Gentlemen?,
For those of us handcuffed to workstations for eternity, a new top end Xeon is always interesting- we can begin the delightful process of imagining how fast animations of 8,000 part CATIA assemblies will run and the painful process of imagining the eventual price and how to pay for it.
But, and I say it reluctantly, the thorough tests and considerations by Chris Angelini fall a bit short as it only included comparison to one processor that resembles a competitor to the E5-2697 V2, which is the 8-core and $1,950 E5-2687W.
I can't think that anyone using a i7-3770K will wake up one day and realize that, instead of a $325 quad core, what they really need is a $2,800-$3,100 (guessing) 12-core- and 256GB of ECC Ram. The E5-2697 V2 will be considered by those with single CPU LGA 2011 workstations using the lowly E5-1650, E5-1660, E5-2XXX six core and the E5-2643 and E5-2687W 8-core, as a way to have 12 cores / 24 threads for without starting over and more probably, by those making the calculations for and against having to buy a dual CPU system with two $1,500-$2,000 CPUs. And, because multi-core applications are still so rare- rendering is the most common, plus scientific like MATLAB and certain custom code scientific modeling, and mathematical / financial analytics, the market will be specialized and limited.
What would be very useful would be to compare the E5-2697 V2 to various dual four, six, and eight-core systems. There are a lot of advantages to dual CPU's as they support more memory and have more PCIe lanes, which can add CUDA or Xeon Phi co-processors, RAID controllers, and etc, but importantly, the fewer the cores on a CPU, the higher the clock speed. The quad core $360 E5-1620 has a base clock speed of 3.6GHz, a similar amount of cache per core, and runs on the first core at 3.8. One could argue that on the majority of applications including most of Autodesk, Dessault, and Adobe, the 1620 at 1/8 the price could perform in the realm of the E5-2697 having a base speed of 2.7 and on one core up to 3.5Ghz. The next step up is the six-core E5-1650 at 3.2 / 3.8 at $600 and it takes the $1,100 E5-1660 to have both six cores and the 3.6 speed of the 1620 at 1/3 the cost. Then, one can move into the realm of dual six-core Xeons which can have higher clock speeds than the 2697 and may be temptingly near in cost to the single one. This is only a sample of the difficulty in judging the usefulness and cost / performance of these very high-end CPU's. It has to be reverse analyzed from the applications and even then the cost may not be justified for only occasional bursts of maximum performance.
The speculation as to the E5-2697 use in the new "dustbin" Mac Pro is interesting. The Dustbin Pro is entertaining and has some excellent thinking behind it in being airflow-centric, but it also has aspects of style over substance and exudes a sense, whether fair or not, of being static- that it can never be changed, or you have to find curved Quadros not more than 14cm long and hang additional drives off USB cables. Of course, it may be very flexible, as Apple must know that in the increasingly specialized and optimized world of high-end workstations, hot-rodding is absolutely essential. Still, it's bit sad to see Apple plodding in the decorative rather than the striking game-changing innovations of the Jobs era.
So, the E5-2697 V2 is very welcome, but to run the complex equation that would determine if it's worthwhile- and the price- there needs to be a comparison with a variety of single and multiple quad, hex and octo-core Xeon systems running multi-core benchmark tests or better, the applications of the kind for which this category of CPU's was intended. And, all of that is academic and still limited until the cost is known.
Oh, and one more thing, how well does a pair of E5-2697 V2's do? I need to make a dynamic 3D model of the entire atmosphere,...
Cheers,
BambiBoom
)
Facebook
Twitter
Instagram