Could someone please tell me if GPUs crunch the same work units as their CPU cousins? Is there any difference in speed between the two?

Could someone please tell me if GPUs crunch the same work units as their CPU cousins? Is there any difference in speed between the two?

The work is the same. There is no such thing as CPU tasks and GPU tasks. There are only tasks to be completed. When a computer asks for work they are assigned to the device that requested the work. Be it a CPU or a GPU.
Often a GPU and a CPU will both end up being assigned the same tasks. So the "wingman" for your GPU tasks could be a CPU or the other way round.

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[

...
Is there any difference in speed between the two?

Usually a lot of difference, though some of the weakest GPUs are slower than the strongest CPUs. This project has their "GPU multiplier" setting at 8, indicating that GPUs are on average 8 times faster than CPUs, but that hasn't been updated in a couple of years. That multiplier is reflected in several places, notably that quotas and limits for GPUs are 8 times higher than CPUs.

                                                                   Joe

Often a GPU and a CPU will both end up being assigned the same tasks. So the "wingman" for your GPU tasks could be a CPU or the other way round.

Can I ask, is that significant?

I guess I am curious if it is more cost effective to buy a good CPU or GPU.

Often a GPU and a CPU will both end up being assigned the same tasks. So the "wingman" for your GPU tasks could be a CPU or the other way round.

Can I ask, is that significant?

I guess I am curious if it is more cost effective to buy a good CPU or GPU.

The easy answer is a GPU.

For instance, a $250US GPU might have a daily credit in the 18,000-20,000 "credits" range. As far as I know, there is no such thing as a $250 CPU that comes close.

Same would be true of a $150 CPU / GPU comparison.

I can't say about a $600 CPU vs GPU. I don't run in those financial circles, but I'd be willing to wager that the GPU is still running rings around the CPU.

Now, a $100 GPU against a $700 CPU? The CPU might win that race.

FWIW, my GTX460's are capable of about a 16,000-18,000 RAC each. My Intel Core i7-950 overclocked to 4ghz is capable of about 10k RAC. GTX460's new at the time were around $200, and besides making sure the motherboard would support it (PCIe 2.0x16 slot) and having a large enough power supply, it was plug and play. A big rowdy 6 core i7-990x is around $1000, and still will just barely compete with one of my single GTX460's. A GTX570 is worth about 25-30k RAC alone, and the GTX580 is worth I'd say around 35K RAC.

In the end, they crunch the same work, and have the same result, the GPU's just tend to be much faster about it. The best thing to do would be to analyze your current setup and choose the best path to take. With your computers being hidden, it's hard to offer any advice.

---

Is there any difference in speed between the two?

My E8400 Core 2 Duo takes about 35-40min for a shortie, two at a time.
My GTX 560Ti takes less than 3 min to do 2.

---

Grant
Darwin NT

FWIW, my GTX460's are capable of about a 16,000-18,000 RAC each. My Intel Core i7-950 overclocked to 4ghz is capable of about 10k RAC. GTX460's new at the time were around $200, and besides making sure the motherboard would support it (PCIe 2.0x16 slot) and having a large enough power supply, it was plug and play. A big rowdy 6 core i7-990x is around $1000, and still will just barely compete with one of my single GTX460's. A GTX570 is worth about 25-30k RAC alone, and the GTX580 is worth I'd say around 35K RAC.

In the end, they crunch the same work, and have the same result, the GPU's just tend to be much faster about it. The best thing to do would be to analyze your current setup and choose the best path to take. With your computers being hidden, it's hard to offer any advice.

Things change in a hurry at the end of the "model year."

I just saw a 560ti for about $130. Of course, the 448 "core" models are still fetching big bucks.

I paid $200+ (one much more +++) for OC and SOC 460s (which, BTW, have been doing SETI's work pretty-much 24/7 for a long time now; I hope I can be as big a fan of some model of 600-series because the 500s were never my favorite).

It feels like yesterday that a 9800GT was about $130.

Often a GPU and a CPU will both end up being assigned the same tasks. So the "wingman" for your GPU tasks could be a CPU or the other way round.

Can I ask, is that significant?

I guess I am curious if it is more cost effective to buy a good CPU or GPU.

The easy answer is a GPU.

For instance, a $250US GPU might have a daily credit in the 18,000-20,000 "credits" range. As far as I know, there is no such thing as a $250 CPU that comes close.

Same would be true of a $150 CPU / GPU comparison.

I can't say about a $600 CPU vs GPU. I don't run in those financial circles, but I'd be willing to wager that the GPU is still running rings around the CPU.

Now, a $100 GPU against a $700 CPU? The CPU might win that race.

To add to that.

Generally GPU's have a greater PPW (Performance Per Watt). Not to say they use less power, but you get more output for the power they consume. As such some people choose to only use their GPU for processing work.

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[

Could someone please tell me if GPUs crunch the same work units as their CPU cousins? Is there any difference in speed between the two?

The work is the same. There is no such thing as CPU tasks and GPU tasks. There are only tasks to be completed. When a computer asks for work they are assigned to the device that requested the work. Be it a CPU or a GPU.

Only read this if you're sure you understand everything else that's been said...

They say every rule has an exception, and here it is. There are certain workunits that are identified by the splitters as being "vlar"s. Never mind what this means, I'm not real clear on it myself. The important thing is that they don't run properly on GPUs, to the point that they can sometimes crash the whole computer. Therefore, the scheduling server won't assign these to your computer when it asks for work for its GPU. This situation leads to other seemingly odd occurrences (that don't really hurt anything) and in the ensuing discussions we tend to refer to "CPU tasks" and "GPU tasks." The important thing to know is that this is really the only time that distinction needs to be made, and you generally don't need to worry about it.

[edit]That problem is only with Nvidia, not all GPUs. (As I understand it). However, don't let that factor affect your decision on what GPU to buy, as I believe they are working on revised code so it won't happen any more.[/edit]

---

David
Sitting on my butt while others boldly go,
Waiting for a message from a small furry creature from Alpha Centauri.

Could someone please tell me if GPUs crunch the same work units as their CPU cousins? Is there any difference in speed between the two?

The work is the same. There is no such thing as CPU tasks and GPU tasks. There are only tasks to be completed. When a computer asks for work they are assigned to the device that requested the work. Be it a CPU or a GPU.

Only read this if you're sure you understand everything else that's been said...

They say every rule has an exception, and here it is. There are certain workunits that are identified by the splitters as being "vlar"s. Never mind what this means, I'm not real clear on it myself. The important thing is that they don't run properly on GPUs, to the point that they can sometimes crash the whole computer. Therefore, the scheduling server won't assign these to your computer when it asks for work for its GPU. This situation leads to other seemingly odd occurrences (that don't really hurt anything) and in the ensuing discussions we tend to refer to "CPU tasks" and "GPU tasks." The important thing to know is that this is really the only time that distinction needs to be made, and you generally don't need to worry about it.

It is just NVIDIA GPU's that have, or did have, the issue with VLAR tasks. ATI GPU's seem to run though them without any issues.

Some people ask "Why are they not sending out any GPU work?" or something to that effect. So it is always good to explain that there is no such thing as one or the other. I would think, as you mentioned, saying you have 100 GPU tasks and 50 CPU tasks would be correct as well. As you are identifying the tasks you have assigned to your devices.

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[

Could someone please tell me if GPUs crunch the same work units as their CPU cousins? Is there any difference in speed between the two?

The work is the same. There is no such thing as CPU tasks and GPU tasks. There are only tasks to be completed. When a computer asks for work they are assigned to the device that requested the work. Be it a CPU or a GPU.

Only read this if you're sure you understand everything else that's been said...

They say every rule has an exception, and here it is. There are certain workunits that are identified by the splitters as being "vlar"s. Never mind what this means, I'm not real clear on it myself. The important thing is that they don't run properly on GPUs, to the point that they can sometimes crash the whole computer. Therefore, the scheduling server won't assign these to your computer when it asks for work for its GPU. This situation leads to other seemingly odd occurrences (that don't really hurt anything) and in the ensuing discussions we tend to refer to "CPU tasks" and "GPU tasks." The important thing to know is that this is really the only time that distinction needs to be made, and you generally don't need to worry about it.

It is just NVIDIA GPU's that have, or did have, the issue with VLAR tasks. ATI GPU's seem to run though them without any issues.

Some people ask "Why are they not sending out any GPU work?" or something to that effect. So it is always good to explain that there is no such thing as one or the other. I would think, as you mentioned, saying you have 100 GPU tasks and 50 CPU tasks would be correct as well. As you are identifying the tasks you have assigned to your devices.

It is actually very rare that the servers don't send out GPU work - the vlar situation is about the only case.

More commonly, when people use that phrase, it turns out that their computer isn't asking for GPU work, when the owner thought it should.. For the solution to that problem, they have to look closer to home.

You are entering dangerous relms here, for your wallet that is:-)
Its not only the GPU you need to think about, but your power supply having the oomph to supply it and that your case is cool enough to keep the GPU within reasonable heat range while its crunching non stop.

Which is what a goodly number of us demented seti-ists do, in the endless struggle to wring the most RAC out of the machines being run.

After the rather expensive GPU endevours, comes the additional machines, built for the sole purpose of 'crunching' work units:-)

It can be rather addictive, and if you are'nt carefull pretty expensive..

So either you need deep pockets or the willpower to resist the temptation to snap up that nice juicy GPU that just been advertised at your favourite etailer:-)

Anyway deep pockets or not the process can be fun, looking around for good bits of kit, and slaving away at builing more and more computers:-)

Regards,

---

Cliff,
Been there, Done that, Still no damm T shirt!

Almost everything i´m getting are VLARs on my GPU.

---

With each crime and every kindness we birth our future.

yeah I keep rescheduling and seti keeps sending me more VLAR's. Time for NNT and start crunching Milkyway until the shorties come home to roost

---

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