Use me like a 2 buck floozey!

[Movieman]

Ok, where do you want me?
What's the best use of my 2 DX machines to help you guys?
You've seen what's happenning in Rosetta and although as co captain I have to do what is best for the team as an individual I'm expressing my anger in being blown off by the developers.
Info,links, FAQ, tell me where you want me. It may be only temporary but I'm willing to go where you think I can be of most help.
Use me, abuse me guys!:p:

[[XC] leviathan18]

you can make quick guides to setup for the dogs.... fool proofs thats a good work to invest time :p: and clean the turtle shell so he can keep up with me in rosetta LOL

[[XC] riptide]

Ya! Treat ME like a rented mule!

EDIT: If your not into starting rosetta gain, try SOB or QMC. DX would destroy SOB. Is that what you meant by asking what to do with your DX?

[[XC]melymel]

Seventeen or Bust, it's great for netburst and riptide has a sticky at the top of this section. If you want to stick medical though D2OL could use some help to defend 1st. :toast:

[Movieman]

Seventeen or Bust, it's great for netburst and riptide has a sticky at the top of this section. If you want to stick medical though D2OL could use some help to defend 1st. :toast:
ok, so how about one each on D2OL and one on SOB?
Work for you guys?

[[XC]melymel]

ok, so how about one each on D2OL and one on SOB?
Work for you guys?

That could work :p:

[[XC] riptide]

ok, so how about one each on D2OL and one on SOB?
Work for you guys?
Agreed! Follow the sticky for SOB. As you know we have the Resevoir Dogs team there if you want, or go it alone under your own name. Any questions AFTER ;) reading the sticky, PM me.
Rip

[Movieman]

ok, since I said use me,tell me, which of the 2 needs me the most? will drop mt DX3600 monster( well it was till the invention of Conroe:p: ) and the DX2400 on the other..
Tell me what YOU want..Like I said, abuse me, make me whimper!:D

[[XC] riptide]

ok, since I said use me,tell me, which of the 2 needs me the most? will drop mt DX3600 monster( well it was till the invention of Conroe:p: ) and the DX2400 on the other..
Tell me what YOU want..Like I said, abuse me, make me whimper!:D
SOB on the monster! DX3600!!! Here we come!

[Movieman]

SOB on the monster! DX3600!!! Here we come!
Done..anything I need to know? Links,tips? vasoline since I am being abused?:p:

[[XC] riptide]

Done..anything I need to know? Links,tips? vasoline since I am being abused?:p:
http://www.xtremesystems.org/forums/showthread.php?t=107098
Question... you going as the Dogs or as yourself?

Also since this is a DX run ' sobsvc -o ' as a service install command line, where the -o means to optimize for your multi proc system. Then go into the client after the service starts and modify your settings. or just start the client as normal after install and choose your setting then! And then mess with the services command lines!

Or sobsvc -p:2:1
(forces 2 clients to be started and allows all three to run on any available
CPU)

PS: Vasoline is for noobs :rofl:

[Movieman]

http://www.xtremesystems.org/forums/showthread.php?t=107098
Question... you going as the Dogs or as yourself?

Also since this is a DX run ' sobsvc -o ' as a service install command line, where the -o means to optimize for your multi proc system. Then go into the client after the service starts and modify your settings. or just start the client as normal after install and choose your setting then! And then mess with the services command lines!

Or sobsvc -p:2:1
(forces 2 clients to be started and allows all three to run on any available
CPU)

PS: Vasoline is for noobs :rofl:
Dry? Oh you are one nasty bugger...
and bugger is a bad choice of words in that line!:p:

[Movieman]

Dry? Oh you are one nasty bugger...
and bugger is a bad choice of words in that line!:p:
questions: SOB, wernt to the web page, created account, downloaded and installed, shows anonymous and no team, try to input and says thats already taken( my username and email) what gives????
also showing about 40% cpu usage across the 4 (hyperthreading) cpu's

[Martijn]

questions: SOB, wernt to the web page, created account, downloaded and installed, shows anonymous and no team, try to input and says thats already taken( my username and email) what gives????
Shouldn't use new account, but the options. You don't want to make a new account you've already created before :)

Did you use the service install?

[Movieman]

Shouldn't use new account, but the options. You don't want to make a new account you've already created before :)

Did you use the service install?
nah, I'm a windows guy!:D From what I see it's just using 25% of cpu power

[[XC] riptide]

questions: SOB, wernt to the web page, created account, downloaded and installed, shows anonymous and no team, try to input and says thats already taken( my username and email) what gives????
also showing about 40% cpu usage across the 4 (hyperthreading) cpu's
http://www.xtremesystems.org/forums/showthread.php?t=107098
Go to point 5)
Go into config! Enter your team and username! I don;t see you on the roster yet.

Movieman. After you get your login sorted you will then have to use the sobsvc command line and switches that I mentioned above to take advantage of your DX!

[KaptainBlaZzed]

questions: SOB, wernt to the web page, created account, downloaded and installed, shows anonymous and no team, try to input and says thats already taken( my username and email) what gives????
also showing about 40% cpu usage across the 4 (hyperthreading) cpu's


to get 100% usage on a DX you need to go to the command line, navigate your way to the SOB install folder and enter this command.

sobsvc -p:4

[[XC] riptide]

to get 100% usage on a DX you need to go to the command line, navigate your way to the SOB install folder and enter this command.

sobsvc -p:4
KaptainB, some people might argue that 1 instance across each processor is a more agressive way to use HT.Also it gives more ample room to manouver for cache. Also what cache does 3600 have?

From the readme

"Hyperthreading (present if enabled in the BIOS on Intel P4 Xeon and P4
3.02+GHz), however, provides additional "logical" CPUs which share most
of the resources of the physical CPUs. The optimal pattern for them *appears*
to be assigning one client to Logical Processor 1 on each Physical Processor.

However, another, more aggressive use of hyper-threading seems to sometimes
provide even better results: in addition to the above, assigning an additional
number clients (half as many as the number of Physical Processors), allowing these
to "float" over all the Logical Processor 0's. Running "sobsvc -o2" will perform
this optimization."

So that would be 3 processes!

EDIT: I see from your sig that alot of your machines Kaptain, are 512K bearing DX's. 2 instances on each Proc IMO is not ideal, as each instance of SOB needs 512K cache to be optimal! You can ask Prime95 on free-dc if you need more info!

[KaptainBlaZzed]

i have changed the config of 2 of my DX's to the optimized 3 client version, using the command sobsvc -o2.

the other DX is split 50% SOB and 50% eOn.

we will see what the numbers say.

[[XC] riptide]

i have changed the config of 2 of my DX's to the optimized 3 client version, using the command sobsvc -o2.

the other DX is split 50% SOB and 50% eOn.

we will see what the numbers say.
This will be interesting as I have a DX somewhere in the post! Damn Canada post! :mad: Also note that one should only compare during the same block as different k and n values will change the rate of work plus the ceM 'rate' on the client!;)

[Movieman]

ok, with the help of our good teammate Martijn I have SOB running 4 instances at 100% cpu usage showing over 1 mil so far on the cEMs/sec scale
He tells me I should be seeing close to 4 million on each client.
system is a dual xeon: 3600/2mb/800 on a SM X6DA8-G2 MB
http://www.supermicro.com/products/motherboard/Xeon800/E7525/X6DA8-G2.cfm
2-1 gig Viking/Corsair ECC Reg DDR2-400
Raptors in raid0

[[XC] riptide]

ok, with the help of our good teammate Martijn I have SOB running 4 instances at 100% cpu usage showing over 1 mil so far on the cEMs/sec scale
He tells me I should be seeing close to 4 million on each client.
system is a dual xeon: 3600/2mb/800 on a SM X6DA8-G2 MB
http://www.supermicro.com/products/motherboard/Xeon800/E7525/X6DA8-G2.cfm
2-1 gig Viking/Corsair ECC Reg DDR2-400
Raptors in raid0
4 mil! Might be a little much... but we'll see. All depends on what n/k numbers you are crunching! Did you see the previous post of mine about the ideal number of instances.? Anyways let us know how you get on!

What sub team you on? I don't see you?

http://www.seventeenorbust.com/stats/teams/team.mhtml?teamID=517

[Movieman]

4 mil! Might be a little much... but we'll see. All depends on what n/k numbers you are crunching! Did you see the previous post of mine about the ideal number of instances.? Anyways let us know how you get on!

What sub team you on? I don't see you?

http://www.seventeenorbust.com/stats/teams/team.mhtml?teamID=517
I registered under Movieman on the XtremeSystems team.
I don't think I'm going to see 4 mil on any one client. I'm seeing a TOTAL of all 4 of app 4.2 mil..Is that good,bad?
I don't know anything about the scoring on this.
I need a password to join the XS team..can someone PM it to me..

[[XC]melymel]

password is : xs

:toast:

[KaptainBlaZzed]

I registered under Movieman on the XtremeSystems team.
I don't think I'm going to see 4 mil on any one client. I'm seeing a TOTAL of all 4 of app 4.2 mil..Is that good,bad?
I don't know anything about the scoring on this.
I need a password to join the XS team..can someone PM it to me..


4.2 mill on a 3.6 DX is not so good, i have a 2.4b @ 2.875 getting 3.3 mill. I would guess and say the with all 4 client running you should get about 8 or 9 million on the 3.6 DX.

You have to wait for your rate to top out, which takes about 2 days.

at any rate i get about 600g per day from a DX @ 3.15 so you should bet a little more, maybe 700g per day

[[XC] riptide]

I need a password to join the XS team..can someone PM it to me..I understand you've been busy but all this was in the sticky! xs as mely says! :slap:
You should be ssing a total of 4.2 worth ON ONE processor alone. IE total should be 8mil. What are your k and n numbers?

EDIT: It does take many hours for rates to stabilize anyways!

[PFOUR]

I'd say you should see around 2 million per logical processor at 3.6 ghz.
SOB appears to be very clock speed dependent and that's what I get from 3.6 Ghz.#
So you should be looking at around 8 million all in with 4 logical processors.

[[XC] riptide]

Well i float one instance over one physical proc. 1) if i use it for something else I can still get performance 2) During 2 24 hour tests I have gotten better overall reults this way. However this test was on a northwood with only 512 cache, so i reckon on my other chip with 2.5 cache, 2 instances may work better. But not tested yet. Again all this depends on n & k numbers. i beleive the benchmark for this will come out in a later release of the client.

[Movieman]

I understand you've been busy but all this was in the sticky! xs as mely says! :slap:
You should be ssing a total of 4.2 worth ON ONE processor alone. IE total should be 8mil. What are your k and n numbers?

EDIT: It does take many hours for rates to stabilize anyways!
Client1: K:21181 N: 6859604
Client2: K:55459 N: 6858534
Client3: K:10223 N:6859529
Client4: K: 10223 N:6859565
showing app 4.2 million total on the 4 clients 7 hours into the app..100 blocks done

[PFOUR]

I have 1 running on k10223 which is particularly slow, should finish tommorow though.
Still seems low to me Movieman, I'm not familiar with dx's, are they running 3,6ghz clock?
How much Ram do they have?

[Martijn]

I have 1 running on k10223 which is particularly slow, should finish tommorow though.
Still seems low to me Movieman, I'm not familiar with dx's, are they running 3,6ghz clock?
How much Ram do they have?
3.6 Ghz, 2 gigs of ram :)

And I have no experience at all with HT cpus, so I'm sorry for the error. Doesn't running more clients on a HT cpu make it useless then? If the rates are divided/2?

[njkid32]

OK, I just installed the client and joined XS. I want to make sure I did this right. I just installed once and am running 4 instances.. Is this correct? What is this service trick?

[[XC] DragonOrta]

Did you just start the SoB client four times from the start menu? If so, you're working on the same number 4 times, which does nothing more than running 1 instance. Here (http://www.free-dc.org/forum/showthread.php?t=8086) is a thread at Free-DC about installing as a service.

[Martijn]

OK, I just installed the client and joined XS. I want to make sure I did this right. I just installed once and am running 4 instances.. Is this correct? What is this service trick?
You can also ask MM or me of course :toast:

[PFOUR]

If I run 2 clients on HT P4 I get slightly higher combined score than 1 client.
HT appears to work for me on 3.2 P4 @3.6 with HT and 1Gb Ram

[njkid32]

Ok, this is what I just did and its only using 50% with one client open (yes I started the service)

SOBSVC -i (to install SB service)
net start "Seventeen or Bust service" <enter>
SOBSVC -p:4 for quad cores
SOBSVC -o to opti

Edit: So should I be running 4 sep clients?

Edit#2: It seems the serivce I installed is just sitting there using only 25% cpu... Uhhhgggg I feel like and idiot here. Damnit would someone hold my hand:rofl:

[KaptainBlaZzed]

Ok, this is what I just did and its only using 50% with one client open (yes I started the service)

SOBSVC -i (to install SB service)
net start "Seventeen or Bust service" <enter>
SOBSVC -p:4 for quad cores
SOBSVC -o to opti

you have to restart the service after you change the number of clients

i am not sure which is better 3 or 4 clients, the jury is still out on that one

i have been runnig 4 clients since i started and today i changed it to 3 optimized clients, only time will tell
I used the command sobsvc -o2
this gives you 3 optimized clients

[njkid32]

For some reason it will only let me optimize 2 clients and then shows affinity 0 and 1

If I type SOBSVC -o4 it defaults to 1 instance. If I type SOBSVC -o2 it will opti 2 cores and thats it. Weird! Any other suggestions?

[PFOUR]

I'd go with SOBSVC -p:4
Should get you 100% usage with best output from what I've seen.

[KaptainBlaZzed]

For some reason it will only let me optimize 2 clients and then shows affinity 0 and 1

If I type SOBSVC -o4 it defaults to 1 instance. If I type SOBSVC -o2 it will opti 2 cores and thats it. Weird! Any other suggestions?


since you have 4 physical cores i would run 4 clients. I am not even sure what the optimization does, I think it is for machines with HT.

stop the service
run sobsvc -p:4
then restart the service

[njkid32]

Yes thats what I did but I also wanted to optimize it... This is where I am confused..:)

http://img82.imageshack.us/img82/4870/soblargeos7.jpg (http://imageshack.us)

Edit: oopps big pic:)

[PFOUR]

Looks fine to me.

[[XC] serlv]

ok, since I said use me,tell me, which of the 2 needs me the most? will drop mt DX3600 monster( well it was till the invention of Conroe:p: ) and the DX2400 on the other..
Tell me what YOU want..Like I said, abuse me, make me whimper!:D

Bend over, biotch! :D

[njkid32]

One last question:) What priority do you run on? Idle or normal?

[[XC] riptide]

One last question:) What priority do you run on? Idle or normal?
Well... it all depends on what you want to be able to do with the :banana::banana::banana::banana::banana: while your running. I usually have mine 'Above normal' cos of HT thats ok, but yours you could run 3 this way and leave one 'idle' to let your PC function. (Its 5:35 here just woke up for a coffee and when I see NJ posted in a SOB thread i said 'please make it that clark is here too please'... then I opened and saw the big pic with 4 clients and a smile ensued.) Oh and if you guys are all asking questions, do you think the sticky needs modification?

[[XC] riptide]

Movieman you floozy! I don't see you on the team roster. on seventeenorbust.com. You crunching for Boinc Synergy now!? Or have you got with de Dogs? And in reference to your scores... there is def something wrong here pal! You running anything else on those DX's? Kill 1-2 of those instances and let the others stabilise and tell me whats the score. Remember this app needs 512K Cache each! I've seen Cellys >4Ghz get poor results on this because and I guess cache problems.

EDIT: And for all you boincers, this project works great with Boinc. Running this at 50% of processor will ONLY drop boinc performance by 40% or thereabouts. You'll still be pulling in 60% of your usual Boinc credits! So thats like 110% DC performance with this combination of projects. Not sure but is this just an Intel netburst thing. Kaptain Blazzed can update you on this.

EDIT: KB will also let us know if that info on 3 clients on a DX is better than 4? ;)

[njkid32]

He is running SoB but has it set so that it doesnt update till after a full test is done. I changed mine to that as well. I did notice if it updates between blks the cpu usage goes down....

Also is this based on Ghz? I am running at 3ghz but I could run it at 3.5ghz instead. Will that make a diff in SoB?

[[XC] riptide]

He is running SoB but has it set so that it doesnt update till after a full test is done. I changed mine to that as well. I did notice if it updates between blks the cpu usage goes down....

Also is this based on Ghz? I am running at 3ghz but I could run it at 3.5ghz instead. Will that make a diff in SoB?
A very good. i also mentioned that in the sticky that uploading blocks momentarily stop production. NJ if you can run at 3.5, do it! It should scale sharply with Ghz. We should see some nice dumps then. And we'll WRECK the 40's ranking too.

[njkid32]

Ok, I am going to run at 3.3ghz for now cuz I can do it at a low voltage. You have no idea how hot this CPU gets and my room is hot enough:)

[Martijn]

Ok, I am going to run at 3.3ghz for now cuz I can do it at a low voltage. You have no idea how hot this CPU gets and my room is hot enough:)
:slobber:

[ShootStraight]

I'm guessing there is no OSX client for SOB?

-SS

[njkid32]

I just looked around and I found this post---

What about running an emulator like qemu (http://fabrice.bellard.free.fr/qemu/) or bochs or wine?

[[XC] riptide]

Ok, I am going to run at 3.3ghz for now cuz I can do it at a low voltage. You have no idea how hot this CPU gets and my room is hot enough:)
I'll be waiting in great anticipation for the 4 dumps off this one! ;) Are you getting a time to complete estimate. (Want to put it in my diary:D )

[njkid32]

I turned the upload back on cuz it doenst seem to hurt at all. IDK how well its doing though maybe you can show or tell me... Where would the time to complete estimate be? It also says I have about 5700 blocks left in the test is this right?

[Martijn]

I turned the upload back on cuz it doenst seem to hurt at all. IDK how well its doing though maybe you can show or tell me... Where would the time to complete estimate be? It also says I have about 5700 blocks left in the test is this right?

Have a look at this:

[njkid32]

Mine is only telling me 43s per block. Its not giving me an overall EST.

[[XC] riptide]

Mine is only telling me 43s per block. Its not giving me an overall EST.
GRRRR! Turn off the intermittent reporting of blocks for a few seconds and your predicted time to finish the entire test will now show.

[njkid32]

Ahhh 77:10:00. Hows that? Bad?

[[XC] riptide]

Ahhh 77:10:00. Hows that? Bad?
Not really. Good I'd say. mine takes about 3 days or there abouts. Ahhhh your so cute. Isn't he guys! He's all worried little clark isn't up to scratch. Rippy picks up eric and cradles himIt's ok NJ everything's gonna be alright:rofl:

[Martijn]

Ahhh 77:10:00. Hows that? Bad?
I have done about 20% of the full test, and I'm still @ ~80 hours left. ~100 hours/test on the D820 sigged below.

Yours is better than mine ;)

[njkid32]

Not really. Good I'd say. mine takes about 3 days or there abouts. Ahhhh your so cute. Isn't he guys! He's all worried little clark isn't up to scratch. Rippy picks up eric and cradles himIt's ok NJ everything's gonna be alright:rofl:

Your starting to sound like Mamma Turtle:rofl:

Well I was just worried it wasnt running the right way.... Hey you never know this project might not be good for Clark. He isnt great at everything (at least till he turns into superman:))

[[XC] riptide]

I have done about 20% of the full test, and I'm still @ ~80 hours left. ~100 hours/test on the D820 sigged below.

Yours is better than mine ;)
Thats not bad Martijn! Your stats don't reflect this, you running other stuff aswell?

EDIT: NJ - you're already making a nice impression on the stats page. Lets see how she goes! http://stats.free-dc.org/new/teamstats.php?proj=sob&team=XtremeSystems

[Martijn]

Eric, you might want to take advantage of this: SoB improves a lot when lowering your mem timings. When I'm running my memory at stock timings (5-5-5-15), I get no more than 2.4 million cEMS/sec. However, when I lower the timings to 3-3-3-12 (where I'm currently running at), I get the scores you saw above ;). If your ram can give a little more than what it's giving now (some people are lazy, in Rosetta it didn't make a difference whatsoever), go for it :toast:


Thats not bad Martijn! Your stats don't reflect this, you running other stuff aswell?

EDIT: NJ - you're already making a nice impression on the stats page. Lets see how she goes! http://stats.free-dc.org/new/teamstats.php?proj=sob&team=XtremeSystems

I have only been running R@H the past few days (see sig, just dumped ~600 points again). I'm now turning towards SoB. Rosetta is over for me now, at least for the coming months.

[PFOUR]

I have done about 20% of the full test, and I'm still @ ~80 hours left. ~100 hours/test on the D820 sigged below.

Yours is better than mine ;)

How many clients are you running on that 820?

I have an 805D @3.88ghz running 2 clients (no HT on 805) at around 5.2 million each

[Martijn]

How many clients are you running on that 820?

I have an 805D @3.88ghz running 2 clients (no HT on 805) at around 5.2 million each
2 :)

[PFOUR]

Maybe the high Multi on the 805 helps too.

[njkid32]

This is where I am at now. Its weird though client 2 current k: 24737 is much lower than the other three clients.. Hmmmm

http://img247.imageshack.us/img247/9097/sob2largeug7.jpg (http://imageshack.us)

[Martijn]

There are different K-levels that still need processing, no worries :)

EDIT: bottom of the page.

http://seventeenorbust.com/stats/

[PFOUR]

That's impressive from 1 machine!
The lower k numbers do appear slower.

[[XC] riptide]

That's impressive from 1 machine!
The lower k numbers do appear slower.
The highest k number 67607 I had gave me the slowest! 15-18% slower!
I think Moddolicious (maybe wrong) had posted a conroes firgures of >6XXXXXX. Can't remember what speed though!

I'm starting to see a lot of variation between all these k/n number combinations and tests. I think I should heed what has been said on the other forums about these that they are not accurate and are only ballpark figures.

So lets extrapolate this argument... I wonder is there a way to to get the server to send us k/n combinations for fast tests? hmmmm. Because if you think about it, our credits maybe associated with these speeds. I'm sure it averages out over time though. NJ, have you had a chance to look at our stats and where we are. Check out 'Ukraine's history! Unbeliivable!

[Martijn]

One of my 2 clients is running a 55459/11650366, the other one a 10223/11650349, and guess what: the lower k value is faster :)

Seems logical, forumla is k^n+1. Lower numbers should give higher scores...

[PFOUR]

EDIT: bottom of the page.

http://seventeenorbust.com/stats/

So it's not the lower k numbers it's the k numbers with the larger n bounds

[njkid32]

OH NO!! A cricket just commited suicide and jumped into a 90mm 110CFM fan. He was on my foot and I kicked and he jumped at the same time right into the fan. It was KEWL kinda like a bug zapper.:rofl:

[Martijn]

OH NO!! A cricket just commited suicide and jumped into a 90mm 110CFM fan. He was on my foot and I kicked and he jumped at the same time right into the fan. It was KEWL kinda like a bug zapper.:rofl:
:rofl:

Don't stick your fingers into it, btw :stick:

[njkid32]

Well the client with the lower K # is staying about the same as the others percentage wise. So IDK!


Edit: I have stuck my fingers in worse things than that:rofl:

[PFOUR]

The highest k number 67607 I had gave me the slowest! 15-18% slower!
I think Moddolicious (maybe wrong) had posted a conroes firgures of >6XXXXXX. Can't remember what speed though!

I'm starting to see a lot of variation between all these k/n number combinations and tests. I think I should heed what has been said on the other forums about these that they are not accurate and are only ballpark figures.

So lets extrapolate this argument... I wonder is there a way to to get the server to send us k/n combinations for fast tests? hmmmm. Because if you think about it, our credits maybe associated with these speeds. I'm sure it averages out over time though. NJ, have you had a chance to look at our stats and where we are. Check out 'Ukraine's history! Unbeliivable!

My E6300 @ 450x7 3150mhz is running 2 clients at approx 6200000

[[XC]thewildblue]

I have always run 4 instances on my DXs. Would I gain anything on 3 ?

I have put this on a 3200+ @ 2500 and its doing 11g per update which isnt too bad. Its out producing a p4@2.26 by quite a big margin on the cmes thingy.

I only have 4 machines on this at the moment and Im going quite well.
p 4 2.26
Dx 2.8 @2.8 (1mb cache)
Dx 1.6 @2.94 (512k cache)
a64 3200+ @ 2500

Thats giving me 66g per update.

[PFOUR]

I've tried With a HT P4 and 2 instances gives more than double the output of 1 so the HT appears the more efficient way.
I'd imagine 4 on a DX to be best.

[njkid32]

Jeez I cant wait till my second kents gets here:)

[[XC]thewildblue]

Jeez I cant wait till my second kents gets here:)

:slobber:

I can only dream of such things.....................I would love one of those

[[XC] riptide]

I've tried With a HT P4 and 2 instances gives more than double the output of 1 so the HT appears the more efficient way.
I'd imagine 4 on a DX to be best.
When i get home I will investigate 2 instances on >512kb cache p4 chip also. As Ive said 2 instances on my northwood (512kb cache) gave slightly less than the 1 instance. Althought the readme file said 3 instances would be 'agressive' use of the HT on a DX.! (I wonder was that readme written with a 512 or 1mb cache based DX?)

[[XC]thewildblue]

When i get home I will investigate 2 instances on >512kb cache p4 chip also. As Ive said 2 instances on my northwood (512kb cache) gave slightly less than the 1 instance. Althought the readme file said 3 instances would be 'agressive' use of the HT on a DX.! (I wonder was that readme written with a 512 or 1mb cache based DX?)


Both my DXs have 2gb of ram. Wonder if that makes any difference ?

[[XC] riptide]

Both my DXs have 2gb of ram. Wonder if that makes any difference ?Yep people at Ars technica said that over 512MB on a HT would be slighly beneficial to have 2 instances, and less would be better off with one instance. Theres a lot of things seemingly at play here cache/ram/timings(according to Martijn)/HT. And a lot of things to be able to adjust from the client side Affinities, number of instances, HT 'float over' effect etc etc. One would need a host of configurations and a good weekend to get a definitive picture. I can only read and think so much!:( Even the readme's use vague language when suggesting things. there does not seem to be any definitive talk anywhere in any guides for DX's except of course what users report back. I'm still waiting for Kaptain Blazzed to come back to me on his 3 instances per DX machine. That will tell a tale!

[Movieman]

Movieman you floozy! I don't see you on the team roster. on seventeenorbust.com. You crunching for Boinc Synergy now!? Or have you got with de Dogs? And in reference to your scores... there is def something wrong here pal! You running anything else on those DX's? Kill 1-2 of those instances and let the others stabilise and tell me whats the score. Remember this app needs 512K Cache each! I've seen Cellys >4Ghz get poor results on this because and I guess cache problems.

EDIT: And for all you boincers, this project works great with Boinc. Running this at 50% of processor will ONLY drop boinc performance by 40% or thereabouts. You'll still be pulling in 60% of your usual Boinc credits! So thats like 110% DC performance with this combination of projects. Not sure but is this just an Intel netburst thing. Kaptain Blazzed can update you on this.

EDIT: KB will also let us know if that info on 3 clients on a DX is better than 4? ;)
Yes, looking at my scoring so far it's way off. Still seeing a million per client
These are the 2mb cache Irwindales in this pig..The best that money could buy last August..3600/2mb/800's running at 3600..and thats all thats running on this machine besides me using IE,ICQ and MSN..nothing else pal..
and yes, I am on the team..I have the box checked to not upload the individual tests? and just when the entire unit is down..Is that a mistake?
Just unchecked that box so you stats whores will see some numbers!

[KaptainBlaZzed]

after watching my stats for quite a few hours now, i believe that if you run 3 optimized clients on a DX you will get slightly more points than if you run 4 regular clients.

the increase is only about 5%, but every little bit counts.

the command is "sobsvc -o2" for 3 optimized clients.

[[XC] riptide]

after watching my stats for quite a few hours now, i believe that if you run 3 optimized clients on a DX you will get slightly more points than if you run 4 regular clients.

the increase is only about 5%, but every little bit counts.

the command is "sobsvc -o2" for 3 optimized clients.
PWNED.WHAT DID I SAY GUYS! WHAT DID I SAY. KB I owe you a pint for that. Movieman, PFOUR!! ITs official, 3 instances of SOB on a DX are better than 4!! This is in the sticky!

[[XC]thewildblue]

Looks like I need to tweak then. Thats an extra 5% on 2 systems then.

[[XC]thewildblue]

Changed on the DXs lets see if we get an improvement.

[njkid32]

Clark is the worse cpu on SoB.. He isnt doing $hit.... Yes I have had alot to drink LMAO...