Subject: Re: The annoying way that BOINC "throttles"..
From: "~misfit~" <misfitnz@yahoot.com.au>
Date: 12/11/2007, 11:08
Newsgroups: alt.sci.seti

Patrick Vervoorn wrote:
In article <4737ac8e@news2.actrix.gen.nz>,
~misfit~ <misfitnz@yahoot.com.au> wrote:
Patrick Vervoorn wrote:

No idea what these problems were, hopefully you solved 'em (I gather
from some text below you did).

I did indeed. My new motherboard, an Asus P5K-E WiFi-AP, with the
P35/ICH9 chipset, doesn't give enough vcore. Even with the BIOS
defaults set my machine kept random rebooting or blue-screening.
After much gnashing of teeth and rending of garments (also using of
monitoring apps) I discovered that, if I increase what the BIOS
tells me the vcore is by 0.15v then it's fine. It's delivering 0.15v
too little. Took a while to work it out as, with any new build, any
number of things could cause the symptoms I experienced. It's the
first time I've struck that particular problem.

Was this before or after the overclocking? :)

Before. I always like to get a new mobo stable, which normally means 
defaults, before updating to the latest BIOS and downloading Windows 
updates. OCing comes after established stability.

It's either this, or try to find out what exactly the BOINC folks
had in mind when they implemented the xx% CPU usage option.

Yeah. Or give up SETI altogether, in which case I'd have to
immediately stop reading or posting in this newsgroup. <g>

Well, perhaps leave the computer on, running SetiBOINC at 100% every
night before you go to bed, and set the computer on some sort of
timer, where it switches off automatically after xx hours? If you set
that to 6 hours, for instance, it will then effectively use 25% of
the available CPU power for SetiBOINC?

Good point.

I got the demo the minute it came out (well, minus some download
time of course ;). I haven't really monitored CPU usage during the
demo, but it in any case detects a quad core CPU without any
problems. Very impressive game BTW, the Demo is well worth the 1.9G
download.

So I'm told. It's on my download list, if I can't find it on a cover
DVD in the next few weeks.

By that time the final game should be about out. ;)

By which time I still probably won't be able to justify the expenditure of 
buying it.

[Writing something yourself to throttle SetiBOINC]

Ahh, but is the reward worth the work? Something I have to ask
myself often. ;-)

Fulfillment? Satisfaction? And the continued fuzzy happy feeling of
contributing to SetiBOINC? I think the reward is well worth the work!

Ahhh, the reward systems in individuals varies so much doesn't it? Also, so 
is the base knowledge, from which the "work" starts. As my base knowledge of 
writing code is _zero_ the hill is steep indeed.

Or perhaps try to warm up the BOINC people to a new kind of
implementation of the CPU usage parameter...

[Linux Live CDs]

And?

And I think that I'm not yet ready to leave my Microsoft security
blanket.

'Microsoft' and 'security' in one sentence!? Daring!

LOL! Yes, you have a point. Of course I mean security in the sense in which 
one finds security in familiar things. Certainly not in the sense of 
computer security. <g>

Your ISP doesn't have an unlimited plan?

Yes, however it's expensive and it's through a different, ummmm,
'pipeline'? You're de-priotitised, throughput decreases.

Very weird. By de-prioritising you mean your ping times increase? So
what ping do you get to the first link after passing through your
ADSL modem? This sounds like ADSL from the stone age!

Ping times increase, throughput drops.. I can't tell you the ping times as I 
on;y stayed on that plan for one billing cycle. However, the ping time that 
goes up is the one across (or, more correctly, under) the ocean, out of New 
Zealand. It's almost as if there's several priority levels for international 
traffic that vary depending on your ISP and what plan you have with them. Of 
course, ultimately all data goes through the hands of the monopolistic New 
Zealand Telecom. The giant of a company that used to be a state-owned 
enterprise, that also owns 50% of the "Southern Cross" cable that carries 
the bulk of data out of and into NZ.

The courts have ruled that NZT must "unbundle" and break their operations 
into three sepeate entities but any flow-down to the customer is years away. 
Going by similar scenarios in other countries, England for one, it's a fair 
bet that NZT will use stalling tactics (they already are) to milk that cow 
for as long and as hard as they possibly can. Millions in lawyers fees can 
be recouped in a matter of days at the rates they are profiting from 
fleecing the New Zealand citizenry.

What speed are you on before you're rate-limited?

256/128kbps. Pathetic huh? I've just decide to pay another $10/month
and am getting upgraded to 2M/128 tomorrow.

If you want an ADSL line with that speed, you'd have to spend quite
some effort overhere. The 2M down is still with the same 1GB / day
limit? If I did my math wrong, you can spend your day's worth of
downloading in about 68 minutes using 2M down.

Something like that. However, I'd then get the data I can download for the 
other 23 hours at 64/64kbps "free". Woohoo!!

Actually, my ISP, the best in the country, does have a window of free 
download time for two hours every night between 6 and 8pm. During this time 
whatever you transfer isn't counted against your daily total. That's where 
the benefit of the 2M over the 256kbps comes in. If I time my downloads 
right I can get quite a chunk of data during that period, even though the 
whole system does slow down a bit then as everyone is thinking the same 
thing.

Amazingly enough, the Shenk death animation still manages to bring
the 8800GTX, using a Glide Wrapper, to it's knees. ;)

I'm a bit curious if/when we'll ever see a GPU capable of running
the Shenk death animation without any hickups. That should be the
day!

Amazing how such an old game can do that. It has to come down to
badly-written code IMO. There doesn't seem to be any other reason
why a game over 5 years old can bring a graphics card 100 times as
powerful as those available in it's time to a grinding halt.

It's probably that, however, my experience with it also indicated it
has something to do with the way the architecture of these GPUs has
evolved.

I've by now played Diablo II on quite a bit of GFX hardware, and that
partical piece of OpenGL stress-test actually ran best on the old
P4-2.4GHz with a GeForce4-Ti4200 card.

When I upgraded it to a 6600GT, the Shenk animation was going slower.
The 7600GT, same CPU, improved it a bit, and now on the
Q6600/8800GTX, it's even better, but still not as good as on the
Ti4200.

When I noticed the lower performance with the 6600GT (totally
unexpected, since the 6600GT blew the Ti4200 on all fronts out of the
water) I researched it a bit, and what I can vaguely remember is that
the 6600GT was an improvement on almost all fronts (more units, each
with more performance), except for one area, where the Ti4200
actually had more units (with of course a lower performance). My
'conclusion' then was that the units which were left out when going
from the Ti4200->6600GT were actually the ones being used a lot for
the Shenk Death Animation.

Interesting that you should mention that. My best cards for Diablo II have 
also been the ti4x00 range. I have thre of them, two 4200s and a "Golden 
Sample" 4400 that will run faster than a 4600 and is stock-clocked at above 
4400 speeds. I was quite disappointed when I fired D2 up for the first time 
after getting this system going with it's 7800GT only to find that it seems 
slower than my old Barton/ti4400 combo. :-(

I can't recall more details at the moment, and I also don't have the
time to look it all up again, so sorry for being so vague. Anyway,
this has very little to do with SetiBOINC anymore. ;)

Indeed. However, I don't see this group being used for much else in the last 
few weeks other than threads that I've started. :-)

Cheers Patrick,
-- Shaun.