Super PI with msec/anticheat mod

[snq]

That's good to hear

So.. does anyone have any more ideas for new features? I'm kinda bored

[scsim]

Hi! I am facing some strange issues. When I OC my P4 3.0E to 4.3GHz, Super PI will post crazy results. At 4.289GHz, my result is 31.562sec. When I bump my FSB 1 MHz higher, the results became 1hr 51min. Of course, it doesn't take that long to complete (slightly more than 30sec). Just that the time shown for each iteration is wrong.

I used to face this issue when WinXP couldn't detect anything more than 4.3GHz problerly. SP2 seems to solve that problem but it appears that no all programs are working. Any work around?

Thanks!

[bachus_anonym]

@snq

I've just seen some CRAAAZZYYY time in SuperPi 16M with the latest MOD. Checksum turnes out OK Something that might be similar to above post... Just look for yourself, bud:

11:30 - P4 3.0EGHz @4501MHz (15x300)

09:06 - P4 530 3GHz @ 3890MHz (15x259)

The author of that first score says that he was not cheating and numerous runs show pretty much same results...

Something must be wrong with that MOD. Can you take a look at it, please ?

[snq]

Hmm.. that's odd.
An explanation could be that your timers are off at such high clocks. I still use windows timing functions in this version, but I believe this timer relies on clock speed and afaik it's calibrated at boot time. Now I'm not sure exactly how this is done but it could be that sometimes (?) the timer is not recalibrated when you change the clockspeed using clockgen or some other tool?
Obviously it's a problem tho, I'll see what I can do about it! I just hope I don't have to fall back on the old 15-msec-jump-timing :/

[bachus_anonym]

Hmm.. that's odd.
An explanation could be that your timers are off at such high clocks. I still use windows timing functions in this version, but I believe this timer relies on clock speed and afaik it's calibrated at boot time. Now I'm not sure exactly how this is done but it could be that sometimes (?) the timer is not recalibrated when you change the clockspeed using clockgen or some other tool?
Obviously it's a problem tho, I'll see what I can do about it! I just hope I don't have to fall back on the old 15-msec-jump-timing :/

Well, this person says that with Original SuperPI he gets around 14:00, which is about right... So you might wanna try to poke around and see what could be the problem.... And I think so far this happened only on Intel / Prescott...

[snq]

The problem is I changed the function used to get the time, so the problem has to be there. This is really bad news because I really loved this new function without the weird jumps. But if it's acting like this it will have to go

Unfortunately I don't have a prescott to test with but I will try starting with a heavily underclocked system and throw up the FSB while in windows, that should show me if this function adjusts properly or not.

[snq]

I don't know anything about clocking Intel CPUs but do those things support HT and if so do you have it enabled?

If so that could be causing the problem.

On a multiprocessor computer, it should not matter which processor is called. However, you can get different results on different processors due to bugs in the basic input/output system (BIOS) or the hardware abstraction layer (HAL). To specify processor affinity for a thread, use the SetThreadAffinityMask function.

[snq]

Can you guys test this one?
It has the old timing functions, still msecs but with small jumps. I guess for calculating 16M/32M 15 msecs more or less wont be a big deal

[scsim]

Thanks, the new version is working properly now. My CPU is running at 4.35GHz instead of 4.305GHz, thus its no longer 1hr 51m, but

http://home.no/trac3r/super_pi_mod.zip

I downloaded from this mirror above and I gotten:



After downloading the 1.4:



Its working! Thanks!

[Sucka]

Good work on the PI updates. Now once everyone can use SSE3 we won't have anything to complain about :p

[bachus_anonym]

thanks snq,

I posted this information along with a new link on my native forums... I will let you know how it worked out

[snq]

Its working! Thanks!

Great I still want the really precise timing back tho. I'm thinking of what I could do to work around this problem. On my own system(s) the amount I clock doesn't affect the timer but I guess windows or the bios freaks out at that speed

[ArcTan]

hmm
doesn't seem to work with clockgen(nf3 one)?
It gives out the same time as the speeds in which I boot at

[snq]

ArcTan, that's what I was afraid of. The timer is not reinitialized
Try this one: http://superpi.radeonx.com/super_pi_mod-1.4.zip

[ArcTan]

thanks
works now

[Sons_of_Piru]

impresive reverse engineering work, good job

[snq]

Thanks Sons of Piru

[kiko]

same settings. :/





Ciao

[detonator]

same the guys abouve, my ptime is still 0.500msm differetn from the originall....
anti cheat=35.629
originall=35.140....
inprouved but still wrong

[TEDY]

what is NOT EXACT IN ROUND ?

280*9....prime stable and superpi not ?

works now with 1.4

[eva2000]

didn't know you had a new version out heh but here's what i managed

max pifast and superpi 1M able

[kiko]

small improvement w/ previous version.



Ciao

[mcnbns]

Thank you, SNQ! I might do some old school Athlon XP benching and see if I can't beat my old 36s record.

[C3]

Click on the pic to get the highres version!

What do you think?

[fordf250]

Dfi 875p-t with intel 640 on air