When I overclock, I do it not to have the best and fastest system, but to maximize the bang for my buck. Don't get me wrong, I'm interested in having a fast sytem, but I would never for example, by an Athlon FX just to have the unlocked multiplier.
So I have recenlty begun to think about building a decent system around based on an Athon 64 value overclock. On the other hand, I really would like to look into a 939-based Nforce 4 system because I want to get a Geforce 6600 GT AND I like the idea of Native Command Queuing for my HD. Are the desires for overclocking value and platform features incompatible? I feel that the 754 vs. 939 complicates the situation when it comes to overclocking as well. As we have seen, the 90nm 939 3000+ overclocks very nicely and I would be pretty happy to get it to a nice and conservative 2.4 running at 266 fsb. I also understand that I can probably get the same speed out of a 130nm 2800+ on 754. Furthermore, since pushing HT past 1000MHz seems iffy, I would probably have to set the HT multiplier to 3x which gets me 800MHz. This is perfect for 754 but underclocked for 939; not sure if it matters. But what might matter is the fact that 939 has double the memory bandwith. I'm not sure how much of a difference it makes at clock speeds as low as 2.4, but I'm sure it will be more important as clock speeds scale up. Also regarding memory, I have been planning to get value PC3200 at CL2.5 to keep costs down, and just use a memory divider to get ~200 MHz. I figure that I will get at a minimum, the same performance as a stock system clocked at 2.4. I understand that spending more money on faster memory helps because latency if effectively reduced (bandwith is maybe not so important). One more thing I'm worried about is that for 754, I think that means that I will have 2 memory modules (2x512MB) loading the memory bus so I may have to set the command timing to 2T and I know that this kills performance. This wouldn't be a problem with 939 because there is only 1 module loaded on each channel.
So, what do I do? I'm pretty set on getting a value processor, but the other variables are 754/939, AGP/PCIe, value/fast memory. I'm not exactly sure how I want to prioritize bang for my buck vs. platform features and future upgrade potential. I am ALWAYS torn between the two. There never seems to be a good combo that hits both of these goals. What is recommended for each of these goals? IS there a reasonable compromise?
So I have recenlty begun to think about building a decent system around based on an Athon 64 value overclock. On the other hand, I really would like to look into a 939-based Nforce 4 system because I want to get a Geforce 6600 GT AND I like the idea of Native Command Queuing for my HD. Are the desires for overclocking value and platform features incompatible? I feel that the 754 vs. 939 complicates the situation when it comes to overclocking as well. As we have seen, the 90nm 939 3000+ overclocks very nicely and I would be pretty happy to get it to a nice and conservative 2.4 running at 266 fsb. I also understand that I can probably get the same speed out of a 130nm 2800+ on 754. Furthermore, since pushing HT past 1000MHz seems iffy, I would probably have to set the HT multiplier to 3x which gets me 800MHz. This is perfect for 754 but underclocked for 939; not sure if it matters. But what might matter is the fact that 939 has double the memory bandwith. I'm not sure how much of a difference it makes at clock speeds as low as 2.4, but I'm sure it will be more important as clock speeds scale up. Also regarding memory, I have been planning to get value PC3200 at CL2.5 to keep costs down, and just use a memory divider to get ~200 MHz. I figure that I will get at a minimum, the same performance as a stock system clocked at 2.4. I understand that spending more money on faster memory helps because latency if effectively reduced (bandwith is maybe not so important). One more thing I'm worried about is that for 754, I think that means that I will have 2 memory modules (2x512MB) loading the memory bus so I may have to set the command timing to 2T and I know that this kills performance. This wouldn't be a problem with 939 because there is only 1 module loaded on each channel.
So, what do I do? I'm pretty set on getting a value processor, but the other variables are 754/939, AGP/PCIe, value/fast memory. I'm not exactly sure how I want to prioritize bang for my buck vs. platform features and future upgrade potential. I am ALWAYS torn between the two. There never seems to be a good combo that hits both of these goals. What is recommended for each of these goals? IS there a reasonable compromise?
Facebook
Twitter