Oh its possible Mark.

I fried one maybe 10 years ago.
You just need to try very hard LOL

Older CPUs had no inbuilt thermal protection, stop the fan watch the CPU cook.
All current CPUs will self throttle their clock speeds once they reach their rated themral maximum.

---

Grant
Darwin NT

As is, even though the program reports 71°C, the fan runs at only 34% of available speed.

The hysteresis for some automatic fan controls on GPUs has been noted in some reviews as less than optimal.
Nobody likes a noisy video card, so some manufacturers will opt to run their GPUs as hot as possible to keep the noise down. If noise isn't an issue, just set it to run at 100%.

For my systems i'm happy for them to run the GPU at 70°. No matter how hot it gets in my room, the fans will ramp up their speed to keep the GPU at 70°.

---

Grant
Darwin NT

As is, even though the program reports 71°C, the fan runs at only 34% of available speed.

The hysteresis for some automatic fan controls on GPUs has been noted in some reviews as less than optimal.
Nobody likes a noisy video card, so some manufacturers will opt to run their GPUs as hot as possible to keep the noise down. If noise isn't an issue, just set it to run at 100%.

For my systems i'm happy for them to run the GPU at 70°. No matter how hot it gets in my room, the fans will ramp up their speed to keep the GPU at 70°.

The kitties don't rely on hysterics.....err.....hysteresis.
All GPU fans are locked at 100% all the time.
Better safe than sorry.

---

"Time is simply the mechanism that keeps everything from happening all at once."

I have nVidia GTX 460 with 2 GB of RAM.

I run 7 gpu work units, it consumes 1.8 GB.

GPU-z shows that:
GPU Temp: 71°C
Fan Speed (RPM): 1500 RPM
Fan Speed (%): 34%

As you can see, if I want to tinker with fan speed settings, I can push the fan speed higher (50 or 60% for examples) and bring the temperature down a few degrees. As is, even though the program reports 71°C, the fan runs at only 34% of available speed.

Are you saying you run 7 Wu's at a time?
You might want to knock that down a bit. It's not the amount of RAM used its the % of GPU used that is important. If you watch GPUZ you will see that you are probably maxxed out on GPU with 2 WU's anything over that is actually slowing you down.

For example, I have 3 GB Graphics card. I run 3 WU's with 2 CPU cores idle to feed the GPU. now check my computers and you'll see that the AMD/ati 7970 Computer is able to generate around 30k credit per day.

Now look at your production. 12k and I assume that is your only PC. That is very low if running 7 WU's at a time is really what you are doing

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

Are you saying you run 7 Wu's at a time?
You might want to knock that down a bit. It's not the amount of RAM used its the % of GPU used that is important. If you watch GPUZ you will see that you are probably maxxed out on GPU with 2 WU's anything over that is actually slowing you down.

Yeah, after reading Fred's Optimise your GPU thread, i bumped my GTX560Ti up to 3 WU's at a time, but left my GTX460 at 2. The GTX460 generally sits at around 97% load. The GTX560Ti was generally around 95-99% load. Since upping it to 3 at a time it sits at 99%.
Keep in mid that as long as the GPU is at 92% or higher load, the temperature won't drop. High 80%s, then it's temperature will drop off.

---

Grant
Darwin NT

I have nVidia GTX 460 with 2 GB of RAM.

I run 7 gpu work units, it consumes 1.8 GB.

GPU-z shows that:
GPU Temp: 71°C
Fan Speed (RPM): 1500 RPM
Fan Speed (%): 34%

As you can see, if I want to tinker with fan speed settings, I can push the fan speed higher (50 or 60% for examples) and bring the temperature down a few degrees. As is, even though the program reports 71°C, the fan runs at only 34% of available speed.

Are you saying you run 7 Wu's at a time?
You might want to knock that down a bit. It's not the amount of RAM used its the % of GPU used that is important. If you watch GPUZ you will see that you are probably maxxed out on GPU with 2 WU's anything over that is actually slowing you down.

For example, I have 3 GB Graphics card. I run 3 WU's with 2 CPU cores idle to feed the GPU. now check my computers and you'll see that the AMD/ati 7970 Computer is able to generate around 30k credit per day.

Now look at your production. 12k and I assume that is your only PC. That is very low if running 7 WU's at a time is really what you are doing

I am not disagreeing. However, I wish to point out that BOINC is using only 0.04 of the CPU.

I actually ran an experiment.using Einstein@home. Einstein uses 0.2 CPU. So when I was running less than 5 tasks (1, 2, 3 or 4), the CPU was not fully loaded. When I set to run 5 task, one CPU core became fully loaded. 5 x 0.2 = 1, so one core became 100% engaged.

Since SETI@home only takes 0.04 CPU, my bottleneck is not CPU, it is amount of RAM on my video card.

Purely for demonstration purposes:

---

Since SETI@home only takes 0.04 CPU, my bottleneck is not CPU, it is amount of RAM on my video card.

What people are pointing out, is that by running 7 tasks at a time you are actully processing less work per hour than if you ran 2 tasks at a time.
If you want to make full use of a CPU core, use it for crunching, or just leave it spare to feed the video card (if you had one capable of doing that much work).
Running a WU on that CPU core, and 2 on the GPU would produce far more work per hour than you are presently doing.
By all means, there's no real problem with what you're doing, other than being very inefficient. It's taking 2 hours to process a WU that if you were doing 2 at a time would take about 15-20min.

EDIT- you are doing 3.5WUs per hour.
doing 2 at a time you would process 6-8 (closer to 8) per hour.

---

Grant
Darwin NT

just for giggles try running 1 WU at a time. Note the speed that it completes the work in and the & gpu utilization. bump it to 2, repeat. then 3. Notice that it takes longer and longer to complete work. THe idea behind running multiple WU's is to complete the WU's faster than if they were running 1 at a time. you've actually done yourself a disservice by overkilling the multiple WU's.

You are certainly utilizing your video RAM and your math skills are working correctly, However you are overloading the GPU not CPU. WU's shouldn't take 2 hours to complete on your GPU when they should take 3-5 minutes. I

f you would have paid attention when the Gpu apps were introduced they explained that 4XX series can handle 2 at a time. 5XX and 6XX can handle 3 and sometimes 4. Nowhere did anyone say that any GPU can handle 7.

again My 3GB 7970 can only run 3 at a time. Do you really think your 460 is faster than mine or the guys running other bleeding edge video cards.

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

Since SETI@home only takes 0.04 CPU, my bottleneck is not CPU, it is amount of RAM on my video card.

What people are pointing out, is that by running 7 tasks at a time you are actully processing less work per hour than if you ran 2 tasks at a time.
If you want to make full use of a CPU core, use it for crunching, or just leave it spare to feed the video card (if you had one capable of doing that much work).
Running a WU on that CPU core, and 2 on the GPU would produce far more work per hour than you are presently doing.
By all means, there's no real problem with what you're doing, other than being very inefficient. It's taking 2 hours to process a WU that if you were doing 2 at a time would take about 15-20min.

EDIT- you are doing 3.5WUs per hour.
doing 2 at a time you would process 6-8 (closer to 8) per hour.

I will try it.

---

Since SETI@home only takes 0.04 CPU, my bottleneck is not CPU, it is amount of RAM on my video card.

What people are pointing out, is that by running 7 tasks at a time you are actully processing less work per hour than if you ran 2 tasks at a time.
If you want to make full use of a CPU core, use it for crunching, or just leave it spare to feed the video card (if you had one capable of doing that much work).
Running a WU on that CPU core, and 2 on the GPU would produce far more work per hour than you are presently doing.
By all means, there's no real problem with what you're doing, other than being very inefficient. It's taking 2 hours to process a WU that if you were doing 2 at a time would take about 15-20min.

EDIT- you are doing 3.5WUs per hour.
doing 2 at a time you would process 6-8 (closer to 8) per hour.

I will try it.

My GTX-670 is doing 3 at a time and it replaced a GTX-460 that was doing 2.

The times remained fairly consistent.

---

I have noticed that running more than 2 WUs/GPU core returns little additional benefit.
As long as GPU utilization is in the high 90's........you are running about as fast as you can, and trying to stuff more work into the pipe just does not result in more work actually being done.

---

"Time is simply the mechanism that keeps everything from happening all at once."

For those of you that like to read, but hardly ever (or never) post:

a) You should post!:)
b) Keep in mind that all the above advice is for tower/desktop PCs. Laptops can run a lot hotter and still live a healthy & long life. A crude rule of thumb is if you here your laptop fan going into turbo/lift-off mode, it's time to clean the vents.

https://picasaweb.google.com/118232178280313837741/Tech#

---

For those of you that like to read, but hardly ever (or never) post:

a) You should post!:)
b) Keep in mind that all the above advice is for tower/desktop PCs. Laptops can run a lot hotter and still live a healthy & long life. A crude rule of thumb is if you here your laptop fan going into turbo/lift-off mode, it's time to clean the vents.

https://picasaweb.google.com/118232178280313837741/Tech#

The max/default temp for my HD5870 GPUs is 100C, according to SmartDoctor.
Higher will throttle them down. Also speed up the fan.

And I never run more 2 WU per GPU as it hardly ever more effective then 1 or 3 or
even more. Even on high end or # of compute units > 10 CUs.
F.i. an ATI HD5870 GPU has 20 Compute Units.
And an NVidia GTX480 has 15 Compute Units.

---

Since SETI@home only takes 0.04 CPU, my bottleneck is not CPU, it is amount of RAM on my video card.

What people are pointing out, is that by running 7 tasks at a time you are actully processing less work per hour than if you ran 2 tasks at a time.
If you want to make full use of a CPU core, use it for crunching, or just leave it spare to feed the video card (if you had one capable of doing that much work).
Running a WU on that CPU core, and 2 on the GPU would produce far more work per hour than you are presently doing.
By all means, there's no real problem with what you're doing, other than being very inefficient. It's taking 2 hours to process a WU that if you were doing 2 at a time would take about 15-20min.

EDIT- you are doing 3.5WUs per hour.
doing 2 at a time you would process 6-8 (closer to 8) per hour.

Running so many GPU tasks at once also causes bottlenecks in places like the system memory. You may have lots of memory, but all that data has to be shunted in and out which takes time, leaving the CPU and GPU idling while waiting for the next chunk of data to come down the pipeline.

I've done a lot of testing with Win XP and have found that maximum efficiency is achieved running 2 tasks per card. Once GPU usage hits 99% there is nothing more to be gained by upping the number of units per card.

In answer to your original question. Depending on which socket the card is installed in my GPU's run between 50 and 70 deg C. with the fans set at 100%

T.A.