Poor OC with FX-53 S939

See specs in sig.

I can only get the cpu to 202x12.5 (2.52 GHz), stable. 205x12.5 is not stable, and i have yet to test between 202 and 205. Anyway, anandtech has gotten theirs to over 2.6 stable. i can get it to 2.6 but it gives bsods after about 30 mins of benchmarking.

any suggestions? voltage just increases temp. tried decreasing HT to 800, no avai.

how was anandtech cooling there chip. Also not all chips OC equally. You could have the exact same stepping as anandtech and still not overclock the same as anandtech.

very true. it's all dependent on the individual chip. it's not really such a huge difference anyway. i would think something was wrong if they were getting 3 ghz and you were only able to add a few mhz...

what was the fsb and mult on anandtech's chip? they might have a lower bus and higher mult to get the 2.6, or they might have a lower mult and a higher bus.

see about that, research it fully, and then see if you can figure out a way to get past that. if not, then a few mhz fsb overclock won't really get you much performance extra. that chip is already powerful enough.

Yes it could very well be the board or your mem or PSU or many other things. I would start with lowering bus & higher multi just to verify that it is mabye the CPU that is holding you back.


Jess1313

My 3000 is running almost 2.6 stable.. .. ..

One thing to realize, btw, is that your memory is running at only 194MHz. NEVER use half multipliers with Athlon64, all it does is cripple your memory, because CPU/memory half dividers don't exist. Your CPU/memory divider is currently 13.

That said, try out 13x200, because all in all, half multipliers just do strange things. Most FX53's on air that I've seen did roughly 2.7GHz, and used around 1.6v, usually no more, so its not surprising that more isn't doing anything.

Make sure you have the latest BIOS and then make sure your AGP/PCI lock is set to 66/33.

The lock is set, that was actually the first thing I set. And, yes, I know that not all CPUs are created equally, but I figured I could get close. In effect, I'm only 75MHz short of 2.6ghz, which, when you think about it, isn't bad, but in the overall picture, it's nice to say "hey, I have 2.6GHz" versus 2.52. It's more of a mental "illusion" thing that thinks 2.6 is much faster than 2.52 when it really isn't. With that being said, Anandtech had 13x204 (http://anandtech.com/mb/showdoc.aspx?i=2128&p=5) with HyperTransport of 800MHz (they reduced from 1000) and stock voltage.

I tried 13x200, just to get 2.6GHz (I also reduced my HT to 800). I used 1.6v, the comp crashed after about 30 mins and was running at about 155 during the stress testing. I have to wonder whether or not it crashed because of the temps.

Or, would it be more beneficial to jack up the FSB (such as anadtech, who were able to get 289). Not that I could get 289, as my ram would be the limiting factor as it's PC3700 (466).

Well, in the meantime, I'll be running some more tests, I'll try more combinations of the multi and fsb.

155 f means a mind-boggling 68 C. You should not even be thinking about overclocking with temps that high. I'm nearly postitive that that's whats holding your overclock back, and in fact, its impressive that you managed to get even a 100MHz overclock in those conditions. What fan are you running? It seems like your heatsink is not mounted very well, perhaps even improperly.

Or your case cooling is inadequate. What are your case temperatures?

Case temp (rather high actually): 90F
Hard drives: 85F

Case = Lian Li PC-65. The cooling definitely has a lot to be desired, especially with only one exhaust fan port. I'll have to modify that when I get a chance....then probably enlarge the top fan port to a 120mm fan. One of the good things that my previous case had was a fan on the side window that would blow right onto the motherboard.

I'm running a Thermalright SLK-948U with a 92mm fan pushing roughly 40CFM...I like a quiet case. I think it's not cooling as well as the SLK-800U (in my old Barton rig) because this 948U doesn't cover the entire CPU surface, which was very scary when I powered up, but it works. Oh, by the way, I overclocked because with stock speeds, my CPU was running around 105F idle and 120F load.

I managed to up the OC, btw. I got it to 12.5x203 for a stable 2.54GHz, I'm testing at 12.5x204 (2.55ghz) right now. By the way, at 2.55ghz, the CPU temp doesn't go above 130F.

I KNOW it's the voltage which makes it heat up as much as it does. The clockspeed makes it heat up more, but not nearly as much as the voltage does. For example, right now, my CPU has been testing for about 10 minutes it's at 126F (this is at 12.5x204=2.55) at 1.500v. At as little as 1.550v I notice a very large jump in CPU temps, instead of not going above 130, it's usually at 125 by the time i boot up, then jumps to 135-140 during load at the same CPU speeds.

Oh, by the way, I took off one of the side panels and low and behold, the cpu temp dropped by about 15 degrees and CPU temp may still drop, so one problem is definately Lian Li's crappy cooling. I love the case, but the cooling is amazingly substandard. Time for some modifications.

After about 5 minutes, CPU is running at 113F ***FULL LOAD*** (from a previous 128 with side panel on)

My ram is running kinda hot to the touch, is that how OCZ normally runs?? The ram's not even overclocked!!! My older corsairs (XMS 3200LL) would be mild to the touch at same conditions.

Case temps dropped to a mere 84F with side panel off.

Don't worry, it is covering the full die, just not the whole heatspreader. Same sorta scare happened with the Intels when people found that the Thermalright HS's didn't cover the full heatspreader. Rest assured, the actual die is much smaller than the IHS. Your memory running hot is perfectly typical. I would say its time to add blowholes or something of the like, because your case really does seem to be holding your temps back.

well there is one....rather small...80mm blowhole. I'm thinking about getting a 120mm template and use some sort of tool (any suggestions? maybe dremel) to cut it out. Then also add a blowhole in the back (for a total of 2). I donno what was going through Lian Li's minds when they decided to have only ONE exhaust fan port.

Only scary thing would be screwing up, there goes $100. I can do this with wood or some other thing, but metal...while maintaining a nice finish (w/o scratching it) will be difficult.

I've used a dremel many times to cut holes, but be careful because it's one way..........

Per Gautam's advice, I changed the multi to 12. I was able to boot up to about 220FSB, but windows would shutdown. I'm still in the stability testing phase, but it has lasted the longest so far at 12x212=2544. I have not changed the voltage (it's been at 1.5) which is probably why it wasn't stable at 220 fsb. Thus, I have a feeling 2.54 GHz is the maximum stable yield of this CPU, remember earlier it wasn't stable at 12.5x205. The only thing I haven't done is increase the voltage. Since I know that raising the voltage is the major cause of my heat problems, I will save that for last. If I find that the CPU runs stable at the current speed (12x212), I will increase the voltage and try 12x215 for starters. If that is actually stable, then time to increase fsb until its unstable and then increase voltage; and then stop when I am happy, it maxes out, or voltage goes above 1.65v.

Anything above 1.65v makes the temps skyrocket to an unbelievable 150 idle cpu temp. Even 1.55 gets very, very hot. At this rate, I should probably go with water cooling, but too risky....water + computer = not good. I suppose I could coat everything with some special barrier thing that I heard somewhere.

I said try out 13x. But I think that the first thing you should do is remount your heatsink, as your temps are still very high. I run at 1.7v, and idle at around 28C or 66ish f. Load is just around 40 or 100 f, with nothing more than a 948U. I think your TIM just isn't applied as well as it could be, and this can be a matter of 10 to even 30 C.

For instance, when I put my mom's computer together lazily, I got load temps of around 70C. Redid the Ceramique and they plummeted to around 44C.

Watercooling is incredibly safe as long as you take your time with it, though.

I agree those temps are way too high. You are using thermal paste right? Might be a stupid question, but Ive seen it before...
I also had to mod the cooling on my LianLi PC-60. I cut an extra 80 hole on the back directly below the factory one (the fans look stacked), its close but it will fit. Also i did a side panel fan blowing straight onto my ram. My case runs in the low 30's (c) now and I can hit 2.6 Ghz which was tough prior too the mod's .....

I know you said 13, but 13 doesn't work because the minimum fsb i can use is 200; and that would make my CPU run at 2600, which is not stable.

Yes, Arctic Silver 5 is being used.

Gautam, you're using a Termalright SLK-948U? Wow, if that's the case, then my HS is, without question, mounted incorrectly, which is odd...it installed pretty nicely; I probably just need to put more AC5 on there...or maybe I have too much. Eh, I doubt I have too much.

Well, I'll fix the heatsink and give you an update later on.

You should understand that silence and normal high overclocks don't exactly go hand in hand. I allways have silence as a main criteria, and have to keep my overclocks about 10% lower then they would be with a normal soundlevel setup. My experiences with my older rig with a 1700 dlt3c processor and slk800 heatsink:

Processors fan 80mm panaflo L1A @ 7v
case fan 80mm panaflo L1A @ 7v

Max overclock 2.1ghz

Raising the fans to run at 12v directly allowed me to gain about 150mhz and add 0.5vcore. Changing the processors fan to something louder that moved more air and adding another panaflo to the case gave me another 0.5 vcore and gain almost 100mhz overclock.

Alot of people getting good overclocks have far more extreme cooling methods then I had on my processor even at the loudest. Actually the highest overclocks on air usually come from computers that sound like aeroplanes.