Wednesday, December 29, 2010

Testing Throttled Intel Westmere X5650 CPUs in an HP blade

This is a continuation of our search to find the cause of slowness with our Dell M610 blades using dual Intel X5650 Westmere processors. (It has since been resolved).  Please see the other articles I have written, especially Flaws with Intel Westmere X5650?  The other relevant articles are Diagnosing Throttled or "Slow" Systems (Processors to be Precise) and Diagnosing Throttled Processors - Part 2.
Fortunately we were able to borrow an HP blade to test with (ProLiant BL460c G6).  We swapped in our Westmere CPUs and our RAM.  We had the blade for a very limited time, so this testing was about as unscientific as you can get.  We did find some interesting results but I definitely do not consider them conclusive.

We did see strong correlation between benchmark results of "good" and "bad" CPUs.  They were different between HP and Dell, but the benchmark results were predictable.  I tried making sure that relevant BIOS settings (e.g. hyperthreading disabled, turbo mode toggled back and forth, etc) were all the same.
  1. Good CPUs we tried were ~120 gflops in Dell and ~115.5 in HP
  2. Bad CPUs we tried were ~105 in Dell and 113.7 in HP
  3. Toggling turbo mode had almost no measurable effect in the HP blade
  4. No throttling was detected at all in the HP blade under any test scenario
So what does this mean?  I'm not sure.  I wish we could have done more thorough testing but we needed to return the borrowed HP blade.

Noting that is test was extremely limited and should not be considered conclusive, points to consider are:
  1. Dell blades get better performance than equivalent HP blades when using "good" CPUs
  2. HP blades get better performance than equivalent Dell blades when using "bad" CPUs
  3. Dell blades get extremely poor performance when using "bad" CPUs
  4. The HP blade has mediocre (but different) performance with "good" and "bad" CPUs
  5. Turbo mode didn't seem to have much of an effect on the HP blade
  6. The mediocre performance of HP blades using "good" CPUs is almost exactly the same as a Dell blade with turbo mode disabled.
  7. HP blades with "bad" CPUs show slightly reduced performance but are much faster than when those same CPUs are in a Dell blade (113.7 gflops in HP compared to 105 in Dell)
  8. No throttling was detected at all in the HP blade
  9. "Good" CPUs we tested in the HP blade were 115.5 gflops and "bad" ones were 113.7.  It's not the same spread as in Dell systems, but it was measurable and repeatable with my limited testing
Though I find the above results to be interesting, I want to reiterate again that it was not a scientific test at all and should not be used to draw any real conclusions.  The only semi-conclusive statement I can make is that the CPUs do seem to behave differently in HP and Dell blades.  The benchmark results are vastly different but the "good" CPUs are always measurably faster than the "bad" ones when tested under similar conditions (in a Dell blade vs. an HP blade).  So the problem does follow the CPUs but manifests itself in different ways.

Please also see the other articles I wrote on this subject that are linked in the first paragraph.

No comments:

Post a Comment

Please leave any comments, questions, or suggestions below. If you find a better approach than what I have documented in my posts, please list that as well. I also enjoy hearing when my posts are beneficial to others.