Jun 22
Hedged Bets
posted by: Player0 in computering, projects on 06 22nd, 2008 | | No Comments »

I finally finished installing my Asus P5E3 Premium.  I finally made the switch to DDR3.  I took a big risk in this not doing anything for my performance.  I just wasn’t able to make the Q9450 stable at over 420MHz on the Maximus Formula and suspected this was a chip limitation.

I didn’t have high hopes for the P5E3 in solving this issue.  But, in fact, I gained 30MHz FSB on this board.  450MHz is rock solid while I have problems at 460 which I can’t seem to resolve.  However, 450-460MHz is the sweet spot on the X48 because of latency and tRD etc so I’m happy to be here.

The serious limitation is the fact that I can’t use a 9x multiplier with this chip.  3600MHz is definitely respectable but I bet this could ramp up to 4GHz easy with a better multiplier.  It makes the E8400 look better and better since it can hit insane speeds.

The question remains… is a really fast dual core better than a slightly fast quad core?

As for DDR3, it’s really nice not to have to worry about memory holding me back.  I’m not sure it’s gaining a lot of real world performance.  But overall I’m running about 5-6FPS higher in Crysis which makes it playable at the highest settings.

More info, pictures, and the new watercooling set up can be found here.

I’m not completely happy with the P5E3 Premium.  It comes with 802.11n and I wonder how much money I paid for that complete waste.  Is Asus really that out of touch with their users?  Anyone who’s going to blow that much money on an overclocking motherboard is probably using it for games.  And if those are online games then you really don’t want to be using a wireless connection.  The added latency and instability just isn’t worth it.  Also, coming from the Maximus or Rampage formula with the POST code reader, the on board power/reset switches, the CMOS clear on the back panel, etc are all missing features on the P5E3 Premium.

But my biggest complaint is the lack of voltage and temperature reporting.  This makes overclocking REALLY difficult since the board will over or undervolt any particular setting you make.  If you’re considering upping something like PLL or Termination voltages you really need to know what they are set to at default.  For example, on the Maximus the NB voltage defaulted to about 1.61V.  If I didn’t know that then I would have considered the Auto to a 1.55v NB setting an overvoltage when in reality it would actually be adding less voltage to the mix.

Jun 12
Q9450
posted by: Player0 in computering, projects on 06 12th, 2008 | | No Comments »

I finally broke down and purchased a Q9450.  MicroCenter had one at better than online pricing so I couldn’t resist.  Unfortunately, it is overclocking poorly for me.  Something is limiting my FSB to under 428MHz.  The 1333MHz FSB of this chip only gives me a 8x multiplier.  So I can only pull about 3.4GHz out of it.

Not sure if it’s the chip or my Maximus Formula.  When I see other modern Intels breaking 4GHz without a sweat I really start to wonder what I’m doing wrong.

I’m going to pick up a Rampage Formula I think and see if I do better with that.  If that doesn’t work than I know this chip is garbage and I’ll get an E8400 or something.

I’m also upgrading my watercooling system with an Apogee GTX Cu, a Laing D5 pump, and a set of NB/SB/Mosfet waterblocks from Slovenia.  I’ll be removing the Laing DDC pump from my system but I’ll still have the D4 in there.  So dual 12v pumps which are not bad to go with my roughly 3×120mm RAD set up, a whole bunch of chipset waterblocks and my watercooled 8800 Ultra.  Why the big upgrade?  My NB is running too hot because it’s getting heated up from the CPU and GPU and i need better detla T’s out of the waterblocks to counter act this summer heat.  I never replaced the stock thermal goop on the Maximus Formula with some AS5.  So my NB temperatures crawl up to 45c easy which is the exact point that X38 becomes unstable.  It might account for my shitty FSB, it might not.  But if I’m going to water cool I’m going to do it right.

Jun 12
DCSig 4.5 Update
posted by: Player0 in dcsig, php, projects on 06 12th, 2008 | | No Comments »

I’ve been doing some PHP coding at home this past week and have made some new progress on the new version of dcsig.com.  DCSig has needed a full rewrite for a long time now.  The current iteration, version 4, only supports Folding@Home and does that extremely slowly.  The website and user interface are terribly old fashioned and clunky.  The PHP code itself has been massaged and tweaked since I first started learning PHP many years ago.  It takes quarter of a day to process data that should take less than an hour.  There’s no use in salvaging 4.

But 4.5 is a monster.  Nothing is salvageable from the old code base.  My major slowdown has been in determining the framework I’m going to build 4.5 on.  I’m on my third framework since beginning the rewrite.  I’m also on my third Folding@Home stats parser since beginning the rewrite.  This has been the cause of my delay as I have really been learning a lot of new technologies.

But I have it set now.  I know exactly what I’ll be using.  For the MySQL database layer I’ve decided on PDO.  It’s a native MySQL driver with prepared statements, parameter binding and most importantly it’s very speedy.  For JS I’ve decided to go with YUI.  It’s heavier than other frameworks but it brings a lot to the table in terms of modules from color picking to chart generation.

Where am I right now?

I have the new DB schema mostly done.  This was an important step since the DB layer has been my primary source of woe in the old code.  I’ve grown a lot with MySQL optimization in the past few years.  It’s probably my biggest improvement.

I have a brilliant F@H parser now.  It’s faster than I could have ever imagined.  It may even be fast enough to support team threat/overtake for the top 1000 teams.  The dynamic F@H stat page scraper is done with my cutting edge code but the static F@H page is still using my slightly older code.  I’m debating whether or not to rewrite the static stat page scraper again.  I’d rather put some of my time in to getting other things done.

The move to PDO from my own DBI framework I’ve just started.  The F@H parser needs to be updated to PDO which will take a few extra hours up but in the end it’s worth it.

I have a working sessions class for login and account handling.  Its loaded with modern security techniques.  The current version is riddled with security holes and just doesn’t make sense in general, even for ancient PHP code.  I have a solid Exception handling class as well as my general Status/Debug logger class working great.  I have my Registry class working great.  I’m poised with all sorts of UTF-8 support.  I have a decent start on my User/Account class now.

What’s next?

Finishing up the account data situation and moving on to service data where you can enter F@H teams, ids, etc.  I have some new features here which should be great.  I *PROMISE* to support more than F@H this time, I just need to add some extra classes.

Skin and Sig Image generation is a bit of a mess.  This is a big change from the old version although I think I’m stuck using GD2 still which is slow and memory leaks all the time.  I may consider the switch to ImageMagick.

Then comes the fun part.  Using AJAX to build the sig creator.  Then all the UI stuff for dealing with accounts and sigs.  Then building YUI graphs from the parsed data and such.

Then DCSig 4.5 will be open for beta testing.  Only new sigs and new accounts will be supported at that time.  I have not decided to import old accounts and skins from DCSig 4.0 yet.  It will happen after open beta.

How much longer is this going to take?

Forever, as usual.  Someone might have something cooler by then.  Or they already do.   As much as I would like to be productive on this real life has other priorities for me.  But don’t give up hope!  Some cool things will happen eventually here!