DragonClaw

I remember the days when Linus was still working for NCIX. It's great that the chance he took paid off and now he's here to bring us top-quality tech content!

Nice to see yet another international giveaway! TechQuickie favourite video (one of many): Jay's video:

Here's a real blast from the past. These are parts from my old socket A (462) PC that were just lying around, so when I had some free time I decided to put them back together, see how far I can overclock the CPU ('cause at the time I was using it I didn't know what overclocking was) and benchmark it. Specs: Athlon XP 2200+ Thorton core (1.8GHz) @ 2.45GHz - pretty nice overclock for that chip, I did try unlocking the extra 256KB L2 cache, but it didn't work :[ Radeon 9550 768MB of DDR RAM AsRock K7VT6 motherboard And here are some Cinebench 2003, Super PI and wPrime results: Not that these numbers should mean anything to most people. After all it's been more than 10 years since the last socket A CPU was released.

Them back buttons! Back when the G2 launched, all I needed was 5 minutes to play around with it to establish that it was a brilliant idea.

Anybody?

I've been running my PowerColor 7870 Myst at 1200MHz core clock (1.262v) for quite some time. It's was 100% stable. Then I decided to change the thermal paste because I hadn't done it for a while. I cleaned both surfaces with alcohol, applied the new TIM (which is the same as what I had been using), assembled the card and ran some Unigine Valley to make sure that the temps were OK (i.e. the paste was applied properly). And they were, the GPU never goes above 70C under load. However, my card just won't run at 1200MHz anymore. Artifacts appear on the screen seconds after I start Unigine Valley and if I let it run for a little longer - it crashes. I took off the radiator again to check if there were spots that are not covered with paste - there weren't. Since then I've reapplied the TIM three times with no success. I should also say that the card is stable at it's stock clocks. Could I have done wrong? Thanks! P.S. Those Unigine artifacts are hilarious, though. Check them out:

I have a 4-pin fan from a stock AMD cooler plugged into the 4-pin CPU_FAN header on an Asus Crosshair III Formula. I've enabled Q-Fan function in the BIOS and installed AI Suite to be able to control things from Windows. So here's the problem - whatever I do, my CPU fan is running at a constant speed. I think the problem originates from the temperature sensor AI Suite uses. Right now it's reading 44C CPU temp, but it's actually running at 33C (confirmed by AOD, Core Temp and HWInfo). How do I fix this? I don't want to run the fan at 100% all the time and if I don't, my CPU would fry whenever a more intensive program/game is running. Thanks!

Oh, man! Forget about it, everything works like a charm now. As it turns out I need higher LLC to make the CPU stable at higher frequencies. I did try 'High' before posting this topic but it didn't help at all so I just ignored it and never tried an even higher LLC setting. I guess I was afraid of how much more voltage it would feed the CPU. Many thanks to all who made time to help me solve this and to Priller for suggesting a higher LLC setting. Currently running 3.6GHz overclock, which passes Cinebench, wPrime and SuperPI. Prime95 stability test coming up. Then all that's left for me to do is buy a better cooler and see how much more I can squeeze from my unlocked CPU.

OCCT reports 1.41V at idle and 1.42V under load. Given that the figures are rounded to two digits, I'd say OCCT and CPU-Z report identically. By the way, 3.1GHz was stable before for all six cores with just a multiplier bump (before I started messing around with FSB, RAM and other speeds). So I don't think there was any point in what I've just done. Did you mean 3.3Ghz instead of 3.1?

How do I measure the actual voltage? The board doesn't have Vcheck points. AOD, Core Temp and HWINFO report 1.4V all the time, while CPU-Z says 1.408V at idle and 1.424V under load, if that is of any use.

Done. Now what? EDIT: I don't have APM. I used to have it under Advanced BIOS Features when I had an FX CPU, but now it's gone. I don't have HPC either.

No. The test in the picture above was done at 8-8-8-23 timings and 1.5V. Are you saying that the RAM might be what causes crashes when all six cores are enabled, while being perfectly stable with 4 cores or less?

I have 2x2GB of A-Data DDR3 at 1333MHz. Since I had to raise the bus speed (because the CPU doesn't seem to like multiplies over 16x), I set it to 1066MHz in the bios to make sure it's not running out of spec.

So I've had some more time to mess around with different settings and I've discovered a relation that I'm not sure is normal: the more cores are enabled, the lower the max frequency I can achieve. Before that, though, I found which were the worst cores by booting the PC with Core 0 (because it can't be disabled) and one more core enabled. For each combination of two cores, I ran SuperPI 1M, wPrime 32M and Cinebench R15. The only combinations that crashed during these tests at 4GHz 1.45V were the Core 0 + 4 and Core 0 + 5 (the only two combinations with unlocked cores). I know - probably not the most scientific way to do this, but one would think that the least stable cores would be the locked ones and my test confirmed that. For my next test, I decided to run Core 0 + Core 4 + Core 5 (the default core and the two unlocked ones) at 3.6GHz 1.4V (I know, quite a lot of voltage for such frequency, I just wanted to make sure the CPU has enough juice) - a setting which is no where near stable when all six cores are enabled. Here's the result: The other three cores that are disabled in this picture are three out of the four which are enabled on the X4 960T. They would do 3.6GHz at 1.4V just fine so I'm not going to bother to test them separately. So clearly, the fact that I cannot go above 3.2GHz with all six cores and have it be Prime95 stable for more than one or two hours is not due to bad/defect cores. I'm running out of ideas, please help! I really want to run this bad boy at 3.6-3.8GHz. EDIT: The VRM and CPU temperatures are not causing this. I'm monitoring them with thermocouples and they're both at reasonable temperatures.

Whoa, you got yourself a great deal! I knew I was taking a bit of a chance by buying a CPU that I intended to be unlocking so my hopes for a good OC were not too high. However, I did find it very curious that the CPU outright didn't want to go over 3.2GHz at any voltage so I posted this topic to ask if was doing something wrong. At least now that I'm at my max frequency, I don't have to worry about buying a better cooler, hehe. (unless I wanted to lock it again and OC the hell out of it, which I dont, 'cause I like seeing more colors on my All CPU Meter gadget :lol: ) Fingers crossed that the new x86 architecture they're designing from the ground up would be awesome.