Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

jmcube

macrumors member
Jun 6, 2013
58
10
I was thinking of the Opterons. They are the one to compare to Xeons.

It seemed to me like Apple does a pretty good job of using multiple cores, so Opterons could be strong?

Also, MacPro is for serious work and Opterons are too.

This would also be a big change and not that crazy modular idea.

Jon
 
Last edited:

deconstruct60

macrumors G5
Mar 10, 2009
12,224
3,823
none Piledriver is single socket desktop

Not. Piledriver is a micro-architecture code name similar to "Sandy Bridge" , "Ivy Bridge" , "Haswell" on the intel side. Just like Intel the basic micro-architecture, with some minor adjustments, is used across the whole CPU product lines. For example,

http://www.anandtech.com/show/6430/amd-launches-opteron-6300-series-with-piledriver-cores

Similar to Intel the architecture rolls out slower to the Server variants. That is mainly a issue of demand, process technology yields , and size of the respective products. ( growing mobile, smaller dies have better yields, fewer cores leads to smaller dies ).


and Opteron's while most having lors of core are really geared toward servers. The also need to compile an AMD OSX which means two versions of OSX. It'd be a bad call unless the whole line changes to AMD

And yet Linux , Windows , etc. all manage this with little to no difficulties. Talking a very small subset of privileged mode instructions. Normal apps (which is a large bulk of what folks sweep up into the umbrella of SO ) don't really have any differences ( can play with optimizer settings. and Intel's compilers won't do a good job but this is not a "loose sleep" issue).

Apple is far more likely not to switch to AMD because they can't switch all of the x86 CPU purchase orders over to AMD. Apple's buying cloak with Intel helps them influence things like a more aggressive stance on integrated graphics, but far more it gets them discounts. Apple is Scrooge McDuck when it comes to saving money.

----------

I was thinking of the Opterons. They are the one to compare to Xeons.

It seemed to me like Apple does a pretty good job of using multiple cores, so Opterons could be strong?

AMD cores don't have as high of throughput. They generally have more in the Opteron space but not necessarily better ones. They also tossed SMT/hyperthreading so matched in contexts were this is significant memory latency generated by branching/varying memory access driven by workload.

Many computational benchmarks have Intel generally out in front in terms of $/performance .


If just counting cores , without respect to performance , then the GPGPUs have both of them beat. If an embarassingly parallel float workload then both on performance, core count, and $/perf .
 

GermanyChris

macrumors 601
Jul 3, 2011
4,185
5
Here
Not. Piledriver is a micro-architecture code name similar to "Sandy Bridge" , "Ivy Bridge" , "Haswell" on the intel side. Just like Intel the basic micro-architecture, with some minor adjustments, is used across the whole CPU product lines. For example,

http://www.anandtech.com/show/6430/amd-launches-opteron-6300-series-with-piledriver-cores

Similar to Intel the architecture rolls out slower to the Server variants. That is mainly a issue of demand, process technology yields , and size of the respective products. ( growing mobile, smaller dies have better yields, fewer cores leads to smaller dies ).




And yet Linux , Windows , etc. all manage this with little to no difficulties. Talking a very small subset of privileged mode instructions. Normal apps (which is a large bulk of what folks sweep up into the umbrella of SO ) don't really have any differences ( can play with optimizer settings. and Intel's compilers won't do a good job but this is not a "loose sleep" issue).

Apple is far more likely not to switch to AMD because they can't switch all of the x86 CPU purchase orders over to AMD. Apple's buying cloak with Intel helps them influence things like a more aggressive stance on integrated graphics, but far more it gets them discounts. Apple is Scrooge McDuck when it comes to saving money.

----------



AMD cores don't have as high of throughput. They generally have more in the Opteron space but not necessarily better ones. They also tossed SMT/hyperthreading so matched in contexts were this is significant memory latency generated by branching/varying memory access driven by workload.

Many computational benchmarks have Intel generally out in front in terms of $/performance .


If just counting cores , without respect to performance , then the GPGPUs have both of them beat. If an embarassingly parallel float workload then both on performance, core count, and $/perf .

Opteron's don't use piledriver arch. piledriver is for their desktop line. Opteron's are not workstation processors for the reasons you stated below but they are great server processors
 

.:Aleph:.

macrumors member
Jun 4, 2013
74
4
Does it matter which display/display's?

Were you going to make a pricing argument or a display quality argument? If so make it...

I was more going for what is going to happen with Apple's Thunderbolt Display line.

Would you care to guess?

And if there are no Apple HQ monitors in the near future, which way will you go?
 

goMac

Contributor
Apr 15, 2004
7,662
1,693
The also need to compile an AMD OSX which means two versions of OSX. It'd be a bad call unless the whole line changes to AMD

No they don't. AMD licenses Intel's instruction set, and all AMD chips are instruction compatible with Intel chips. There is no such thing as "compiling for AMD" in since they're the same instruction set. There is just x86.

People run the existing version of OS X on AMD all the time no problem.
 

GermanyChris

macrumors 601
Jul 3, 2011
4,185
5
Here
No they don't. AMD licenses Intel's instruction set, and all AMD chips are instruction compatible with Intel chips. There is no such thing as "compiling for AMD" in since they're the same instruction set. There is just x86.

People run the existing version of OS X on AMD all the time no problem.

when you build a hackintosh on AMD you need a custom kernal..
 

KaraH

macrumors 6502
Nov 12, 2012
452
5
DC
I was more going for what is going to happen with Apple's Thunderbolt Display line.

Would you care to guess?

And if there are no Apple HQ monitors in the near future, which way will you go?

Still a bit lost in how that relates to what you quoted me saying. It is a complete non sequitur.

What apple does with their displays does not factor into things, at least for me. If they do not make a monitor plenty of other people do. In fact since their monitors (well, docking stations) cost more than ones with comparable/better specs I probably would not have gone with them anyways unless they came up with a wow factor.
 

jmcube

macrumors member
Jun 6, 2013
58
10
Opteron's don't use piledriver arch. piledriver is for their desktop line. Opteron's are not workstation processors for the reasons you stated below but they are great server processors

Pretty sure the Opteron 6300 are Piledriver.

I thought that Intel being better at performance/$ was due to most programs only being written to utilize a couple of cores. Opterons are the ones that go over 12.

I know that GPUs do better than CPU at massive parallel, but Apple doesn't traditionally have great support for those either.

It is just that Apple is one of the main supporters of OpenCL and have GCD (both are very nice), so I expect that maybe they could utilize the improved parallel ability of the Opteron.

Jon
 

ThisIsNotMe

Suspended
Aug 11, 2008
1,849
1,062
I'll just leave this here.......

From MAY 22, 2013
Intel's product database has been updated, and it now shows five new Xeon Phi co-processors. These five are followups of the original Xeon 5110P, SE10P, and SE10X models. Two lighter Xeon Phi 3100 parts have shown up: a mid-end part, the 5120D, and two premium 7100 series parts.

For those who don't know what a co-processor is, in the case of these Xeon Phi co-processors, it is simply an x86 based processor slammed onto a PCIe 8x expansion card. The purpose of them is to increase processing power for desktops and servers, specifically for tasks that have to be executed by a processor, not a graphics card.

2ry4y2q.png


http://www.tomshardware.com/news/Intel-Xeon-Phi-Coprocessor-CPU,22700.html

Thunderbolt 2 is PCIe 8x compatible ;)
 

goMac

Contributor
Apr 15, 2004
7,662
1,693
That Xeon Phi article is a little misleading...

Xeon Phi is much more like a GPU than a CPU. It's NOT ideal for tasks that are best for CPUs, it's actually best for OpenCL tasks. It can't be used like a normal processor. In fact, it used to actually be a GPU with display output.

Also, I'm pretty sure Thunderbolt 2 is still 4x.
 

spunkgarLEWII

macrumors regular
Jun 1, 2013
100
0
I just got done telling that to a friend of mine. AMD is a hard one to make a hackintosh out of simply because it requires not only a custom kernel, but also the right firmware so that the OS can understand it.

Really, those with AMD PCs should not even bother creating hackintoshes as they are not easy to make. Intel is the way to go in terms of hackintosh.

when you build a hackintosh on AMD you need a custom kernal..
 

wallysb01

macrumors 68000
Jun 30, 2011
1,589
809
I thought that Intel being better at performance/$ was due to most programs only being written to utilize a couple of cores. Opterons are the ones that go over 12.

I think the modular core thing gets a lot of people confused. Every Pilerdriver module is 2 cores, but for a lot of workflows, even those that scale greatly past 12 cores, the useful integer to count are AMD modules or Intel cores. And once you do that, you find the gap in core count is gone, and all you're left with is Intel's greatly superior performance per core.

I've seen this in my own tasks having switched between E5620s, E5-2630s, AMD 6220s and AMD 6320s. The E5-2630s run laps around all of the rest, at least for my workloads.

The place AMD has an advantage is top end 4 socket servers. The reason it is only the top end, is because 2 socket Intel systems are generally still faster than bottom end 4 socket AMD 6300s, which will be similar in price. But, a 4 socket AMD 6380(or similar) system sits in a nice spot, that intel just doesn't cover. The E7-4600s at decent clock rates are super expensive.

I know that GPUs do better than CPU at massive parallel, but Apple doesn't traditionally have great support for those either.

Nor do a great many things...
 

ThisIsNotMe

Suspended
Aug 11, 2008
1,849
1,062
That Xeon Phi article is a little misleading...

Xeon Phi is much more like a GPU than a CPU. It's NOT ideal for tasks that are best for CPUs, it's actually best for OpenCL tasks. It can't be used like a normal processor. In fact, it used to actually be a GPU with display output.

Also, I'm pretty sure Thunderbolt 2 is still 4x.

I am sure Apple can figure it out with Grand Central Dispatch.
Anyways, form what I have read about the platform, each core is an x86 compatible chip minus some features found in "consumer" chips. No additional programming required to take advantage of Xeon Phi co-processor.
 

goMac

Contributor
Apr 15, 2004
7,662
1,693
I am sure Apple can figure it out with Grand Central Dispatch.

GCD doesn't work with expansion cards. There are a lot of reasons why that is true that aren't really fixable. GCD isn't magic, it's pretty much just an abstraction of the same threading API we've had all along. It's not some new technology.

OpenCL is the API used with the Xeon Phi.

Anyways, form what I have read about the platform, each core is an x86 compatible chip minus some features found in "consumer" chips. No additional programming required to take advantage of Xeon Phi co-processor.

It's not the language required, it's because it's an expansion card. When cards are not directly hooked to main system memory or the hard disks there are a lot of special things required to load software onto the card.

It's also not very much like a normal x86 processor. A lot of cores at a low clockspeed, again, just like a GPU.

The Xeon Phi is far more like a GPU because, again, it came out of a project to build an x86 GPU.
 

luffytubby

macrumors 6502a
Jan 22, 2008
684
0
Also they would never word it like Gigantic.


No internal expandability and no FW800 would make it unsellable to such a huge segment of the pro market (audio, video etc) that it won't happen. Maybe some time in the future we can all go 100% Thunderbolt but at this point there are too many FW/PCI reliant expansions that it doesn't make sense. If Apple does this now, they have indeed killed the Mac Pro.

Being Pro has nothing to do with being expandability. And of course they won't kill the machine because of that. Nobody gave a crap about a proprietary system like FW, which is why it's now dead. Give Apple more money and use their new formats like you are supposed to.




What is the whole point of calling it a MP if it has no internal expandability? You may as well just call it an iMac at that point.

If it means they are filled, well, hopefully BTO allows me to select all HHDs and take one out. For all the hype about SSDs the HDD is still better for data drives, time machine, and system snapshots. For the system drive I would prefer to use a Viper SSD from OWC.

the point of the iMac was to make a consumer friendly all-in-one. Not being able to upgrade it has nothing to do with the characteristics of it at all.

Apple makes TONS of money off their upgrades so forcing people to pay extra for drives, ram and such makes perfectly good sense. They are a business trying to make good looking computers. Apple is not about choice or getting the best deal.
 

KaraH

macrumors 6502
Nov 12, 2012
452
5
DC
the point of the iMac was to make a consumer friendly all-in-one. Not being able to upgrade it has nothing to do with the characteristics of it at all.

Apple makes TONS of money off their upgrades so forcing people to pay extra for drives, ram and such makes perfectly good sense. They are a business trying to make good looking computers. Apple is not about choice or getting the best deal.

Re-read what you quoted me saying. I was talking about expandability, not upgradeability. Actually, the only way you could enforce upgrades would be to force you to do them as a BTO option. Anything an Apple store can do to upgrade a part will be figured out by some third party and you can not void the whole machine's warranty if someone upgrades another component.

What I am looking for is something like empty 3.5 drive bays and free slots, both would allow people to do as they will with their machines. If you were making it a consumer device you would have neither.

An iToy user will not go through the hassle of upgrading, oh, their drive. Even if the hassle is minimal (it is a royal pain with the 2012 iMac, which is how Apple minimizes the issue) most of their users will just make all of the upgrade decisions on the BTO page at Apple's prices. Then buy a new iToy in a few years when they outgrow it (even though it will still work for years after).
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.