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

What config are you likely buying?

  • 4-core, D300s

    Votes: 14 12.5%
  • 4-core, D500s

    Votes: 5 4.5%
  • 4-core, D700s

    Votes: 1 0.9%
  • 6-core, D300s

    Votes: 21 18.8%
  • 6-core, D500s

    Votes: 27 24.1%
  • 6-core, D700s

    Votes: 12 10.7%
  • 8-core, D300s

    Votes: 2 1.8%
  • 8-core, D500s

    Votes: 5 4.5%
  • 8-core, D700s

    Votes: 9 8.0%
  • 12-core, D300s

    Votes: 2 1.8%
  • 12-core, D500s

    Votes: 0 0.0%
  • 12-core, D700s

    Votes: 14 12.5%

  • Total voters
    112
Choice B
Your work is intermittent, such as programming, with occasional high multi-core use. In this case don't go for cores. The 12 cores would put you at 2.7GHz which is where you'd be most of the time, versus 3.7 GHz for the quad. Clearly the quad would be faster most of the time. Sweet spot; the hex core. Only drops 200 MHz for 50% more cores, and basically the same stepping.

Mmmmm programming is highly parallel. Developers ideally want the 12 core, not the 4 core. The biggest time suck for developers is compiling, which scales evenly across cores.

Sure, most the time you aren't compiling, but those other times you are using a word processor to type code. You aren't going to notice the 1.0 ghz loss in a word processor, but you will notice the loss at compile time, and it'll be annoying if your projects are large.

Now of course not all developers can afford a 12 core, and a 4 core is certainly ok for developers. But in no way is a 4 core better than a 12 core for developers.

(Citation: I'm a developer)
 
Mmmmm programming is highly parallel. Developers ideally want the 12 core, not the 4 core. The biggest time suck for developers is compiling, which scales evenly across cores.

I haven't run tests too recently but in the past compiling was I/O bound, once CPU's had gotten above a few GHz and a few cores. I'm doubtful that with 12, no make that 24 cores wouldn't bottleneck on reading and writing the SSD.
 
I haven't run tests too recently but in the past compiling was I/O bound, once CPU's had gotten above a few GHz and a few cores. I'm doubtful that with 12, no make that 24 cores wouldn't bottleneck on reading and writing the SSD.

Code files are usually a few kilobytes to maybe a few hundred. Outputted binaries per class are usually a few hundred kilobytes. Even if you throw some indexing and intermediate files in there, code compiling is not very IO bound. It can be IO latency bound, which is very much solved by SSDs. The reason you tend to see latency is that code files are very small, but there are a lot of them, so a traditional disk usually has to thrash to get all the data, but an SSD doesn't have that issue.

I very much doubt at that scale 12 physical cores will bottleneck based on IO. Code compilation these days is very lopsided to CPU with all the intermediate code representations and optimization that goes on in the compiler.

I can easily tell the difference between compiles on a dual core, quad core, and eight core machine (I work with all three.) Dual core is usable, quad core is fast, and my eight core is a monster, especially if I'm compiling large libraries or apps.
 
Code files are usually a few kilobytes to maybe a few hundred. Outputted binaries per class are usually a few hundred kilobytes. Even if you throw some indexing and intermediate files in there, code compiling is not very IO bound. It can be IO latency bound, which is very much solved by SSDs. The reason you tend to see latency is that code files are very small, but there are a lot of them, so a traditional disk usually has to thrash to get all the data, but an SSD doesn't have that issue.

I very much doubt at that scale 12 physical cores will bottleneck based on IO. Code compilation these days is very lopsided to CPU with all the intermediate code representations and optimization that goes on in the compiler.

I can easily tell the difference between compiles on a dual core, quad core, and eight core machine (I work with all three.) Dual core is usable, quad core is fast, and my eight core is a monster, especially if I'm compiling large libraries or apps.

Fair enough, I haven't run these tests with SSD's, maybe it's different now. Also the most cores I have access to is four Xeon.

Of course the problem is that the 12 core is horrifically expensive compared to the hex. Octo might be an option. Can you do a test, of compiling the same library on all three and post the results here? Hopefully everything else is approximately the same - SSD and CPU speeds and such.
 
Given the front page news that the D700 upgrade will be a relative bargain at $600... is anyone figuring on changing their load-out? :)
 
Assuming the MR prices are correct, I'll be going with the

6-core
D700
32GB RAM
512GB Storage

I was on the fence about 6 vs. 8 core, D500 vs D700, and 16 vs 32 GB. For $400 I might as well get my additional memory when I but the nMP. Since I can afford it and I'll be keeping this machine for a few years, I'll go with the faster GPU even though I have limited need for it at present.
 
I was looking at an 8 core, but that's put it out of my reach for now.

A 6 core with dual D700's, and 500GB would be perfect! I'll take the stock 16GB RAM, and I'll upgrade to 32GB myself, depending on the price I can get.

That way I can play this all day long without lag. :p

 
What is that funky lookin game?

Starbound, by Chucklefish!
http://youtu.be/OXCYpADxyGE

Terraria/Minecraft in an infanite randomly generated galaxy with several playable races, and many more. The game is in Beta, and they're significantly expanding to it weekly.

In that screenshot you can see me ( the human spec ops chap ), preparing to murder a Glitch ( sentient robot ) King.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.