Jump to content
HWBOT Community Forums

The official Old School is Best School Round 5 thread.


Massman

Recommended Posts

  • Replies 79
  • Created
  • Last Reply

Top Posters In This Topic

Good point. So basically any slower SuperPi 1.5 calculation that 24h is need, right? That should be tested easily w/o overclocking - the AMD K5 PR75 is slow-enought... And the machine seems stable enought, right?

 

Super_Pi_1_M_6_32_705.png Super_Pi_32_M_5_36_25_606.png

 

...

 

Yet still I failed to manage SuperPi 32M on same mobo with just exchanged the CPU to AMD K5 PR75 :( Overclocked to 90MHz (60x1.5). Last attempt I had to stop at loop 21 or 22, because K6 arrived and I wanted much more that the K5 score to know, if it will run at 7.14MHz FSB. It does not run... :(

 

So I put K5 there back again, clean the CPU legs well, as ludek suggest ... and run SuperPi 32M w/o overclocking. That will be 27-28h I presume...

Edited by trodas
Link to comment
Share on other sites

But do not put contact cleaner inside socket. Tested it and it works baad. Best way is to put contact cleaner directly on CPU pins and RAM pins and then put it on tissue to dry.

 

Putting c-c to socket 370 caused problems with tualatin CPUs :(

Link to comment
Share on other sites

Ok, I cleared thoroughly (even that they looks perfect, gold-plated and very clean - compare with the K6 description bellow) only the pins. Put the CPU back in action and under WinNT it is now calculating the SuperPi 32M. Expect 27-28h to finish, witch will answer the 24h question.

 

But it is true, that I never managed the 32M run to finish on K5 before. Now it is NOT overclocked (50x1.5), voltage is maximum the board can deliver (3.5V) and 16k SuperPi tooks 15sec...

 

BTW, the legs of the K6 is silver only, bendt and looking bad and unclean (some seems to be gold plated sometimes, but now the gold is gone...?) ... yet the CPU works great, stable and fast. Yes, I overvolt it a little (minimum the board support is 1.8V, the K6-III+ 400MHz want 1.6V by default), but I also overclock it a bit, using the x6 multiplier (remaped on x2 settings) and 83MHz, witch gives 500MHz in turn.

 

We see.

 

Up next is to obtain the Socket 3 mainboard - I want the FIC 486-GAC-2, put a Pentium Overdrive in it and test that combo in action. Anyone have FIC 486-GAC-2 mobo to sale/spare...? :D

Link to comment
Share on other sites

I have bad news, guys. I was reluctatnt to admit, that there might be a 24h limit in SuperPi 32M test, however now I have to rethink that. The K5 PR75 run slowly so it took the time:

 

Super_Pi_32_M_24h_failure.jpg

 

As you can see, the 23h 44min 11.743sec is very, VERY close to the 24h mark. Now let's consider this - the machine run 100% stable for first 20 loops. Loop 20/21 is not the problematic "saving" loop, where is serious HDD activity and even most CPU/mem stable machines could fail. However very soon after I saw this:

 

Super_Pi_32_M_failed_24h.png

 

So, unless someone produce 24h+ result with SuperPi mod1.5 XS, then I do believe that there is 24h limit. I tried the SuperPi 32M with my K5 (50x1.5) for 5th time and it always crashed on machine, that run stable highly overclocked K6 (83x6): http://hwbot.org/submission/2996958_

 

SuperPi 32M is therefore 24h limited.

 

...

 

However I have to believe that we can still have some low-end scores, if there is 1) patch that remove the 24h limit, while not inflicting damage to the calculating process 2) if we use the older SuperPi - but w/o the hash it is just a quessing game, if user cheated or not...

 

Also we can try getting SuperPi 32M score "as close to 24h, as possible."

 

...

 

Bugs like that suxx bad :( :( :( This is depresing. We never can have a SuperPi score on slow CPU's... The question is only if the Pentium Overdrive can score under 24h on Socket 3. I believe, that there IS a possibility.

 

...

 

Subsequent findings:

Pentium 1 60MHz have no SuperPi 32M score: http://hwbot.org/hardware/processor/pentium_1__60mhz/

 

Pentium 1 66MHz have no SuperPi 32M score: http://hwbot.org/hardware/processor/pentium_1__66mhz/

 

Pentium 1 75MHz have SuperPi 32M score at lowest at 112.6MHz: http://hwbot.org/benchmark/superpi_-_32m/rankings?hardwareTypeId=processor_897&cores=1#start=0#interval=20 (18h 52min worst time)

 

Pentium 1 90MHz have SuperPi 32M score at lowest at 102.8MHz: http://hwbot.org/benchmark/superpi_-_32m/rankings?hardwareTypeId=processor_898&cores=1#start=0#interval=20 (worst time at 19h 2min)

 

Pentium 1 100MHz have SuperPi 32M score at lowest at 112MHz: http://hwbot.org/benchmark/superpi_-_32m/rankings?hardwareTypeId=processor_899&cores=1#start=0#interval=20 (with 16h 50min time)

 

 

...

 

Conclusion - with some luck and very well optimized run, Pentium Overdrive at 83MHz (at least) can get the SuperPi 32M bench under the critical 24h limit of SuperPi 32M. The score is, IMHO, possible.

 

Who does it, should be, however, granted a pretty big crown for being socket 3 hero andking of kings for olscholl overclocking ;)

Edited by trodas
Link to comment
Share on other sites

ObscureParadox -

there is no submission on the bot that goes over 24H

 

It just cannot be, I believe that I already proved it. 5 crashes at 24h mark is, IMHO, enough.

 

the benchmark has a max run time of 24Hours which is a little strange

 

Not that much, wPrime 1024M run suffer the same problem. When time flip at midnight to the next day, it write just "cheat detected" and stop the calculation. So this is another bench, besides SuperPi 32M, witch is limited to 24h.

I recently encountered that with... you quessed what - AMD K5 PR75. So it is not all that strange for me, it is "normal behaviour" from my point of view. Of course that *SHOULD* get fixed. Who will bug the autor/autors or anyone who can help out?

 

 

lanbonden - more verifications is need, agreed. But I'm quite sure already.

Link to comment
Share on other sites

ObscureParadox -

 

It just cannot be, I believe that I already proved it. 5 crashes at 24h mark is, IMHO, enough.

 

 

 

Not that much, wPrime 1024M run suffer the same problem. When time flip at midnight to the next day, it write just "cheat detected" and stop the calculation. So this is another bench, besides SuperPi 32M, witch is limited to 24h.

I recently encountered that with... you quessed what - AMD K5 PR75. So it is not all that strange for me, it is "normal behaviour" from my point of view. Of course that *SHOULD* get fixed. Who will bug the autor/autors or anyone who can help out?

 

 

lanbonden - more verifications is need, agreed. But I'm quite sure already.

 

The thing i do differnt is runing it on a "modern" rig with winXP, not that it should mater but still nice to see.

 

wonder if we get this as a stage in the country cup, longest superpi32M run :D

Link to comment
Share on other sites

wonder if we get this as a stage in the country cup, longest superpi32M run :D

 

That would basically means to slow it down till almost 24h and then "fire it up" and hope not exceed the 24h mark, but get to the 23:59.999 :D That would be funny compo, tough SuperPi did not like to be slowed down too much (at least for sure it hate disabled caches - it always crash). Possibly another bug?

Link to comment
Share on other sites

when using gear older than socket 5 - then OCers can use "the old" SPI version.

 

Agree, the ms is not necessary for these long runs

 

Why older that Socket 5? On Socket 7 mobo you cannot run the K5 PR-75 SuperPi 32 bench, because it will not made it before 24h...!

Even overclocked to 90MHz (60x1.5) it does not stand a chance. Losing 8-10min per loop to 1h, witch with 24+ loops mean sure crash at the end!

 

 

TerraRaptor -

So is it confirmed you cannot have >24h in pi32m?

 

I would say at 100%.

Link to comment
Share on other sites

probably wont be home when it happens, but first referens round will finish in a bit over 4hours at an estimated runtime of 23h 53mins to see that my slowdown dont affect superpi in any wierd way. Next part after that is to start a new round with the goal to finish just after 24h to see if its possible with a "modern" system with winXP or if its actually superpi thats to blame instead of a bug with winNT and superpi or something.

Link to comment
Share on other sites

There is no SuperPi 32M result over 24h:

http://url.hwbot.org/1R1cwYN

However, there are SuperPi 1M results over 24h:

http://url.hwbot.org/1jQOwNH

 

It is allowed in the case of very old hardware to use the old version of SuperPi (the one that doesn't have ms)

That would have been great for the comp.

So that everyone is on the same page; would it be possible to list what hardware exactly (to the best of your knowledge) would be allowed to run on the older version of SuperPi and can be submitted to Bot.

Link to comment
Share on other sites

lanbonden -

with a "modern" system with winXP or if its actually superpi thats to blame instead of a bug with winNT and superpi or something

 

I get crashes on my K5 PR75 in WinXP too. Just WinNT can be optimized better, so I have higher chance of making the run bellow 24h critical crash-mark...

 

 

 

Christian Ney -

is allowed in the case of very old hardware to use the old version of SuperPi (the one that doesn't have ms)

 

Witch one that is, where to download it and who is responsible for the mod1.5 XS anyway? Maybe a proper way would be to get "mod1.6 XS" that will allow over 24h calculation time...?

Also - what define "very old hardware"? Basically any HW, that cannot finish the run under 24h apply?

 

 

mr.paco -

So that everyone is on the same page; would it be possible to list what hardware exactly (to the best of your knowledge) would be allowed to run on the older version of SuperPi and can be submitted to Bot.

 

Exactly my point. But list will be always incomplete and subject to chages, so I suggest the 24h rule: Under 24h - you have to use mod1.5 XS, over 24h - you have to use older SuperPi.

 

Best solution will be a patch...

http://www.xtremesystems.org/forums/showthread.php?292870-SuperPi-32M-mod-1-5-need-patch-that-make-32M-calculations-over-24h-works!

Link to comment
Share on other sites

Where to get the v1.6...?

 

Okay, googled it there:

http://www.techpowerup.com/wizzard/super_pi_mod.exe

 

From: http://www.systemshock.co.za/forums/topic/30628-superpi-mod-1-6-available-for-download/

 

...

 

The executable is notably smaller (92 528 bytes compared to 104 960 bytes for the "classic" SuperPi mod1.5 XS) and after this predicably end in yet another crash:

 

AMD_K5_PR75_at_90_MHz_Super_Pi_32_M.jpg

 

(Asus TXP4-X, AMD K5 PR75 @ 90Mhz (60x1.5) )

 

...I could give it a try. The results from v1.6 will be accepted as these using v1.5...???

Edited by trodas
added info
Link to comment
Share on other sites

As predicted, SuperPi 32M run fail, when it should show new finished loop with more that 24h:

 

Super_Pi_32_M_failed_24h_again.jpg

 

...but my GF was faster and captured the fail on her cellphone camera, witch the SuperPi window still in background (looks like that there is some autoclose wait in WinNT...):

 

Super_Pi_32_M_moment_of_crash.jpg

 

So I started the SuperPi v1.6 and in just 24h we will know, if it can surpass the 24h limit or not for the 32M test. Yet still I repeat the question: The results from v1.6 will be accepted as these using v1.5...???

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • Create New...