Need for speed, but hit a wall

CPU: Intel E5200
Motherboard: DP35DP
DDR2: 4GB 4-4-4-12
Graphics Card: NVIDIA GeForce 8400GS

I picked this combination because I need something to “crunch numbers” at home. But I need something faster than the machine’s current operating speed - something around 4 is what I need.

However, I’m stuck. The BIOs does not have a voltage option and I’d really like a nonintensive solution.

Does anyone have a "work-around" for this combination?

Thanks

4ghz might be stretching it a bit, but check this thread: http://www.ocforums.com/showthread.php?t=586596

Thank you.

I'll definately read it.

Just read it. It's definately a creative solution but I think I'll use it as a last resort.

Thanks again,
Dale

I don't think that there are too many other choices with your motherboard, unless you decide to get a motherboard designed to overclock that kind of cpu.

Thanks for your input Glussier. Looks like you are right. My options are limited.

But just as I thought my ship was on the shoal, I realized that the value of a stable machine when crunching data can mean the difference between loosing or keeping your job.

It’s better that I just “take a deep breath” and stay up an extra hour or so than try to explain why the analysis is incorrect.

Thanks for your insight.

Best Regards,
Dale

Have you considered upgrading to a quad-core CPU? It's not a free upgrade like an overclock, but if the apps you're using are multithreaded (and CPU limited at the moment), you could effectively double the rate at which you're calculating. Newegg has a Q6600 with heatsink for $190.

Yes, the quadcore was considered.

Please correct me if I’m wrong , because my academic background is statistics not CS and I don’t have a EE . . .but the main performance obstacle is the firmware and software, not the hardware. Specifically, the OS is the main limiting factor here.

If someone came up with an OS written for multicore processors that would handle serious algorithms, I believe my concerns would be addressed.

As a footnote, you’re probably familiar with the MIT “poor student option” of the 80’s. Seems access to the Cray was limited and several postgrads were worried about being able to complete their research projects.

So they contacted Intel and cut a deal for processors that were installed on some CB breadboards. These students wrote an OS specifically for their system and wham-bam they ended up with a one-of-a-kind that performed almost as good as the Cray . . for their purposes at least. Although I’m simplifying this illustration I believe the solution they had was and is very real.

It's not possible to magically make something multi-threaded. Multi-threading introduces all kinds of nasty issues, race conditions, data dependency (from previous instructions), synchronization/locking, etc. Some algorithms/problems are easy to multi-thread (some sorting algorithms like mergesort, pixel operations, video encoding), since they can be easily divided into independent sub-tasks (for image operations, each thread can take care of a part of the image, for video encoding, alternating frames by different threads, etc), while some algorithms are next to impossible to parallelize (alpha-beta algorithm used in zerosum game AI, emulation, etc), since they are inherently "serial" algorithms, where later parts of the algorithms need to use data calculated in earlier parts, making it impossible to do both parts at the same time.

Modern OSes already handle multicore processors quite well and for quite some time already (Linux kernel 2.0, circa 1996, introduced SMP support), but it is up to applications to be written to take advantage of it.

There is no "magical" solution.

Thank you Cyberfish for the overview. But I wasn’t looking for magic.

I just didn’t want to spent $4,000.00 for the “warp drive” software written specifically for the AMD and Intel quadcores.

Most importantly - thanks everyone for the help and advice.

Best to all,
Dale