SuperPI 2.0 on the horizon?

[5-Clicks]

Hey Fugger (or anybody else who's in charge of sPI database and upkeep)

Is anything going to happen to sPI? I mean, now that the 1M is barely a benchmark anymore and doesn't signify very much stability at all (9.7s by coolaler and the PC still wasn't stable enough to capture a screen!). So now that 32M is getting into the 14 minutes time zone, wouldn't it just make sense to add more digits to the super pi calculations? Maybe 64M and even 128M? Is that possible? As CPUs and RAM get faster and faster, the time it takes to calculate PI decreases and thus doesn't put as much stress on the system...meaning no proven stability.

So to sum it up, would it be too much to ask for a SuperPI 2.0?

[Cordovader]

To prove stability you have prime, Sp2004, Rosetta...

IMHO Spi is more like a competition than a stability program, althought it can be used to prove some stability

[5-Clicks]

I understand what you're saying, but it'll eventually get to the point where it's just the same as who gets the best SuperPI 16k socre.

What I like about SuperPI is that I can let it run for a half an hour or so and if it passes then it's something that's accomplished and regarded as stable and a worthwhile run. But because the computers now are faster than last year, the 32M stability rating for a conroe is now equal to the 16M rating of an FX-55.

[NickS]

Well until it gets to that point, I think it's fine for now .

Nick

[lawrywild]

eh?? why would you create a new version of superpi when there's sop many more time-consuming tests available??

If 1M doesn't test stability than why not run 32M???? Doesn't sound like you thought this post through.. =/

[5-Clicks]

If 1M doesn't test stability than why not run 32M???? Doesn't sound like you thought this post through.. =/

Because 32M doesn't test full stability either. It's equal to 16M on an A64.

[lawrywild]

Because 32M doesn't test full stability either. It's equal to 16M on an A64.

What are you expecting to happen in a new superpi??

it calculates Pi to a specified amount of digits.. what exactly can you improve? LOL

[NickS]

The length of the test? Like calculating Pi longer than the current "specified amount of digits."

Nick

[don_vercetti]

So surely we would just add 64M, and 128M, yes?

[sai]

SuperPi is perfect to test ram stability in windows. But since we're going 2GB and eben higher we need more digits (like 64M, 128M). That'll use more ram. BTW actually there was a version with 64M, but dunno anymore how it was named

[NickS]

PiFast maybe?

[Fixxxer]

I'd like to see a new version of SuperPi modded by XS, with 64 and 128M, just to empty the 10L dewar.

With the C2D processors, a LN2 session can be made with very few LN2.

[kiwi]

Lol, SuperPI is not a true stability tester

SuperPi is perfect to test ram stability in windows. But since we're going 2GB and eben higher we need more digits (like 64M, 128M). That'll use more ram. BTW actually there was a version with 64M, but dunno anymore how it was named

Sorry, it is not, not even 32M

Example,
1 bad memory stick passes memtest #5 and #8 but gives 1000s of errors is other tests and immediately fails to run 3DMark. Yet, it passes SuperPI 32M. This is ridiculous



SuperPI is only competition, not a stability tester

Try Rosetta, 3D and real gaming, BF2, UT2004 etc

[sai]

@Nicks

No SuperPi with 64M option

@kiwi

memtest is just to get a rough estimation what the ram can do in windows. and i had many games running flawless but superpi 32m would fail. a 64M version would need about 1GB of ram so if 1GB is unused it's possible that this unused part has errors, BUT I never had any problems after testing with 32M and 64M would even be more accurate

[Janchu88]

what i suppose to be much more interesting...

1.The option to activate Multithreading @superPI (disabling multithreading as well)
2. SuperPI with 64bit extensions

[nn_step]

ANd I hope to god they add a feature that makes 1Mb 10s validation different than 16kb 10s Validation. so that we don't have to mess with that bull, should someone try to fake it.

[ewitte]

suggestions:

- Multithreaded

- Offer an option to loop for x number of cycles. Running 32M Pi for 2 days straight while utilizing all processors would be a better indication of stability. Although still nowhere near perfect.

[farksy]

I think superpi is more or less a ram tweaking program than a stability benchmark.

I've always regarded it that way ...

[cirthix]

A better idea would be to run superpi for a specified amount of time and record/compete with the number of digits that could be calculated in that amount of time. 30s, 1m, 2m, 4m, 8m, 16m, and 32m would be good places to start.

[NiCKE^]

Can't we just start to bench 4M instead?

[SMa]

A better idea would be to run superpi for a specified amount of time and record/compete with the number of digits that could be calculated in that amount of time. 30s, 1m, 2m, 4m, 8m, 16m, and 32m would be good places to start.

That's a great idea.
That would make it worth to call the new SuperPi "2.0"

Also adding multithreading would be great.

[Cornelious0_0]

To prove stability you have prime, Sp2004, Rosetta...

IMHO Spi is more like a competition than a stability program, althought it can be used to prove some stability

Couldn't have said it better, that's exactly how I was gonna be approaching this thread.

[JuanFlaiter]

What about making standard SSEX versions standard?

[@_dud]

Multithreaded would be sweet! then the scores would all cut in half. Here we come 5sec 1M.

Like others I think Super pi is more of a competition than stabillity, but it would be cool to have a 64m, and 128m, and to be able to loop it.

personally I think superpi is fine for now.

[Tulatin]

Systool can do 64-128M, but it says perfect stability, when Prime/OCCT take s...