Overclocking Results
It is important to state that in all cases we are using the stock retail heatsinks (except with the simulated X2 3800+, which used a Zalman CNPS9500). Again, adding a nice aftermarket cooling solution might improve overclocking results a bit, but for the extra $30-$50 you could simply upgrade to a faster processor instead. Here are the overclocking results we achieved as well as the settings that were used. We tried for 100 MHz increments on the CPU overclock, so if we could get a system running at 2.5 GHz but not 2.6 GHz, we stopped at 2.5 GHz. With additional time and effort, we could improve the final results and increase performance a few percent, but searching for the elusive last 3%-5% is beyond the scope of this article. (CPU speeds are approximate to within 10 MHz -- slight differences between the specified bus speed and the actual bus speed can affect clock speed slightly.)
Starting with the percent overclocks achieved, the two Sempron offerings tie for first place with a 39% overclock when using the budget motherboards (we were able to get a 50% overclock with the Sempron AM2 using the Gigabyte nForce 590 SLI motherboard), while the Pentium D 805 falls behind with "only" a 35% overclock. However, getting a 35% overclock out of a budget dual core platform is in some ways more impressive than a 40% overclock of a single core CPU. If you happen to run applications that take advantage of multiple processors, you can probably already guess that the Pentium D 805 is going to easily beat anything that the two Sempron platforms can put up in SMP-aware benchmarks. It's also important to note that the Pentium D would actually load Windows at 3.8 GHz and POST at up to 4.0 GHz. Better cooling (and a better power supply) would have almost certainly allowed a higher overclock, but as mentioned for the price you may as well start with a better CPU.
The Sempron systems were far less likely to POST at higher overclocks, at least with the budget motherboards. The Biostar TForce AM2 motherboard in particular hit a brick wall with a HyperTransport bus speed of 280 MHz; below that it would POST, but above that it required the use of the CMOS clear jumper to get the motherboard back online. Given our experience with other AM2 motherboards, we would venture to guess that a BIOS update could improve the Biostar motherboard. However, we are currently using the latest Biostar TForce 550 BIOS, and the motherboard has been available for about six weeks now. Hopefully, Biostar will release an improved BIOS for the motherboard that addresses performance and overclocking, but we don't make purchase recommendations based on hope.
In terms of ease of overclocking, the Biostar TForce4U-775 was actually the most difficult system to deal with. The difficulty wasn't so much in actual capability, but more in terms of understanding the BIOS options. We ended up using "expert" mode for the bus and memory settings, then we specified a front side bus value of 720 MHz and a memory speed of 540-720 MHz dependent on which memory was used. CPU-Z doesn't always report the correct memory speed on the motherboard, which caused some confusion on our part. For example, the 720 FSB speed and 720 MHz memory speed shows up in CPU-Z as a 1:1 memory ratio and a 360 MHz memory speed. However, benchmarks confirm that the memory is indeed running at a 1:2 ratio.
So now we've set the stage with the processors used and the overclocks achieved. Let's find out how things shake down in the actual benchmarks. We have run most of our typical motherboard benchmarks, and all benchmarking was conducted at 1280x1024 resolution with settings appropriate for a midrange graphics card, with the exception of 3DMark03/05 which were run at the default 1024x768.
It is important to state that in all cases we are using the stock retail heatsinks (except with the simulated X2 3800+, which used a Zalman CNPS9500). Again, adding a nice aftermarket cooling solution might improve overclocking results a bit, but for the extra $30-$50 you could simply upgrade to a faster processor instead. Here are the overclocking results we achieved as well as the settings that were used. We tried for 100 MHz increments on the CPU overclock, so if we could get a system running at 2.5 GHz but not 2.6 GHz, we stopped at 2.5 GHz. With additional time and effort, we could improve the final results and increase performance a few percent, but searching for the elusive last 3%-5% is beyond the scope of this article. (CPU speeds are approximate to within 10 MHz -- slight differences between the specified bus speed and the actual bus speed can affect clock speed slightly.)
| AMD Sempron 754 System | ||
| Stock Setting | Overclocked Setting (39%) | |
| HyperTransport Speed | 200 MHz | 278 MHz |
| HyperTransport Multiplier | 5X | 3X |
| CPU Multiplier | 9X | 9X |
| CPU Speed | 1800 MHz | 2500 MHz |
| Memory Ratio | DDR-400 | DDR-400 |
| Actual Memory Speed | DDR-400 | DDR-556 |
| Memory Timings | 2-3-2-10-2T | 2.5-3-3-10-2T |
| CPU Voltage | 1.350 V | 1.550 V |
| Memory Voltage | 2.800 V | 2.800 V |
| AMD Sempron AM2 System | ||
| Stock Setting | Overclocked Setting (50%) | |
| HyperTransport Speed | 200 MHz | 300 MHz (278 MHz Biostar) |
| HyperTransport Multiplier | 5X | 4X |
| CPU Multiplier | 9X | 9X |
| CPU Speed | 1800 MHz | 2700 MHz (2500 MHz Biostar) |
| Memory Ratio | DDR2-800 | DDR2-533 |
| Actual Memory Speed | DDR2-720 | DDR2-772 (DDR2-714 Biostar) |
| Memory Timings | 4-4-4-12-2T | 4-4-4-12-2T |
| CPU Voltage | 1.350 V | 1.550 V |
| Memory Voltage | 2.100 V | 2.100 V |
| Intel Budget System | ||
| Stock Setting | Overclocked Setting (35%) | |
| Front Side Bus Speed | 533 MHz | 720 MHz |
| CPU Multiplier | 20X | 20X |
| CPU Speed | 2667 MHz | 3600 MHz |
| Memory Ratio | DDR2-667 (2:5) | DDR2-720 (1:2) |
| Actual Memory Speed | DDR2-667 | DDR2-720 |
| Memory Timings | 4-4-4-12-2T | 4-4-4-12-1T |
| CPU Voltage | 1.125 V | 1.400 V |
| Memory Voltage | 2.100 V | 2.100 V |
| Simulated AMD X2 3800+ AM2 System | ||
| Stock Setting | Overclocked Setting (30%) | |
| HyperTransport Speed | 200 MHz | 260 MHz |
| HyperTransport Multiplier | 5X | 5X |
| CPU Multiplier | 10X | 10X |
| CPU Speed | 1800 MHz | 2600 MHz |
| Memory Ratio | DDR2-800 | DDR2-533 |
| Actual Memory Speed | DDR2-720 | DDR2-650 |
| Memory Timings | 4-4-4-12-2T | 4-4-4-12-1T |
| CPU Voltage | 1.350 V | 1.500 V |
| Memory Voltage | 2.100 V | 2.100 V |
Starting with the percent overclocks achieved, the two Sempron offerings tie for first place with a 39% overclock when using the budget motherboards (we were able to get a 50% overclock with the Sempron AM2 using the Gigabyte nForce 590 SLI motherboard), while the Pentium D 805 falls behind with "only" a 35% overclock. However, getting a 35% overclock out of a budget dual core platform is in some ways more impressive than a 40% overclock of a single core CPU. If you happen to run applications that take advantage of multiple processors, you can probably already guess that the Pentium D 805 is going to easily beat anything that the two Sempron platforms can put up in SMP-aware benchmarks. It's also important to note that the Pentium D would actually load Windows at 3.8 GHz and POST at up to 4.0 GHz. Better cooling (and a better power supply) would have almost certainly allowed a higher overclock, but as mentioned for the price you may as well start with a better CPU.
The Sempron systems were far less likely to POST at higher overclocks, at least with the budget motherboards. The Biostar TForce AM2 motherboard in particular hit a brick wall with a HyperTransport bus speed of 280 MHz; below that it would POST, but above that it required the use of the CMOS clear jumper to get the motherboard back online. Given our experience with other AM2 motherboards, we would venture to guess that a BIOS update could improve the Biostar motherboard. However, we are currently using the latest Biostar TForce 550 BIOS, and the motherboard has been available for about six weeks now. Hopefully, Biostar will release an improved BIOS for the motherboard that addresses performance and overclocking, but we don't make purchase recommendations based on hope.
In terms of ease of overclocking, the Biostar TForce4U-775 was actually the most difficult system to deal with. The difficulty wasn't so much in actual capability, but more in terms of understanding the BIOS options. We ended up using "expert" mode for the bus and memory settings, then we specified a front side bus value of 720 MHz and a memory speed of 540-720 MHz dependent on which memory was used. CPU-Z doesn't always report the correct memory speed on the motherboard, which caused some confusion on our part. For example, the 720 FSB speed and 720 MHz memory speed shows up in CPU-Z as a 1:1 memory ratio and a 360 MHz memory speed. However, benchmarks confirm that the memory is indeed running at a 1:2 ratio.
So now we've set the stage with the processors used and the overclocks achieved. Let's find out how things shake down in the actual benchmarks. We have run most of our typical motherboard benchmarks, and all benchmarking was conducted at 1280x1024 resolution with settings appropriate for a midrange graphics card, with the exception of 3DMark03/05 which were run at the default 1024x768.
Facebook
Twitter
RSS
56 Comments
View All Comments
Calin - Wednesday, July 12, 2006 - link
We were promised reviews of micro ATX mainboards a while back (hopefully with integrated video performance too, dual and single channel :) ). The article should be in the works now.Anyway, none of the integrated graphics/microATX boards I looked upon (from MSI, Asus etc) did have anything in the way of overclocking/overvolting/undervolting/...
I'm waiting, thanks
Rix2357 - Monday, July 10, 2006 - link
Most of these motherboards are decently cheap, but I wonder if they are as stable as some more well established brands. Of course, I go way back and a mantra that I've heard over and over, for stability Asus and Intel motherboards are the only way to go. While the DFI board in the test could be reasonably expected to be relatively stable, I am unsure about the biostar motherboard.My case in point from way back. The ECS K7S5A motherboard for a Socket A Athlon motherboard was supposedly one of the best, but it still had it's quirks. It's no where near as stable as the Asus Williamete motherboard at that time. It also has cold boot issues that numerous different bioses have never fixed. Hard drive detection with a third party IDE controller could be problematic at best.
Calin - Wednesday, July 12, 2006 - link
I have an ECS K7S5A mainboard, and its stability was good. Anyway, that was around the time when even top mainboards had one or two errors in 72 hours of benchmarks/stability checks.I would say Gigabyte boards are good too - I've had not so positive experiences with Foxconn boards.
JarredWalton - Monday, July 10, 2006 - link
The Biostar + PD805 has been in use for roughly six months, without any noticeable problems. The DFI Infinity has also been in use for a while without problems. AM2 is too new to say how it is in terms of stability, but I definitely enjoy the extra features (because I am one of those people that has GbE running through the house). If your primary concern is stability, you'd probably want to skip out on overclocking anyway. One of the biggest difficulties in overclocking is determining long-term "stable" settings. It can take months to get what you feel is an optimally stable system, and minor tweaks to voltages, memory timings, bus speeds, etc. can turn a crash-prone system fully stable (or vice versa).Avalon - Monday, July 10, 2006 - link
If these are budget considerations, why is 2-2-2 DDR memory and an XXX XFX listed in there? You can get an eVGA 7600GT CO for less than $150AR from time to time, and 1GB of CAS 2.5 memory will only set you back $90 and be practically as fast. That's $60 saved.Gary Key - Monday, July 10, 2006 - link
Jarred explained this on page 2. The memory selected for the S754 system is what we had available at the time for 2-2-2 operation. You are right that additional money could be saved for a minor performance delta. Since this article was looking at budget gaming performance, the XFX video card was chosen since it was still in our price range and is factory overclocked allowing a small increase in performance for those uncomfortable with overclocking the GPU. We mentioned that with rebates just about any of the 7600GT cards could be had for around $140~$160.
This article was designed to provide an umbrella look at what is available in the $650 price range and how it performs with today's game titles. We certainly understand that additional money could be saved or could be spent depending upon the individual's objective in building a system. As an example, on the S754 system you could have spent a little less on the memory but could have taken those savings and applied them to a better audio solution resulting in the same end figure. Some people would rather have the improved memory performance and live with the on-board audio. In the end, for about $650 you can build a decent machine today and there are an incredible amount of component choices that allow you to do this.