Rockdpm :
If money is not a issue, why not just get a new platform? 4 years has been long enough imo. And some one else also mentioned you can run into bottlenecks... not so much at 4GHZ. but if you have to high voltage feed your cpu for a long time just to cut a corner that you could be saving a nice components life well.... For the sake of the enthusiast... Please up your game!.
I agree if money is not an issue. However check this out, made an interesting discovery. If you use the passmark CPU benchmarks, I found that within the 900s Nehalems, the clock speed to passmark score was pretty proportional. For example:
The Intel Core i7 920 @ 2.67Ghz does a 5518
The Intel Core i7 975 @ 3.33Ghz does a 6890
The Percent increase from 5518 to 6890 in passmark is exactly 24.864%. The percent increase from the clockrate of 2.67Ghz to 3.33Ghz is exactly 24.719%. The difference between these two *differences* is just 0.583%...of which the factor of a different computer parts can make. Theoretically, the value of the 975 should have been 6882 to be directly proportional, but small changing circumstances can account for that. If we use the 920 as a base to predict its passmark score if we increase clock from 2.67Ghz to 3.33Ghz, the value of direct proportion or theoretical is 6882...which is a mere 0.116% difference. I bet you make that number smaller if you use the exact same system. So within a certain architecture it appears clock to performance is very proportional. Another example is the 930 and 960...two different CPU's of one architecture:
The Intel Core i7 930 @ 2.8Ghz does a 5832
The Intel Core i7 960 @ 3.2Ghz does a 6632
Percent Difference in Clock: 14.286%
Percent Difference in Passmark: 13.717%
Percent Difference: 4.148%...surely a larger difference, but then again circumstances like other hardware may/may not account for that.
So I'm sure you can get a more accurate value if you take the 920/930/940/950/960/975 and average out all the possible differences in clock performance between all the combinations of those. However, if we use the first example with the two opposite sides of the range: 975 and 920 which had just 0.583% then a *nehalem* quad core at 4Ghz (overclocked) can give you theoretically:
Intel Core i7 960 @ 3.2Ghz does a 6632
Intel Core i7 960 @ 4.0Ghz does a 8290 + or - 0.583%
I've heard people pushing these cpu's well past 4Ghz and if the passmark represents any real-world value then 8290 is faster than a Ivy Bridge i5-3450 that many recommend for gaming, faster than a i7-2600...etc. So LGA1366 is still absolutely no slouch...it has great CPU's, overclockables very well, and provides good performance. I'd save money and get a 960, or if you can find a 965/975 of similar price, extreme editions.
However, I could be wrong in my calculations/perspective of passmark as a good benchmark or how it relates to clock to clock performance.
JackNaylorPE :
I'd grab a 950/960 and add a Mushkin Chronos Deluxe 120 GB for $99.99.......
.....nothing touches it performance wise *
.....nothing else has its long life *
.....nothing else costs $0.80 per GB
* except other SSD's with premium toshiba toggle mode flash
http://www.newegg.com/Product/Product.aspx?Item=N82E16820226318
http://www.tweaktown.com/reviews/4328/mushkin_chronos_deluxe_120gb_solid_state_drive_review/index6.html
To sum it all up with a bow on top, you get amazing performance, extremely long service life and a hassle free low price point on a drive that literally has very little competition in the marketplace.
I can't find any good benchmarking data. Is a chronos faster than a Vertex 4?