hi, im back into seti and i was wondering, how many points for a astropulse unit ? thx

The exact number depends on the individual unit. However, they average around 750 credits each.

---

Cruiser
Gateway GT5692 L-f-S Edition
-Phenom X4 9650 CPU
-4GB 667MHz DDR2 RAM
-500GB SATA HD
-Vista x64 SP1
-BOINC 6.2.19 32-bit client
-SSE3 optimized 32-bit apps

Hi, yes average between 720 and 750, most off them about 720 , my experience, too.
On my QX9650 they took 22 hours, without the potimized app., but the CPU runs @ 3,410GHz and uses PC8500, OCZ-REAPER(with heatsink), runs at 410MHz (@ the moment);FSB=410x4x(6/5x)=~1355MHz
Haven't tested the new app. , also because I'de like a 64BIT version ;).
[Maybe I should compile one , myself!?]

---

Have they worked out the credit equity issue, where the MB wu's were getting more credits than an equivalent cputime applied to AP?

NO
MB gets more credit my RAC has taken a beating! Due to the AP WU credit deficiency they should be at least 1200 credits if not more!

Have they worked out the credit equity issue, where the MB wu's were getting more credits than an equivalent cputime applied to AP?

I think the fix is in place, it's just going to take a while for enough folks to crunch AP for it to finally "dial in."

[edit]On my primary workstation, the optimized MB doesn't give much leverage, while optimized AP is twice as fast as stock.[/edit]

---

With the AKV8-SSSE3 apps for MB my time is around an hour for MB WU for about 45 Credits or a RAC of 3600 at 80 WU/20hrs and with the AKV8-SSSE3 apps for AP my time is around 20 Hours for an AP WU for about 750 Credits for a RAC 3000 at 4 WU in 20 hrs. So for the same time less credit for AP as compared to MB.
So I am running at a deficiency by running AP WU’s, which does not include a whole 24 hrs in which case my MB RAC would be even higher.
Not only that but my pending credit is close to 16000! So I am doing work and will have to wait till who knows when I will get the credits. Sounds like the Soviet/Chinese work system do the work and maybe I will get my credits.

With the AKV8-SSSE3 apps for MB my time is around an hour for MB WU for about 45 Credits or a RAC of 3600 at 80 WU/20hrs and with the AKV8-SSSE3 apps for AP my time is around 20 Hours for an AP WU for about 750 Credits for a RAC 3000 at 4 WU in 20 hrs. So for the same time less credit for AP as compared to MB.
So I am running at a deficiency by running AP WU’s, which does not include a whole 24 hrs in which case my MB RAC would be even higher.
Not only that but my pending credit is close to 16000! So I am doing work and will have to wait till who knows when I will get the credits. Sounds like the Soviet/Chinese work system do the work and maybe I will get my credits.

One of the problems with credit is that it will vary by CPU type and instruction mix. Makes it hard to determine the proper credit multiplier.

Eric Korpela came up with a method of comparing the average granted credit to the predicted credit for a median machine, and slowly adjust multipliers to bring credit in line.

I believe that process is still running.

However, when you compare "Optimized" Multibeam to "Optimized" Astropulse, one of the things you're comparing is how well the optimizers have done in optimizing.

The disparity is much higher between Optimized Multibeam and stock AP, of course.

---

It is similar to work pay at job MB for 20 hrs I get 3600 Credits while at job AP for 20 hrs I get 3000 Credits. MB is a better paying job than AP. I should quit AP and get a second job at MB.
Another way of looking at it is less science is done for AP than for MB for the same amount of time or electricity

It is similar to work pay at job MB for 20 hrs I get 3600 Credits while at job AP for 20 hrs I get 3000 Credits. MB is a better paying job than AP. I should quit AP and get a second job at MB.
Another way of looking at it is less science is done for AP than for MB for the same amount of time or electricity

Except, when you work, your "credits" (paycheck) can actually be used for something. Cobblestones aren't worth the paper they're printed on.

---

It is similar to work pay at job MB for 20 hrs I get 3600 Credits while at job AP for 20 hrs I get 3000 Credits. MB is a better paying job than AP. I should quit AP and get a second job at MB.
Another way of looking at it is less science is done for AP than for MB for the same amount of time or electricity

You are still comparing apples and oranges. The MB opti apl. is SSSE3 while the AP opti apl. is SSE3. The AP opt. still as a way to go to match the MB optimization.

That is not to say that there isn't a disparity in the credit issued. Personally I don't care. I just see a new and possibly better way to search for ET.

It is still an achievement and I achieve less with AP than I do with MB. I can process more signals-WU with MB than I can with AP such as 80 signals-WU with MB versus 4 signals-WU. With AP for the same amount of very real electricity and wear and tear on my equipment so it looks like more science is done with MB as compared to AP.
To put it another way it cost me about 20USD a month to run my computer 24/7 so for that 20 dollars for the month and in 20 hrs during the month I can do 80 WU of MB VS 4 WU of AP. So is a signal = a WU? If so Than I can process 80 signals vs. 4 signals for the same amount of electricity or money.
To put another way I get fewer rewards in the form of SETI Credits for AP WU than I do for a MB WU. The AP WU should have more value than a MB WU because it uses more resources to do an AP WU. But instead I receive fewer credits for the same resources.

It is still an achievement and I achieve less with AP than I do with MB. I can process more signals-WU with MB than I can with AP such as 80 signals-WU with MB versus 4 signals-WU. With AP for the same amount of very real electricity and wear and tear on my equipment so it looks like more science is done with MB as compared to AP.
To put it another way it cost me about 20USD a month to run my computer 24/7 so for that 20 dollars for the month and in 20 hrs during the month I can do 80 WU of MB VS 4 WU of AP. So is a signal = a WU? If so Than I can process 80 signals vs. 4 signals for the same amount of electricity or money.
To put another way I get fewer rewards in the form of SETI Credits for AP WU than I do for a MB WU. The AP WU should have more value than a MB WU because it uses more resources to do an AP WU. But instead I receive fewer credits for the same resources.

A work unit is the recorded output of a channel on the multibeam receiver on the telescope in Puerto Rico.

The "multibeam" application looks for narrow band signals of three different times.

The Astropulse application looks at a larger sample of data, and looks for broadband signals.

If it helps, think of multibeam as looking for old-style Morse Code dots and dashes, while Astropulse is looking for video or WiFi.

If I remember correctly, Multibeam will look for 20 signals, and if it finds 20 that is the -9 overflow "error" -- some work units have thousands of signals.

Astropulse has a slightly higher threshold at 30 signals.

If you are interested in the "science" it is hard to say which is more valuable. Ultimately, we're sifting through this huge field of hay, and somewhere in there is one very tiny needle.

... and since we know the "needle" is of extraterrestrial origin, we're not even sure what it is going to look like, or what it is made from.

But that's science, and BOINC projects do not pay for science, they pay for work.

... and they pay in Cobblestones, which are fun to talk about, but they're worth less than Canadian Tire Money.

All of the machines crunching for BOINC here are running anyway.

---

Thank you Ned for the help with the difference and in the signals we are working with. I just know that when I go to the SETI market shopping for oranges (MB) and apples (AP) I can buy more oranges (MB) than apples (AP) with my $20 in electricity each month.

Thank you Ned for the help with the difference and in the signals we are working with. I just know that when I go to the SETI market shopping for oranges (MB) and apples (AP) I can buy more oranges (MB) than apples (AP) with my $20 in electricity each month.

Not really.

When you grab a box of SETI-oranges, it weighs a certain amount, and is a certain size. The SETI-apples have a different weight, and they come in a bigger box.

... but we're still comparing apples and oranges, and they aren't the same.

You're talking about "science" and we aren't technically paid for "science" we're getting Cobblestones for work. That's an important distinction -- if a work unit is completely quiet, no signals at all, you still get "paid."

When the SETI@Home project actually finds a signal that represents intelligence, that is when we find out who actually did "science."

I know, that isn't as exciting, but we need to be honest with ourselves as to what we're actually doing.

Milo Bloom said "The first sign of a nervous breakdown is when you start thinking your work is terribly important."

---

I believe I have made my point if I do AP my RAC will suffer. Any more comment is like talking to a wall. AP WU is not fair and operates at a deficit.

I believe I have made my point if I do AP my RAC will suffer. Any more comment is like talking to a wall. AP WU is not fair and operates at a deficit.

As usual, "Your mileage may vary."

When I look at a recent work unit you completed, I see this:

Windows optimized S@H Enhanced application by Alex Kan
Version info: SSSE3x (Intel, Core 2-optimized v8-nographics) V5.13 by Alex Kan
SSSE3x Win32 Build 41 , Ported by : Jason G, Raistmer, JDWhale

A version of multibeam that completes a work unit significantly faster than the "stock" client.

I have no idea how much time Alex Kan spent finding ways to do the same work in less time, but I know it took a lot of time, and more than a little skill. Kudos also to Jason G, Raistmer, and JDWhale, who spent a lot of time making it work under Windows.

So, absolutely, the first-pass at optimizing Astropulse is not going to be as fast as several generations of optimized Multibeam -- but I'll bet it catches up.

... and the only argument that fits is "my RAC will go down."

Mine (while never very large) has actually gone up crunching only (optimized) AP. The optimized Multibeam apps. aren't that fast on my CPU.

---

...I have no idea how much time Alex Kan spent finding ways to do the same work in less time, but I know it took a lot of time, and more than a little skill. ...

I picture it, perhaps in my overly romanticised way, as among the thousands of lines of code, a particularly sticky 'hotspot' is analysed, white-boarded in arcane illegible scribbling, cigarettes smoked and bourbon consumed, tested, then rejected in favour of another approach (perhaps vodka). rinse and repeat. which means that thousands of lines may sometimes take mere hours, but key optimisations may take days or weeks for even a few measly lines of code.

Of Alex's code the particularly arcane sections are the ones that require a little lie down and a cup of tea (in my case) after working out what they actually do. Those are usually the really effective ones from which there can be learned a lot about the hardware, when read in tandem with CPU manufacturer optimisation manuals anyway.

Now consider that hardware is right in the process of some fairly major evolutionary steps, and you can multiply that image by a few people headed in different exploratory directions... exciting times.

Jason

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

...I have no idea how much time Alex Kan spent finding ways to do the same work in less time, but I know it took a lot of time, and more than a little skill. ...

I picture it, perhaps in my overly romanticised way, as among the thousands of lines of code, a particularly sticky 'hotspot' is analysed, white-boarded in arcane illegible scribbling, cigarettes smoked and bourbon consumed, tested, then rejected in favour of another approach (perhaps vodka). rinse and repeat. which means that thousands of lines may sometimes take mere hours, but key optimisations may take days or weeks for even a few measly lines of code.

My picture is much the same, except there are no cigarettes, Jolt! instead of Bourbon, and you left out the double-stuffed Oreos (which would work no matter what beverage is involved).

---

I have no idea how much time Alex Kan spent finding ways to do the same work in less time, but I know it took a lot of time, and more than a little skill. Kudos also to Jason G, Raistmer, and JDWhale, who spent a lot of time making it work under Windows.

Looks like Alex released his first one on May 27, 2006, then he just kept working in more optimizations and speed as he went.

---