Investigations into Athlon X2 Overclocking
by Jarred Walton on December 21, 2005 12:00 PM EST- Posted in
- CPUs
Introduction
Note: This is the second article in a series of overclocking articles. Our first article looked at overclocking a Venice 3200+ chip using two different types of RAM, and it laid the groundwork for what we're doing in this article. We've updated some of our component choices as well as our benchmarks. The major difference is that we're now using an X2 3800+ for the processor, and we also upgraded to a 7800 GTX graphics card. We'll be looking at performance with several types of RAM as well. These articles are not targeted at the advanced overclockers, nor are we going for extreme performance at any cost. The main objective is to use some cost-effective setups and show what sort of performance level can be achieved.
As we mentioned before, there is a risk in overclocking and we do not recommend it for everyone. A system that appears stable for weeks or even months can suddenly have problems, so we definitely wouldn't recommend overclocking for casual computer users who may not be able to recognize or deal with such issues. We take no responsibility for any difficulties or losses that you may experience by using the information in this article, and we certainly take no responsibility for any damage that may occur to any person, place, or object. The risk is yours alone, and a little bit of caution won't hurt. Since we are only using a small subset of parts from the available options, this is merely intended as a baseline performance measurement. Finally, there is no such thing as a "guaranteed result"; you may or may not match the results that we achieve.
We started our recent look at overclocking platforms with the Venice 3000+ and 3200+ processors. Using a $130-$180 processor, we showed that it was possible to come near the performance offered by the extremely fast and extremely expensive FX-57. Actually, we came closer to FX-55 performance, but for about 1/6 the price, that’s still very impressive. Having looked at the Venice core, the next worthwhile upgrade to the processor has to be the move to dual cores. The change also comes with a major increase in price, unfortunately, and depending on the task, it may or may not matter. Current games show no benefit from multiple cores, so unless you plan on running some other CPU intensive tasks in the background while gaming, the upgrade may not be worthwhile.
We’re looking at the Athlon X2 3800+, as you can probably already guess. However, we started work on the benchmarks quite a while back and there are actually a couple of newer options that are potentially higher performing. The new Opteron 165/170 chips use the Denmark core, which is basically the workstation version of the Toledo. While the 165 comes with a lower default clock speed, we would venture to say that it has the potential to overclock just as high as the X2 3800+ that we’ll be using in this article, and likely even higher. (The reasoning stems from the way CPUs are binned and tested. Workstation/server parts undergo much more rigorous validation processes, and typically, this means that the parts have more overclocking headroom. Most server chips are rated extremely conservatively, as component failure is far more undesirable – and uncommon – than in desktop computers.) Combined with the increased L2 cache, you potentially end up increasing performance without spending any more money.
Rather than abandoning all of the benchmarks that we’ve already run, though, we’re going to present the results along with some commentary on the overall experience of overclocking the Manchester core. Also note that retail supplies of the Opteron DC parts may dry up in the near future. So, while we have confidence that the X2 3800+ will be available for purchase six months from now, we can't say the same of the Opteron 165.
Our last article contained a massive amount of introductory material, covering the various components that you need to consider when building a system designed for overclocking. We can safely skip all that this time, though newcomers might find it helpful to review the material. Most of our setup remains unchanged, but let’s explain that in more detail.
Note: This is the second article in a series of overclocking articles. Our first article looked at overclocking a Venice 3200+ chip using two different types of RAM, and it laid the groundwork for what we're doing in this article. We've updated some of our component choices as well as our benchmarks. The major difference is that we're now using an X2 3800+ for the processor, and we also upgraded to a 7800 GTX graphics card. We'll be looking at performance with several types of RAM as well. These articles are not targeted at the advanced overclockers, nor are we going for extreme performance at any cost. The main objective is to use some cost-effective setups and show what sort of performance level can be achieved.
As we mentioned before, there is a risk in overclocking and we do not recommend it for everyone. A system that appears stable for weeks or even months can suddenly have problems, so we definitely wouldn't recommend overclocking for casual computer users who may not be able to recognize or deal with such issues. We take no responsibility for any difficulties or losses that you may experience by using the information in this article, and we certainly take no responsibility for any damage that may occur to any person, place, or object. The risk is yours alone, and a little bit of caution won't hurt. Since we are only using a small subset of parts from the available options, this is merely intended as a baseline performance measurement. Finally, there is no such thing as a "guaranteed result"; you may or may not match the results that we achieve.
We started our recent look at overclocking platforms with the Venice 3000+ and 3200+ processors. Using a $130-$180 processor, we showed that it was possible to come near the performance offered by the extremely fast and extremely expensive FX-57. Actually, we came closer to FX-55 performance, but for about 1/6 the price, that’s still very impressive. Having looked at the Venice core, the next worthwhile upgrade to the processor has to be the move to dual cores. The change also comes with a major increase in price, unfortunately, and depending on the task, it may or may not matter. Current games show no benefit from multiple cores, so unless you plan on running some other CPU intensive tasks in the background while gaming, the upgrade may not be worthwhile.
We’re looking at the Athlon X2 3800+, as you can probably already guess. However, we started work on the benchmarks quite a while back and there are actually a couple of newer options that are potentially higher performing. The new Opteron 165/170 chips use the Denmark core, which is basically the workstation version of the Toledo. While the 165 comes with a lower default clock speed, we would venture to say that it has the potential to overclock just as high as the X2 3800+ that we’ll be using in this article, and likely even higher. (The reasoning stems from the way CPUs are binned and tested. Workstation/server parts undergo much more rigorous validation processes, and typically, this means that the parts have more overclocking headroom. Most server chips are rated extremely conservatively, as component failure is far more undesirable – and uncommon – than in desktop computers.) Combined with the increased L2 cache, you potentially end up increasing performance without spending any more money.
Rather than abandoning all of the benchmarks that we’ve already run, though, we’re going to present the results along with some commentary on the overall experience of overclocking the Manchester core. Also note that retail supplies of the Opteron DC parts may dry up in the near future. So, while we have confidence that the X2 3800+ will be available for purchase six months from now, we can't say the same of the Opteron 165.
Our last article contained a massive amount of introductory material, covering the various components that you need to consider when building a system designed for overclocking. We can safely skip all that this time, though newcomers might find it helpful to review the material. Most of our setup remains unchanged, but let’s explain that in more detail.
Facebook
Twitter
RSS
46 Comments
View All Comments
JarredWalton - Wednesday, December 21, 2005 - link
This may seem like a stupid question, but you did copy the SuperPi executable into two separate folders, right? Otherwise, the two running processes overwrite each others' data and one will always fail. Anyway, I don't find SuperPi to be a very useful stress test compared to Folding@Home, Prime95, and several other utilities; it just doesn't stress the system out that much IMO.Yianaki - Wednesday, December 21, 2005 - link
Yes of course it is in two folders. I realized that the SECOND time I did it Heh.Leper Messiah - Wednesday, December 21, 2005 - link
Thats actually a good thing, my X2 3800 does 2.65 at 1.425 vcore stable a rock. Looks like this x2 test is a good average indicator instead of most reviews which have the nice cherry picked silicon.JustAnAverageGuy - Wednesday, December 21, 2005 - link
Yeah, My Opteron 165 seems to top off at around 2.6GHz with the stock cooler.Araemo - Wednesday, December 21, 2005 - link
If you buy a socket 939 opteron, will it work in a normal NF4/etc mobo?A dual core opteron is tempting if it will work in the standard enthusiast motherboards. Get a nice heatsink and get it nice and toasty, I could turn my heater back off. :) And I hope 2GB RAM sticks go down in price within the next 9 months.. I'm still debating between a sweet laptop or a good overclocker desktop for my next computer, the desktop would be much cheaper, for sure, but it is a pain to take to LAN parties.
JustAnAverageGuy - Wednesday, December 21, 2005 - link
Most NF4 motherboards support the S939 Opteron, yes. Check the manufacturer's site to confirm though.Googer - Wednesday, December 21, 2005 - link
I found it a bit humourous that this http://images.anandtech.com/reviews/cpu/amd/athlon...">graphresembles a tent. It reminded me of the days in high school when kids would get fill in the bubble tests and use the answer sheets to do connect the dot drawings.
I wonder if Jarred had too much time on his hands?
kleinwl - Wednesday, December 21, 2005 - link
If you are going to start testing various cooling systems and how they affect max overclock... go ahead and throw in a Seasonic PSU as well. The Seasonic should be rejecting less heat into the case which may make as much difference in overclock as a more efficent Heat Sink. In any case... try it out please!<Note I have a XP-90, with a Antec SmartPower 2.0, on a venice... and I'm curious how such a case temp difference could affect the overclocking potential>
BigLan - Wednesday, December 21, 2005 - link
quote: Looking at the different RAM options, it's difficult to make a good case for spending tons of money on memory.I've always thought that spending a lot of extra cash on memory was a bad idea. It pretty much shows no improvement in Fear. It's nice to see a review of the everyday stuff.
Puddleglum - Wednesday, December 21, 2005 - link
The results for Fear looked bizarre. After reading some of the charts where 4xAA is used on games like Battlefield2 and FEAR, which would be a nice feature to show off on a high-end system, the numbers reveal marginal performance.I confess, I'm still using a Ti4200, which is only performing well in games because it's not drawing the DX9 stuff, and I've truly been waiting for an ideal video card to come out that's worth purchasing; but the new cards that are out right now are making it easy to sit back and wait for the hardware/software ratio to become a little more price-competitive.
Also, why is the OCZ PC4800 freaking out with BF2 when the CPU is overclocked to 2.1GHz?