Dr Donald Kinghorn (HPC and Scientific Computing)

Xeon E5v3 All Core Turbo and Amdahl's Law

Written on January 8, 2016 by Dr Donald Kinghorn
Share:

This is a re-write of an older post that was done as a guide to the (then new) Xeon E5v3 processors. In this re-make all of the data has been updated to use All-Core-Turbo as the CPU clock speed for the theoretical performance calculations. That changes things significantly. All-core-turbo is a much better performance measure than CPU base-clock frequency!

If you don’t feel like reading then just scroll down to the second chart and click some of the buttons and then realize what you are looking at :- )

I used to conservatively just use Intel's CPU base-clock in performance estimates but then started noticing that with newer Intel processors I was ALWAYS seeing  all-core-turbo CPU frequencies in /proc/cpuinfo when I had a system under heavy load. (... with the exception of laptops. There you see thermal induced frequency throttling under load, as expected.) Check out Matt's interesting article on thermal throttling

An Intel Haswell based workstation with proper power and cooling WILL run at all-core-turbo clock frequency under full load. So why doesn't Intel just report all-core-turbo as the processor clock frequency? I really don't know but I'm guessing it has something to do with Engineering and Marketing butting heads. It's actually even difficult to find all-core-turbo frequency information. Most spec lists just report base and max-turbo clock frequencies, both of which are not very useful. The official Intel document with the good information is here (pdf file),  Intel Xeon Processor E5 v3  Product Family Processor Specification Update August 2015.  That is a really interesting document!

Amdahl’s Law

The most important consideration when configuring a system for optimal parallel performance is the process scaling of your program. You need to have some idea of how many cores can be effectively utilized in order to make an informed decision about your system configuration. If you only want to run one job at a time using all cores on your system then you need to know how many processes your program will scale to before parallel scaling degradation limits your performance gains. If you know that your code only scales well to 8 process you need to decide if you want to just configure an 8 core machine or configure a machine that will let you run several of these 8 core jobs at the same time. ... see "Other Considerations".

To get the idea of Amdahl’s law consider this: If you have a single threaded program and you can find a section of the code that uses 90% of the time and you can make that part of the code run in parallel (perfectly), even though that sounds good, your program will never be more than ten times faster, no matter how many cores you use! [You might want to take a look at Matt's article about estimating performance with Andahl's Law.]

If your code scaling is not great then you are likely better off with fewer cores running at higher clock frequencies. If you code scales really well then you will likely benefit from a higher core count.

The following chart shows the Amdahl’s Law curves up to 36 cores for 7 different parallel fractions, P, ranging from 1 to 0.95, i.e. from perfect linear scaling to 95% of execution time in parallel (maximum speedup = 20).

speedup = 1/( (1-P) + P/n )

where P is the parallel fraction and n is the number of processes (cores)

Notice how the speedup falls off with increasing core count. Just because your program runs almost 4 times faster with 4 cores does not mean it will run 36 times faster with a dual 18-core system.

To evaluate processor performance under the influence of Amdahl’s Law, observe that the speedup is the “effective” core count. If we calculate the the theoretical performance of a systems using this "effective core count" we get a much better picture of potential "real world" performance.

performance = "Effective core count" * All-core-turbo Clock (GHz) * Special ops. i.e. AVX2, FMA3 (16)

for a dual E5-2699v3 system with perfect parallel scaling, P=1,that would be

performance = 36 * 2.8 * 16 = 1612.80 GFLOPS

Now at a parallel fraction of .95 Amdahl’s law gives us;

effective number of cores = 1/( 1-.95)  +  .95/36)  = 13.1

this give a performance at P = .95 of

performance(P=.95) = 13.1 * 2.8 * 16 = 586.47 GFLOPS

In the following chart 27  E5 v3 processors are listed in decreasing cost order (cost of two 26xx CPUs or one 16xx CPU) The bar length corresponds to the theoretical peak performance UNDER THE INFLUENCE OF AMDAHL’S LAW!

 

click button to change Amdahl's Law parallel fraction for performance scaling

* Ordered by price from high to low. 26xx is price for 2 CPU's 16xx is price for 1 CPU

When I first looked at this chart I was shocked! It doesn’t tell the whole story though. There are are other general usage considerations. Also note that some of the processors have a larger “smart cache” per core and that can have a big influence on codes that are slowed down by cache misses etc.. I have a table of the processors with some of their features listed at the bottom of this post.

Other considerations

The three primary ways to utilize a multi-core system are;

  • Run single parallel jobs with all available cores
  • Take advantage of the increased core count to facilitate larger problem sizes
  • Run multiple, “single” or “few” process jobs

These three use cases are governed by the following;

  • Parallel performance characterized by Amdahl’s Law ( we looked at this above)
  • Parallel performance according to Gustafson's Law
  • Efficient job scheduling

 

Job Scheduling

You can treat a high core count workstation as a replacement for a small cluster. Set it up with a job scheduler, create a queue and load up your jobs. You may have some job runs with single threaded code and some jobs that really can’t take advantage of more than a couple of parallel processes. Let the scheduler balance the load. This can be a great way to get good utilization out of your system. In this case your choice of processors may be dictated more by your budget than anything else. If you have the resources you can go with a dual 18 or 14 core processor and get to work. Modern job schedulers are “parallel aware” so you can run mixed job types. Setting up a job scheduler is not always trivial but can certainly be worth the effort. Examples are SLURM, Grid Engine, Torque, PBS etc..

Gustafson’s Law

The next case that can be facilitated by a many-core workstation is running "larger" problems than you could with a less capable system. This is the realm of Gustafson’s Law.

The ideal case for Gustafson’s law on a workstation is when having twice as many cores means you can run a job that is twice the size in the same amount of time. (You will likely need at least twice the memory too!) You have to be careful when considering this type of scaling (weak scaling) on a single node workstation since larger problems can be limited by memory performance. On a cluster distributing a larger parallel job over several nodes will benefit from a more even distribution over cache and memory controllers and this can sometimes make a big difference in parallel performance and can occasionally result in “super linear” scaling because of the better memory utilization. On a single node many-core workstation you do get the extra cache associated with the cores but the number of memory controllers is fixed.

On a many-core workstation you are more likely to be limited by the Amdahl’s Law performance of your code regardless of the problem size. However, if you are looking to increase the size of the problems that you look at, lots of cores and lots of memory are your friends!

Intel Xeon E5 V3 CPU Specs

Processor Base
Clock
All-Core
Turbo
Cores Smart
Cache
Cache
Per Core
TDP Price*
E5-2699 v3 2.3GHz 2.8GHz 18 45MB 2.5MB 145W $4,109
E5-2698 v3 2.3GHz 2.8GHz 16 40MB 2.5MB 135W $3,220
E5-2697 v3 2.6GHz 3.1GHz 14 35MB 2.5MB 145W $2,702
E5-2695 v3 2.3GHz 2.8GHz 14 35MB 2.5MB 120W $2,424
E5-2687W v3 3.1GHz 3.2GHz 10 25MB 2.5MB 160W $2,141
E5-2690 v3 2.6GHz 3.1GHz 12 30MB 2.5MB 135W $2,090
E5-2667 v3 3.2GHz 3.4GHz 8 20MB 2.5MB 135W $2,057
E5-2683 v3 2.0GHz 2.5GHz 14 35MB 2.5MB 120W $1,846
E5-2658 v3 2.2GHz 2.5GHz 12 30MB 2.5MB 105W $1,832
E5-2680 v3 2.5GHz 2.9GHz 12 30MB 2.5MB 120W $1,745
E5-2670 v3 2.3GHz 2.6GHz 12 30MB 2.5MB 120W $1,589
E5-2643 v3 3.4GHz 3.6GHz 6 20MB 3.33MB 135W $1,552
E5-2660 v3 2.6GHz 2.9GHz 10 25MB 2.5MB 105W $1,445
E5-2650v3 2.3GHz 2.6GHz 10 25MB 2.5MB 105W $1,166
E5-2637 v3 3.5GHz 3.6GHz 4 15MB 3.75MB 135W $996
E5-2640 v3 2.6GHz 2.8GHz 8 20MB 2.5MB 90W $939
E5-1680 v3 3.2GHz 3.5GHz 8 20MB 2.5MB 140W $1,723
E5-2630 v3 2.4GHz 2.6GHz 8 20MB 2.5MB 85W $667
E5-1660 v3 3.0GHz 3.3GHz 8 20MB 2.5MB 140W $1,080
E5-2623 v3 3.0GHz 3.3GHz 4 10MB 2.5MB 105W $444
E5-2620 v3 2.4GHz 2.6GHz 6 15MB 2.5MB 85W $417
E5-2609 v3 1.9GHz 1.9GHz 6 15MB 2.5MB 85W $306
E5-1650 v3 3.5GHz 3.6GHz 6 15MB 2.5MB 140W $583
E5-2603 v3 1.6GHz 1.6GHz 6 15MB 2.5MB 85W $213
E5-1630 v3 3.7GHz 3.8GHz 4 10MB 2.5MB 140W $372
E5-1620 v3 3.5GHz 3.6GHz 4 10MB 2.5MB 140W $294

* Price from Intel ARK

Happy computing! --dbk

Tags: Xeon E5v3, Parallel Perfromance, Amdhal Law, All Core Turbo