Best CPU Stressor Out There

[mcrumiller]

I've been overclocking my E8500 for a little while now and attempting to find the best ways to stress my system. I had been using Prime95 up until now as the best system stressor--using small FFTs I could push my cores (water-cooled) up to around 80-85C, and I didn't really want to extend beyond that without drastically reducing the lifetime of my CPU.

So I decided to test out how MATLAB could handle some large matrix multiplication. I booted up and entered the following commands to multiply two random double-precision 10,000 x 10,000 matrices together:

>> x = rand(10000); y = rand(10000);
>> z = x * y;

Lo and behold, I watched my cores skyrocket all the way up to 120C! I dove for the power button of course and managed to save my CPU before permanent damage.

I've since, unfortunately, lowered my speeds and voltages so as not to run into this situation. But it makes me wonder about the ability of programs like Prime95--which basically run separate, non-communicating threads on each core--too fully stress the system. I'm guessing that the addition of the communication between cores is enough to cause the heat surge here, something you don't see in Prime95. I was hoping someone could suggest another program other than Prime95 or Orthos that might stress my system to the same degree as Matlab--one that would report a failure rather than giving me a BSOD.

On a side note, I'm running a watercooled dual-core intel E8500 on an Asus P5N72-T Premium (same thing as Striker II Formula).

 

[Kraeoss]

and at what speed and voltage did you have the chip at ? if you dont mind me asking ?

 

[error8]

He probably has that chip at 1.9 V.

Best cpu tester is Linpack, which you could find it in the latest OCCT version. But before going there, you have something bad on your hands, maybe the water cooling system is faulty or incorrectly installed, or you're just using way too much voltage.

 

[Cheex]

OCCT 3 (built-in Linpack test).

/thread

 

[VirtualLarry]

80-85C under water? Somethings wrong there.

 

[ShawnD1]

Originally posted by: Cheex
OCCT 3 (built-in Linpack test).

/thread

I checked this out after reading this thread, and I'm very impressed. It can somehow make my computer draw 480W of power. The most I have ever seen, while doing GPU and CPU folding, was only 350W. That's one hell of a stress test.

 

[error8]

And your cpu temp is? 150 C?

 

[Idontcare]

Originally posted by: VirtualLarry
80-85C under water? Somethings wrong there.

+1

And 120C under any kind of cooling whatsoever is problematic.

 

[exar333]

Originally posted by: VirtualLarry
80-85C under water? Somethings wrong there.

Could be a low-flow (Think Seinfeld here) watercooling setup. Considering the CPU is only a dual-core, thats a lot of heat for a water-cooled rig....

 

[mcrumiller]

Hey all, thanks for the input--

I hit 85C when both cores are fully loaded. I'm around 50-55C when idle. On stock settings I'm around 40C when idle, which is reasonable considering my configuration. Here's some of my specs:

FSB - 440
CPU - 9.5 * 440 = 4180 MHz
Vcore = 1.525v (my chip is VID'ed at 1.225v for 3.16 GHz, but it doesn't OC too well)
NB - 1.44v
VTT - 1.26v

My mem is perfectly stable--I have it unlinked at 850MHz with 4-4-4-10/T1, trc=16 trd=7
I also have dual GTX 280's in the water loop (in addition to NB and SB) with two rads. The graphics cards bumped my idle temps up a few degrees when I installed it.

I was willing to pump up to around 1.55v before I ran the Matlab test--now 1.5v is my max. Anything under 1.5v and I can't break 4GHz, and below that...what's the point of buying all this WC? =P If I can't get high stable speeds with lower vcores I might consider upgrading to an E8600, but I really can't justify the price right now.

 

[n7]

Welcome to the forums.

I'm calling BS though, sorry.

Your E8500 will shut off due to thermal protection well before it hits 120C, unless your sensors are wrecked.

I hate to be the hater, but until i see screenies showing over 100C with CoreTemp or RealTemp, i don't believe.

Run LinX or IBT set to max RAM to see high temps.
http://www.xtremesystems.org/f...howthread.php?t=201670
http://www.xtremesystems.org/f...howthread.php?t=197835

 

[exar333]

Originally posted by: mcrumiller
Hey all, thanks for the input--

I hit 85C when both cores are fully loaded. I'm around 50-55C when idle. On stock settings I'm around 40C when idle, which is reasonable considering my configuration. Here's some of my specs:

FSB - 440
CPU - 9.5 * 440 = 4180 MHz
Vcore = 1.525v (my chip is VID'ed at 1.225v for 3.16 GHz, but it doesn't OC too well)
NB - 1.44v
VTT - 1.26v

My mem is perfectly stable--I have it unlinked at 850MHz with 4-4-4-10/T1, trc=16 trd=7
I also have dual GTX 280's in the water loop (in addition to NB and SB) with two rads. The graphics cards bumped my idle temps up a few degrees when I installed it.

I was willing to pump up to around 1.55v before I ran the Matlab test--now 1.5v is my max. Anything under 1.5v and I can't break 4GHz, and below that...what's the point of buying all this WC? =P If I can't get high stable speeds with lower vcores I might consider upgrading to an E8600, but I really can't justify the price right now.

I am not that experienced with water cooling, but it sounds like there is too much in the cooling loop to keep everything optimally cool. Have you considered splitting the loops to have one cool the SB + GPUs and the other for the CPU + NB or something along those lines? Mid 80s for CPU load for a water cooled rig is way to high....

Can you load some temp screenshots?

 

[mcrumiller]

Sure, there's a lot in the loop. I suppose I could have bypassed, the southbridge, but the SB barely produces any heat anyway that it wouldn't have changed much. Additionally, two separate water loop wouldn't be any better anyway: the delta through each radiator is greatest at highest temperatures, so optimally you want to have the hottest water going through the radiators to create maximal cooling. I have my radiators back to back in the loop before the CPU so the CPU is getting the coolest water in the system. The ASUS boards have notoriously hot NBs, so I definitely need that under water as well.

As for n7...here's a screenshot where I'm hitting 93C:
http://www.neddasty.com/misc/core_temps.jpg

And yes, I suspect that Core Temp was misreading the temps before when I hit 120C. I'd rather not crank up my system to that level again for fear of doing permanent damage.

I don't think these temperatures are too ridiculous consider the voltage, speed, and components in the loop. It's possible that I could screw my heatsink on a little tighter. I might take the system apart and check the radiators and what not to make sure all the fins are straight. Remember...this is an entire extra billion clock cycles per second that I'm achieving here, so I'm not complaining too much.

 

[mcrumiller]

Also n7...thanks for that LinX program tip...it's amazing! Prime95 would take hours to indicate an error...this hits one in.....37 seconds!

 

[Dadofamunky]

I agree with stress testing, particularly if you're an OCer, but it only takes you so far. If your system passes things like OCCT and Prime and runs everyting you do without blowing up, I consider that good enough. I don't really care about Linpack and such because nothing I do will EVER come close to stressing out the CPU like that. Also, Linpack does not tell you what else is going on in your system - memory and chipset come to mind.

I must note, idling at 50-55C is too high for watercooling, much less the chip-frying temps at peak. Even basic aircooling like my system doesn't come close to that. Check to see if you have air bubbles in your setup. You may need to bleed the system some more. This is quite possible since you've indicated you have a fairly complex loop.) It's similar to bleeding the filter lines in an aquarium. Also, check your radiator. My guess is even if you have several devices in the loop it should never get that high at idle. I think some troubleshooting is in order.

 

[wwswimming]

or ... download trials of 3D Max 2009

http://usa.autodesk.com/adsk/s...eID=123112&id=10083915

and set up a simple particle simulation
create - particle systems - PF Source

using geometry instead of tick-marks.

then assign a raytraced material to the geometry.

a great way to peg the CPU.

on the other hand, who wants to pay the power bill ?

 

[Gillbot]

Originally posted by: mcrumiller
Sure, there's a lot in the loop. I suppose I could have bypassed, the southbridge, but the SB barely produces any heat anyway that it wouldn't have changed much. Additionally, two separate water loop wouldn't be any better anyway: the delta through each radiator is greatest at highest temperatures, so optimally you want to have the hottest water going through the radiators to create maximal cooling. I have my radiators back to back in the loop before the CPU so the CPU is getting the coolest water in the system. The ASUS boards have notoriously hot NBs, so I definitely need that under water as well.

As for n7...here's a screenshot where I'm hitting 93C:
http://www.neddasty.com/misc/core_temps.jpg

And yes, I suspect that Core Temp was misreading the temps before when I hit 120C. I'd rather not crank up my system to that level again for fear of doing permanent damage.

I don't think these temperatures are too ridiculous consider the voltage, speed, and components in the loop. It's possible that I could screw my heatsink on a little tighter. I might take the system apart and check the radiators and what not to make sure all the fins are straight. Remember...this is an entire extra billion clock cycles per second that I'm achieving here, so I'm not complaining too much.

I don't even know where to begin tearing this apart.... Gah, i'm not even going to bother.

 

[mcrumiller]

Gillbot, not sure what you're claiming...you think I should have two separate loops? I think that would be sub-optimal....each loop would have half the water of the original, only now the heat for each loop is only distributed amongst half the water. The loop with the higher heat will be even hotter than it would be shared with everything else.

Also, I'm not saying "it's better to have your water hotter"...I'm saying that you want to set up your system such that the hottest water is passing through the radiator to allow for the most efficient cooling. More heat will be dissipated from the system. Do you think it's better to spread out the rads in the system? I'm open to any suggestions, I'm no expert on these matters.

 

[Idontcare]

Originally posted by: mcrumiller
Gillbot, not sure what you're claiming...you think I should have two separate loops? I think that would be sub-optimal....each loop would have half the water of the original, only now the heat for each loop is only distributed amongst half the water. The loop with the higher heat will be even hotter than it would be shared with everything else.

Also, I'm not saying "it's better to have your water hotter"...I'm saying that you want to set up your system such that the hottest water is passing through the radiator to allow for the most efficient cooling. More heat will be dissipated from the system. Do you think it's better to spread out the rads in the system? I'm open to any suggestions, I'm no expert on these matters.

I'm no water-cooling expert, pm Aigo if you want some one-on-one assistance there, but your temps are what most of us would expect from an air-cooled rig at those frequencies and voltages, not a water-cooled rig.

That is why you see your thread being populated with post after post of gasps and yikes expressions of disbelief.

 

[Kraeoss]

idk if i'm right but if you took 1 big pot or water and cooled it then take that same volume of water break it in 2 and cooled em which would cool first ?.... point being 2 separate loops with 2 rads will yeild better results than 2 loops on 1 rad. and keep your loops separate gpu usually ouputs more heat than cpu but can cause the cpu increase in heat rather than the alternative.


am i right ? $0.02

 

[mcrumiller]

Of course it would...but I have two rads. Two pots of water cool faster than one because double the surface area exists for two pots to cool...not so on my system, where you have the same number of radiators regardless. A more realistic analogy would be the following: what cools faster: two pots at 70C or two pots, one at 60C and one at 80C?

 

[aigomorla]

Originally posted by: Idontcare

Originally posted by: VirtualLarry
80-85C under water? Somethings wrong there.

+1

And 120C under any kind of cooling whatsoever is problematic.

what im wondering is how he managed to bypass thermal shutdown at 100C

LIST YOUR LC SYSTEM.

i need to see what the heck your running for a dual core to get that high.

Originally posted by: mcrumiller

I was willing to pump up to around 1.55v before I ran the Matlab test--now 1.5v is my max. Anything under 1.5v and I can't break 4GHz, and below that...what's the point of buying all this WC? =P If I can't get high stable speeds with lower vcores I might consider upgrading to an E8600, but I really can't justify the price right now.

which is why i want to know what your LC is. At those voltage you shouldnt be breaking 50.

Im guessing Thermaltake or CoolIT. (or dead / nonworking pump)

Im gonna say this as nice as possible, but your majorly messing up somewhere.

 

[mcrumiller]

That's correct, but as I said above, the analogy is flawed because two pots of water provide more surface area for the water to cool than do one pot of water. It's more like if you had two tanks of water, each with their separate openings, or two combined tanks of water with still only two openings.

Anyway, aside from this, what do other people suggest?

 

[Kraeoss]

listen to aigo he know the best techniques for efficent LC

 

[aigomorla]

Originally posted by: mcrumiller
That's correct, but as I said above, the analogy is flawed because two pots of water provide more surface area for the water to cool than do one pot of water. It's more like if you had two tanks of water, each with their separate openings, or two combined tanks of water with still only two openings.

Anyway, aside from this, what do other people suggest?

your analogy doeesnt apply because your system is NOT WORKING.

Watercooling works on equalibrium unless you have an absolute crap system.

if your planing for temp spots, then your planning a crap system.

LIST YOUR PARTS. (without this, NO ONE CAN HELP YOU)

Your DOING SOMETHING MAJORLY WRONG.

Im not even breaking 70 on an i7 @ 1.48Vcore with HT ON with LINPACK.

http://i125.photobucket.com/al.../aigomorla/Final-1.jpg

And your breaking 120? That doesnt sound right now does it?

My system is usually 5-10C better then everyone's, but still dude, your system is SERIOUSLY WRONG.


Also the day an intel dualcore puts out that much heat is the day i retire from intel and go AMD.