Do you dial your system up to 11?

J

jimbowdang

Limp Gawd
Joined
Jul 26, 2008
Messages
237
I was tinkering around with my overclock the other day and bumped it up another 400 MHz to 3.8 GHz and I started to wonder, I know there are plenty of overclockers around these forums but for us DCers what kind of overclock settings and temps do you feel comfortable running 24/7?

My system peaked at about 80C while running LinX and obviously F@H doesn't push it that high but I'm probably going to try and lower my CPU voltage a bit to lower temps or maybe even add another fan.

As far as my GPU goes I leave it at stock because it seems like every OC I've tried on it will start crashing the display drivers.



 
How high? Stock - it's stable. Temps? Under 70 on GPUs, 27 on my 260 folder
Posted via [H] Mobile Device
 
its good to test with linX but ultimately the only temp that matters is what FAH pushes it too.. for me i push everything to the bleeding edge of whats stable.. but thats because i like getting everything possible out of what i have.. i just wish i had a motherboard that would allow me to push my phenom II 940 to 3.8ghz.. but i guess i cant complain since it was a straight drop in on a 70 dollar motherboard thats not even suppose to support this processor to begin with.. thankfully biostar screwed up and released the pre-lease bios for this board which didnt have the 95w processor limit..

in the end gpu overclocking is far easier and much simpler to do then cpu overclocking.. basicly my way of overclocking gpu's when i cant find numbers for them is running Furmark.. then slowly raising the gpu/shaders(linked) in 54mhz steps on the shaders every 20-25 minutes with furmark running.. then watching the spinning donut for artifacts or if the drivers crash i drop it back down 54mhz and let it run for about an hour to see if its stable.. 1 note.. you can go higher with your overclock on your card running FAH gpu client then you can actually game at.. for example my 8800GT was stable folding at 760/1890/1050 but for games i had to drop it down to 735/1836/1015.. same goes with my GTX 260.. i can push the shaders up to 1512 for folding but i have to drop it down to 1456mhz for gaming..
 
I generally don't like my Q6600 over 70. It used to run right around there but I dusted/turned up my fans slightly and haven't checked until now. Right now the highest I've seen is only 59 :eek:


Maybe I should consider upping my overclock a bit...
 
If it's a rig that I tinker with on a regular basis, it will pushed to the max of what it is stable (within reason)

I'm not about taking voltage too far above stock. (I killed a Socket A chip back in the day)

That said, I've got almost a 1ghz OC on both my main rigs.
 
Kendrak said:
If it's a rig that I tinker with on a regular basis, it will pushed to the max of what it is stable (within reason)

I'm not about taking voltage too far above stock.

That said, I've got almost a 1ghz OC on both my main rigs.
Click to expand...


as long as you stay within the limit by intel/AMD recommendes for the max safe voltage anything between thats perfectly fine and wont do any damage to the cpu..
 
sirmonkey is correct regarding the CPU. However, you may want to try a few rounds with StressCPU2. It executes a special version of the Gromacs innerloops that mixes SSE and normal assembly instructions to heat your CPU as much as possible. This is the closest CPU torture test to running F@H that we currently know about, and as such is recommended for stress testing machines that will be running F@H. Available for Linux and Windows.

However, -bigadv might stress CPUs more than StressCPU2 so it might no longer be as close of a torture test as it once was.
 
My Q9450 runs 3.6 24/7 @ 1.38 Vcc + Vtt (450 fsb), doesn't like to go higher without going above 1.4v.
w/TRUE in push/pull (2 S-Flex) and the washer mod I don't go above 52c load. God I love the TRUE.

I may pick up a Q9550 @ MC, relegate the Q9450 to system #2, and move that E8400 to system #3. The E8400 is running @ stock, I've never even bothered trying to oc it. :D Then again, do I need 3.8+ on a quad? Ehh not really.
My work system (#3), the E2200 is @ 2.7 with stock volts; can't go higher because the POS Asus board has almost no ram dividers and the [800] ram in that system doesn't post above 1000. [Grumble grumble crap Asus board]

My new 260 216 was a stock clocked model I flashed to SC speeds to match my other. :) (first time bios flashing a gpu)

Needless to say I'm a big fan of free performance (as long as it's in safe voltage and decent temps)
 
I am currently running my 920 at 3.8 between 55 and 60 degrees and 1.275v. I achieved this without doing much change in the bios. I would like to push it to 4.0. Currently it will go to about 3.86, but any higher at all and it will not even post. More voltage does not help, so it is most likely a memory or other ratio setting causing the problem. I just have not taken the time to find the problem. If I can get 4.0 with it and stay under 80 degrees, I will be happy.

I have run Prime 95 and OCCT to stress the processor. At this speed, it will run either all day and stay below 65. Most people see OCCT giving higher temps than Prime 95, but I do not, about a 3 degree difference the other way, Prime 95 is higher. Both of them give a higher temp than folding. I do not see any measurable temperature difference between regular 8 thread smp and 8 thread bigadv. I have not looked for a difference, I just have not noticed one on my rig.

My GPU stays below 65 mostly, so I am happy there, also.
 
On my Q9550 I run it at its max safe volt with 1.36 actual volts (not the Bios setting) and have it cranked stable 24/7 prime crunching (Gimps PrimeNet/SOB Sieve/PSP Proth Sieve) @ 3.7ghz.
Temps are no problem for me due to water cooling at 51C max full prime load but the max safe on the Q9550 is 72C so I agree to keep it below 70C.
 
Although the Core i7 CPUs throttle down their speed starting at 100C, I am of those who like to keep them at 80C or lower. One could probably go a bit over 80 for short periods of time but anything under 80 would probably be fine for a full-time folder.

Edit: Of course you do not want to try this with C2Q's! :D
 
Q9650 @ 4.2.... right around 65C... I want to get temps lower before I push higher
Phenom @ 14x200 for 2800... temps below 50C.... not stable at 2850
3xGTX260s shaders at 1512, none will clock above that stable
9800GT shaders at 1925, random reboots at 1950
8800GS shaders at 1725, EUEs at 1750

so yea, push to the bleeding edge, but watch the temps and don't go bananas on the over-volting. I'm only ever-volting the CPUs and northbridges, and then I'm keeping them in the green voltages

 
Right now my i7 is at stock settings, although all cores seem to be running at 2.83GHz according to CPUz (I guess turbomode is working for me), but my GTX260 is at 700MHz core (not sure what the shaders are... I'm not at my PC right now, but they are linked if that helps).

I might work on overclocking my CPU again later on. Or just sit tight until the 6 core CPUs come out...
 
My 4870's at stock, pushing them even 1MHz beyond 750/900 produces a series of rather spectacular VPU recovers even at idle.

My Core i7 920 is clocked at 3560MHz with 1.32Vcore. I haven't tried pushing the Vcore more vigorously, so 3600 in Winter and 3560 in Summer is the most I can push it.
 
jfb9301 said:
Q9650 @ 4.2.... right around 65C... I want to get temps lower before I push higher
Phenom @ 14x200 for 2800... temps below 50C.... not stable at 2850
3xGTX260s shaders at 1512, none will clock above that stable
9800GT shaders at 1925, random reboots at 1950
8800GS shaders at 1725, EUEs at 1750

so yea, push to the bleeding edge, but watch the temps and don't go bananas on the over-volting. I'm only ever-volting the CPUs and northbridges, and then I'm keeping them in the green voltages
Click to expand...

9800GT shaders at 1925, random reboots at 1950 <-- check your numbers in rivatuner.. ill bet anything the shaders are actually running at 1890 and not 1925.. since the 54mhz jumps it would either be at 1890 or 1944.. which is why its probably crashing at 1950..


Mr. Pedantic said:
My 4870's at stock, pushing them even 1MHz beyond 750/900 produces a series of rather spectacular VPU recovers even at idle.

My Core i7 920 is clocked at 3560MHz with 1.32Vcore. I haven't tried pushing the Vcore more vigorously, so 3600 in Winter and 3560 in Summer is the most I can push it.
Click to expand...

probably need to check your temps on the 4870..

you should be able to do 3.8ghz at 1.35v if your able to do 3.6ghz at 1.32v..
 
probably need to check your temps on the 4870..
Click to expand...
Not really. I went through this before on another forum. Temps are quite low, mid-60s at load, and feeling the exhaust seems to confirm that the temps aren't wildly off. It certainly is a far cry from the mid-80s people report is normal on 4870s.

you should be able to do 3.8ghz at 1.35v if your able to do 3.6ghz at 1.32v..
Click to expand...
Maybe. I've tried up to 1.34, and it won't budge past 3.6. Though, having said that, I had to bump voltage 0.02V past stock to get past 3.0GHz, and after that it was a clear run all the way to about 3.4GHz. So what you say could be true. I should dedicate some time to investigating this...
 
Last edited:
Mr. Pedantic said:
Temps are quite low, mid-60s at load, and feeling the exhaust seems to confirm that the temps are wildly off. It certainly is a far cry from the mid-80s people report is normal on 4870s.
Click to expand...
My Sapphire 4870's run at 800/1000mhz stable crunching Milkyway@Home (which pull very [H]ard on GPU's) with max temps at 91-95c with 96% fan speed and 4x120mm fans blowing high speed at them stable 24/7 for the last 6 months, so mid 60's are not the problem.
 
Mr. Pedantic said:
Not really. I went through this before on another forum. Temps are quite low, mid-60s at load, and feeling the exhaust seems to confirm that the temps aren't wildly off. It certainly is a far cry from the mid-80s people report is normal on 4870s.


Maybe. I've tried up to 1.34, and it won't budge past 3.6. Though, having said that, I had to bump voltage 0.02V past stock to get past 3.0GHz, and after that it was a clear run all the way to about 3.4GHz. So what you say could be true. I should dedicate some time to investigating this...
Click to expand...


as long as your not on stock cooling.. you have all the way up to 1.5v.. so theres plenty of room to experiment..
 
I don't care about temps, if it's stable I am happy. Been overclocking all my cpu's for over 10 years, never fried a single one. (And have had many man many of them too)
emthup.gif
 
as long as your not on stock cooling.. you have all the way up to 1.5v.. so theres plenty of room to experiment..
Click to expand...
This CPU has to last me at least 5 years, though. So I'm not so sure that 1.5Vcore is the way to go. I'll see just how much of an improvement I get, though.
 
Mr. Pedantic said:
This CPU has to last me at least 5 years, though. So I'm not so sure that 1.5Vcore is the way to go. I'll see just how much of an improvement I get, though.
Click to expand...


i wasnt saying go directly to 1.5v.. but you have plenty of room between stock and 1.5v to mess around with different options.. as long as you stay within the limitations that intel says for that processor.. it will last 10+ years at 1.5v.. but just mess with different voltages and see what you can get..
 
I've put it up to 1.4Vcore, and I'm testing with StressCPU V2. I've put it back up to 3.6GHz for the moment, and I've done about 1.75x10^14 FLOPs.

How long should I run it for? A couple of hours? A set number of FLOPs? Temps have stabilized at about 70 degrees C.
 
No good. At 1.4Vcore I can only get it to 3740MHz stable on the SMP client for more than a few seconds. I can boot up to, and run StressCPU V2 for hours without problems, but once the SMP client starts up, within seconds I either get a reboot or a BSOD.

I've moved it back to 3.56GHz on 1.3Vcore, because I don't think the extra 0.1V is worth the 200MHz I can get out of this.
 
Mr. Pedantic said:
No good. At 1.4Vcore I can only get it to 3740MHz stable on the SMP client for more than a few seconds. I can boot up to, and run StressCPU V2 for hours without problems, but once the SMP client starts up, within seconds I either get a reboot or a BSOD.

I've moved it back to 3.56GHz on 1.3Vcore, because I don't think the extra 0.1V is worth the 200MHz I can get out of this.
Click to expand...

normal SMP or -bigadv?
 
Mr. Pedantic said:
No good. At 1.4Vcore I can only get it to 3740MHz stable on the SMP client for more than a few seconds. I can boot up to, and run StressCPU V2 for hours without problems, but once the SMP client starts up, within seconds I either get a reboot or a BSOD.

I've moved it back to 3.56GHz on 1.3Vcore, because I don't think the extra 0.1V is worth the 200MHz I can get out of this.
Click to expand...

Did you check your RAM? Maybe that's what causing the problems.
 
No, I bumped the Vmem all the way up to 1.65V. For some reason it won't let me change the multiplier for memory. Well, it does let me, but it won't POST if I do.
 
Mr. Pedantic said:
No, I bumped the Vmem all the way up to 1.65V. For some reason it won't let me change the multiplier for memory. Well, it does let me, but it won't POST if I do.
Click to expand...

It could be that your RAM is running too far out of spec, either by speed or timings.
 
It could be that your RAM is running too far out of spec, either by speed or timings.
Click to expand...
I haven't tried changing timings yet, but there's little I can do about the speed it's running at, seeing as I can't seem to be able to change the multiplier and still get a POST.
 
crazjayz said:
It could be that your RAM is running too far out of spec, either by speed or timings.
Click to expand...


that would be my guess too.. probably need to drop the multiplier down to ddr3 1333 or ddr3 1066 if its ddr3 1333 stock.. kinda odd that it wont let you do that though..
 
that would be my guess too.. probably need to drop the multiplier down to ddr3 1333 or ddr3 1066 if its ddr3 1333 stock.. kinda odd that it wont let you do that though..
Click to expand...
I'm running memory that's specced at DDR3-1333 at 8-8-8-24 timings. I would have thought the memory would run fine at 1500-odd MHz...

What sort of effect does slower timings have on performance? I'll try raising the timings tomorrow, I'm a bit too tired for today to be getting frustrated with BSODs and random restarts I think...
 
has almost 0 effect on performance.. you might be able to get it to run at ddr3 1600 @ 1.65v and 9-9-9-32..

back in the days of ddr1 ram timing was very important.. but now with ddr2 and ddr3 theres almost little or no difference in performance except in benchmarks..
 
I have been getting 1600 RAM and under running it for my i7. It's still faster than what the 920 is paired with but the RAM is under spec and should be stable.

I've been doing that since the socket A days.
 
I've always had a habit of dialing up my system to 11 :)...with my i7, I've cranked it up to 4.1Ghz so far...and I'm happy with that...As far as I'm concerned, if it's not throttling, it's not overclocked enough!...Well, I do try to stay just under the throttle threshold ;)


live [H]ard!
 
As far as RAM goes, higher memory bandwidth actually does make a noticeable impact in performance when running -bigadv workunits. You would want to get your memory running as fast as possible to get the most PPD out of your system.
 
We should find someone who doesn't mind losing the points and have them run to 5 frames with different ram times at the same OC so we can find how well it scales
Posted via [H] Mobile Device
 
Vaulter98c said:
We should find someone who doesn't mind losing the points and have them run to 5 frames with different ram times at the same OC so we can find how well it scales
Posted via [H] Mobile Device
Click to expand...

Meh, I'll do it. I switch the i7 over to normal smp because I wanted to game this weekend and I finished a -bigadv WU.

Currently I'm running at:
664ram.jpg

I'm getting 7438ppd on a -smp 8 p2669 WU

I'll brb with new RAM settings
 
Keep in mind that RAM speed probably doesn't make as much of a difference with normal SMP units as it does with -bigadv units.
 
Zero82z said:
Keep in mind that RAM speed probably doesn't make as much of a difference with normal SMP units as it does with -bigadv units.
Click to expand...

It's worth the test though..... let see what happens.

I couldn't go up to the 10X ratio, 8 is where I was, so I dropped it to 6. Just have to wait for a few frames to finish.
 
Well that was fun.

Like Zero said, little to no difference. I'll keep this RAM setting and when I load up -bigadv again I'll revisit it.

With triple channel, I just can't see it making that big of a difference however.
 
Ehh... :(

I tried running at looser timings...and still no luck. I went all the way up to 10-10-10-30, and it wouldn't go any further than it did last time. I think I'll just leave it where it is...
 
You must log in or register to reply here.
Back
Top