I have a system (retired server that was not being used) that can be configured to have hyperthreading (Hyper-Threading Technology or HT Technology, and abbreviated as HTT or HT) enabled or disabled.

With hyperhtreading disabled:

Intel(R) Xeon(R) CPU X5660 @ 2.80GHz [Family 6 Model 44 Stepping 2]
Number of processors 12
Memory 131059.09 MB
Cache 256 KB
Measured floating point speed 3554.51 million ops/sec
Measured integer speed 8297.16 million ops/sec

With hyperthreading enabled:

Intel(R) Xeon(R) CPU X5660 @ 2.80GHz [Family 6 Model 44 Stepping 2]
Number of processors 24
Memory 131059.09 MB
Cache 256 KB
Measured floating point speed 2913.02 million ops/sec
Measured integer speed 7240.22 million ops/sec

I have read somewhere that turning off hyperthreading actually helps in some applications . . . the idea being that the memory can't feed the 24 cores or some such. And with the hyperthreading enabled, you can see that the individual logical processors have reduced performance, but not by half.

So my question is: Has anyone done a definitive test on whether SETI performs better on Intel Xeon processors with hyperthreading enabled or disabled?

---

Highly dependent on the exact configuration of your system. With plenty of memory (2Gb/thread) gives a better throughput that with HT off on my only Intel system. The key thing to know is many of the processor speed tests only measure the performance of one thread, so, while the figure may look lower (and indeed are lower) than HT off you have twice as many threads working so the overall performance is better.

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

128GB / 24 cores = 5GB per logical processor, plenty I think :-)

---