Hi all,

Lots of people wonder which of Tetsuji's optimized seti clients is the quickest on their systems. Theoretically, the most recent versions (YAOSCW-x-r8.1) should be the quickest. But that's not always the case.

So with help of you all I'll make a listing on my website for easily picking the right client for your system.

What do I ask from you? Well, to crunch the reference workunit using different optimized clients and send the result to me (or post in this thread) along with as much info on your processor as you can find. (CPU-Z is good to find that out).

Here's how to do it:

1. Download and install the reference unit in an empty folder

2. Download an optimized client (see link in signature) and unzip in the same folder

3. Shutdown as much CPU consuming applications on your system as possible
(don't forget Boinc :) )

4. Run the setiathome_4.11_windows_intelx86.exe file
(try not to use your system intensively while the refence unit is processing). You won't see the workunit running anywhere but in your taskmanager.

5. When done, please send the init_data.xml file by e-mail (or the number in the wu_cpu_time field) along with as much info on the used CPU as possible to speedy67 at marisan.nl, or post the info in this thread.

Up to now only my home computer is listed, but hopefully with your help we can make this list longer.

Greetings,
Speedy67

---

So with help of you all I'll make a listing on my website for easily picking the right client for your system.

Feel free to take any timing info from this table. Most of times are for Linux clients but there's one box with timings for Windows clients (the 3.2GHz P4 Prescott box).

---

Metod ...

Feel free to take any timing info from this table. Most of times are for Linux clients but there's one box with timings for Windows clients (the 3.2GHz P4 Prescott box).

Thanks.
The one you describe as 'Tetsuji's windows 4.11' is that the seti-p4-sse3?

Hopefully we will be able to build a nice listing for the windows clients.

Greetings,
Speedy67

---

Speedy67 my AMD 64 mobile Athlon 3700+ has 3dnow, sse, sse2, and sse3. I've downloaded 10 different optimized clients: P r8.1, W r8.1, K r8.1, P r7, W r7, K r7, Plainvanilla, P3 zip, SSE2 zip, and P4 SSE3 zip. How would I be able to tell if the results were any good or not?

tony

Speedy67 my AMD 64 mobile Athlon 3700+ has 3dnow, sse, sse2, and sse3. I've downloaded 10 different optimized clients: P r8.1, W r8.1, K r8.1, P r7, W r7, K r7, Plainvanilla, P3 zip, SSE2 zip, and P4 SSE3 zip. How would I be able to tell if the results were any good or not?

tony

Hi Tony,

Do you mean whether the results will be valid or not? I can't remember reading anything about people having trouble getting their results validated using these optimized clients. They seem to be working good.

As for your processor, theoretically the YAOSCW-W clients or the seti-sse2 should be the most suitable. I'm not sure if the YAOSCW-P or seti-p4-sse3 can be run on an Athlon64, because some of the clients are using specific Pentium optimization. (read here)
If you try to run a client not compatible with your processor it will error out before crunching the workunit.

And as for the speed part.. that's what we're trying to accomplish.. a nice listing in which you can see what client is the fastest for your processor.

Greetings,
Speedy67

---

OK, then if it's an incompatible client it will/should just error out. The only display I'm seeing is the Explorer window and the Taskmanager showing 97-99%cpu. Is that what I'm supposed to see? and when it's done, I assume that file will just appear in the explorer window and my CPU usage will drop. is that right?

OK, then if it's an incompatible client it will/should just error out.

If I understand correctly it will. Looks to me that any messages from the client will appear in the stderr.txt that is created.

The only display I'm seeing is the Explorer window and the Taskmanager showing 97-99%cpu. Is that what I'm supposed to see?

As long as you're not running your regular boinc app or anything else that could cause such a cpu load, yes.

and when it's done, I assume that file will just appear in the explorer window and my CPU usage will drop. is that right?

The init_data.xml file will be created in the folder with the seti app and the reference unit once the client finishes processing the reference unit.

Greetings,
Speedy67

---

Intel P4 3.2GHz(Northwood) 512KBytes L2 Cache
Bus Speed 800MHz

YAOSCW-N-r8.1

4903.187500
4895.703125


YAOSCW-N-r7

4866.250000
4864.609375



YAOSCW-W-r8.1

4899.031250



seti-p4-no-sse3

5042.921875

---

Intel P4 2.4GHz(Northwood) 512KBytes L2 cache
Bus Speed 400MHz

YAOSCW-N-r8.1

7298.366606


YAOSCW-N-r7

7493.216787

---

AMD Athlon 64 FX-53 (Sledgehammer) 2.4GHz 1024KB L2 Cache

YAOSCW-W-r7
4966.968750

seti-sse2.zip
5527.390625

jo-athlon64-sse2_1105-1
5586.562500

I also benchmarked YAOSCW-W-r8.1 to see if it was faster on my machine than r7. I believe it's around 6 minutes faster, but the results are no longer on my system, so I don't know for sure. I thought I posted them somewhere here on the fora after running the benchmark, but after looking, apparently not. Due to the heat, I can't bench it at the moment. Sorry. :(

I also benchmarked YAOSCW-W-r8.1 to see if it was faster on my machine than r7. I believe it's around 6 minutes faster, but the results are no longer on my system, so I don't know for sure. I thought I posted them somewhere here on the fora after running the benchmark, but after looking, apparently not. Due to the heat, I can't bench it at the moment. Sorry. :(

Maybe you can bench it again sometime when it cools down? :) Thanks anyway for the results you gave so far.

I'm now benching my old reliable Pentium3 450, but that process will take a little while.. don't know why I had YAOSCW-W-r7 running on that one, since it doesn't even support sse2, but the results were validating ok.

Greetings,
Speedy67

---

The one you describe as 'Tetsuji's windows 4.11' is that the seti-p4-sse3?

You've got me on this one. I don't remember for sure, but my guess would be that it's a P4-SSE2 with FFTW.

---

Metod ...

Maybe you can bench it again sometime when it cools down? :)

Will do, but the forecast is hovering right around too %$@# hot. Hopefully I'll be able to run again soon.

Are the given results for presscott cpus with HT off or am I in trouble?

P4 3.2 Prescott HT on = 6686s YAOSCW-P-r8.1

mobile AMD Athlon 64 3700+
mmx, 3DNow, SSE SSE2 SSE3 X86-64
128 L1
1024 L2
1024MB ram
2406 Core speed

all speeds listed in increasing order.
P4 SSE3.zip, P r8.1, and P r7 won't run

W r7 (5267 seconds)
W r8.1 (5338 seconds)
SSE2.zip (5779 seconds)
K r8.1 (6181 seconds)
Plain Vanilla (6303 seconds)
K r7 (6436 seconds)
P3.zip ( 6552 seconds)

---

Are the given results for presscott cpus with HT off or am I in trouble?

P4 3.2 Prescott HT on = 6686s YAOSCW-P-r8.1

The times are with HT enabled (at least for the 3.0GHz Prescott) but with only the reference unit task running (meaning the CPU is 47-49% idle when running the reference unit.)

When crunching away 2 WU's at a time with boinc they normally compute in about 4500-6900 seconds depending on the WU.

Greetings,
Speedy67

---

Are the given results for presscott cpus with HT off or am I in trouble?

P4 3.2 Prescott HT on = 6686s YAOSCW-P-r8.1

The times are with HT enabled (at least for the 3.0GHz Prescott) but with only the reference unit task running (meaning the CPU is 47-49% idle when running the reference unit.)

When crunching away 2 WU's at a time with boinc they normally compute in about 4500-6900 seconds depending on the WU.

Greetings,
Speedy67

I think for HT enabled machines (or dual cores), it would be best to run 2 separate copies of the reference WU at the same time thus mimicking what would most likely be used in real life. ie, most people with HT machines are going to run 2 WUs at a time, not one.

I think this would give a more meaningful result in terms of processing times.

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

I think for HT enabled machines (or dual cores), it would be best to run 2 separate copies of the reference WU at the same time thus mimicking what would most likely be used in real life. ie, most people with HT machines are going to run 2 WUs at a time, not one.

I think this would give a more meaningful result in terms of processing times.

Ned

That is absolutely true, but on the other hand I don't think it will make any difference in regard to which client is the fastest for that CPU. I will add that extra info for my own Prescott 3.0 soon.

Greetings,
Speedy67

---

Yes, it won't affect which client is fastest, but in terms of what sort of performance you can expect from that particular processor, it would be nice to see how it performs under real world conditions - ie when hyperthreading 2 proceses.

But you're right - firstly we should establish which client is fastest on each processor.

Anyone benchtested an Athlon XP yet?

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

Are the given results for presscott cpus with HT off or am I in trouble?

P4 3.2 Prescott HT on = 6686s YAOSCW-P-r8.1

The times are with HT enabled (at least for the 3.0GHz Prescott) but with only the reference unit task running (meaning the CPU is 47-49% idle when running the reference unit.)

When crunching away 2 WU's at a time with boinc they normally compute in about 4500-6900 seconds depending on the WU.

Greetings,
Speedy67

I think for HT enabled machines (or dual cores), it would be best to run 2 separate copies of the reference WU at the same time thus mimicking what would most likely be used in real life. ie, most people with HT machines are going to run 2 WUs at a time, not one.

I think this would give a more meaningful result in terms of processing times.

Ned

This was my thinking when I ran two together. I just reran YAOSCW-P-r8.1 on its own and got a 5115s result. But I still think the 6686s result is more valid since thats how it will crunch normally.