 There are a few ways to choose a processor for a no-nonsense power users PC, and it all depends on whom you ask. Some will say go with the first step below the flagship processor in the companys CPU lineup. As of this writing, that would mean the Athlon 64 X2 5000+ or the Core 2 Duo E6700. The second fastest processor in Intels or AMDs current product offerings is usually sold at a considerable discount when compared to the top of the line, but they arent what one would typically consider cheap. Others will recommend buying the cheapest processor thats in the same product family as the companys current flagship. They argue that these processors come off the same manufacturing line, and with a little overclocking, theyll perform almost as well. |  After polishing the Core 2 Duo E6400 processors integrated heat spreader and our coolers base, we applied a smooth layer of thermal paste to the CPU. | We would like to take yet another approach: The second processors from the top are a bit too expensive for us to consider them practical for this build, and the least expensive processors usually have the lowest multipliers. Low multipliers can make overclocking difficult because some motherboards cant hit FSB or HyperTransport frequencies high enough to excessively tweak the CPU and remain stable. So, for our no-nonsense Intel- and AMD-powered systems, we decided to go with middle-of-the-road processors and overclock them as far as possible. The chips we chose have higher multipliers and are only marginally more expensive than the most affordable processors in their respective lineups, yet they still only cost about a third of the price of Intels and AMDs current flagships.
Intel Core 2 Duo E6400 There are currently two types of processors in Intels Core 2 Duo lineup, those with 4MB of L2 cache and those with 2MB. The processors with 4MB of L2 cache are the higher-clocked, more expensive offerings. When clocked at similar frequencies, however, theres only a slight performance difference between the two processors. For our build, a Core 2 Duo with 2MB of cache would suffice, so we went with the Core 2 Duo E6400 ($250) and sprung for a high-performance CPU cooler: the Arctic Cooling Freezer 7 Pro. In its stock configuration the Core 2 Duo E6400 runs at 2.13GHz. You can get that speed by multiplying the CPUs respective stock multiplier (8) by its FSB frequency (266MHz). By raising either of those values, the processor would run at a higher frequency. The Core 2 Duo, however, isnt unlocked like the Core 2 Extreme, so its not possible to overclock the chips by raising the multiplier. The only way to overclock a Core 2 Duo is by raising its FSB frequency. |  Our processor and motherboard combination peaked at a 360MHz FSB frequency, producing a 2.88GHz CPU clock and a DDR2-900 memory speed. | Before we got to the business of overclocking our processor, however, we modded it slightly to help it run a little cooler. We followed essentially the same procedure you can find in the October issue of CPU on page 40. Because the CPUs integrated heat spreader and our Freezer 7 Pros copper base were already nice and flat, all we did was use a wadding metal polish to eliminate small peaks and valleys from their surfaces. Doing so makes the surfaces mate better, which aids in heat transfer. This is a quick and easy mod that will usually shave a couple of degrees off a processors operating temperature, so theres really no reason not to do it—well, except for that voided warranty thing.
Make The Grade Once we sufficiently polished our CPUs heat spreader and coolers base and installed everything, we began overclocking the system via the BIOS. We overclocked the system in multiple steps: First, we disabled whatever integrated peripherals we didnt intend to use and then we locked the motherboards PCI Express frequency at 100MHz. Doing this will keep the PCI-E links from becoming unstable and potentially prevent an integrated peripherals failure from hindering our overclocking efforts. Next we increased all pertinent voltages for the processor, memory, and motherboard chipset. At their default voltages, each of these system components would only overclock to a certain point, but with a little extra juice we could overclock them even further. Because we didnt take any elaborate cooling measures, though, we only bumped up the CPU voltage from 1.35 to 1.41V, our memory voltage to 2.3V, and our MCH (northbridge) and ICH (southbridge) voltages by 0.1V.
  Hit The Virtual Wall After making the necessary voltage changes in the system BIOS, we then began overclocking our processor by increasing its FSB frequency. At first we raised the FSB frequency by about 10MHz at a time until the system was no longer stable and then backed the frequency down 1 to 2MHz until the system stabilized. When all was said and done, we settled on a 360MHz FSB frequency with a final CPU clock speed of 2.88GHz. Our CPU probably had some more in the tank, but the affordable Abit P965 motherboard we used for this experiment got flaky once the FSB hit speeds above 3.45MHz or so. There have been reports of the Core 2 Duo E6400 processors hitting above 3GHz, so we may revisit this CPU in the future and couple it with a more overclocker-friendly motherboard. |  We polished our Athlon 64 X2 4600+ processors heat spreader before applying a high-quality thermal paste and installing the Thermaltake Mini Typhoon CPU cooler. | AMD Athlon 64 X2 4600+ As we already explained in the first half of this article, the components we chose for our practical power user systems had to meet certain criteria. Like our Intel processor, the CPU for our AMD build needed to be significantly more affordable than the companys current flagship Athlon 64 FX-62 processor, but it also had be a good candidate for overclocking. After a bit of searching, we decided to go with AMDs Athlon 64 X2 4600+. The price cuts AMD made prior to Intels introduction of the Core 2 Duo put the standard edition Athlon 64 X2 4600+ into sub-$250 territory, which made this 2.4GHz CPU a good candidate. We should note that AMD also offers an energy-efficient version of the X2 4600+ that would have worked just as well for this build, but its about 20% more expensive. (You can tell the difference between AMDs energy-efficient processors and their normal counterparts by their model numbers. The third digit of the processors part number details the maximum power consumption of that model: A denotes normal power, O indicates a 65W energy-efficient processor, and D signifies a 35W energy-efficient, SFF processor.)
Prep The Hardware To get our Athlon 64 X2 4600+ ready for tweaking, we gave it the same treatment as our Core 2 Duo E6400. Its integrated heat spreader was nice and flat, so we simply used a wadding metal polish to buff it to a mirror-like shine. The Thermaltake Mini Typhoon that we used to keep the processor cool already had a perfectly polished base, however, so we didnt have to touch it. Although it works well, we decided to forego AMDs stock PIB (processor-in-a-box) cooler in favor of the Mini Typhoon for a couple of reasons. For one, the Mini Typhoon performs better than AMDs PIB. The Mini Typhoons also quieter than the stock cooler and its design helps blow plenty of air over our motherboards VRM and the system RAM—two things that should help stabilize the system while overclocking.  |  Raising the motherboards HyperTransport and chipset voltages will usually help them remain stable at higher frequencies. | The Overclock Unlike the Core 2 Duo E6400, the Athlon 64 X2 4600+s multiplier is slightly unlocked. You can specify multipliers lower than stock; AMD does not allow higher multipliers, though. This opens up some interesting possibilities for tweaking overall system performance by utilizing a lower multiplier with higher HyperTransport and memory frequencies, but thats a discussion well have to cover someday in another article. Our goal for this build was to hit the highest, fully-stable processor frequency possible. That meant sticking with the X2 4600+ processors 12X multiplier and raising the HT reference clock as high as possible. To help us achieve our goal, we had to alter a couple of BIOS settings before tweaking the CPU. As youd expect, raising the HT reference clock results in a higher-clocked HyperTransport link. Some processors and motherboards will no longer be stable at a certain point, though, so before taking the processor any higher, we decided to drop the HT multiplier from 5X to 4X. Changing the HT multiplier to 4X set the link speed to 800MHz, giving us some room to work. We also dialed the PCI-E clock to 100MHz to prevent pushing the PCI-E interface out of spec. And finally, we cranked the processors voltage up from 1.35 to 1.5V and set the memory, HyperTransport, and chipset voltages to 2.3, 1.3, and 1.3V, respectively. |  By increasing our processors core voltage and tweaking memory and chipset voltages, we were able to raise the HT frequency to 237MHz with a peak CPU clock of almost 2.9GHz. | All She Wrote With all of the preparation done, we slowly began raising the 200MHz default HT reference clock speed until the system became unstable. We increased the speed a few megahertz at a time and ultimately settled upon a 237MHz reference clock with a final CPU clock speed of 2.84GHz. The system actually booted at higher speeds but couldnt make it through our suite of benchmarks without crashing, so we backed off a bit and stabilized the system at 237MHz. A CPU clock speed of 2.84GHz is about an 18.5% increase over the Athlon 64 X2 4600+s default 2.4GHz frequency; enough of a boost to put it past the flagship Athlon 64 FX-62. Not bad for a $250 processor. With more exotic cooling and perhaps a bit more voltage, we suspect frequencies higher than 2.9GHz are a possibility with the Athlon 64 X2 4600+, but were pleased with these results. Our processor and motherboard combination played very well together and it definitely showed in our benchmarks. by Marco Chiappetta
Performance: Before & After In its default configuration, the Core 2 Duo E6400 cant hold a candle to Intels current flagship Core 2 Extreme X6800 processor. The E6400s lower clock speed and smaller L2 cache result in significantly lower benchmark scores in every application. When we overclocked the E6400 to 2.88GHz, however, the combination of higher core and FSB frequencies made the E6400 perform similar to the Core 2 Extreme X6800. Also, a side effect of a higher FSB frequency is additional memory bandwidth, which helps minimize the performance delta between the two processors. In our benchmarks, which we ran on the Abit P965-based motherboard coupled to a GeForce 7900 GTX and running Windows XP SP2, the overclocked Core 2 Duo E6400 came very close to catching the Core 2 Extreme X6800, which was running on a 975X Express-based motherboard. As you can see from our results, if we were able to squeeze a few more megahertz out of our P965 motherboard and CPU combo, surpassing the performance of the X6800 would have been a possibility. Perhaps with a different motherboard at the heart of the system we would have had better luck, but were not complaining. | | Stock Clock Speeds Core 2 Duo E6400 at 2.13GHz | Overclocked Speeds Core 2 Duo E6400 at 2.88GHz | Stock Clock Speeds Core 2 Extreme X6800 at 2.93GHz | | SiSoft Sandra 2007 Pro | | Processor Arithmetic | | Dhrystone ALU | 19,773Mips | 25,963Mips | 27,166Mips | | Whetstone iSSE3 | 15,241MFlops | 17,921MFlops | 18,438MFlops | | Processor Multimedia | | Integer x4 aEMMX/aSSE | 117,694itps | 153,346itps | 159,627itps | | Floating-Point x4 iSSE2 | 63,443itps | 82,847itps | 86,392itps | | Memory Bandwidth | | Integer Buffered iSSE2 | 5,404MBps | 6,424MBps | 5,607MBps | | Floating-Point Buffered iSSE2 | 5,401MBps | 6,419MBps | 5,602MBps | | LAME MT MP3 Encoding (minutes:seconds) | | Single-threaded | 1:04 | 0:49 | 0:47 | | Multithreaded | 0:45 | 0:33 | 0:32 | | Cinebench 2003 (seconds) | | Single-threaded | 69.9 | 54.2 | 51.8 | | Multithreaded | 39.4 | 29.3 | 28.4 | | KribiBench v1.1 | | Sponge Explode Model | 4.48fps | 5.98fps | 6.14fps | | Ultra Model | 1.48fps | 1.89fps | 1.97fps | | F.E.A.R. v1.07 | | Low-resolution CPU test (640 x 480, low quality) | 221fps | 261fps | 272fps | | Quake 4 v1.2 | | Low-resolution CPU test (640 x 480, low quality) | 162.4fps | 201.8fps | 203.5fps | |
Performance: Before & After Our Athlon 64 X2 4600+ processor proved to be a decent overclocker when coupled to Abits nForce 570 SLI powered KN9 SLI motherboard and the Thermaltake Mini Typhoon. With some mild voltage increases and a higher HT frequency, we took our X2 4600+ to almost 2.9GHz, which was enough of a boost to overtake AMDs current flagship Athlon 64 FX-62 processor in our benchmarks. We tested the processor on Abits KN9 SLI with a single GeForce 7900 GTX in one of its graphics slots and Windows XP Professional with SP2 installed on its drive. Despite having a smaller L2 cache (512 vs. 1MB per core), the overclocked Athlon 64 X2 4600+ outpaced the FX-62 in every benchmark we ran, thanks to a slightly higher core frequency and an increase in memory bandwidth resulting from the higher HT links speed. | | Stock Clock Speeds Athlon 64 X2 4600+ at 2.4GHz | Overclocked Speeds Athlon 64 X2 4600+ at 2.84GHz | Stock Clock Speeds Athlon 64 FX-62 at 2.8GHz | | SiSoft Sandra 2007 Pro | | Processor Arithmetic | | Dhrystone ALU | 17,321Mips | 21,265Mips | 20,411Mips | | Whetstone iSSE3 | 14,632MFlops | 17,864MFlops | 17,230MFlops | | Processor Multimedia | | Integer x4 aEMMX/aSSE | 45,126itps | 53,593itps | 52,936itps | | Floating-Point x4 iSSE2 | 48,824itps | 58,076itps | 57,642itps | | Memory Bandwidth | | Integer Buffered iSSE2 | 8,174MBps | 9,438MBps | 8,500MBps | | Floating-Point Buffered iSSE2 | 8,164MBps | 9,422MBps | 8,464MBps | | LAME MT MP3 Encoding (minutes:seconds) | | Single-threaded | 1:23 | 1:09 | 1:11 | | Multithreaded | 0:54 | 0:44 | 0:47 | | Cinebench 2003 (seconds) | | Single-threaded | 77 | 66.1 | 66.3 | | Multithreaded | 41.3 | 35.2 | 35.4 | | KribiBench v1.1 | | Sponge Explode Model | 4.46fps | 5.22fps | 5.14fps | | Ultra Model | 1.27fps | 1.51fps | 1.47fps | | F.E.A.R. v1.07 | | Low-resolution CPU test (640 x 480, low quality) | 213fps | 252fps | 248fps | | Quake 4 v1.2 | | Low-resolution CPU test (640 x 480, low quality) | 145.2fps | 178.2fps | 169.1fps | |
|