Yeah. That is exactly what they are saying. This is why they make any Mac soldered/unupgradeable. They want to keep you in buying new mode rather than keeping the same mode.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
WWDC wishlist
- Thread starter AidenShaw
- Start date
- Sort by reaction score
One thing that you have to realize is that currently AMD gets 227 GPU dies that are ready and working. On average, every single die costs AMD... 87$. That is how high cost of 14 nm wafer is currently. 20k Dollars. Nvidia has exactly the same situation, however, TSMC has a bit higher prices, it costs them 15K to buy a wafer and they sell each wafer for 18-20k depending on volume. Nvidia did not stacked a lot of those GPU dies, that is why there are huge shortages in the retail. Few reasons for this: Very high risk on production, secondly, very high prices of wafers, thirdly, very limited factory slots on TSMC. The production costs or rather wafer costs are presumed to lower over time, however we are talking not 5K$ but rather 12-15K end price for every wafer on 14/16nm node.
Why I write all this? Because production costs on 10 nm will be 4-5 times higher than are currently. That is why AMD says that 14 nm is the last node where we can see 600mm2 GPU. And multiple GPUs but smaller on smaller nodes are the way to go in future. Today we see something like this:
~115mm2 die size - low-end.
~230mm2 die size - mainstream.
- 300-400mm2 - midrange.
>400mm2 - high end.
In future:
<100mm2 - low-end.
<150mm2 - mainstream.
<250mm2 - MIdrange
<350mm2 high-end.
With extreme emphasize on multiple GPU setups.
Unless you want to pay as a consumer 40k$ for Titan X GPU class, that is reality.
More performance you will get not exactly with every node, but with architecture changes, and every single silicon vendor has to focus on it. That is why AMD changed their approach, and Vega is 14nm new architecture, from AMD, despite the fact, that Polaris this year is also new/updated architecture.
One more thing: I have many, many times written on this forum that future market is 90% BGA. And the shoes are starting to drop again, despite many people laughing about this possibility.
I think they solder things down, e.g. in laptops, because that's the way the entire electronics industry HAS to go to continue to make progress. (Yes, I'm sure there are some other motivations too, like dodging the support of people who have hacked their computers.)
Past some point, there is NO WAY to get the higher performance, lower power, and more compact packaging that people want and expect without using ball-grid-array or chip-scale packaging, etc. There's no chance that this can be upgraded by users.
[doublepost=1465766728][/doublepost]
Yes, the wall, in various forms, is approaching.
Why oh why would you run linux on a nMP? Why overspend and restrict yourself to an un upgradable machine when you can go with an HP or Dell workstation that can grow as your load expand or better yet go with a VM server cluster instead of waisting your company money on fashion statement instead of on the right tool for the job.
The short answer is: I don't know anything better.
The long answer is: I still don't know anything better than what I have working now.
If I can scale better, easier, and effectively? Yes, makes sense to me.
The VM server cluster may be what would work for me?
➰
You left one off:
Yawn, Nothing, not even waiting for Godot, and I don't really care anymore.
Apple will be off in the weeds until they fail and almost go out of business (again), or just actually go out of business. But this time Steve will not be there to reboot them. I've accepted the fact that it is going to be a slow and painful downhill ride for them, but if you can't see it coming you obviously have your head so far up your fanboy arse there's no hope for you.
I use 'you' in the general sense and am not directing it at the OP.
I for one am NOT planning to spend any $$ on Apple products for the foreseeable future.
Yawn, Nothing, not even waiting for Godot, and I don't really care anymore.
Apple will be off in the weeds until they fail and almost go out of business (again), or just actually go out of business. But this time Steve will not be there to reboot them. I've accepted the fact that it is going to be a slow and painful downhill ride for them, but if you can't see it coming you obviously have your head so far up your fanboy arse there's no hope for you.
I use 'you' in the general sense and am not directing it at the OP.
I for one am NOT planning to spend any $$ on Apple products for the foreseeable future.
Last edited:
Just get a nice server-grade machine focusing on RAM and CPUs (e.g. completely ignore the gpu) with an external optical storage if it's in your budget and disks on raid. Install VMware ESXi on it and import your existing VMs in it. You'll be amazed of the managing features and the flexibility you are going to have compared to client-grade vmware fusion running on nMP. I'd also bet that keep scaling out by adding nMPs is way more expensive, in the limits of wasting money, considering your virtualization needs.
Of course, also keep in mind that flexibility means that you need to take under consideration a few things beforehand. Like, projecting at the near future, how many VMs are you going to end up with.
This may be the exact migration solution that would fit for me.
http://www.vmware.com/products/vsphere-hypervisor/
I do want to stay with VMware because of the familiarity with its products.
They have a trial period that would allow me to use an existing server.
This would address resource allocation in a multiple server environment and continue to scale as I need more VM in the future. It's difficult to estimate the VMs needed as we grow with clients.
Thanks for this heads up as I missed looking further within VMware.
➰
Last edited:
I've Ubuntu 16.04 on my mac pro as alternate boot, but this is an condition I use primarly OS/X for development and most task, and sometimes I switch to Ubuntu to run certain tools not available or restricted on OS/X (as Intel PE), but in no way I'll choose an mac for Linux production, not only its cheaper, also easier to manage an PC Based Linux WS/Server, and actually I'm planning one to re-use the ECC ram from my next mac pro on a Xeon-D Server MB.
Consider Most Mac's logic board are sourced from Foxconn, even the best case you'll never get the quality/features for a server or a WS you'll find on brands a Gygabyte or Supermicro.
It's no sense.
[doublepost=1465826755][/doublepost]
At least on part u'll got served today.
I'm not as radical, but certainly Apple should segregate Services/Computers/Consumer Devices (iToys) operations and let it to be managed by peopler with real grounds on their sectors, Macintosh are being managed as Vacuum Machines like business, while Services seems Apple's black hole, and iToys seems land of Gods ruling them all...
IToys r us it appears to be.
Yes...they need sectors...as matter fact....because they are having identity crisis by having too many mixed categories, they should temporary shut down their business and do soul searching.
Actually the rumors and speculation of the bubble in R&D being the "apple car" .... I would say the car ( if Apple is actually working on a whole vehicle as opposed to just automotive electronics consoles ) is a bigger black hole that the services or iOS.
iOS is making money. Lots of it. To dismiss it as "iToys" is at best naive and at worst delusional. Likewise services is solidly on track to making more money than Mac system sales do. It is growing.
Macs have the same problem the iPhones are about to run into. Namely a growing segment of folks not buying anything because the old stuff is "good enough". [ There are several active buy/upgrade/drop money into my Mac Pro threads in the top 20 most active in this forum on a weekly basis. ] Apple is making fewer newer Macs designs but people are also not buying also. ( It isn't just a one-way street).
