It's been over a year since I built my last PC, and all those killer Core 2 Duo benchmark and overclocking results were making me anxious. I just pulled the trigger on the following Core 2 Duo upgrade:
Ben Hollis: You might try CPU Burn?
http://users.bigpond.net.au/cpuburn/
Jeff: I’m a big fan of Zalman’s heat pipes, as well. They’re a bit more beautiful. I used the same rationale as you did for spending the extra money for the E6600; the extra cache made the deal so much sweeter. You can overclock all you want, but cache misses are still going to destroy your performance. I went with a gig of OCZ 800, because it was on sale 
From scratch, the system I bought came out to be $1200, without a monitor. Not bad for a system which overpowers my previous computer (a P4 2200 laptop) by an order of magnitude!
What video card and power supply are you running Jeff? Your purchase is what I was looking at with the exception of the memory and heatsink/fan. I typically stick with Corsair and was going to buy the DDR 2 800 XMS 2GB set. Also, I’m not familiar with the heatsink/fan combo and would have went with a nice Zalman (which have never let me down). I’m a huge Asus fan and believe that is a great board. The processor is the best bang for your buck. Nice choices.
Excellent post - since my workplace won’t purchace a proper development box for me, this looks like just the frugal upgrade needed to run Lotus Notes!
Hm, I wonder about these heatsinks, one kilo hanging by the socket?
The Scythe Infinity is overkill for this CPU and pricey to boot. I wanted overkill in this scenario, but you may not. Here are some cheaper CPU coolers that will work just as well on the Core Duo 2:
Scythe Katana, $24.99
http://www.newegg.com/Product/Product.asp?Item=N82E16835185123
Arctic Cooling Freezer 7 Pro, $29.99
http://www.newegg.com/Product/Product.asp?Item=N82E16835186134
(the latter is a good suggestion from Geoff’s entry linked above!)
Out of curiousity, did you notice the “wrong chipset” comment about your motherboard on newegg?
G965 / P965 – G965 has integrated graphics but is otherwise identical per this article:
http://www.digit-life.com/articles2/mainboard/i965-chipset.html
I’m not sure why that person would expect integrated video… maybe he thinks the serial port is supposed to be a VGA connector?
Will an upgrade as described in this post make a substantial difference?
A Core 2 Duo at ~3.0 GHz would be close to 100% faster than a P4 3.0 GHz. Plus the whole dual core thing (I assume your P4 is single core). I’d call that substantial!
Does that mobo have an AGP slot?
Nope, all modern mobos are PCI express only. But there are lots of good inexpensive PCIex video cards. I can recommend the nVidia GeForce 7600gt as a good mid-range performer:
http://www.codinghorror.com/blog/archives/000662.html
What video card and power supply are you running Jeff?
Antec Neo 480w PSU and ATI 1900XTX video card. I recommend Seasonic S12 power supplies because they’re nearly 85% efficient, which means less heat and less waste.
Anything in the 330w-380w range is fine.
It’s a common misperception that you “need” 400w or more power for a computer. Unless you’re running dual high-end video cards and an overclocked CPU with multiple hard drives, it’s unlikely you could even crack 300 watts.
If you’re supposed to be running Notes, that’s a good thing. Notes has no place on a real box.
Out of curiousity, did you notice the “wrong chipset” comment about your motherboard on newegg?
Heh - very funny! I just did the same thing and overclocked mine as well! Love that E6600, it has a lot of room for overclocking!
http://www.recursivecoder.com/blog/2006/09/dual_core_processing.html
Do you know what the linux equivalent to Prime95 is? It might be nice to be able to take a clean system and stress-test it with a Live CD instead of installing Windows first - I still worry about what all the hard crashes you’ll incur while testing overclock settings will do to Windows.
I’m generally suffering from slow performance in Visual Studio 2005 on a P4 3.0, 2GB RAM. Will an upgrade as described in this post make a substantial difference?
Hi Jeff,
Nice setup. I curious what you did with your old parts? EBay?
The Athlon X2 4800+ that you’re replacing is better than my Athlon X2 3800+.
G965 / P965 – G965 has integrated graphics but is otherwise identical
My reading of the comment was that this person did NOT expect integrated graphics (expected P965) and was very upset (judging by his low rating) when the Intel chipset ID software told him that his P5B had the G965.
His mobo works just fine, but he sees a number he didn’t expect and he’s irked. Geeks. Go figure.
Some possibilities (aka wild, unfounded speculations) which occurred to me. (1) The Intel chipset ID s/w he used has a bug? (2) User error of some kind? (3) Intel could be using G965s with the video disabled to meet demand for P965s?
Oh well. Thanks for the gracious response to my probably not all that pertinent question.
-john
Jeff, did you just burn up your computer?
http://forums.silentpcreview.com/viewtopic.php?t=34826
Hope you get it running again!
I’m going to disagree with this overclocking love-in and say that any overclocking is idiotic. Intel et al design their components to work at certain speeds. They test the components coming off the assembly line to confirm that the required levels are reached, and to mark up the parts appropriately according to the speeds achieved by each part, or the amount of cache that actually tests OK on the processor (for example). I expect that all Core 2 Duos in the E6x00 range (and possibly the Extreme Edition X6800 too) are coming off exactly the same production line, made with exactly the same masks; E6300 and E6400 parts are those where some of the cache doesn’t work and the cache has to be restricted to 2MB, or where the cores can’t hit the clock speed to be labelled an E6600. E6600s couldn’t hit the 2.66GHz to be labelled an E6700, and E6700s can’t hit 2.93GHz to be labelled X6800.
Now, the actual results of production may not match the demand curve, so there may be some parts that are marked down from their actual test results. But you’ll never know which those are. There’s a certain amount of safety margin too, but again, you don’t know how close to the safety margin your particular processor and other components can handle.
What makes a part top out at different speeds, despite being manufactured with the same masks, is that there will always be slight alignment problems in applying the different masks, and slightly different amounts of chemical etching, which will make some features slightly larger or smaller than actually designed. This affects the resistances and capacitances in the circuits, which affect the time taken to charge (or discharge) to a particular voltage, if a change in voltage is applied. This RC circuit is often not desired in digital electronics, but comes from the inherent resistances and capacitances; it cannot be eliminated. A change in these values - an increase in the resistance and/or capacitance - increases the charging or discharging time, which affects the propogation delay through the circuit, leading to the possibility that the wrong value is sampled on the clock edge. This leads to all sorts of erroneous results, including some instructions failing that should never be able to fail.
People get around this by increasing the supply voltage. With more supply voltage, the RC circuit reaches the reference sample voltage more quickly. In the other direction, though, I think it’ll take longer to drop below the reference voltage required to switch a transistor off. There is also more current flowing in the circuit and therefore more power being consumed and dissipated (as heat). More cooling can help, but there’s still a greater possibility of getting hot spots which burn out a particular section of the chip (generally not the silicon, simply melting through the metal interconnect between features).
did you just burn up your computer?
Yep, unfortunately. Northbridge temperatures of 120c (as measured by a temperature gun) is something I’ve never seen before. I suspect a defective motherboard, as things were working just fine until that point. Even under load I never saw the NB go over 70c.
Anyway I returned the mobo and ordered a replacement.
Intel et al design their components to work at certain speeds
Well, if you don’t like overclocking, don’t do it. It’s the same reason people trick out and hot rod cars. It’s not for everyone, but to argue that it violates some fundamental law of physics is silly.
" I can bask in the glory of a system that’s 50% faster than my old Athlon X2 4800+"
But how much faster is it than it would be if you didn’t go through the hassle of overclocking it? In other words, what did you REALLY gain by going to the trouble of overclocking? Or did you forget to test the base configuration… 