the VM's are running one application you don't need a VM you need a container.

Look up docker it will do what you want without the resources you need to run a whole machine. If you still insist on running full VM's drop them on top of Linux (since most are Linux) and use the native KVM then manage with the Mac if that's where your comfortable.

The tubes are not meant to do what it is you want to do they're core, memory, and PCI starved even clustered.

 

[rockyromero]

If the VM's are running one application you don't need a VM you need a container.

Look up docker it will do what you want without the resources you need to run a whole machine. If you still insist on running full VM's drop them on top of Linux (since most are Linux) and use the native KVM then manage with the Mac if that's where your comfortable.

The tubes are not meant to do what it is you want to do they're core, memory, and PCI starved even clustered.

Now I know what you meant by Docker.

https://www.docker.com

This is a new concept to me and I will consider this since it also interfaces with VMware.

I'm using VMWare fusion 8 now on MacOS and it runs well.

Docker may have scaling opportunities that I'm not aware of yet.

Would you suggest Red Hat on a nMP with Docker?

➰

 

[Ph.D.]

You could..buy a Mac Pro 2016.

I think that's Apple's message, like it or not. Pro users who want to stay with Apple need to get settled on a 2-3 year upgrade cycle. Buy the new one after about 3 years, sell the old (prices should stay reasonable), and you have new graphics, updated CPU, faster ports, etc., for roughly what it would cost to upgrade and update everything at once in a tower anyway.

Not so bad if you think about it. The only thing you lose is the ability to do minor incremental updates. But those are slowing down these days. Think the wait for 14nm was bad? The wait for 10nm will be a long one too.