SuperPi 1MB is now dead ! => what will replace it ?

[zytrahus]

Hi,


Just some words to say what I think about "the new superpowerfull superpi104" ...

The new tweak has completely kill superpi. Before it, it was already difficult to make a right comparative between the different realases (japanese superpi ; english release... ), but now there is really no interest to use it...

I could understand that somebody wants tweak to his own use, but I really don't understand how this "somebody" could use it to run for overclocking results.


Superpi didn't have any "optimisation" for current cpus so it was a good way to compare each cpu : AMD or Intel ; frequency ; memory ; cache... Each parametre had its own influence.




I know I'm the only one who find this tweak as poor as s*** and I just want give my thanks to "the superpi killer". I don't want to know who is it, I just hope a way to replace this small bench (with maybe more accuracy in result)


../..

[Kunaak]

I know I personally won't be giving a damn anymore when I see super Pi shots from intel. but Super Pi is still very valid for AMD Xps and A64's, and Intel Northwoods.

but honestly, I always perferd Hexus Pi-fast myself.

[Mini]

Is it the prescott optimizations???

The 1 that lets the prescott use all of its powers instead of limiting it???


A bit lame that a prescott is only faster than the NW-C cuz of the bigger cache... It has ALOT more features that now can be used

This is like saying that dual channel ram is cheat

Cuz not every1 has it... Only socket 939 A64 have it ;( *BUHU*

NOTE: If it is not the prescott optimized thingy then disregard this reply

[bias_hjorth]

Is it the prescott optimizations???

The 1 that lets the prescott use all of its powers instead of limiting it???


A bit lame that a prescott is only faster than the NW-C cuz of the bigger cache... It has ALOT more features that now can be used


This is like saying that dual channel ram is cheat

Cuz not every1 has it... Only socket 939 A64 have it ;( *BUHU*

NOTE: If it is not the prescott optimized thingy then disregard this reply

Dude why are you even answering this thread. The reply tends to disagree with your own word. You´re talking about we should stop whining and in the note you´ say its okay. That like I said that you shouldnt answer my reply cause of ignorrence than less its about the reply.

[st0nedpenguin]

Is it the prescott optimizations???

The 1 that lets the prescott use all of its powers instead of limiting it???


A bit lame that a prescott is only faster than the NW-C cuz of the bigger cache... It has ALOT more features that now can be used

This is like saying that dual channel ram is cheat

Cuz not every1 has it... Only socket 939 A64 have it ;( *BUHU*

NOTE: If it is not the prescott optimized thingy then disregard this reply

Well then, surely if it's fair to use a Prescott only "optimisation" to enhance performance, there should be some alternative tweak for the Athlon64 platform too, and one for the Northwood, cause I'll guarantee both those CPU's could be optimised to run SuperPi faster too, considering the programs age?

It's not about the Prescott gaining performance with the patch, the problem is that no other CPU is gaining performance too, which to me, makes it an unfair benchmark now, especially as there's no way to visibly tell the difference between a patched and unpatched score, unless you know exactly what to look for.

And that, is why the poster is saying SuperPi is dead, and why i agree.

[bias_hjorth]

I agree with you zytrahus. Whoever did the patch should take a standing applause. :" Thank you for destroying a great benchmark"

[saaya]

I agree with you zytrahus. Whoever did the patch should take a standing applause. :" Thank you for destroying a great benchmark"

yeah...


imo benchmarks should not get special optimizations at all, they are supposed to show the raw performence of hardware...

so lets assume one piece of hardware can run applications faster than the other, but with optimizations the latter becomes faster than the first one. then the optimized benchmark will mislead people into thinking the latter hardware is faster, wich is true with optimizations, but the fact is that 9 out of 10 applications wont get an optimization, so in those apps the first piece of hardware will actually be faster.

benchmarks are supposed to indicate the performence of a piece of hardware it will give us in a range of applications, as soon as you optimize a benchmark it will no longer show you what performence we can expect from hardware ina wide range of applications.

a good result in that benchmark will only mean that, that this piece of hardware performs well with that benchmark app, it will no longer serve as a benchmark to get an idear of the performence of the hardware in other applications.

thats the odd world of benchmarks... as soon as they become popular there are optimizations for them wich nullifies their function as a general performence indicator...

[Skeet]

Imho it would be best if you could in the benchmark say, have two sorts of modes: RAW performance and Optimized performance. The first would be the "official" comparable benchmark and the second more of a tech perk useful for testing new CPUs and also satisfy the bunch who wants to have such things.

[kyosen]

Hi, I'm a japanese. Yes, I love SuperPI, of course

At first, I've noticed this thread yesterday.
So, Nazaar,
> they are probably not thinking, they are just pouring ln2 on their cpu's
--
Wall of language is higher than you expect, for our Asian people

And I already picked up this SuperPI SSE3 patch affair on my own BBS.
http://219.113.251.125/c-board/c-boa...ne;no=3045;id=
Sorry, of cource in japanese.

zytrahus, I appreciate you for starting this thread and effort to make
new benchmark program!

CodeRed(thanks for your A64Tweaker),
> I wonder what the Japanese overclockers think about this prescott cheat?
--
SSE2/SSE3 patch to SuperPI itself is interesting,
but
"Registering score by patched SuperPI without mention to patch"
is another thing...it's a cheat, I think, as you almost all think same.

zytrahus,
> I mailed the japanese lab last week to know if source code is open or not.
> It's not. So difficult to change something (on the interface for example).
>
> A good thing would be to identify (graphicaly) and pack (crypt) the code to > lock any possible change on the code.
>
>
> But it sounds difficult
--
Yes, crypting/locking is difficult.
Today I've created 6 SuperPI binaries, i.e.
Original Solver Version, SSE2 Version, SSE3 Version,
both of English and Japanese versions.
They have identifications, but no crypting/locking.
So they are better than "silent" patched binary, I think,
but not perfect at all...
Anyway I attach several screenshots.

[zytrahus]

Hi, I'm a japanese. Yes, I love SuperPI, of course

At first, I've noticed this thread yesterday.
So, Nazaar,
> they are probably not thinking, they are just pouring ln2 on their cpu's
--
Wall of language is higher than you expect, for our Asian people

And I already picked up this SuperPI SSE3 patch affair on my own BBS.
http://219.113.251.125/c-board/c-boa...ne;no=3045;id=
Sorry, of cource in japanese.

zytrahus, I appreciate you for starting this thread and effort to make
new benchmark program!

CodeRed(thanks for your A64Tweaker),
> I wonder what the Japanese overclockers think about this prescott cheat?
--
SSE2/SSE3 patch to SuperPI itself is interesting,
but
"Registering score by patched SuperPI without mention to patch"
is another thing...it's a cheat, I think, as you almost all think same.

zytrahus,
> I mailed the japanese lab last week to know if source code is open or not.
> It's not. So difficult to change something (on the interface for example).
>
> A good thing would be to identify (graphicaly) and pack (crypt) the code to > lock any possible change on the code.
>
>
> But it sounds difficult
--
Yes, crypting/locking is difficult.
Today I've created 6 SuperPI binaries, i.e.
Original Solver Version, SSE2 Version, SSE3 Version,
both of English and Japanese versions.
They have identifications, but no crypting/locking.
So they are better than "silent" patched binary, I think,
but not perfect at all...
Anyway I attach several screenshots.

sounds good

[kristos]

Yes, crypting/locking is difficult.
Today I've created 6 SuperPI binaries, i.e.
Original Solver Version, SSE2 Version, SSE3 Version,
both of English and Japanese versions.
They have identifications, but no crypting/locking.
So they are better than "silent" patched binary, I think,
but not perfect at all...
Anyway I attach several screenshots.

could you give the links for those plz?

[Sucka]

benchmarks are supposed to indicate the performence of a piece of hardware it will give us in a range of applications, as soon as you optimize a benchmark it will no longer show you what performence we can expect from hardware ina wide range of applications.

On that same note, should everyone use the same driver set for benchmarking? If you'll notice, each new set of drivers offers up a couple hundred more points in 3dmarks than the previous set, those to can be considered optimizations as well? I really likes PI as a raw benchmark, and will continue to run it without patches.

[zytrahus]

I don't think Dual Channel and the prescott optimization have a way to be compared

Saying this optimization is not a bad thing because it works only on prescott, is not looking very far : All future Intel cpus and certainly AMD ones could work with this optimization and so there won't be only one cpu working with the cheat...

[bias_hjorth]

I dont even think its the Prescott that have some sort of optimization, I honestly think that its the program thats optimized for a larger cache combined with a longer pipe. But I could be wrong :/

[Crankster]

It's a wholly different benchmark.
Really Mini, do you belive what you post?

[Sons_of_Piru]

i agree
this patch has killed superpi

[Mini]

I think its fine.. Seriously... I dont even have a prescott...

I have heard that it lets the prescott use some of its instructions and stuff... Thats what i am thinking is ok...

If this is a cheat identifying the prescott and running another or faster bench by cutting something away then it sux...

But its like 3dmark... ATI optimizes 3dmark05.. So does Nvidia... If its the same benchmark they run and no quality is lost then its IMO not a cheat...

Edit: Then again... If Futuremark made a patch that ONLY improved ATI or Nvidia or AMD/intel performance that would suck...

But 3dmark05 is the same... Nvidia has SM3.0... ATI doesnt...

Prescott has new instructions... A64 and NW dont?

[zytrahus]

Right of 3dmark 2k3 2k5, but CPU influence is not important and these bench are made for GPU now... At the first 3Dmark (2k1 we already found some optimizations.)

The most important interest of superpi was is "non optimized" way, then :
- overclocking
- tweaking mobos
- killing task in windows to reduce task and memory

are not cheat, but modding the "code" of the *.EXE is.

[CodeRed]

I wonder what the Japanese overclockers think about this prescott cheat?

[funkflix]

That patch really sucks!

@ mini

I'm sure u can optimize SuperPi for all newer proz. but we dont want that!!
U get it?

@zytrahus

I thin in ur SuperPi-List are two faked nforce2/XP results. Look @ 35sec,
there are one with 11,5 x 200 2,5 3 3 6 and one with 12 x 202 2 3 2 6,
thats wont work! U need 11 x 250 2-2-2-11 to get 35sec SuperPI, and i need
11,5 x 247 2-2-2-11 to get it!

[zytrahus]

thanks for message.
The DB was a temp one... I wanted add screen to valid it, but I was out of time.... And now, I think I'm going to find an other bench (maybe pifast) and rewrite a DB directly on http://www.cooling-masters.com

[funkflix]

thanks for message.
The DB was a temp one... I wanted add screen to valid it, but I was out of time.... And now, I think I'm going to find an other bench (maybe pifast) and rewrite a DB directly on http://www.cooling-masters.com

Anyway, thanks for adding so fast

@ Mini

U dont get it, but np its ur opinion!

[LardArse]

I wonder what the Japanese overclockers think about this prescott cheat?

I think they have their own code of benching that they observe... I mean seriously,,, would it be that hard to photoshop a superpi score? If it can be done all this while with a software, then why is this patch gonna destroy something that is already not there in the first place? :the infallibility of a Superpi screenshot as a legit performance indicator

[CodeRed]

I think they have their own code of benching that they observe... I mean seriously,,, would it be that hard to photoshop a superpi score? If it can be done all this while with a software, then why is this patch gonna destroy something that is already not there in the first place? :the infallibility of a Superpi screenshot as a legit performance indicator

I agree, the Japanese oveclockers have a strong sense of pride in their work ... too strong to cheat by photochopping. But there are many people that wouldnt hesitate to cheat, now it has been made easy for them.

[TodB]

Sorry for the offtopic, but is there a link that I can use to download that prog?? couldn't find it anywhere on the web...