It's a Prescott 4E @ 2.8GHz, motherboard uses 915G chipsets and have PCI-E x16.

It haven't been use for a long time...

I knew, I won't use the CPU to do the project, only GPU.

If you want to make a "dedicated device" out of it, why not?

It haven't been use for a long time...

Why not? It's embodied grid energy.^^ http://en.wikipedia.org/wiki/Embodied_energy

---

My reasons to participate in grid computing:
1. I'm using true renewable energy (German accreditation; no certificate purchasing)
2. Production and disposal (will) amount to about 95% of the energy "used" during the lifetime of my PC
3. Helping

Yes, so I think maybe it's better to wake it up, but the P4 CPU is hundreds times slower than a tesla, but use nearly half of electricity, so it's not efficient.

As far as your original question goes: Yes, a graphic card with GPGPU would be ideal for you. May that be AMD FireStream or Nvidia CUDA.

This is more a question for a hardware board but 915G was introduced in June 2004. I'm guessing you have PCI Express Version 1.x but it's backward compatible: http://www.tomshardware.co.uk/answers/id-2013016/pci-express-gpus-backwards-compatible-slots.html
So version-wise you can put a PCIe 2 card into a PCIe 1.x slot.

Size-wise you can also put a PCIe x1 card into a PCIe x16 slot:
http://www.tomshardware.co.uk/forum/205984-33-card-slot

Look also at this post and thread for "efficiency":
http://setiathome.berkeley.edu/forum_thread.php?id=65362&postid=1149951

Microarchitecture is another question:
http://en.wikipedia.org/wiki/CUDA#Supported_GPUs
I think you could even go for Fermi ...

As you can see I thought about upgrading myself but I'm also accounting production, transport and the additional power consumption into the efficiency formula ... YMMV.

---

My reasons to participate in grid computing:
1. I'm using true renewable energy (German accreditation; no certificate purchasing)
2. Production and disposal (will) amount to about 95% of the energy "used" during the lifetime of my PC
3. Helping

And for that matter, there is no need to go for a Tesla. That may be a dedicated co-processor, but then you pay for that as well. You can get quite the same result by just adding a cheap AMD or Nvidia GPU. All Nvidia GPUs from an 8300GT and higher will be able to do CUDA. There's also no real need to buy a new card for that computer, second or third hand is good enough.

---

All Nvidia GPUs from an 8300GT and higher will be able to do CUDA.

You probably mean 8500GT or 8400GS. GPU model should be G80 upwards (Tesla is the microarchitecture. Cards called Quadro could even be Kepler.).
http://en.wikipedia.org/wiki/CUDA
https://developer.nvidia.com/cuda-gpus
http://en.wikipedia.org/wiki/List_of_NVIDIA_Graphics_Processing_Units

Also note that some projects need a certain "Compute Capability":
http://boinc.berkeley.edu/wiki/GPU_computing

There's also no real need to buy a new card for that computer, second or third hand is good enough.

Now there are other problems with second hand. One counter-argument is the encouragement of the throw-away society. Sorry.

---

My reasons to participate in grid computing:
1. I'm using true renewable energy (German accreditation; no certificate purchasing)
2. Production and disposal (will) amount to about 95% of the energy "used" during the lifetime of my PC
3. Helping

All Nvidia GPUs from an 8300GT and higher will be able to do CUDA.

You probably mean 8500GT or 8400GS.

No, I meant an 8300. But I didn't check what letters it had before or after on the Nvidia CUDA list, instead I wrote that from memory. Doesn't really matter, most all Nvidia models, from Geforce 8xx0 models onwards, support CUDA 1.0 or above.

Tesla is the microarchitecture.

When you tell me you have a Tesla in your computer, I'll think of this thing:

The rest of them are Geforce models, with a number. To me. To most of the people using them, I think. But if you want to be politically correct on these things, go ahead.

Next I'll be talking about the electric sports car Tesla.

---

Tesla is the microarchitecture.

When you tell me you have a Tesla in your computer, I'll think of this thing:

The rest of them are Geforce models, with a number. To me. To most of the people using them, I think. But if you want to be politically correct on these things, go ahead.

Next I'll be talking about the electric sports car Tesla.

:P Relax, that was for Sam and my own clarification. I didn't want to better you. Peace?^^

---

My reasons to participate in grid computing:
1. I'm using true renewable energy (German accreditation; no certificate purchasing)
2. Production and disposal (will) amount to about 95% of the energy "used" during the lifetime of my PC
3. Helping

Or talk about ol' Nick:

---

No you!

---

My reasons to participate in grid computing:
1. I'm using true renewable energy (German accreditation; no certificate purchasing)
2. Production and disposal (will) amount to about 95% of the energy "used" during the lifetime of my PC
3. Helping

All Nvidia GPUs from an 8300GT and higher will be able to do CUDA. There's also no real need to buy a new card for that computer, second or third hand is good enough.

Motherboards with integrated 8200 GPUs can also do Cuda:

http://www.nvidia.co.uk/object/geforce_8200mgpu_uk.html

Message 8756

I have a feeling there are 8100 integrated GPUs, although I can't find reference of them on Nvidia site any longer, I did report about them some years ago:

http://setiathome.berkeley.edu/forum_thread.php?id=70821&postid=1338421

Edit: found it:

GeForce 8100 mGPU

Claggy

Yes, we've been there already. :-)

I wrote:

Doesn't really matter, most all Nvidia models, from Geforce 8xx0 models onwards, support CUDA 1.0 or above.

(source)

---

Yes, we've been there already. :-)

I wrote:

Doesn't really matter, most all Nvidia models, from Geforce 8xx0 models onwards, support CUDA 1.0 or above.

(source)

Ah yes, you have, I saw the pictures, then nothing else. ;-)

Claggy