Database Server CPU Comparison: Athlon MP vs. Hyper Threading Xeon

AnandTech Website Database Performance Transactions per Second

Intel Xeon 2.2GHz x 2 (HT Enabled) Intel Xeon 2.2GHz x 2 AMD Athlon MP 2000+ (1.6GHz) x 2 AMD Athlon MP 1.2GHz x 2 Intel Xeon 1.7GHz x 2 Intel Xeon 2.2GHz (HT Enabled) AMD Athlon MP 2000+ (1.67GHz) Intel Xeon 2.2GHz AMD Athlon MP 1.2GHz Intel Xeon 1.7GHz 734.8 708.6 699.8 544.0 537.6 469.2 468.6 417.4 393.9 331.0 | 0 | 147 | 294 | 441 | 588 | 735 | 882

There is a lot of information to digest here so let's start at the bottom and work our way to the top; first, let's tackle all the single CPU comparisons. The Athlon MP 1.2GHz vs Intel Xeon 1.7GHz is the same comparison we made last June. When looking at single processors alone, the Athlon MP 1.2GHz can process 19% more transactions per second than the 1.7GHz Xeon. What's also interesting to note is that the single Athlon MP 1.2 falls within 6% of the new 2.2GHz Xeon with 512KB of L2 cache. In fact, the Athlon MP 2000+ is over 12% faster than the 2.2GHz Xeon, but flip on the Hyper Threading option in the BIOS and the picture changes quite a bit.

Enabling Hyper Threading turns one physical CPU into two logical CPUs, each with their own architectural states (including registers, program counters, etc...). So a single 2.2GHz Xeon with Hyper Threading enabled appears to be two CPUs under Windows 2000 Server. Enabling Hyper Threading boosts the performance of the single Xeon CPU by just over 12%, giving it the same transaction processing power as the single Athlon MP 2000+.

Moving up to the multiprocessor configurations illustrates that as powerful as single CPUs can get, a minimum of dual CPUs should be considered for any serious database serving tasks. Even the dual 1.7GHz Xeon comes out faster than the single 2.2GHz Xeon with Hyper Threading enabled by a decent 14% margin. What's also interesting is that the dual Athlon MP 1.2GHz setup is only 1% faster than the dual 1.7GHz Xeon system in this test. You'll recall from above that there was a much larger (19%) difference when looking at single CPUs alone but that advantage has been erased when moving to dual CPUs. There's a bottleneck at work here but it's difficult to say based on these results alone. Remember that the 760MP North Bridge uses a point to point connection to both processors, meaning each one gets a dedicated 2.1GB/s FSB to the North Bridge. However, there is only a single DDR266 channel to the memory banks, meaning that the two CPUs still have to share the same 2.1GB/s of memory bandwidth. Whereas the E7500 platform has 3.2GB/s of FSB and memory bandwidth that the two Xeons share. It could very well be that the Athlon MPs are becoming memory bandwidth limited.

But enough with those "slow" CPUs; although we were blown away by the dual Athlon MP 1.2GHz setup last year, we were stunned by the performance of the dual Athlon MP 2000+. A 28% performance improvement over the previous champ is not something to scoff at, and it also puts it just about on par with the new 2.2GHz Xeon processors which feature much larger L2 caches. Enabling Hyper Threading on the dual CPU 2.2GHz Xeon setup improves performance by another 3.6% but not nearly the boost we saw when enabling it on the single CPU.