Investigations into Athlon X2 Overclocking
by Jarred Walton on December 21, 2005 12:00 PM EST- Posted in
- CPUs
System Configuration
Since we're only looking at one CPU this time, we added and changed the memory options a bit. Last time, we looked at using value RAM vs. performance RAM based on Samsung TCCD memory. OCZ has moved from the old Rev2 Platinum to their new EL Platinum rated to PC-4800 speeds, and the older Rev2 platinum is no longer being manufactured. We're also adding in some results using OCZ VX Gold, Patriot 2x1GB, and we'll try to run some tests using 4x512MB of the OCZ EL Platinum. The VX RAM will be somewhat limited in performance by the maximum 3.2V that the DFI Infinity provides, but it will also serve as a reference point for tighter timings and lower bandwidth in comparison to the PC-4800 RAM.
Along with the more expensive CPU, we felt that it only made sense to upgrade the graphics card, so we're running with a GeForce 7800 GTX - an XFX model with default 450/1200 clock speeds. Note that for games, the results from the Venice processor will basically match the results that we would achieve with the Manchester, so you can refer to the earlier article to see how this system would perform in games with an X800 Pro. The reverse is also true; if you're only running games without any background tasks, the results here (in games) will show how well the Venice setup should perform with a more expensive graphics card. Here's the complete list of components used.
A complete set of benchmarks for a given configuration requires around 6-8 hours to complete, provided that nothing crashes. Unfortunately, crashes are an all-too-common occurrence when overclocking. A few driver updates (in order to run FEAR optimally) also invalidated earlier results. There are many, many factors that were not fully examined due to time constraints. This is not meant to be a fully comprehensive treatise on overclocking the X2 3800+ or any other processor. Different motherboards are certainly viable - in fact, quite a few motherboards should actually overclock better than the DFI Infinity that was used. This is really intended as a look at performance scaling, as well as some information for those looking to get the best bang for the buck. Building a faster system should be quite easy, but building a system with better performance per dollar (using a dual core processor) will be much more difficult.
I was also contacted by Asetek about testing out their MicroChill and WaterChill cooling solutions. This seemed like an appropriate place to add those to the list of benchmarked configurations, but they really deserve a separate review, so look for that in the very near future. Besides, article bloat and readability can be a problem.
Since we're only looking at one CPU this time, we added and changed the memory options a bit. Last time, we looked at using value RAM vs. performance RAM based on Samsung TCCD memory. OCZ has moved from the old Rev2 Platinum to their new EL Platinum rated to PC-4800 speeds, and the older Rev2 platinum is no longer being manufactured. We're also adding in some results using OCZ VX Gold, Patriot 2x1GB, and we'll try to run some tests using 4x512MB of the OCZ EL Platinum. The VX RAM will be somewhat limited in performance by the maximum 3.2V that the DFI Infinity provides, but it will also serve as a reference point for tighter timings and lower bandwidth in comparison to the PC-4800 RAM.
Along with the more expensive CPU, we felt that it only made sense to upgrade the graphics card, so we're running with a GeForce 7800 GTX - an XFX model with default 450/1200 clock speeds. Note that for games, the results from the Venice processor will basically match the results that we would achieve with the Manchester, so you can refer to the earlier article to see how this system would perform in games with an X800 Pro. The reverse is also true; if you're only running games without any background tasks, the results here (in games) will show how well the Venice setup should perform with a more expensive graphics card. Here's the complete list of components used.
| AMD Overclocking System | |
| Hardware | Component |
| Processor | Athlon 64 X2 3800+ Manchester 2x512K 2.0GHz (OEM) |
| Heat Sink/Fan | Thermalright XP-90 with generic 92mm 3000 RPM fan |
| Motherboard | DFI nF4 INFINITY nForce4 AMD 6.70 drivers |
| Memory | Patriot/PDP 2x1024MB ELL 2-3-2-5-1T OCZ EL Platinum PC4800 2-2-2-5-1T/2.5-4-4-7-1T OCZ Gold VX 2x512MB 2-2-2-8-1T@3.2V Generic PC3200 2x512MB 2.5-3-3-8-1T |
| Video Card | XFX 7800 GTX (450/1250 clocks) ForceWare 81.95 drivers |
| Hard Drive | Seagate SATA 250GB 7200RPM 8MB 7200.8 |
| Optical Drive | NEC 3540A |
| Case | Antec Performance II SX635BII (2x80mm Vantec Stealth fans installed) |
| Power Supply | OCZ PowerStream 600W |
A complete set of benchmarks for a given configuration requires around 6-8 hours to complete, provided that nothing crashes. Unfortunately, crashes are an all-too-common occurrence when overclocking. A few driver updates (in order to run FEAR optimally) also invalidated earlier results. There are many, many factors that were not fully examined due to time constraints. This is not meant to be a fully comprehensive treatise on overclocking the X2 3800+ or any other processor. Different motherboards are certainly viable - in fact, quite a few motherboards should actually overclock better than the DFI Infinity that was used. This is really intended as a look at performance scaling, as well as some information for those looking to get the best bang for the buck. Building a faster system should be quite easy, but building a system with better performance per dollar (using a dual core processor) will be much more difficult.
I was also contacted by Asetek about testing out their MicroChill and WaterChill cooling solutions. This seemed like an appropriate place to add those to the list of benchmarked configurations, but they really deserve a separate review, so look for that in the very near future. Besides, article bloat and readability can be a problem.
Facebook
Twitter
RSS
46 Comments
View All Comments
TheHolyLancer - Wednesday, December 21, 2005 - link
i that since this is an OC thread, they should have used a DFI NF4 Ultra-D or a Expert, they have a 4V jumper that allows you to take DRAM voltage into 4 V (i hope no one does though)JarredWalton - Thursday, December 22, 2005 - link
The higher voltages would have helped the VX RAM a bit. I may shift to a LanParty SLI-DR for the cooling test... or at least try it at some point to see how much of a difference it makes in performance.KingofCamelot - Wednesday, December 21, 2005 - link
I noticed that the BF2 demo file for v1.12 did not work. The bf2bench.demo file needs to be changed for it to work. The bf2bench.demo file can be opened in Notepad, and the following changes need to be made.These lines:
demo.camerafile mods\bf2\Demos\jw112.bf2cam
demo.demofile mods\bf2\Demos\jw112.bf2demo
Need to be changed to:
demo.camerafile mods\bf2\Demos\jwanandtech112.bf2cam
demo.demofile mods\bf2\Demos\jwanandtech112.bf2demo
JarredWalton - Thursday, December 22, 2005 - link
Thanks! I've corrected the file and uploaded the new version.--Jarred Walton
tayhimself - Thursday, December 22, 2005 - link
Hey Jared,This was a very well written article. You were thorough with the benchmarks almost to a fault. I liked your introductory and ending commentary. Your first article was just as good.
Props!
sxr7171 - Thursday, December 22, 2005 - link
Agreed. This was a quality job for sure and the questions he raises at the end are very pertinent. I'm sure he'll come up with the answers.ElFenix - Wednesday, December 21, 2005 - link
i assume you used the stock heat sink/fan unit?JarredWalton - Wednesday, December 21, 2005 - link
Er, sorry I forgot to mention that. I used an XP-90 with a 3000 RPM 92mm fan (generic fan). I'll make a note of that, since that's important information. The followup looking at cooling options will use a retail HSF as well as the XP-90, an Asetek MicroChill, and an Asetek WaterChill. (Why Asetek? Because they asked me to review their two products.)Furen - Wednesday, December 21, 2005 - link
How come the graphs arent zeroed? I suppose it'd be pretty much a bunch of overlapping straight lines if they were but having a graph that shows framerate from 63.5 to 65.0 is not much better.JustAnAverageGuy - Wednesday, December 21, 2005 - link
Yeah, the graphs could be a bit misleading unless you look at what the lines actually represent.The difference between the OCZ PC4800 and everything else looks huge in the http://images.anandtech.com/reviews/cpu/amd/athlon...">Doom 3 graph @ 1600x1200 4xAA, but if you actually look at the lines, the difference is less than 1 frame per second.