Is cpu overclock good for rendering?

General Forums Off-topic Chat
1

I have a question, is it good to overclock CPU, for rendering purposses? i tend to buy new powesuply and new cooler for my quad core processor, and i just wanted to know would that be usefull and now much would that benefit me in rendering when it comes to speed. i would be very gratefull if you could give me an advice, so i could save money, or invest them in some other things if the overclocking of processor isn’t good for that purposses. my config

core 2 quad Q9505 intel procesor 2.83ghz
4gb ram memmory
500gb hard drive
and thats it :stuck_out_tongue:
thanks !:spin:

2

If your processor is running faster your renders will be faster.

3

Unless you’re running a 64 bit OS, any RAM over 3GB is wasted: the OS is simply unable to address any more memory. Overclocking your CPU will speed up rendering - but, by how much? It takes significant overclocking to produce significantly shorter render times. Significant overclocking without managing the resultant CPU heat build up can dramatically shorten the life of your CPU.

4

im running windows 7 x64 , i know that bout the ram memory, i was just courious with this cpu boost thing. but if i put tested

http://www.racunari.net/slike_big/big_lcpower_LC6600GP2.jpg
big_lcpower_LC6600GP2.jpg477×513

cooler on the cpu? i mean, wanted to buy powersupply 650W , and cooler that can easly cool even the i7 intel processor. does that means that life of my processor will be still short? i mean, has anyone tried to clock cpu?? now im in great dubt when it comes to that, since i have to give great deal of money, and that what i can get is nothing seems like so.

just to sum all shortly-overclock under every conditions means fail sooner or later??? or it doesn’t have to be like that if you have great powesuply and cooler on CPU? (im not talking about extreme overclocking, just 400-600mhz max boost)

if that means failing my cpu sooner or later then i won’t invest in that definitely.

5

In other words, what you gain in speed, you lose in stability.
Overclocking will speed up your renders, but if you overdo it, it might crash and/or bluescreen. Hours of rendering at 100% is enough stress to overheat many stock computers, (especially laptops) so be careful. :wink:

6

I agree with callmeismael you won’t find a major difference unless you do some serious overclocking. If you need faster render times, look for an optimised build of blender on graphicAll.org http://www.graphicall.org/builds/ I’ve had 2.49 builds which run 30% faster than the standard build. Now I don’t check. Optimised builds are just faster more stable and safer than overclocking.
Also you ought to go with a good power supply especially if you have a good graphics card.

7

I routinely overclock my Intel Q9400 from 2.66 to 3.2 GHz with no problems at all
I do have a nice oversize cooling setup
BUT, if it’s going to cost you a lot of money I’d say don’t do it
The difference in render time just doesn’t justify spending money that you’re having to worry about
My cooling setup was about $30 and I had the extra money so it was no big deal, I understand bucks are tighter for lots of folks

here’s my cooler

http://www.tigerdirect.com/applications/SearchTools/item-details.asp?EdpNo=2135206&CatId=1588

8

yea i agree, well this cooler i wanna buy is bout 25euros, that would not be the problem, but i think if id wanted to overclock my q9505 i dont think i would be able to perform it. cuz right now at the moment i have some cheap blueberry powersupply 400W http://www.blueberry-eu.com/main.php?action=filter&group=prod&producer=power&subgroup=psb400 so id have to buy new powresupply that would be from 40 to 50 euros + cooler 25, roughly in total about 75 euros, wich i think is too much.for that. i mean to much risk in exchange what i get . so when i sum all this. i guess best thing for me would be to buy some cheap cooler, and to run in with this config without overclock i guess. i say that because i think this stock cooler of mine is going over 70 C in full load without clocking.

9

hey thanks, ill look at that too. thanks for the advice.

10

Really? You’ve got a quad and still “over-clocking” to max cycles? The first place I’d look is process priority and memory allocations; kill off those useless background apps too (even security if you disable internet and run just to render locally).

Not to be condescending, but be sure you’re using max processors in Blender (I set to 8 for quad to max usage - 100% and still able to run gen apps like browser and e-mail).

Sorry, had to reply. I used to over-clock my 7HZ 8086 back in the day…

P.S. comments apply to Windows system, Linux is similar but different :slight_smile:

P.P.S. - Yes, fail sooner when over-clocking; not likely you’ll get sufficient cooling to prevent “heat” damage. But, if you can cool it like it’s over-clocked you might actually extend the life of your CPU.

11

this was also very useful text Jambay, thank you for your time to reply, i didn’t try that with processes but i will definitely have that in mind before starting my next renderer. thanks again :- ) p.s im ordering right now a new cooler, and if the temperatures came out as acceptable, i will consider about new power supply as well but that is really under big question mark i really don’t think i will buy new power supply and overclock cpu though , cause what you pay isn’t equivalent what you get unless i extreme overclock my cpu and get my system to very unstable condition, and i can’t affort right now new PC so i don’t think i’ll risk that much.

12

I´ll throw in my 2 cents.

The days where overclocking posed any risk are long over.
Nowadays processors are all one processor really within one architecture.
They get produced with the highest specs and then they look at the yield they get according to the specification sheet.
Those which run with XY Voltage at XY speed are labeled the SuperHighend Processors.
Those wich run with XY-Z Voltage at XY-Z speed are labeled the Enthusiast Processors.
Those with broken cores, or damaged transistors get units disabled and labeled Triple or Dualcore.

Your chances are at 90% that your processor will still run stable at higher clocks and at even higher clocks if you kick up the voltage a bit.
Take the Q9550 (as I got one) it has no reference voltage. They basically set the reference speed (2.83GHz) and a set voltage below the maximum spec for 24/7 usage (1.3685V)
The Q9550 can hit its 2.83 GHZ from 0.8500V-1.3625V shipped and boxed by intel.
You can go beyond 1.3625 no Problem, but then your in an area where you reduce the lifespan of your chip due to thermal stress.

My Q9550 was shipped with 2.83GHz and default at 1.2750V.
Now it is running with 3.60GHz ar 1.2080V.
27% faster which means in animation terms, render 3 hours instead of 4 for instance, while I reduced the voltage which even prolongs the life of my chip and reduced power loss through heat.
And besides that it runs rock solid, no bluescreens, no freezes, no crashed, but you really got to know what you do and play a bit with the settings.
I am sure I could pop it beyond 4GHz within the intel specs, but my memory wouldn´t comply because its still DDR2-800 (which runs nice with 1000)… what brings me to my next point:

Number 1 reason for unstable systems after overclock are cheapass memory modules.
Only nvidia chipsets can unlock memory clocks from frontsidebus clocks.
AMD and intel chipsets cant. Basically (not general though) the memoryclock = 2*FSB.
If you raise the FSB too high, cheap memory modules can´t handle the OC and those are in 99% of the times the case for a freeze or a bluescreen.

So IF you read into OC and IF your time to tinker around is worth less than buying a faster machine, OC is just fine nowadays.
Processors got a lot of potential, the manufacturers just don´t give a crap to put time into finding its limits it would be too expensive and would have to be kicked down to the customer.

gl!

13

for life expectantcy look at some discussion in this thread

http://blenderartists.org/forum/showthread.php?t=198060&highlight=cpu+overclock

you can also download a free cpu temperature monitor

in summary for every 10C increase in temp you reduce the life by 1/2
so very important to keep it as cool as possible and under the max temp of manufacturer!

salutations

14

arexma

hey thank you for this, this was so so useful to me… i can’t eaven tell how how much!!.. cause you have similar processor like me but not the same i have weaker version of Q9550 which is Q9505(witch 6mb cache memory and not 12 like yours has) and i suppose i could take simmilar settings like you described if would like to overclock my CPU, can you tell me how long are you working with this clock?? are you reseting default clock each time when the rendering is finished?? or you keep it overclocked all the time??

im in great dillema right now.

15

RickyBlender

so you say, if i keep it in manufacturers temperatures , all the time, it doesn’t matters if its overclocked or not it will have normal lifespan?? if that is the conclusion, than that is great!!! that concludes this topic and gives me the answer that i wanted to hear:) and i want to thank again for this great blender communitiy for provideing me a great answer to this question, i learned a lot from you guys.

16

this temp relationship is valid at any temp even under the max manufaturer temp spec!

it’s just that if you work near the max temp then life is gone be be very short
so it mainly means you should make certain that your coolling system fan heatsink are clean and work at maximum efficiency
so that the CPU temp is minimum then you get maximum life!

but if you overclok it’s even more important to make certain that you have enough ventilation to keep the CPU temp at a minimum
might even be a good idea to add a bigger fan to increase the cooling effect

and with the program to measure temp of CPU you can test the CPU temp
and don’t forget the longuer a program works and the higher the temp of the CPU will be when you make a test
up to a certain max temp function of the heatsink ventilator you have on your machine!

but you have to test CPU temp that’s the main indicator before you overclock and after
then if you see a large temp change then may be your cooling sytem is not good enough and needs to be beef up a little!
if not then the life expectancy will be lower !

salutations

17

an obvious question nobody has asked you is do you have a motherboard that will allow you to overclock the CPU?
A lot of the OEM boxes don’t allow this

18

i know all that, i use for that Everest - system stability test, so far with this stock cooler i had 70 C maximum. under 100% and one hour testing. thats why i want new cooler, and for overclock i will have to think about it little more i guess.

p.s btw i think that 60C is ok for overclocked processor.

19

You´r doing it wrong, 60/70°C is too much, its too close to the max. spec. temperature by intel.
Mine runs with a deltaT of 3°C in idle and deltaT of ~20°C under load on aircooling.

idle: air inside case 27°C, cpu 30°C
load: air inside case 30°C, cpu 50°C
Cooler: Noctua NH-U12DX

make sure you got a nice case cooling as well, if the air inside the case got no flow you cook up the cpu for nothing.
and sorry I can´t tell you my settings. For a stable OC you got to find out the timings that mach your cpu, mainboard and memory.
OC’ing the CPU means also OC’ing the memory unless you got a nvidia chipset.
If the system is unstable after OC you want to kick up the voltage of the southbridge just a tad, that usually fixes it. Also check the memory and its temperatures.
Usually you also have to OC the chipset a bit, because the signals sent by the processor get distorted by the OC, you got to adapt the termination voltage (VTT) which is close to voodoo.
The VTT is the cleaner after a signal. When the processor send a signal along the bus you got to think of it as a sine wave although its just 1 pulse. If there are lots of sine waves they start to overlap and the signal gets unclear -> freeze. There´s the termination voltage. You might be aware of the fact that opposite amplitudes of waves erase each other. thats what the termination voltage basically explained does. It clears the signal on the bus so the next signal can go on a clear road. With OC the signals get denser AND the amplitudes higher (due to overvoltage) so you got to play with the termination voltage. And play is literally, set it, run prime95 (my fav), coredamage and whatever to see if it is stable, if not try other setting. It is impossible to even give a rule of thumb for the VTT because it happens on a microelectric basis. You might be able to find it with a multimeter and oscilloscope and measuring voltages and signals on the running machine =)

And before playing around be sure to locate your bios reset jumper on the mainboard =)

And the lifetime stuff… dont worry. As long as you stay below the specified temperatures intel supplies you´r save.
Personally I don´t care if the CPU lasts 20 or 10 years, never had one that long anyways.

A word of advice, be VERY careful tinkering with voltages, if you set one wrong you can fry your system from the graphics card over the mainboard and ram to the cpu. It is a delicate matter.

Personally I think your CPU is so hot because the voltage is on auto in the BIOS.
When you OC, the BIOS also kicks up the voltage automatically (everest is great, it can display voltages up to the 4 digits if you set it in the options) and there is the nasty thing called vDroop.
Whenever the CPU goes under load, the transistors get hot and the electrical resistance changes. To compensate for that, the Core voltage drops, the downside is, on an OC´d system this dropp in the vCore can mean it will freeze because there is not enough power, usually doesn´t happen because the voltage is set too high by the bios.

My advice to start:
Set the reference clock.
and LOWER the vCore until it starts to be unstable.
Then you know the minimum voltage you need to support reference clocks.
Then start to work from there. Kick up the Voltage a bit and start raising the FSB. Keep an eye on the memory. We are actually lucky because up to 400MHz FSB is no problem.
Stock should be 333MHz, up to 400MHz with a FSB:DRAM = 1:2 results in memory of 800FSB which is just nice with DDR2/800. Thats 3.4 GHz with 8.5 multiplier.
Beyond that point you might have to raise the memory voltage to supply the higher clocks AND/OR loosen the memory timings.

And one hour testing… If it runs 4h primestable (prime95) in maximum heat and 4h in memory stresstest you can consider to call it stable and start to run FarCry2 Benchmark or something wich really tortures the system. There are systems that ran 12h primestable and crashed in the first 10 seconds running Futuremark or Crysis Benchmark.

Everest´s great for monitoring and diganose (html logging) but crap for torture tests.
Coredamage, Prime95, Furmark, Intel BurnIn, FarCry2Bench, Futuremark those are nice tests.

Happy frying =D

20

im sorry, but i have to ask you… ARE YOU WORKING IN INTEL? :smiley: :slight_smile: :slight_smile: this is too much for me… for and hour of a gratis rendering :stuck_out_tongue: :smiley: