I am considering an upgrade to an AMD skt 939 3200+ system mildly overclocked (+10%).

Here's my question:
I am looking at an Opteron 165 and I know it's reputation as a great overclocker, but is it worth another $37 to bump up to the 170?

I have a MSI K8N (Nforce 4 Ultra chipset) mobo and Kingston Hyper X memory.

I appreciate your thoughts and comments.

The math is very simple, the better is the one with greater performance/price ratio.
Because both CPUs have the same architecture and the same features and caches, the performance scalles linear to the freqfency. So, you can calculate the factors by dividing the freqfency with the price.
Because we don't know how much each will cost you, we can't make an exact conclusion. But, acording to the www.newegg.com prices
http://www.newegg.com/Product/Prod [...] 819103586R
http://www.newegg.com/Product/Prod [...] 6819103588
Opteron 165 costs 176$, while the Opteron 175 costs 201$. The Opteron 165 has better value acording to the peformance/price factor

If you plan to overclock, take into account that the Opteron 170 has more multiplier options to choose(11 compared to 9 of Opteron 165), but it will OC less in % than Opteron 165 on a decend mainboard.

I think that you are going to make a wrong choice going after Opteron. For 220$ you can have an Athlon64 FX-55 2.6GHz and a 250GB hard disk.
http://www.newegg.com/Product/Prod [...] 6819103527
Only the CPU has better performance/price value than both models of Opteron(165/170), it has fully unlocked multiplier(31 multiplier choices) and you'll get an extra hard disk.

If I were you, for $198( less than the sum of Opteron 170) I would get an Athlon64 X2 4200+ 2.2GHz. It is faster for singlethreaded apps, and much faster(twice as fast maximum) for multithreaded apps.
http://www.newegg.com/Product/Prod [...] 6819103547

I don't quite understand how the x2 4200+ is twice as fast as the Opteron 170 running multi thread apps. Plz explain.

Infact it's not like he said!
Opterons are highly optimezed for multithreaded apps tipical in server environments and the 170 is a dual core like the 4200 X2.
It will be a little faster than the 4200 even if it has a lower clock speed.

But really it will be a little slower in single threaded apps.

This should help some:
http://www.anandtech.com/cpuchipse [...] =2397&p=14

This is a Opteron vs Athlon64 Test:
http://www.pcstats.com/articleview [...] 466&page=6

I think he thought that Opteron 165 / 170 were single-core CPUs.

But if you can buy a x2 4200+ at $198, you should buy it.

Opteron 165: 1.8GHz / 1MB L2 x2
Opteron 170: 2.0GHz / 1MB L2 x2
Opteron 175: 2.2GHz / 1MB L2 x2
Athlon64 x2 4200+: 2.2GHz / 512KB L2 x2

It depends on what you want to achieve in performance as to whether the Opty 165 or 170 or even the 175 are concerned...
All three with matching decent quality mobo/memory will achieve similar overclocks up to around 2.7GHz to 2.8GHz... plain and simple...

But buy getting a 165, which has a 9x multiplier, you have to run your memory at 300MHz to reach 2.7GHz...
Whereas buying an Opteron 170 with a 10x multiplier, you can reach 2.7GHz using memory at 270MHz...
Whereas buying an Opteron 175 with a 11x multiplier, you can reach 2.7GHz using memory at 245MHz... (which can be done with fairly tight timings of 2.5-3-3-8)

Sorry, I made a mistake. The Opteron 170 and 165 are dualcores and I messed the numbering considering them as singlecores.
If you plain to overclock than the Opteron is better choice than X2. If not, the X2 is faster and is better "bang for the buck".

I made a mistake as I explained in my previous post.
The X2 4200 is faster for everything than both, the Opteron 170 and the Opteron 165. The Opteron and the X2 have the same architecture and are performing same at same freqfency and with same cache capacity. The 10% more freqfency boost the performance more than the extra 512kB L2 cache on the K8 CPUs.

He don't need to run the memory at 300MHz(600MHz DDR). He can use the memory divider and reduce the RAM:CPU clock ratio. For example my Venice 3200+ is OC-ed at 2.7GHz(270x10) using the 5:6 RAM:CPU clock ratio. My RAM is set to work at 166MHz(333Mhz DDR) in BIOS and my CPU/HTT clocks are overclocked for 35%, so my RAM is running at 225MHz(450MHz DDR).
166Mhz * 1.35 = 225MHz

He don't need to run the memory at 300MHz(600MHz DDR). He can use the memory divider and reduce the RAM:CPU clock ratio. For example my Venice 3200+ is OC-ed at 2.7GHz(270x10) using the 5:6 RAM:CPU clock ratio. My RAM is set to work at 166MHz(333Mhz DDR) in BIOS and my CPU/HTT clocks are overclocked for 35%, so my RAM is running at 225MHz(450MHz DDR).
166Mhz * 1.35 = 225MHz

True, you can run a divider, but IMO, you get the best performance running at a 1:1 ratio, keeping in synchronously with the processor..

Additional bandwidth almost does not improve the performance on the K8, when the real-time latency remains same. S754 vs s939, s939 vs sAM2 are just proving that. The best case when the memory bandwidth is doubled shows 5% performance increase, so more expencive memory with better bandwidth is not reasonable choice for performanse.
To be sure about your system, reduce the divider and the latencies in proportion and check the performance of your system.