Super Quick ? - 6 or 8 cores?

[wildmac]

On what do you base that? The only thing is has a "big" advantage is double precision openCL...

The amount of VRAM, and the buss speeds on the cards. For the price difference, it's a very reasonable upgrade. Going up to the D700 is more of a question, depending on if you'd really use all that VRAM.

 

[Dranix]

Hmm having no performance-problems with my 5870 with only 1GB of vram I have problems understanding the need for so much vram.

 

[cube]

Super Quick = Supermicro

 

[theSeb]

So in your estimation does that justify spending more to upgrade from the D500s to the D700s? Or, is this just un-needed for this type of work?

I would say that it would be a wise upgrade. You said you want to keep the computer for 5-6 years. The way things are going more and more stuff is going to be off-loaded to the GPU. PS already uses OpenCL for some of the filters and it can utilise a faster GPU for other stuff anyway without the OpenCL magic.

Mercury Graphics Engine

The Mercury Graphics Engine (MGE) represents features that use video card processor, or GPU, acceleration. In Photoshop CS6, this new engine delivers near-instant results when editing with key tools such as Liquify, Warp, Lighting Effects, and the Oil Paint filter. The new MGE delivers unprecedented responsiveness for a fluid feel as you work.


MGE is new to Photoshop CS6 and uses both the OpenGL and OpenCL frameworks. It does not use the proprietary CUDA framework from nVidia.


MGE requires a supported video card and updated driver. If you do not have a supported card, performance is degraded. In most cases, the acceleration is lost and the feature runs in the normal CPU mode. However, there are some features that don't work without a supported video card.

 

[Michael73]

The clock speeds are confusing and can be completely ignored.

All these processors run at different speeds depending on how many cores you use. Use 1, 2, 4, 6, 8, 12 cores, and the clock speed goes down.

Now Intel probably for legal reasons specifies the clock speed if all cores are running. So they publish the clock speed for four core processor running four cores, six core processor with six cores running, eight core processor with eight cores running and so on.

The reality is, that they all run at exactly the same speed if you use four cores. And they all (except the 4 core machine which can't) run at exactly the same clock speed if you use six cores, and so on.

The eight core machine might display a smaller number, but it runs six cores as fast as the six core machine.

All that said, the price difference from six to eight is huge, and spending the money on more RAM and other things will most likely get you more value for money, unless your machine is cracking on 24 hours a day (in which case you should go for 12 cores).

I *think* I knew that but I'd forgotten so that's good to know. Here's the question…if I get the 8 core and then split 2 cores for a Win7 VM and 6 cores for OS X, what speed is the processor running at? Or 4 and 4? Or if I get the six core model split it 2 and 4?

Maybe the better question is between the 6 and 8 core model, what option runs a VM at the highest combination clock speed?

 

[ssls6]

The clock speeds are confusing and can be completely ignored.

All these processors run at different speeds depending on how many cores you use. Use 1, 2, 4, 6, 8, 12 cores, and the clock speed goes down.

Now Intel probably for legal reasons specifies the clock speed if all cores are running. So they publish the clock speed for four core processor running four cores, six core processor with six cores running, eight core processor with eight cores running and so on.

The reality is, that they all run at exactly the same speed if you use four cores. And they all (except the 4 core machine which can't) run at exactly the same clock speed if you use six cores, and so on.

The eight core machine might display a smaller number, but it runs six cores as fast as the six core machine.

All that said, the price difference from six to eight is huge, and spending the money on more RAM and other things will most likely get you more value for money, unless your machine is cracking on 24 hours a day (in which case you should go for 12 cores).

Do you have a reference for this? I know it is true of turbo boost up and until the point timing/thermal/current limits are reached.

 

[theSeb]

I *think* I knew that but I'd forgotten so that's good to know. Here's the question…if I get the 8 core and then split 2 cores for a Win7 VM and 6 cores for OS X, what speed is the processor running at? Or 4 and 4? Or if I get the six core model split it 2 and 4?

Maybe the better question is between the 6 and 8 core model, what option runs a VM at the highest combination clock speed?

The speeds will remain the same. It depends on whether those cores are being used. Just because you have assigned 2 cores to the VM, it does not mean that it is using those 2 cores all the time. The same goes for the remaining cores in OS X. In the lowest usage you would basically be using 2 cores (one for the VM and one for OS X - unless the virtual machine software supports app nap in amazing ways, which I doubt).

So the 6 core will be running at maximum turbo boost for 2 cores active, which is 3.7 GHz

Likewise, 8 core will be running at 3.8 GHz (max turbo clock for 2 cores active)

This assumes that the new Mac Pro dissipates heat well enough to allow turbo boost to be running at all times. This is something that still needs to be investigated and confirmed.

Read this if you haven't yet:
http://www.marco.org/2013/11/26/new-mac-pro-cpus

----------

Do you have a reference for this? I know it is true of turbo boost up and until the point timing/thermal/current limits are reached.

Intel's ark pages are your reference. But the easier reference is the link I have just posted. The information posted that you are questioning is not correct

----------

The reality is, that they all run at exactly the same speed if you use four cores. And they all (except the 4 core machine which can't) run at exactly the same clock speed if you use six cores, and so on.

.

Using 4 cores these will turbo boost up to

4 core - 3.7 GHz
6 core - 3.7 GHz
8 core - 3.5 GHz
12 core - 3.2 GHz

Using 6 cores these will turbo boost up to

6 core - 3.6 GHz
8 core - 3.4 GHz
12 core - 3 GHz

 

[iBug2]

So the 6 core will be running at maximum turbo boost for 2 cores active, which is 3.7 GHz

Likewise, 8 core will be running at 3.8 GHz (max turbo clock for 2 cores active)

That's the thing though. The 8 core is supposed to run at 3.8 with 2 cores active, but will it? Intel only guarantees that it won't run slower than 3.0 ghz. The thermal limits may be reached sooner than expected.

So if you need single or dual core speeds, you gotta get the 4 core or the 6 core and not the 8 core. The 8 core might run slightly faster, or it might run a lot slower on 2 cores, depending on the situation.

 

[theSeb]

That's the thing though. The 8 core is supposed to run at 3.8 with 2 cores active, but will it? Intel only guarantees that it won't run slower than 3.0 ghz. The thermal limits may be reached sooner than expected.

So if you need single or dual core speeds, you gotta get the 4 core or the 6 core and not the 8 core. The 8 core might run slightly faster, or it might run a lot slower on 2 cores, depending on the situation.

Indeed. We won't know until we see very thorough benchmarking and testing. Even the reviews from the usual suspects never cover this area in enough detail.

 

[Michael73]

Here's the thing or I guess where the rubber meets the road...what is the real world performance hit of going from 3.7 GHz to 3.5GHz or 3.6 GHz to 3.4GHz.

It seems academic or at least so minor that the difference would hardly be noticed.

Can someone frame this in terms of a particular render or encoding job e.g. at 3.7 GHz it might take "x" seconds/minutes versus at 3.5 GHz it will take "y" seconds or minutes?

 

[iBug2]

Indeed. We won't know until we see very thorough benchmarking and testing. Even the reviews from the usual suspects never cover this area in enough detail.

It's because it's not measurable. There's no utility on Mac that shows you if Turbo boost is running at the time or how many cores are running at what frequency.

 

[theSeb]

It's because it's not measurable. There's no utility on Mac that shows you if Turbo boost is running at the time or how many cores are running at what frequency.

Of course it's measurable. I have done it on a 2011 MBA and I was going to write a long (and hopefully interesting) article about turbo boost and throttling under load in Macs. Unfortunately I could not get it work on a 2012 rMBP. It would simply kernel panic on start up and then I got annoyed and gave up on the article.

The kext is IntelCPUMonitor and if you don't get kernel panics, you can use fakeSMC to monitor current clock speed in OS X with the K-CPUi plugin. Maybe people will update IntelCPUMonitor for the latest CPUs. I have not been following this.

 

[theSeb]

Apologies for the double post, but I have found the missing piece of the puzzle for OS X....

Bring on the Mac Pro for thorough testing.

No details on the cores, but we can get that from Activity Monitor

 

[Dranix]

How do you get that display?

 

[flat five]

fwiw, 6 core is the best deal in the lineup whereas 8 core is the worst deal.
(talking about cost per core)

and if talking about cost per performance gains, the 6 is by far the best deal.. even moreso than the cost per core comparison.

(edit).. well 4 core is probably the best deal but it's a bundled up item -or- you must buy it and/or use the quad as a launching point for other comparisons.

 

[theSeb]

How do you get that display?

I was searching for details how to get the info out programmatically and found this app on Intel's site

http://software.intel.com/sites/default/files/article/184535/intel-power-gadget-2.zip

 

[Michael73]

Finally pulled the trigger this morning: 8-core, 32GB RAM, 1TB SSD, D500

Probably not the sweet spot but I didn't want to say, "woulda," "coulda," or "shoulda."

I'm sure this machine will keep me humming along for years but in the meantime, the February ship date seems like forever

 

[theSeb]

Finally pulled the trigger this morning: 8-core, 32GB RAM, 1TB SSD, D500

Probably not the sweet spot but I didn't want to say, "woulda," "coulda," or "shoulda."

I'm sure this machine will keep me humming along for years but in the meantime, the February ship date seems like forever

Nice choice. Happy computing.

 

[propower]

I was searching for details how to get the info out programmatically and found this app on Intel's site

http://software.intel.com/sites/default/files/article/184535/intel-power-gadget-2.zip

I get a dmg but it doesn't do anything? Is there a trick?

 

[theSeb]

I get a dmg but it doesn't do anything? Is there a trick?

Once the dmg is mounted, open it and install.

 

[FredT2]

I get a dmg but it doesn't do anything? Is there a trick?

No. Look for Intel Power Gadget image after you open it. It has a standard install package inside.

 

[flat five]

Finally pulled the trigger this morning: 8-core, 32GB RAM, 1TB SSD, D500

Probably not the sweet spot but I didn't want to say, "woulda," "coulda," or "shoulda."

i don't know.. sounds pretty sweet to me

 

[Michael73]

The new system is so far beyond my current rig…I mean I'm still working with 7200RPM SATA drives, 12GB RAM at 800MHz and a 7970.

The SSD on the nMP is stupid fast compared to what I have now as is the RAM and that 1GB of GDDR5 on the 5770 pales in comparison to the 6GB that's coming.

 

[vladi]

As a semi-professional photographer, I work with large RAW files. I shoot with a Nikon D800, the RAW files are in the 60-75MB range, and when editing in PS the TIFFs easily get into the 150-200MB sizes. I merge several of these into HDR pictures, and it puts some strain on my current MP3,1 Octo/2.8, 16GB RAM, 1TB Accelsior PCIe SSD, GTX680 setup. My WorkFlow is via Aperture, with a few 100GB libraries.

Hence I want to upgrade to the nMP. Probably the 6 Core, 32GB, 1TB SSD configuration.

I keep all my files on my ReadyNAS anyway, and I am thinking about getting an external TB2 housing for my Accelsior card.

Only question I have is if I should upgrade the D500s to the D700s? Would that give me anything?

In addition to my photography work, I use my MP for normal CS6 stuff, and I must confess, I game a bit in Windows 7 (Bootcamp).

I'm afraid number of cores will not have a lot of effect on your workflow as much as the actual speed of those cores and amount of RAM. Your "problem" with your last machines is "only" 16GB of RAM.

I work with 5GB .psb files that are mainly raster composites (with adjustment layers and smart objects) for extra large fine print in 1:1 ratio in 150dpi. Every time I hit a bottleneck it was due to RAM and Scratchdisk, it has really nothing to do with CPU. Our conclusion was to boost up the RAM to 32 and get freakin SSDs. Eventually we went and got 64GB of RAM by getting x2 Crucial 32GB DDR3 sticks because we plan to max it out soon. And Quadro K4000 helps to extent as well.

Photoshop work is quite delicate when it comes to utilizing the cores, most of the time PS will run on just one core and sometimes it will utilize 1/3 of all cores and sometimes GPU will take over. So its like all over the place but most of the PS tasks are not build to utilize more than one core so moral of the story is to get the fastest CPU you can get in term of clock speed. If you go with either quad or hexa you will not go wrong. If you are tight on the money take the quad and use extra $500 for RAM or SSD upgrades.

 

[sboerup]

The new 8core is faster than the 6core in both single and multi core uses. I guess the L3 cache helps, not sure though.