Originally posted by: Le Québécois
I'm new to the overclocking business I just OC my Barton2500 to 3000 by doing 12.5x173mhz. At maximum cpu charge with prime95 running the temp reach 50C ( mobo at 21C ). If i try to boost up the fbs the cpu is unstable. Questions:
1- Is there any danger for OC my cpu from 2500 to 3000? By danger I mean the cpu burning itself up in one year or so...I'm low on money, that's why I just deceided to OC, I dont want to have to buy another CPU in 1 year.
2-Is there a way to overclock it more? I see the temp only goes up to 50C....I read around the web it could go up until 60C before getting error. Why cant i go highter? I know it has to do with the Vcore but I have no idea how set it correctly ( how do i know 1.7 is ok for 180mhz and 1.75 for 190mhz for exemple ).
I normally avoid posting on subjects like OC and PC builds, since my views on these matters tend to generate heated debates - in forums mostly populated by young males - that I'm not really interested in. But since you haven't got any other answers:
1: Yes, your CPU will get a reduced lifespan from OC. OC equipment tend to develop various strange quirks over time. However, it's not possible to say how long lifespan. A moderately OC CPU can last for years and years... Or it may not last much at all. (nVidias new FX6800 GPU seems rather sensitive to OC)
1b: 2500+ is an excellent choice for performance on a budget. You made the right choice. IMO you really don't need to OC it. The performance gains you will achieve are, IMO, not something you will notice much. The 2500+ already performs much better than its rating suggests. That is due to AMD somewhat 'inflating' later numbers, as well as the fact that Intel's P4s don't perform as well in 'real life' as popular, custom tailored benchmarks implies.
The experienced OC'ers will carefully check their systems with various benchmarks. You may not really get what you think you get, by OC'ing. One pitfall is the way the memorybus clock depends upon the FSB. If you set the FSB to a nonstandard value, you will get some change on the DDR-bus, and it won't be what BIOS claims, and it might be slower, resulting in a slower system from OC. (Or it might be faster, I suppose something like this could also be the reason for your system's instability, what kind of RAM are you using, and what memory clock do you *think* you're running?) Same goes for AGP-clock. Most straightforward way to OC a locked CPU, is to select a standard FSB clock (or close). Like going for 3200+, 200MHz on the FSB clock, then either 166 or 200 on the memorybus (DDR333 or DDR400).
Another pitfall, for Intel users, is the P4's tendency to 'throttle', when you ask much of it. Happy P4-OC'ers may cheerfully believe they have a kickass system, just because the resilient P4 happily accepts the outrageous clockrates thrown at it. In reality, soon as it gets some longhaul heavy work to do, it will kick down the internal frequency, and run slower. To the OC'ers delight: "It runs real cool too".
2. Well, yes maybe. I've already suggested one could try fool the chipset into believeing it's a 3200+. You will then surely get your 166MHz/DDR333 memorybus, if that's what you've got dialed in, and correct AGP clock. That is not a "moderate" OC, mind you. (But I've tried it briefly, and it worked. I experienced no strong reasons for continuing the OC though.)
But basically, I believe that worthwhile (performance) OC'ing is an expensive practice. Not a budget choice for the poor. If the performance gain is going to be truly noticable in practice, it has to be _real_ (which I strongly suspect is failing some OC'ers), and secondly it has to be greater than any 2-10%. For that, you need a select cpu with unlocked multiplier, and you need to try several, which is expensive. And you need some serious cooling, a freezer that will take the CPU down to -40C. That way you can accomplish 40-50% overclock. But mind you, it's an expensive hobby. Not something for the poor.
My advice to you is thus: Don't OC, run your 2500+ as such and enjoy it. It will cope with any game, provided you partner it with a decent videocard (R9800 128MB, FX5900xt upwards). On serious applications, you will mostly enjoy very, very respectable performance.
Veritest Winstone2003 benchmarks are excellent benchmarks for correlating to application performance that can be expected in real life. Winstone2002 contains less CPU optimizations, than can be expected in recent versions of mainstream apps. But still, relations between family members are valid. Check out the 2500+ vs. the 3200+ on this page:
http://www.anandtech.com/cpuch...c.aspx?i=1834&p=5
Overclocking successfully to 3200+ will gain you 10% increase in performance.
And this page:
<br
>[url]http://www.anandtech.com/cpuc...c.aspx?i=1834&p=6
[/url]
You stand to gain 8%, if you overclock to 3200+.
Is that worth all the possible problems and risks? If you're on a tight budget?
), when it has been fed numerical crunching for 10-15 minutes or more. And, quite frustrating, temperature indicates only around 50C. This too I came to suspect earlier, but the temperature readings reassured me. However, I finally timed a piece of work comparing to an A64 and another P4. Just to clinch the case, I resorted to the drama of blowing compressed air on the heatsink, resulting in half the computing time. That prooves it to me.
Facebook
Twitter