|
|
I spy setispy...
Since Roelof has been good
enough to update SetiSpy
the least I can do is give a quick plug. He has a short explanation
of why he's released another v2 update (2.9.1 to be precise) on the
top of his page. I like this little gem so much that I thought I'd
let you in on the good news now rather than in a week or so...
Max out.
BX, VIA, Oda: registers to
riches...
I do appreciate people sending beta results in but this
page and results table (when it is finally sorted out) is not the
place for them. Rat Bastard and zAmboni
are much more clued up about the vagaries of beta numbers. Plus
eventually at V3 day the beta will fade quietly into the background.
Though how long Berkeley will accept results from 2.70 will be high
on the worry list of those of you running this quick little number!
So once again thanks for submitting beta benchmarks but I will only
be passing them on to RB and let
him work his magic with them if he's feeling chilled enough. Thin
rodent fur - always a problem, especially if you have a touch of
mange.
A gentle reminder that some people really are still trying to
tweak their boards came from the TLC
mail list a few days back. Grizzled veteran Kevin
Carpenter mentioned he was trying out memory interleaving
(enabled in the bios) on his VIA based Microstar board. He also
alerted me to the fact that there's a new set of '4 in 1' drivers
(v4.24, Aug 10, 2000, 800kb) from VIA applicable to all chipsets, so
do yourself a favour and have a look at the VIA download
site. Don't get new busmasters every day of the week. Returning
to interleaving...this is old stuff to many of you but it did make
me think that there are so many VIA boards out there that some info
would be useful (if a little repetitious). So 'memory
interleaving' is all about avoiding wait
states and thereby appearing to access bigger chunks of RAM at
any one go. Since present SETI crunching still depends heavily on
memory access, being able to enable interleaving (2 or 4 way) is
going to help. With some chipsets a bios upgrade will give you
'memory interleaving' as an option but with some this isn't
implemented in bios. So first off go check for the latest bios for
your board (If you know you have a VIA chipset but are not sure what
it is then Wim's bios site
might help). Bios flashing is straightforward but not something to
do without a little procedural knowledge, so be careful and if at
all uncertain find someone who knows what to do. Most tweaking
errors are recoverable by rebooting but a failed flash is in another
'league of woe' to sort out. If a new bios doesn't give you
interleaving as an option or you are already bang upto date then the
next port of call is that tweaking whirlwind H.Oda.
Two utilities, free
to download, are of real interest, WPCREDIT and WPCRSET, the
first allows you to change chipset registers and the second to
retain that information at boot. These utilities are not for the
fainthearted or those unfamiliar with 'safe mode', lockups or blue
screens. It can take several reboots to find out what works. Rather
than explain the procedures on boards I know little about a link to
this overclockers
interleaving article should sort most of you out for details. Gains
due to interleaving can be fairly substantial as with this Soyo
board benchmark piece. Good luck with the tweaking.
This was all well and good for VIA chipsets but what of that other
unkillable mainstream item the BX. If you fancy tweaking the BX with
H.ODa's little toys then this is
the BX
tweak piece to read and inwardly digest. However, in looking at
that BX article it struck me that all the register settings being
mentioned were identical to those found on AGPinfo,
a BX/ZX (only) utility I came across and wrote about many moons ago.
It is far less daunting than WPCREDIT and with a better user
interface. Needless to say you will still lock up with the wrong
settings. The 'host bus fast data ready' and 'timing' section on the
Memory tab are the things to enable/change, leave the rest alone! It
works in Win98 but not W2k so back to WPCREDIT for that, always good
to learn two ways of doing the same thing.
While at H.Oda get the latest WCPUID
(V2.8 as of August) - one of those little gems that gives lots of
details about FSB, processor and chipset. Finally while there if you
thought that the TLC membership
had the edge on high FSB's try the Information
link on the site. That 210MHz bus on a BH6 is damn impressive.
Right, enough of Mr. Bunny
and Co. time for a few of the recent submissions and an opening bid
of 16:46 from Mschoose2 running
the v2 GUI (!) on an IBM Aptiva (K6-2, 475MHz). The command line
client will knock a chunk of time off that, be brave, what have you
got to lose except your sanity and the respect of screensaver
junkies. A big leap up to kickboxer
lashing out with a 6:38 on a W2K/PIII 600 setup, a little
overclocking is in order as it's only cruising on a 100 FSB if you
want to drop the time. The ubiquitous CAS2 bios setting for your
memory is also highly recommended. Ross
has overclocked his PIII 600 to 900 with a 150 bus, most pleasant
for coffee table chit-chat but unfortunately the beast falls very
slightly off the mark at 4:52. But that's probably more down to
Apollo Pro chipset than any character flaw in Ross who has already
explored 4-way interleaving to get this sub 5. Very slightly quicker
is Pale_Rider (Clint's moody
western is on tonight if I get to finish this in time) at 4:49 also
with a 150 bus on an Abit SE-6 (Intel 815 Solano chipset). But
similarities end here as this PIII is on a 4.5 multiplier to
give 675MHz and the OS is the slightly under performing
Win98ME, sorry 'Millennium' to those of you still under the
impression that MS have a new OS out. Might be a little room to
tweak there as neo1999 posted
4:19 & 4:23 also on ME and a 152 bus a while back. Finally
hell-burner of the week goes to Tomslik
at 4:23 whose PIII600EB is also hampered by an Apollo chipset. This
hurdle has been surmounted through the naked aggression of a 166 FSB
with CAS2 Mushkin PC150 RAM (a very impressive combination) and
WPCREDIT to tweak register settings. I knew there had to be a use
for my opening paragraphs. Tomsliks
previous 166 FSB time, 4:43, used CAS3 PC133 RAM so better memory
has clipped 19 minutes off an already solid time. Great to have South
Park Admin Crew back, I was around when they were TLC's
big hitters. The old days were good but the new ones are even
better. Good stuff people, keep the WU's crunching and may all your
benches have four legs.
Max out.
Now that I'm famous...
Seems to have been quiet around here recently. Could be to do with
my modem giving grief, the 128MB of RAM in my computer giving
up on life (RMA obtained and it is going to a better place - dead
chip hell by snail mail), the gerbil (one of several) getting sick
and giving me a worrying couple of days or a combination of these.
Hardened overclockers always have small rodents as pets (unwritten
rule). Of course it could be none of them and more to do with few
submissions and not a lot of news!
Speaking of overclockers, of which there must be a delightfully high
percentage in the TLC crew, this
alcoholic
bunch at Tech-Junkie
just had to go that little bit further to extract a few extra MHz
from a Celeron (due to the pic's it's a fairly hefty page or 4
download). I am constantly impressed by the spirit of human
endeavour displayed by CPU nuts in pursuit of the strangest ways to
cool systems. They managed to show that even at -61C Celerons have
their limits. In more serious vein Roelof
(SetiSpy
man) sent some interesting info about the Rat
Bastard bench unit. Seems that it is still a reasonable
predictor of WU crunching ability in the beta 2.70/2.71 GUI's -
which is good news for us. But since it's beta stuff I'll let RB
fill you in along with some very quick times for a Thunderbird and
Athlon that he was sent recently.
A post (SETI Driver Cache Management...10/8/00) from
Mike Ober on the alt.sci.seti
newsgroup went into quite some detail about how SetiDriver
(at 1.46 and counting) manages its cache of WU's. I was impressed
with the obvious fail-safe thinking behind the routines and the very
logical methodology behind it, the more I use it the more I feel I
can rely on it. Roelof and Mike
keep a close eye on the posts to alt.sci.seti so any problems with SetiSpy
or SetiDriver
(a highly recommended combination) and they'll wade in very quickly.
Wade being the operative word considering how much other wet crap
tends to seep in there.
A few times to keep the numbers people happy...Lucien
has a dual 550 PIII system taking 8:30 which would not be cool
except of course that's 2 WU's polished off not one. The issue of
bus contention keeps the individual times lousy. Better to come out
with more WU's for TLC than keep
your average time down. A couple of tweaks (CAS2 RAM and disabling
shadow caching) to a PII 350/NT4 combo by guslg
have taken 17 minutes off his time, now down to 7:14. Pretty
respectable. A mildly clocked PIII 600E at 672MHz with no tweaks and
a lot of bits running lobbed out a 6:29 for MirageESO.
Just think what it might do if you benched it without Spy/Driver/AV
& the rest running in the background and also did the dirty on
the BIOS. Save another 30 minutes or so at least. Another rare
sighting of Rambus from fsgray,
over-hyped, overpriced and under-performing but if you've got it
might as well use it. A stock 667Mhz PIII in a HP Vectra with PC700
Rambus memory returned a reasonable 5:53. Memory hub translators and
high latency are always going to kill Rambus for Seti crunching, let
alone price and Intel's lack of support. In contrast a final time
from Frank for his Athlon which is also hampered by less than ideal
architecture for Setiage. This OC'd 800 (to 935MHz) manages a paltry
5:54 which is probably 30 minutes shy of its potential. But hell,
what do I know...I just look a the times and let you people do all
the work - it's a wonderful planet.
Finally it's official, I exist. Must do because there's a thread
in the Seti
& RC5 forum (''Who is this Max?')
and we all know that queries about dubious existence are central to
forum philosophy. So I can come clean and let you know that I'm
neither Rat Bastards
illegitimate love child nor a facet of his multiple personality
problem. Keep cranking.
Max out.
Good times, bad times
Better get the funnies out of the way first...is your case cooling
not all it should be well perhaps a big
fan might help! Just love the wheels to move it around on.
Thanks to BP6.com for that
one. A few weeks back I wrote in hushed and misty-eyed terms about
how slow times were just as worthy of respect as the blindingly fast
stuff that sometimes graces this page. It takes stubborn commitment
to wait for tens of hours for your WU to hatch. The Setiathome
project wants Arecibo radio data filtered (remember that stuff -
it's not just a team race after all) by our machines for scientific
curiosity and is relatively unconcerned about our hardware or
processing speeds. A 200 hour completion time is as valid as a 4
hour one to their database, also remember there are two million
(give or take the few hundred thousand inactive accounts) crunching
away. So you can only imagine my delight as a champion of the
snail to come across this posting on the alt.sci.seti newsgroup:
"David T. Wang, Setiathome
Hall of Shame Project 80386DX33, 128KB cache. IIT 80387
MathCo. 32 MB. Win95. Status: 100.000% complete. CPU
Time: 2353 Hours 15 minutes 31.8 sec." That's ninety-eight days
in old money. In case you are wondering the Hall of Shame is another
Carolyns Clinic project. I
guess those who suggested mockingly of dusting down the ZX spectrum
and tape drive better get to work if they want to beat that.
Fortunately for me there are still game souls who are sending
benchmarks in so I shall proceed to the numbers. In the humble K6
league Bjarne has produced a
slowish 23:36 with a 233MHz sluglet, but hey they all count. Another
time from Bjarne was 12:34 using
Win2000 on a 450 K6-2 which is only two minutes slower than his W98
(I assume the same machine - dangerous I know) submission from last
week. Would expect more, curious. Sucking Coppermines (dual PIII
800's) are bitnauts problem, a
really sad 12:14 even allowing for 2 WU's output. But on an 8
multiplier your overclocking options are limited so perhaps you
might just have to accept the inevitable and hope for better things
cachewise from the V3 client. Still in dual vein but this time
Celeron 550's running Mandrake 7.1 with a single client benching Nemo
pulled out a 7:56. Celerons at this bus (100MHz) can go down to 6:30
depending on RAM timings and especially OS, Linux costs time but at
least you are out of the clutches of MS. Your time is very slightly
down on other Linux boxes, see Drew
at 7:55 but 93 bus or Guru at
7:24.
To relieve the fairly under whelming benchmarks so far neo1999
has produced a pedestrian 7:33 that surprisingly has brightened my
day. Nothing special until you realise that it comes from an
overclocked PII 233! Running at an exceptional (for a PII) 140FSB
and 2.5 multiplier this $35 door-stop clocks in at 350MHz with CAS2
Ram. As a strange comparison guslg has
a more normally aspirated (FSB only 112MHz) PII 350 also W98/128MB
Ram equipped, that comes in 2 minutes quicker at 7:31. Just to
remind those of you who disdain fruity hardware Jonnymac
serves a timely reminder of the beauty of G3Macdom with 4:57. Nice
to better the five on a 500 G3 and only 4 minutes adrift of the
fastest G3 on the table. Finally I'll leave you with the words of neo1999,
'please allow me to bump the fastest Katmai' he cries weighing in
with another serious 147MHz bus on a nominal 450 Katmai (aka 662MHz)
to achieve a very pleasing 4:38. Competition and ego linked
inextricably.
While I'm in an expansive mood and relatively pain free, a
visit to the chiropractor helped greatly, my main excuse for this
late update are the days I've spent helping out at a yet another
Triathlon. Not quite in the spirit of TLC but there are other things
in life than crunching WU's (just) - "burn the heretic"!
But the next horrifying step (expensive Chiropractic advice) is to
move my system (cables all over the shop) to give me a better
position to work from so that the neck/shoulder pain continues to
reduce...the joys of constantly checking SETISpy
have lead to terrible posture!
Max
(hunched over keyboard) out.
|
