Slashdot Mirror
AMD Optimal BIOS settings + Overclocking Guide
JMke writes "Here's a step by step guide on how to get the most out of your AMD setup. Overclocking tips and BIOS tweak settings discussed, as well as an overview of the more popular overclocking tools. Start your overclocking here!" Lots of good info here for getting the last bit of performance out of your system while also watching out for dangers that could fry your processor.
... yet one of my systems constantly locks up even if midly overclocked... Do I spend $100 extra or do I spend 12 hours mucking with my BIOS?
Well, here's the thing: first, you have to have good parts. Unless you have good quality RAM, a properly cooled processor, and a decent power supply, you probably won't be able to overclock very much. To your second point, you can get a 10-20% increase in speed by overclocking, and once you know most of the tricks to overclock, you can probably fix your machine up in about an hour.
The point is though, that if I spend $500 on a (ficticious) 2GHz processor and $100 on cooling so I can run it at 2.1GHz, does it really make sense to do this when the 2.1GHz processor is available with a crappy (but adequate) fan for $550?
The chips are tested at a DIE temperature of 85C. However, the temperature sensor on you MB isn't at the die, it's at the packaging (or maybe just near the packaging), so it's not going to read as hot. A rule of thumb I've read is a 20 degrees difference, so that would mean 65C.
:)
Unfortunately, this is all academic because motherboard temperature sensors are notoriously non-calibrated, and I've seen them be 10C off or more! It's more useful to use MB sensors to monitor relative changes than absolute temperature. (In other words, the difference between temperatures running at idle and at load is probably accurate...but the absolute values are probably fuzzy).
Additionally, the fact that the chip was tested at 85C doesn't mean that it will be harmed by temperatures over 85C; it just means it's likely to not work (a bit won't flip somewhere on the chip). Your system will eventually lock up if this is happening. They used to test some chips at 95C or more, and military use chips get tested even higher, so that shows that the silicon chip itself won't be damaged at those temps...it's just functionality that will suffer.
If your chip is marginal, running for long times at higher temps can cause it to die quickly (atom migration, etc, can be more pronounced at higher temps) so some people will warn of running at very high temperatures. But this is very unlikely, and if it does eventually happen, hey...now you need to upgrade. Sweet.
What if you get a $100 processor (barton 2600+) spend $50 on cooling (Slk 947 U plus fan) and overclock it to the performance of a $250 (barton 3200+) processor with no problems? The same can be said of video cards. It seems that cooling has a higher ROI than just buying the better chip.
Photos.
And alot of people that don't do heavy gaming or graphics processing are right. Think about this . A majority of us use 56k dial up to get on the net but what would happen if the dream of 'Fiber to the home' came true ? We would get 100 megabits of data a second ,enough for HDTV on demand and an alternative to the other monopoly that is cableTV(great evil) . That would be THE day that would signal most of our computers as extinct. Alot of our present computers couldn't even come close to handling that kind of data. That would be a HUGE reason to upgrade to 64 bit and faster computers and a reason to overclock. The computer makers should be organizing to roll out 'fiber to the home' along with municipal governments in the best interests for themselves so they can sell new generation MEDIA pc's and for the public.
The disc is the slowest point, add plenty of RAM as buffer, but nothing makes up for having a fast disc.
15,000 rpm, 3.6ms access time, 8Mb onboard buffer. And an HBA to match.
I find it ironic that people buy cheap systems with slow discs, slow network and insufficient RAM and then try to make it faster by overclocking the CPU.
Government of the people, by corporate executives, for corporate profits.
Because once the application is loaded from hd to ram it's the cpu & ram speed that matters, not the hd speed.
Let's not forget that very few apps even need fast load speeds, even games don't need it. A modern IDE hard drive offers 50+ mB/sec speeds, that's fast enough to load even the largest games quickly, only when video editing would faster speeds be desireable. And hard drives are not like CPUs, their are no hard drives that offer double or triple the transfer rate of the inexpensive drives, your fastest 15k rpm drive might offer 20mB/sec more, which isn't worth the extra $200 and 80+ gig sacrifice to many people, especially when that money could be better spent on faster cpus, video cards or memory.
A fast cpu plus lots of ram paired with a slow hd is like a speed reader who took a few minutes to find the book vs a "metally challenged" person who found the book immediately. Which would you rather be?
my karma will be here long after I'm gone
15,000 rpm, eh? Bet it sounds like a jet taking off. Honestly, yes, the disk is the slowest thing in your system. In most of my usage, however, I don't use the disk very much. What do you need it for? Well, there's launching programs, loading the binary into memory. I do that about once and then leave the programs running all day. I'm satisfied with the launch times, especially given how easily and quickly I can switch between processes that are already in memory.
What else is there? Every now and then you need to load or save a data file. If you are doing multimedia where you have a lot of raw data to read and write off the disk, having a super fast one would doubtless make a huge difference in performance (provided the system bus could handle it, but that's another matter). But not for me, my files aren't very big and I don't perform these operations nearly enough to care.
Now, what would make my system appear much slower would be if I had less RAM, causing virtual memory (swapping fast RAM to much slower disk) to be necessary. That would really make things grind to a halt. Yes, I'm trying to make a point here. Maybe for you that 15,000rpm disk is important enough to be able to justify the cost, but the first performance upgrade most users are going to need and will be able to see immediate results from is adding another stick or two of RAM. It's cheaper too!
I think for most people a disk that fast is overkill, just trying to shave a second or two (maybe a lot less) off an infrequently performed operation. Sure, it feels faster using it, and maybe that improves your mood or something, but perceptions aside, it really isn't that much faster. It would be nice if you had a computer that could do everything you could ever require of it all at once, instantly, but back in reality you have to ask yourself how much of a premium you are willing to pay for a tiny pinch of time.
The only people who would find this article noteworthy are people with AMD processors on Abit motherboards (only company with the softmenu bios option to my knowledge) that have the via KT266a or KT333 chipsets.
Many of these bios settings are based on these boards in particular, most of the time other motherboards don't have these settings, or they're called something else. It's a shame the article is written for Abit boards, because generally if you've gone out of your way to buy one, you already know a bit about overclocking.
There is some info here if you've got an Nforce2 board, but some of it is wrong. The article says to run memory at 166mhz, but these boards can run at 200 (or higher).
Are overclocked CPUs more sensitive to heat, or is this just a "to be safe" recommendation? AMD says they're good up to 85C.
Overclocking is all about temperature.
So long as you don't mess with the Vcore, overclocking does not make a CPU more sensitive to heat. It just will get hotter. If you can keep the temperature of the die under control by dissapating the extra generated heat, there's no problem with overclocking, but it's instrumental that you have accurate temp readings.
I'm only familiar with Intel CPUs and they have an on-die temp sensor which generally is fairly accurate. The embedded CPUs I work with are rated up to 100C and generally work fine up to that temperature. I wouldn't want to run it continuously at that temperature for a very long time though.
Overclocking may result in a loss in stability but that generally doesn't mean you are about to damage the CPU (so long as the temp is not out of spec), it is more likely related to timing issues.
Now increasing the Vcore sounds like a very bad idea to me. I'm not an EE by education, but I would think that you could seriously shorten the lifespan of the CPU by doing this.
btw. What I don't understand in the "NEVER LET IT GET ABOVE 55C IN MY OPINION ON AIRCOOLING, OR 42C USING WATERCOOLING." statement is why the hell would it be depending on what type of cooling is used? This implies that they are not talking about die temperature, because for the die 55C is 55C, doesn't matter how it got there...
Overclocking is stupid. Especially since it started costing more than buying a faster CPU.
That depends entirely on what you're doing.
I very much doubt that most overclockers are clueless enough to have too little RAM (although I do know that many people upgrade their entire system when they start to feel it is slow when getting more RAM would probably solve the problem), so I'll just ignore that part of your argument.
If you're running games or number crunching, you're almost certainly bound by CPU speed and the speed of the memory bus (and for games, the GPU).
If you're compiling software...it depends, but anything big seems to be cpu/memory bound (especially in languages such as C++ that compile slowly).
Interactive use is often disk-I/O bound for program launches, but once it's up and in RAM...most of the time, only if it depends on network I/O.
If you're doing video editing, then yes, you're probably disk-I/O bound.