Yep, I used the v0.37 master installer in December.

I note confusion between the stats of Fred W and Sutaru Tsureku for Seti@Home



Fred W: 8m, rank 869, 99.925%
ST: 60m credits, rank 118, 99.798%


Why is Fred's % higher even though his credits are lower? Maybe one is individual rank and the other team rank?

Those are not ST's stats, but those of the his team.

---

...
By the way, if there is an update to Lunatics, how does that work? Will my Boinc just stop processing because the code will not match the new required version? I presume Lunatics updates when the official Seti@Home does.

We Lunatics do attempt to have updates ready when they're needed. The project normally tests new versions at SETI Beta before bringing them here, so we can follow the changes well.

"Required version" is something the project hasn't used here, though they could. As of now any application for setiathome_enhanced or astropulse_v505 is eligible to get work, though if it does not turn in some results which validate the amount of work it gets will be reduced to one task per day soon enough to protect the project.

There's an RSS feed of Lunatics release announcements, and those are also posted here in the Optimized Applications and Other Binaries - Read Only thread. We hope all our users will at least use one of those methods to keep informed, and of course a new release usually generates new threads here with user experiences noted.

                                                               Joe

By the way, if there is an update to Lunatics, how does that work? Will my Boinc just stop processing because the code will not match the new required version? I presume Lunatics updates when the official Seti@Home does.

Nope - you have to keep checking back here for updates. What you have won't stop processing anyway but it is one of the burdens of running Lunatics builds that you are locked-in to the Boards ;)

F.

That is only if Seti comes out with a new application that's compatible with the old application,

If they were to come up with a new application that's incompatible, and use a new app name, say setiathome_enhanced_plus then Lunatics would need need to produce new apps & Installer targeting the new app name,
users using the old app name may or may not run out of work depending if Seti continued to split work work for the old app name,

Claggy

Hi,

I do have bought a gtx460 today. It runs in my amd six core and it replaces the gt9800 (there is no label on the card, might be a 9600) card. Now the questions:

- what do i have to set up for the 460 ? (cuda+boinc is running fine) ?
- how can i benchmark the system (compare to my intel quad + gtx460se) ?
- does it make sene to run the gtx460 and the 9800 together (the MB can do) ?
- last ... am i crazy (i bought those cards only for seti) ?

Has someone run tests to determine if it's worth it for 3 on a 1GB GTX 460?

On my EVGA GTX460 FTW Edition (1GB memory) running @ 850/1700/2000 (core/shader/memory) 2 x .41 AR WU's complete in about 19 - 20 mins and 3 x .41 AR WU's complete in about 29 - 30 mins so either way I complete about 6 WU's per hour with the fan running at 55% and the temp staying just below 60 deg C.

F.

---

Yep, I used the v0.37 master installer in December.

I note confusion between the stats of Fred W and Sutaru Tsureku for Seti@Home



Fred W: 8m, rank 869, 99.925%
ST: 60m credits, rank 118, 99.798%


Why is Fred's % higher even though his credits are lower? Maybe one is individual rank and the other team rank?




> You could let run (if you don't do it already ;-) 3 WUs/GPU simultaneously (GTX4xx-5xx 1024 MB RAM cards) for max RAC.

Has someone run tests to determine if it's worth it for 3 on a 1GB GTX 460? :)

My own system, due to other uses, would not be a good test system since I don't "leave it alone".

If you look up in the thread you'll see my detailed findings on the difference between 2 WU's vs 3 WU's, and a good discussion on the different aspects of its performance per other GTX 460 owners. If you don't leave your PC alone, or if you run a separate project on the CPU then 2 WU's on the GTX 460 is optimal.

But if you dedicate 1 core of your CPU to feed the GTX 460, then it is my understanding that 3 WU's are optimal(I remember seeing a 10-15% gain in throughput being mentioned by another GTX 460 owner).

This is a lengthy thread, but the information available on it will let you know what this specific card is capable of under varying circumstances.

Cheers, and happy crunching!

---

If you don't leave your PC alone, or if you run a separate project on the CPU then 2 WU's on the GTX 460 is optimal.

Perhaps it's the mix of projects I run on the CPU that gives the performance that I quoted - I run S@H on 2 cores and Climate Prediction on 2 of my overclocked Q9450; perhaps the demands of CPDN are such that there is sufficient headroom for those cores to service the GPU without degrading the performance but certainly I get approx 6 S@H CUDA (.41 AR) WU's per hour whether I run 2 or 3 at a time.

F.

---

If you don't leave your PC alone, or if you run a separate project on the CPU then 2 WU's on the GTX 460 is optimal.

Perhaps it's the mix of projects I run on the CPU that gives the performance that I quoted - I run S@H on 2 cores and Climate Prediction on 2 of my overclocked Q9450; perhaps the demands of CPDN are such that there is sufficient headroom for those cores to service the GPU without degrading the performance but certainly I get approx 6 S@H CUDA (.41 AR) WU's per hour whether I run 2 or 3 at a time.

F.

It was definitely you who I remembered said that 3 was better for you. You make a good point, i'm sure it has to be CPU demand. i'm not sure how the Q9450 stacks up to a Phenom II 940, but mine is OC'd to 3.4 ghz (yeah not much I know) :P

Like I mentioned earlier, I only use the 460 for SETI, whilst all 4 of my CPU cores are 100% dedicated to Einstein@home (and those are some heavy WU's for sure)

I'm not sure if my setup is 100% optimal, or if there is room to mess around still (putting aside the fact that I still have OC headroom on my CPU and GPU). And of course, as it is common to say with these components, your mileage may vary!

---

Recently I have switched over to Windows 7 Professional x64. Now if you recall or care to read my earlier postings on this thread, you will remember that I was claiming a steady APR of 128-132 WU's per day on my computer. Previously, it was running Windows Vista Home Premium x86, but otherwise is running with the same exact configurations as before, and now i'm getting the following:

Number of tasks completed: 967
Max tasks per day: 1028
Number of tasks today: 171
Consecutive valid tasks: 967
Average processing rate: 146.77914758267

So is the gain of around 15 WU's per day attributable to better OS resource management on Win 7 Pro x64?

---

Maybe, maybe not.

Shouldn't the line read "number of tasks validated today"?? When I started doing AP's last year I would turn in one or two daily, but IIRC they were not counted until they validated.

If that's the case, then the number is influenced by your wingmates (your pending credit) as they validate.

It could fluctuate a lot from day-to-day, even if your 460's output does not vary much at all.

Martin

Actually, I think it means "Number of tasks sent today." Looking at one of my hosts, the APR entry for Astropulse v505 says:

Number of tasks completed 40
Max tasks per day 150
Number of tasks today 1
Consecutive valid tasks 49
Average processing rate 9.934678039004
Average turnaround time 2.04 days

The 1 task today is this one, which as of this posting is still in progress. (My host is #223772)


-- Tony D.

---

And the 'average processing rate', or APR, is a measure of the device speed in GFlops.

And the 'average processing rate', or APR, is a measure of the device speed in GFlops.

Well I thought it meant average amount of WU's processed per day. I see how you have a point in it being GFlops, but then it would be much much higher, a 460 is near a teraflop if i'm not mistaken. I'm just trying to clear up some confusion :) thanks for the replies!

---

Actually, I think it means "Number of tasks sent today."

I don't know for sure (I suppose it is detailed in the code somewhere...),

but 10 mins before 00:00 UTC,my app details were:

MB CPU = 1 , MB GPU = 150 , AP = 0

and those "tasks today" numbers weren't even close. My actual numbers, gathered from counting tasks completed and uploaded were:

       Crunched:     Uploaded:
 AP:          3             3
 MB:        207           207

Martin
[edited]: and downloaded was 0 ap and 333 mb.

Average processing rate: 146.77914758267

That APR seems quite low.

My Gigabyte GTX 460: "Average processing rate 278.65838210031 ".

The clocks are 800/1600/2000 for core/shaders/memory. Obviously oc'd.

A ~130 (GFlops) difference!

Martin

I am not really sure you can get much from the numbers that this "detail" is showing. I have 1 system with a HD 5830 in it, the APR for that GPU for MB is:
Average processing rate 153.56997433895
I also have a couple of systems with 2 x HD 5870s in them, those numbers show:
Average processing rate 93.521530090379
and
Average processing rate 68.619074242772

Obviously the HD 5870s have more processing power than the HD 5830. Also, I see this number:
Average processing rate 92.299416344084
For one of my systems (core i7 2600) for AP V505 CPU processing. This would lead one to believe that the CPU is as fast as an HD 5870, no?

Take all the numbers with a grain of salt, I say.

-Dave

---

Take all the numbers with a grain of salt, I say.

It's taken several days for my APR numbers to slowly build up.

135.86388936581
GTX460 (GPU 815MHz, Memory 1GHz, Shaders 1.630GHz), Win7 64bit, i7 2600.
BOINC reports 363 GFLOPS peak when starting up.


EDIT- running 2 WUs at a time.

---

Grant
Darwin NT

[quote]
BOINC reports 363 GFLOPS peak when starting up.

Really with my GTX 470 boinc reports GFLOPS 1120.

Interesting, well thanks all for the observations!

---

Seems like lots of GFLOPS around here. My startup messages:

25/02/2011 23:30:40 OS: Microsoft Windows Vista: Home Premium x64 Edition, Service Pack 2, (06.00.6002.00)
25/02/2011 23:30:40 Memory: 4.00 GB physical, 9.87 GB virtual
25/02/2011 23:30:40 Disk: 20.00 GB total, 15.45 GB free
25/02/2011 23:30:40 Local time is UTC +0 hours
25/02/2011 23:30:41 NVIDIA GPU 0: GeForce GTX 460 (driver version 26099, CUDA version 3020, compute capability 2.1, 993MB, 91 GFLOPS peak)

But still performing OK.

F.

---