Update!
The latest version with cheat protection, more accurate timing and checksum is available here!
------------------------------------------
Well here it is!
My Super PI mod now also includes anti cheating code. Any attempt to cheat will now horribly fail. I changed the text in the titlebar so you will know from screenshots wether or not a cheatfree version was used.
If anyone does happen to find a way to cheat it please send me a PM telling me how you did it so I can add protection. I doubt people with the skills to cheat this version are seriously interested in faking results
I can verify that this version is the the only Cheat Free PI program that I know of.
I tested it as much as possible, and theres little to no cheats possible on this version.
hopefully we don't get some dumb programmer trying to cheat on this version, like we did with CPUZ 1.26...
__________________
Quote:
"The command and conquer model," said the EA CEO, "doesn't work. If you think you're going to buy a developer and put your name on the label... you're making a profound mistake."
I made few test, on a prescot 3200@3600 (second pc)
classic superpii I score 35.890 process timer cheak
with the new anti cheat i score 36.493...
and Im not the unic, on a italian forum few peeps found the same problem...
any idea?
This version is based on original unpatched Super Pi 1.1E.
The code I added only uses CPU time every now and then and totals to no more than say 2-300 instructions over a whole run, so the results should be exactly the same as with the original, or maybe 1 msec slower if you're really unlucky and cross a msec boundary during those 2-300 instructions
What version are you guys using (url please)? Maybe I based my mod on the wrong version :/
well here in italy im not the unic, we opend a topic about this new Super pii and they all have the same problem....is give 0.800 more from the classic , personaly I use the english classic version of the superpii Pi 1.1E
if you watch here ,there are some peeps posting even screen showing different scores,I know is not english., but you can undestend http://forum.hwupgrade.it/showthread...hreadid=891382
Thanks, I'll look into it.
I just did soime quick oc-ing and benching and I came very close to my personal record (like 100 msecs difference). The difference can be explained by that I was running a slighly lower FSB this time. At this moment I can't think of any other explanation than that you guys are comparing it to a different version. But as I said I'll look into it
Edit:
I compared the one you uploaded with the modded, results are in the screenie. They give identical results for me :/
Detonator, I took a look at the screenshot in the thread you posted (this one)
and I think the difference there was at most 0.3 secs. If you look at the results in the right the final time has to be a very high 35. So the results are not off that much. Also I seriously doubt that processtimer can give more accurate times than Super PI itself. I haven't tried it out tho but theoretically it's just impossible.
If you could test with this version. That's the old modded one, the actual time measuring hasn't been changed only displaying it has been changed in that one. And please compare it to both this new one and the results you get with processtimer. The results in that version *are* completely accurate with the original Super PI. The original Super PI records the time in msecs but just doesn't display them.
Let me know what you find out.. Preferably a screenie with the 3 of them run right after each other. Thanks This is impossible to test for me as I get the exact same results for every version.
I get varying results, for example on 64K I get a variation of about half a second.
And for 1M, I get a variation of about 1 sec.
__________________
"Strive for perfection in everything you do. Take the best that exists and make it better. When it does not exist, design it." - Sir Henry Royce
Half a second on 64k? Uhh..
Could you check with this version and see what kind of results you get? I'm starting to doubt my timing code works flawlessly on all systems..
Hmm..
Is the 35 sec the result from this run or from a previous run?
I have to agree tho the timings seem to be off on your system, looking at your 16K score. My system (XP3200+) does 16K in 0.219s. But... As the time with the orignal mod is over 35 secs as well I think the problem is with your system, not with my code. The new one which has more code for timing is even faster on your sys than the one with original timing.
I really don't know what to do about it.. But the results in left top cannot possibly be wrong, unless they are wrong in the original super pi.
is not just me, if you cheak on the link I posted ,almoust all the pepes over there had the same problem, a different btwin 0.700/800 msm
those test were made on a 3200 prescot D0 on abit ic7max 1024 ram.did you try your super pi on an intell sistem?
I haven't tested with an Intel system no, I'll test it on my P4 2.8 but I doubt the results will make any difference.
It still is just as impossible to get wrong results. I've been thinking about it. The anti cheating code I added actually prevents wrong times from happening.
I've used babelfish now to translate the thread for me..
XstasY (you I assume?) wrote:
"on the classic superpii I make 35,890 controlato from process timer, on the new one I make 36,498.."
ProcessTimer is far less inaccurate than Super PI itself. It can in no way tell with the same amount of accuracy when the calculating process is started or stopped. Basically it lags. Of course it will lag both when starting and when stopping so you might think it'll equal out in the end. But when calculating is started, all cpu power goes to super pi so it will lag more. It will notice faster that the process was stopped, so it will give a better result than you have actually achieved.
Next, Mo3bius
From looking at the screenshot I'd say results are off by max 0.3 secs. But look at all the crap he has on his setup and what he's running (the toolbar for example)! I wouldn't expect any stable results from that setup, with or without mods.
Next, Jok3r88
super devout: 34 sec (whatever devout means
super devout mod: 34.857 sec
Results match.
The rest of the posts I cannot make any sense of because babelfish apparently isn't very good at Italian
But.. I suspect that if you'd run the modded program on a clean install and close anything that might affect the time (including explorer, mirc, etc) you will get roughly the same times.
I know for a fact that my code doesn't cause this so the problem must be somewhere else. I'm sorry, I dont wanna sound like an but there can't possibly be anything wrong with the code. My code is only used 20 times during one run. Even if it would be slow as hell, it would not give a raise of even 0.1 sec. The better results you or others may have had before were either caused by inaccuracy or plain luck.
Now I'm off to test with the P4, lets hope it wont make me eat my words
btw those test made in the italian forums are made with the proces timer, they all say the same, a different of 0.800 like I told you before.
I want just to help you out ,becosue wuld be a must to have a spii with msm
and my own test are made on a clean istallation no twick at all, fresh Xp istallation.
I just finished testing on my P4 2.8. 512 MB PC333, pretty fresh XP SP2 install. It's a Dell
First I tested regular Super PI, result was 53 secs
Next I tested mine, result was 53.641 secs
It seemed the first test would've been a low 53 tho (some of the times in between were faster), so I tested again. This time I got the same second values for all of them as I got in mine.
I can't do anything about this. My version uses the exact same amount of memory as the original, uses the exact same code for calculating, the only difference is that when time is taken (20 times per run) it will cost a few more clockcycles. And believe me that will not cause a change of even 10 msecs over a whole run.
The error must be in italian PCs or software because I just cannot get it to make a difference here no matter how many times I try
Also please don't rely on ProcessTimer when comparing results. It might be good for seeing the differences between different memory timings and clock settings but it does not give an accurate time. Mine shows the actual time it took to calculate the digits, plus or minus around 15 msecs depending on your timer resolution (depends on OS).
I understand your frustration tho so I'll keep testing and thinking about it and hopefully I'll find an explanation for it. It's just too weird. I still believe its impossible that the little code I added could have anything to do with it.
AHA!
I just figured something out.
To make a long and boring story short, the problem might be memory related. This is the only thing that has been changed for all of the code. It might be the case that on Intel systems this slows down everything a bit. Actually it MUST be that because if not then I really don't know
I'll try and whip up a new version tonight so we can see if it makes a difference.
Super PI with msec/anticheat mod
Linear Mode
