Looking for some ideas for a new high end workstation build. Used mainly for the Adobe Suite (Photoshop & After Effects CS6, CS6 when out), and 3D (Maya, Zbrush, Realflow) I work mainly in print illustrations so static large prints with lots of detail is what I usually need to create. I also need this machine to render fast and efficiently in 3D.
I will also game a bit on this, but any high end build should be able to withstand gaming anyway. Only game I play now is guild wars 2.
Quality is more important to me than cheaper parts, so feel free to splurge.
Approximate Purchase Date: ASAP
Budget Range: $2000-$4000 (flexible)
System Usage from Most to Least Important: Adobe Suite working with hi-res large files, 3D (everything from modeling, rendering, liquid simulation), Gaming, Multimedia.
Parts Not Required: Just tower ideas, no need for peripherals, etc.
Preferred Website(s) for Parts: Newegg
Parts Preferences: Maxed out quality ram (64gb) for multi-tasking and large files, full-sized case?
Overclocking: Sure, though E5 xeons can't overclock.
SLI or Crossfire: No (3D apps don't use SLI)
Monitor Resolution: Need gfx card capable of doing dual monitors @ 2560x1600 with power to spare. (Flexible but more is better for option to upgrade monitors in the future)
Additional comments: I am open to both single core and dual-xeon ideas, as long as it has all my requirements. I am thinking of a dual xeon rig or a i7-3930k rig. Again, I need this to be about quality and performance..the budget is flexible. I want 1 pc! Please don't tell me to make 1 for gaming and 1 for working...
You seem to already know what you need from your other threads. Nothing new has been released. Only took me a couple minutes just to check the current cheapest prices on newegg. I didn't include any hdd as I didn't know how much extra storage you wanted. For a high end workstation, you should go with a workstation card. Although they will game, it will be less performance than a gaming card at half the price. You could go with a gaming card which will still handle these software well. Just an fyi, 64gb ram is not the max, this mobo supports 512gb ram, but newegg only has 16gb sticks so 256gb max. You meant "when cs7 comes out" right? You typed cs6. Also single core is not the same as single cpu.
Seagate Constellation ES.2 2TB 3.5" SAS 6Gb/s Internal Hard Drive -Bare Drive
Return Policy: Standard Return Policy
LG Black 14X BD-R 2X BD-RE 16X DVD+R 5X DVD-RAM 12X BD-ROM 4MB Cache SATA BDXL Blu-Ray Burner with SW, 3D Play Back - BH14NS40
Return Policy: Standard Return Policy
Grand Total: $3,922.90
> To this I would add the OS, and a competent but not outrageously priced soundcard like the M-audio 2496 "Audiophile" for about $85, making this a bit over $4,000.
The CPU choice of a single i7-3960X is the result of learning that the applications I use the most > Autocad, Revit, Inventor, Sketchup are single-threaded and do not take advantage of multiple cores- they can use only a small percentage of each core. Inventor can only use a maximum of 15% of each of a 6-core CPU. As the threads have to divided among cores and then reassembled, in some operations, it appears, more cores may harm performance. I was surprised that on the Passmark Benchmark, the Precision T5400 performs as well as some i7 6-core and multi-CPU machines. It appears that Maya is gradually introducing some multi-core use, but at the moment multi-core advantages are limited. In this way, the clock speed becomes more important than core count, and as compared to Xeons, the i7 may be overclocked with confidence (and good air-cooling) to say 4.2GHz. Cache seems to be important too, and the 3960X has 15MB. By the way, I'm not very knowledgeable about CPU cooling, so that choice is based on ratings and overall beefiness,..
Another discovery has been that for Autodesk applications and 3D generally, Quadro video cards are the only way to go. I bought the gaming Geforce GTX 285 (1GB) mistakenly thinking it could be soft-modded into a Quadro 5800 (only Nvidia GPU's up to G92 can though), but gaming cards are frame-rate oriented and a Quadro FX580 (512MB) with less than half the CUDA cores actually runs Sketchup at least as fast! The Quadro 4000 appears to be great card, but if you're adventurous, for a bit more you could buy 2- used Quadro 5800's ($3,200 new, now about $450 each) with 4GB (DDR3) memory each and run one monitor on each at 2560 x 1600. I would think for Maya animations that would work nicely!
Autocad, Revit, Inventor, Sketchup are single-threaded and do not take advantage of multiple cores- they can use only a small percentage of each core. "..." the threads have to divided among cores and then reassembled in some operations, it appears, more cores may harm performance.
You contradict yourself saying it doesn't take advantage of multiple cores yet work is divided among cores. While current versions of autocad, revit, and inventor do have multithreaded features, I don't like to compare software that he is not using. Apples and oranges. But to clarify, threads do not get divided into each core, although 1 core can handle 2 threads with ht, 1 thread goes through 1 thread of 1 core. And a single threaded app can use 100% of said thread. More cores does not harm performance, they are just idle. A loss in performance can be caused by a number of other factors, but dormant cores is not one.
All of his software, ps, ae, most other adobe products, maya, zbrush, and realflow, are all multithreaded for the most part. Viewports tend to be single threaded but this is largely gpu dependent. The parts that are cpu dependent are multi threaded namely rendering. More ghz does not mean more performance, some ghz do more work than others. I'd recommend not looking at the individual specs, and focus on end result performance. Who cares if one has more cache or more cores, all that matters is performance. The same goes for gpus.
Workstation cards (quadro and firepro) are recommended for better viewport performance. Any gpu is "frame rate oriented:" workstation cards get more fps in content creation software, gaming cards get more fps in games. More vram does not mean more performance. The higher end cards tend to have more vram but a lower vram card can have better performance. And now with kepler, more cudas also does not mean more performance. Throw that specsheet out the window and bring in the real world benchmarks.
You wrote > " You contradict yourself saying it doesn't take advantage of multiple cores yet work is divided among cores. While current versions of autocad, revit, and inventor do have multithreaded features, I don't like to compare software that he is not using. Apples and oranges. But to clarify, threads do not get divided into each core, although 1 core can handle 2 threads with ht, 1 thread goes through 1 thread of 1 core. And a single threaded app can use 100% of said thread. More cores does not harm performance, they are just idle. A loss in performance can be caused by a number of other factors, but dormant cores is not one."
In my experience it is best to identify the most demanding application /use of a computer and for the workstation for realmisr, this is Maya 3D animation. If you know Autodesk applications, they are purposely structured internally, and externally aspects of the interface are very similar for inter-compatibility and platform performance consistency, and Inventor, due to it's mechanical animation, thermal and stress capabilities, is the application very, very similar to Maya, are identical.
From the Autodesk site >
> "In order to fully benefit from multi-core processors, you need to use multi-threaded software. Unfortunately, Inventor is currently a single-threaded application.
On a dual-core computer, a CPU-intensive operation that uses 100% of the resources of a single-core processor will only use a maximum of 50% of the CPU for that same operation on a dual-core computer, and only 6% of each CPU on a 16-core computer. ... Due to the lack of multi-threading, Inventor is not capable of using more than 50% of the CPU on a dual-core computer, so there is no significant performance gain over a single CPU computer." < END I have been told by systems administrators at engineering and architectural offices that they have seen performance loss in Autodesk applications with and/or lower clock speed 8 and 16 core machines , explained as the way they currently divide and then reintegrate the specific multi-thread streams CPU tasks. This is explained as the result of Autodesk tacking on multi-core capabilities to originally single threaded applications. No doubt they will -probably soon- restructure/integrate everything for multi-core advantages.
You are incorrect that > "All of his software, ps, ae, most other adobe products, maya, zbrush, and realflow, are all multithreaded for the most part. " Maya did add in 2009 some additional features that are multi-threaded, but these are specific and subsidiary operations to an essentially single-threaded application. In this respect, the clock speed, responsible for the fundamental vector/polygon recalculations in animations will make more of a difference than additional cores.
Concerning Adobe CS and 2D graphics application> "Photoshop has been able to take advantage of multi-core processors since the mid-1990s, when it was first modified to take advantage of the rare high end computers with multiple CPU chips in them (each with only a single core at that time). But it gets harder and harder to take advantage of each additional CPU core. Photoshop takes great advantage of 2-4 cores. 8 cores doesn’t give you anywhere near twice the performance of 4 on most operations in Photoshop (or in most other programs either). Partly for that reason, the increase in the number of cores per processor is slowing. We don’t expect to see mainstream processors with 16 cores on them, because almost nothing would use them — they would just sit idle most of the time.
For this reason, concerning the 2D Adobe applications, ( I use CS4) I believe that the 6-core machine described would work very well for realmisr's use and accommodate the coming better use of multiple cores.
Each new feature in Photoshop has to be written to take advantage of multiple cores, and some features that were written when even 2 CPUs were a luxury must be modified or significantly rewritten to get any benefit from a 4- or 8- core machine." <END
You mention that "More vram does not mean more performance," however, realmisr's most demanding application, Maya, must be considered, in which animations can generate huge, huge files, because each animated figure can have 100,000's of polygons. This is why the Quadro 6000, oriented towards real-time video editing, has 4GB and the new top end Quadros have 6GB and (I think) a 512 bit bandwidth. In Maya, more CUDA cores, vram,and CPU speed definitely improves Maya 2011 (which I am learning) performance as I found when upgrading a Dell Optiplex 740 > AMD X2 dc from 2.3GHZ to 3.0GHZ (6000+) and changing from a Quadro FX570 (256MB, 16 CUDA cores) to FX580 (512MB, 32 CUDA cores). Plus, realmisr specifies that he eventually wants to run 2 large monitors at more than 2560 X 1660 resolution. While I don't have figures, it was obvious to me that both higher CPU and HD cache also help significantly with 3D applications as I found with large Inventor and Sketchup files- there is a surprising amount of HD read/write activity during regens.
I'm going with a similar view, ignore the lower end software like ps, which is why I just keep mentioning maya but you should also look at realflow. It is true you will not notice a difference from a capable quad or even dual core to 12 cores in ps. The fact that maya and inventor exist separately should be reason enough not to compare them. Sure most 3d software is similiar but maya is what he uses so maya performance is what should be considered. As you said, yes it still has many single threaded functions. And as I said viewports will tend to be singlethreaded but this is gpu dependent. Even a core 2 duo will still handle modeling easily with a capable gpu well over 5m polys. But then if you ever do dynamics (maya or realflow) or a final render, these are multithreaded and will take advantage of all the cpu power you can get. What's the point of more singlethreaded performance if it doesn't make a difference? You get more multithreaded performance where cpu power is needed; where it counts.
Autodesk Inventor 2009 and later versions support multi-core technology in some very specific areas of the product, including:
Animations and renderings in Studio
ASM (Autodesk Shape Manager) kernel
Since 2012, Inventor uses multiple CPU cores to calculate precise drawing views.
The part that you quoted of the page is talking specifically about inventor.exe which is as stated single threaded. But as a whole inventor in all of its packed features, it would be incorrect to call it singlethreaded. For any software, not all functions are multithreaded, but to say it is single threaded implies nothing is multithreaded. But that's enough of inventor, he's not using it, stop mentioning it.
Again ghz is not a measure of more performance. Different architectures have different ipc. Similar with cuda cores, a card with more cudas can be weaker than a card with lower cudas. The same way higher ghz (same core count) can be lower performance. Or even a cpu with more cache can have lower performance. You need to look at the big picture, end result performance. I will repeat more vram does not mean more performance. If the gpu is too weak to even handle 100k polys, adding more vram does not make it more powerful; it still can't handle 100k polys. And repeating, the higher end cards will have more vram. Of course more vram can help a capable gpu that's why the 6000 has more than the 4000. This isn't really an issue though as none of the suggested cards are weak. It was just worth pointing out.
Also no point in mentioning hdds, all builds contain an ssd.