Wah, I don't online for one day and buta already 5-6 pages from nothing..

The world record is 1THz, but its using a prototype processor builts using Si-Ge, and it cannot run normal program code. It was cooled with liquid helium (FYI, its much cooler than LN2, and several times more expensive). Read about this in in-tech quite a long while back. Maybe the article can still be found in google.

Not really, it also depends on the environment. If the air is perfectly dry, condensation will not happen.

Did this once, but had no camera to take pics..OCed my Pentium 4 Northwood to 3.7Ghz on almost stock voltage. Wanted to go higher but ice melted super fast-damn pentiums are so hot..

The PCB can easily tahan 200C, but the ram chips..no way man..if they're too hot to hold on to continuously (at least 10 sec), then its too hot (~60C).

About heatspreaders..if you actually google up comparisions, most heatspreaders are actually not effective at all, in fact most of them actually trap heat resulting in higher temperatures. The real advantage of heatspreaders is actually cosmetic, and also prevents physical damage to the chips. BGA chips are pretty easy to break.

You can mod the pump with barbs that match the tube you're using.

Played with a peltier once too..but couldn't pump enough power to make it effective at full load. But even then, I had hell with condensation the moment I stopped prime, temps dropped faster than you can say oops..

6000L/H..damn you can wash car with that..

The TLB errata applies to all Stars microarchtecture chips, not just in virtualisation (though its particularly problematic in that area). The errata is not caused by clockspeed, its due to circuit design and it will cause hangs at any clockspeed. Currently AMD has employed a microcode fix that works around the errata, but at a noticable performance penalty. My guess as to why AMD held back their highest end chips is because it makes no sense to release high end chips when you know you cannot get them to run at their fullest potential due to the errata workaround.

Try everest. Never failed me in identifying sensors.

Not likely. PWM temp is usually listed as AUX. System temp is almost always the ambient temp of the air near the motherboard.

That should be it. First line is probably the ambient temp measured at the surface of the motherboard (aka system temp). Your reading looks weird because of sensor inaccuracies and also because your system is idling. Normally processor temp and PWM is higher than system temp, but sometimes it just happens the area near the sensor has poor airflow. Also, the monitoring chip on most motherboards is not well known for accuracy.

It will be hard for a brisbane to compete with a 6000 even at 3.0Ghz, but maybe with a 5600. The problem is that brisbane is at a 0.1Ghz disadvantage due to cache latency, and also those 2 processors have 1mbx2 cache while brisbane only has 512kbx2. Clocked 0.2Ghz higher than the 5600, brisbane might just pull up even with it.

There is a difference, depending on how flat the IHS and heatsink surface originally was. Shininess comes secondary, not much effect between a perfectly flat, unpolished surface and mirror finished, also perfectly flat surface. I did a comparision a few years back while using my first 478 system, got a 3-5 degree drop. With the addition of a direct intake duct I could keep the processor within 5C of ambient at all times. And that was a processor with power consumption similar to the lowest end AM2 X2, 65W version.

I will be doing a long term comparision using some K8 stock heatsinks, will put up my results in a few months time (have a lot of comparisions to do).

Yeah, with AIR-con

Hmm, I'm actually considering removing the IHS..but cannot..even though have no warranty to worry about..
Using XP-120, it will surely crack or chip the core when I mount it..if waterblock then no worry la..

Well, the 1mb cache of the 6000 is nice..but price
Its cooler and has lower wattage cos its only single core..believe this or not, you can actually run most K8 single cores without a HSF, as long as you have a fan blowing on the IHS and do not load the thing for more than 15mins at once (enough to boot up cleanly). On idle it will not overheat, especially if CnQ is on. How I know this? It took more than 5 mins for my former F2 to reach 65C on idle when I removed the HSF during an experiment, and I didn't even have a fan blowing on it. Proves that the IHS is really doing its job, in absence of a HSF it prevents the processor from overheating rapidly, it becomes sort of a mini heatsink.

Something like that..don't try it on those Athlons without IHS though..

Acecombat is selling his off, from dblooi's bulk..
Last chance to ever get one..

It still sort of works out, but the image quality thing now a bit hard to say if gaming. But in video & HD decoding, ATi pawns Nvidia big time.

Boss, you didn't correct my stats in the list

3800+ X2 Windsor F3 | lohwenli | Abit NF-M2 nview | Crucial 10th Anniversary | Thermalright XP-120

I've never used the NF-M2S in any of my own PCs and probably never will.

This post has been edited by lohwenli: Dec 22 2007, 09:47 AM

I didn't receive your reply, guess it got lost or something. But thanks anyway.


Never mind..we'll remind them for you.
Right guys?

I'm not sure, but I have an experience which points to the possibility. Back when I was underclocking my F2, I pushed the vcore as far as I could go, but the processor quickly got unstable at anything but the lowest speeds. The strange thing was, at the unstable point, the ram was heavily underclocked, but the stability tests that it kept failing had everything to do with ram. Stability tests that didn't stress ram could pass for days with no problems.

Here's a quick breakdown. At stock speed, I could run at around 0.950V and at least boot to windows, but only around 1.025v would it pass CPU stability tests happily BUT it would still fail memory tests regardless of what vDimm I pumped (I didn't touch timings though, ram running at stock timings & frequency). Only at 1.125 would it be 100% stable.

Dunno about temp, but the power consumption is
When both are at 3Ghz, a Brisbane G2 takes about half the power of a Windsor (less than 80W). Use stock heatsink also can overclock quite high for G2. Only one downside, must overclock 0.1Ghz more than Windsor to match Windsor's performance.

Supposedly yes, the higher latency is allowing it to clock higher, but it doesn't really seem to be like this for AMD's 65nm procs. Brisbane doesn't overclock better than Windsor, even G1 loses to F3 (only G2 pulls up even). I suspect it was due to design issues, the original cache controller design couldn't work properly under the different characteristics of smaller transistors. Its most likely, since Phenom's main issue is also the cache controller.

Its true for C2D & C2Q, but for AM2 its more like one Phenom = 1.5 Athlon X2 power consumption. Power consumption per core dropped quite a lot in G2, then Phenom.

Not really, PI is used to compare single core performance using basic instruction sets. Many applications still aren't multithreaded (especially games) and a fair number of apps don't really benefit from SSE so PI reflects performance in those applications.

While you're at it, you might as well remove the IHS..save all that trouble lapping..and get better temps too..
But don't do it if you're gonna use a clip on air cooler later on..only screw on coolers..clip on coolers are core crushers.

If you're looking for a value for money setup by overclocking, then you're planned setup has some major mistakes..for a start, no wise overclocker will buy a top end chip unless there's something special about it, which for most cases including this one there is nothing special. All chips of the same revision and core code have the same design, so in your case any chip using a Windsor F3 with 1mbx2 cache will overclock similarly to the 6000, while being a lot cheaper (check here for a list). Board, Asus has traditionally been overpriced, go for something like gigabyte or biostar. Biostar is a good choice for overclockers, almost as good as Abit in features & price but without the quirkiness of Abit (Abit boards usually have some weird 'perangai' somewhere, but are the best value for money in terms of OC functions for their mid to high range boards).

Another thing, don't dream of overclocking a 6000 on stock cooler, its already quite hot at its stock speed of 3Ghz (~100W). And the typical limit for a Windsor F3 using decent third party aircooling in normal, non-aircon room is 3.2-3.4Ghz, there will not be much noticable difference from the stock speed.

It will run in single channel mode. You must pair all your ram to run dual channel, and anything more than 2 sticks (1 pair) will impair overclocking.

I knew that the new Stars architecture has a damned good memory controller, just that I wondered why it was losing in memory benchmarks. Now I know why.
What's really scary about it is when its on a dual processor system..Ram benchmark more than 10000 MB/s..

Its all part of the fun..speed trills..some OCers are like that, others do it to get high end stuff at cheap cost.

A single core 3.4Ghz might actually be a good choice for gaming, since quite a lot of games in the past year are starting to get processor bottlenecked again after a number of years of graphic bottlenecked games (thanks to gameplay physics). Many games don't use multi cores at all, and many of those which do it pretty inefficiently, not much improvement from single core.

I had that problem too..realised why after reading anandtech review on Penryn, which includes an explaination of why vdroop is good (shocking, isn't it?). My mobo (NF-M2 nview) has vdroop disabled, which I've come to realise messes up CnQ and underclocking, because on idle the voltage regulators stop boosting the voltage (it boosts on load to counter vdroop) and the voltage sometimes drops to an unstable level without me noticing it.

Read the article here
http://www.anandtech.com/cpuchipsets/intel...aspx?i=3184&p=5

Its too bad, but at least they've learnt their lesson. I think Phenom 65nm will not be around for long, and AMD will quickly move to Deneb (45nm), which is already in initial production. Deneb will have more cache, and should be able to clock higher. If AMD doesn't mess this up, then AMD will have a fighting chance end of 2008. Only problem now is funding, AMD is running a loss for quite a while. There's been talk of companies possibly buying up AMD like how AMD bought ATI, since AMD/ATI current value is actually less then what AMD bought ATI for.

I don't think it would make much difference in memory benchmarks, but maybe graphics benchmarks. Memory operations don't use the HT link.
Btw its 8 bit vs 16 bit, not 8x and 16x.

I expected as much. AMD's heatpreaders are extremely flat compared to Intel's 775 procs.

No. Its 16x only if there are 2 GPUs or less, if more then it will split one or both of the 16x lines into 2 8x lines.

Kind of sayang very few 790FX mobos actually have slots for 4 GPUs. But then again, very few games can make use of even 3 GPUs. There's an interesting article in anandtech

The 8-pin CPU power connector is different from the 8-pin PCI-E connector, you won't be able to force it in even. However it is possible to rewire it to be useable for PCI-E. Just change the CPU plastic plug with the PCI-E plug.

Well no, if running Crossfire with 2 cards then its 16x 16x PCIE 2.0. If crossfireX with 4 cards then it drops to 8x 8x 8x 8x PCIE 2.0. This is explained on AMD's website (on the page about the 790FX chipset).

10C..at first I thought, no way..
But then I saw you're using watercooling..no wonder la..

@meno managed to overclock one to 2.9Ghz on a mobo without voltage adjustments. No screenshot cos didn't have time to install windows before he left (he's staying in penang, the system is at his hometown).

Wow, yours is pretty uneven before lapping (based on the difference caused by lapping). None of the X2s I've used have been that uneven.

Damn it man, those board makers are really lazy..its not that the board can't run Phenom, but they're taking their own sweet time to release bios updates allowing Phenom to be used on AM2 boards..

I've seen the OCZ reaper do some crazy overclocks..boy does that ram love voltage..2.6v is common, and some have even run it as high as 2.8-2.9v.

Oh btw, 1350 is a typical overclock for the reapers, as long as you're willing to pump the voltage (but still at a 'survivable' level). Timings can be fiendishly tight if you're willing to settle for a lower but still 1000+ overclock.

One was done by one of the forumers, its inside the same subforum (OC united). I don't think it was a review of the ram itself, I think it was a mobo review. I can't remember where I saw the others, but they were all done by extreme OCers which did not mind running damn high voltages. What's surprising is that the reapers survived..

Btw, I cannot guarantee that the current reapers use the same chips as the ones I've seen, so current results may vary.