Thank you for the advise. I suspended and then restarted the stalled WU and it is moving again. It has progressed more than 1% in only 2 minutes. Thank you again.

---

Hal900 I noticed your stats don't have Einstein so I'm asumeing you havn,t done it and it's been a while since I did all I can remember they did take a long time just can't remember how long so he mite not have a problem wish I could member how long they took .

Your get more credit per unit as they do take a while to complete my Rac doubled I think when I did it and why it's at 1.9 mill and seti is only just got passed 1 mill

So if HT isn't the prob maybe he don't have a prob now I think back AP's mite just take some 36hrs the Gpu I think took 9hrs oh well

Change to Seti and then he can get a better comparison

---

Thank you for the advise. I suspended and then restarted the stalled WU and it is moving again. It has progressed more than 1% in only 2 minutes. Thank you again.

Jim, all of the advise you've been given is trying to point you toward something.

A GPU work unit still requires CPU work. With AstroPulse this is a sneaky percentage of a core's ability to crunch. Why sneaky? Well, there are work units that run with very little CPU time, and then there are these other work units that take almost all of a CPU's time in addition to the GPU.

If you are running one of those *and* you have your CPUs completely busy running other things, the GPU work unit is "fighting" for CPU-time. That's really bad news because it means the CPU task is also having to start and stop and move to RAM and move back from RAM and pick-up where it left-off. It's also causing the GPU work unit to do that. All of that swapping-around is taking-up a lot of time.

In general, then, you won't get the most work done in the least amount of time by running every possible work unit all at once. If the computer is capable of running five, you might want to run four, or even three.

If you keep-up with it for a while, you may find that the average time that four work units take to complete is faster if you do them two at a time.

That's going to depend on the specific work units and the specific projects that you have running.

Don't let some of the numbers you see fool you. If it says, "0.24 CPU and 1 GPU" that doesn't mean it is "only" going to use .24 of the CPU's capacity or that it will use 100% of the GPU's capacity. If it were that simple, trying to be helpful would be a whole lot easier.

It's a balancing act and the balance-point changes for each set of circumstances.

Hal900 I noticed your stats don't have Einstein so I'm asumeing you havn,t done it and it's been a while since I did all I can remember they did take a long time just can't remember how long so he mite not have a problem wish I could member how long they took .

If we're talking about Einstein GPU Perseus Arm Survey tasks, depending on the GPU they might take anywhere from 2 hours, 2-at-once, on a R9 270X to nearly 8 hours, one at a time, on a GT240, to almost 5 hours, 2-at-once, on a GTX 750Ti.

What everyone who does Einstein should know is to leave one "core" (or possibly a thread) vacant for the considerable shuttling of data to-and-from RAM that Einstein requires. The GPU needs more CPU resources at Einstein than it does here at SETI.

EDIT: If you want to know some of these crunching times, a look at my machines will give you some ideas. I'm running the following: GT240, 460, 550Ti, 560, 560Ti, 660Ti, 670, 750Ti, 770, HD 6770, HD 7770, R7 260X, R9 270X at Einstein and SETI.

I've been playing-with settings so you'd have to ask me how I have "X" machine configured (2 at once, 1 at once, etc) but other than a few accidental CPU tasks, all of my work is GPU with the CPUs only feeding the GPUs.

AMD CPUs just aren't that great at crunching and the FX series with shared FPUs are even not-as-good "per core and clock cycle" as the Phenom IIs were. But they can move data around pretty fast. I can keep four GPUs busy running eight total tasks for less than 50% of the processor's capacity. Sometimes much less.

Jim, all of the advise you've been given is trying to point you toward something.

Hi tbret, Jim wandered into this thread with a stuck work unit question. Robert actually started the thread and it was he that is having the issues. Jim, the messages tbret posted are actually for Robert. That being said, it is good information for you to have as well since you are starting out with us and it helps explain a lot of how Seti works.

Zalster

Tbret I'm running 7 of the 8 cores and 2 on the GPU and work's great GPU times approx. 26mins per unit on the GPU and aprrox 2.5-3.5hrs on the CPU units how's that compare to the Phenom IIs or Intel

---

Lunatics 64-bit SSSE3 app & HT enabled, while running 8 tasks, provided 31.2% greater output & consumed 11.1% more power. Compared to just running 4 tasks.

Hal9000 I am thinking that I'm not that far off about HT as you said you only get a 30% increase but your doing 100% more work so maybe it's just the fact your doing so many that gives a increase at all .

I would expect there to be at least 80% increase not only 30% which is fine for the better chips but a 2 core I3 with 4 threads mmmmm

---

Tbret if what your saying about the AMD is correct then AMD can get in big trouble I think the box that it came in says 8 cores and 8 threads not 4 cores and 4 threads all very confusing

Is AMD lie'ing about what is on there boxes ? mmm

Can I get some cash sueing them for false claims on there package ? mmm

I need cash please someone tell me yes you can there lie'ing

---

Im running stock MB only this host I can definately state That if you dont a have a GPU running and have HT on you will have a bigger RAC.
Currently I have 4 cores running HT off MB only with the GPU. ( An I7 920 with a GTS 250 ) RAC is still rising, At 4,539. Thats only 800 average less than running lunatics with 8 cores and GPU running MB and AP.

I will drop a core to three and see what the RAC ends up at. I am doing it that way as its faster to see RAC drop then rise.

I dont agree with the dogma of no HT and freeing a core to feed a GPU. Maybe for the higher end cards running two at a time. But not for all machines, Thats why Im testing.

---

[/quote]

Old James

Tbret if what your saying about the AMD is correct then AMD can get in big trouble I think the box that it came in says 8 cores and 8 threads not 4 cores and 4 threads all very confusing

Is AMD lie'ing about what is on there boxes ? mmm

Can I get some cash sueing them for false claims on there package ? mmm

I need cash please someone tell me yes you can there lie'ing

Glenn,

If you haven't kept-up with all of this since the original Bulldozer was first released, I really almost don't know where to start except to suggest you do a search for "Bulldozer architecture."

If you read through that, or "Piledriver architecture" you will discover that a "core" doesn't mean what it used to mean, or more precisely it is beginning to mean what it used to mean a very long time ago.

The FX-8350 has four "compute modules" which are made-up of 2 integer modules and one floating-point module. So it has 8 integer cores, but only four floating point units.

The Phenom 1100T has six integer cores and six floating point units, one for each core.

So not merely in theory, but in actual practice, a Phenom II 1100T can run six threads each simultaneously executing FPU calculations which is 50% more than an 8 core FX processor can do.

BUT, if the FPU isn't required in every thread every time, then the Phenom is "wasting" space the same way a car with a v8 that only needs a v6 is wasting space and power and is heavier wasting gasoline, etc.

This was a trade-off that AMD decided to make: Run the whole thing much faster and share the bits that are not always in use.

I decided it is faster for me to look it up than it is for us to argue back and forth if you decided not to look it up:

http://www.tomshardware.com/reviews/fx-8350-vishera-review,3328-3.html

Notice under "Putting it all together" you will find this:

"In iTunes, which we know to be single-threaded, the FX-8350 demonstrates significant gains over the Bulldozer-based FX-8150. But a Phenom II X6 1100T operating at the same frequency is still faster. And that's before we look at the Sandy and Ivy Bridge architectures, which jump way out in front of anything from AMD."

EDIT: By the way, Glenn, and anyone else reading out there; yes the 8350 at 4GHz does many things faster than a Phenom II at 3.3GHz. Who would ever have thought that??

Jim, all of the advise you've been given is trying to point you toward something.

Hi tbret, Jim wandered into this thread with a stuck work unit question. Robert actually started the thread and it was he that is having the issues. Jim, the messages tbret posted are actually for Robert. That being said, it is good information for you to have as well since you are starting out with us and it helps explain a lot of how Seti works.

Zalster

Yeah, I actually knew that. Look at the advice Jim was given regarding his stuck work unit and you'll see I was responding to that, not to the fact that a restart fixed it, or didn't fix it.

The advice he was given may have nothing to do with his problem, but the advice was nevertheless sound and should result in him having better luck and perhaps fewer stuck work units in the future.

Just absolutely for anyone who might accidentally happen across this entire thread: It isn't about the total number of tasks you can make run at one time. It is, and it always has been, about the total number of useful calculations you can get done per-hour. The two things are never the same.

Hal900 I noticed your stats don't have Einstein so I'm asumeing you havn,t done it and it's been a while since I did all I can remember they did take a long time just can't remember how long so he mite not have a problem wish I could member how long they took .

Your get more credit per unit as they do take a while to complete my Rac doubled I think when I did it and why it's at 1.9 mill and seti is only just got passed 1 mill

So if HT isn't the prob maybe he don't have a prob now I think back AP's mite just take some 36hrs the Gpu I think took 9hrs oh well

Change to Seti and then he can get a better comparison

I have not run Einstein. It is just one I haven't chose to play around with yet.

As I have a machine rather similar to the OP's machine I can see there SETI@home performance is a bit lower than expected. So I suggested changes for improving it. Having and older OpenCL runtime with a newer driver can be a source of problems. Changes to improve their SETI@home performance I imagine would carry over to Einstein.

Lunatics 64-bit SSSE3 app & HT enabled, while running 8 tasks, provided 31.2% greater output & consumed 11.1% more power. Compared to just running 4 tasks.

Hal9000 I am thinking that I'm not that far off about HT as you said you only get a 30% increase but your doing 100% more work so maybe it's just the fact your doing so many that gives a increase at all .

I would expect there to be at least 80% increase not only 30% which is fine for the better chips but a 2 core I3 with 4 threads mmmmm

Your expectations for HT are to to high then. The normal expected gains from HT are 15-35%. When using HT tasks run a bit slower, but running twice as many = more work done over time. This machine Will run 8 AP at a time with times around 16 hours vs running 4 at a time in around 12 hours. So in 48 hours HT on = 24 tasks & HT off = 16 tasks. When running a reduced number of tasks with HT on times are somewhere in the middle.

So for SETI@home and what others have said about running Einstein on the GPU. It would seem they would want to run 3 CPU tasks & 1 GPU task to improve their performance.

---

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

Thanks tbret still confused but I guess I can't sue them grrrrr

only had 1 chip that had HT a Pentium 4 in a 775 socket with 1 core but the m/b had HT so I thought it was a Duo as it started up and had 2 core show up in the task manager and was so dam slow with seti so I turned HT off and it speed up so that's what I have used as a reference

---

So for SETI@home and what others have said about running Einstein on the GPU. It would seem they would want to run 3 CPU tasks & 1 GPU task to improve their performance.

As you know, that's going to depend on the whole system.

AMD CPUs aren't great crunchers. So not-crunching on a core penalizes the machine's throughput less than would be the case with a roaring i7.

And we all know that if you throttle the GPU's work by over-tasking the CPU, there is no way the CPU's additional productivity can be as great as the declining productivity on the GPU.

Doing a GPU work unit at Einstein requires more CPU resources than a GPU work unit requires here at SETI.

Depending on the specific tasks you are running at Einstein, you can usually do very well by doing two-at-a-time on a moderate to high-end card, but you might have to leave another CPU "core" or "thread" open. That's fine since the GPU out-works the CPU by a factor of 10x or more.

But, like you and I both know, everyone's mileage will vary depending on their RAM configuration, the speed of there board, the speed of their card, what components they are mixing, the mix of work on their machine, etc. The only *real* way to know is to try.

I've found that a few of my cards and a couple of my processors are REALLY slow at Einstein compared to how well they crunch SETI. So, I'm going to discontinue running Einstein on those. I've found that a few of my cards are really good and running Einstein and are nothing special at SETI, so I'm going to let those run Einstein.

A lot of that has to do with how the programs were written and how old the code is.

For my own part, I'm finding that my predictions about how my own machines are going to act across projects are pretty bad. A piece of gear that impresses me here seems destined to disappoint me over there and vice-versa, unless for some reason it doesn't.

Take the case of one of my <$230 AMD GPU and my >$350 NVIDIA GPU. The AMD *smokes* the NVIDIA at Einstein, but the AMD is disappointing to the point of embarrassment at SETI. It all seems to be in the coding and the calculations being done. One fits the architecture here better, the other there.

And a lot of that has to-do with the age of the code. Imagine running CUDA 6 capable cards on CUDA 3.2 programs, or vice-versa again. You just can't generalize and make it count.

A lot of why the NVIDIA cards do so well here probably has to do with not-only the calculations, but also because of our great volunteer coders and tweakers who constantly endeavor to make sure SETI@Home code is optimized and up-to-date.

Not every project has those talented people involved.

It makes you sick to think of the FLOPS-undone just because nobody wants to update the programs at those "other projects."

I updated my video driver, suspended SETI@home and Milkyway@home, and gave exclusive processing power to Einstein@home and I'm getting about 0.001% per second. Its at 98% now with only 45 minutes left.

It was so slow because I tried to run it with four, sometimes even five, SETI@home tasks.

I appreciate all the feedback. I think I'll suspend Einstein@home for a while and let SETI and Milkyway run.

It was so slow because I tried to run it with four, sometimes even five, SETI@home tasks.

On your video card?
For MB (Multi Beam), 2-3 (usually 2) is the optimum amount. Any more than that slows things down, significantly. For AP (Astro Pulse) I would suspect 1-2 would be the most you would want to run at one time, depending on the card.

---

Grant
Darwin NT

Good to here you have it sorted sorry I have not posted but I had been a naughty boy and did not wish to get banished so I stayed away from the thread

---

Here is what my computer is doing right now:


http://imgur.com/u0HVM6d

One thing I've learned over my years of fixing computer problems is to know when to leave something alone. Of course I'd like to control how each and every task uses my computer. But I don't know how yet and I don't want to screw anything up or upload false or corrupt data. For now I'm just letting BOINC take over.

I've stopped Einstein@home and instructed Milkyway@home to not download new tasks.

I really appreciate all the responses.

My other questions were about the project itself. If other intelligent life doesn't in fact use radio signals, is capturing pulses and triplets at least providing something of value to science?

If other intelligent life doesn't in fact use radio signals, is capturing pulses and triplets at least providing something of value to science?

Probably not.

---

My other questions were about the project itself. If other intelligent life doesn't in fact use radio signals, is capturing pulses and triplets at least providing something of value to science?

That's specifically why I gave you a link to Eric's video.

That question was asked at the end of his presentation.

I thought his answer was honest. His inflection and body language when giving the answer are important and a transcription of what he said will not communicate the entirety of his meaning.