Is possible to re-enable disabled L1/L2 caches in bios?

[trodas]

Guys, I wonder, if is possible to re-enable L1 and/or L2 caches that are disabled in bios under Windows XP. Specificaly for AMD Athlon / Duron CPU's.

 

These CPU's run deadly slow w/o them (imagine GPU-Z starting 15min - you get the idea what "deadly slow" means) - specifically w/o L1 - and therefore for taking screenshots of the test result I look for means to do it with more "normal" speed and not speed, where just few fast random clicks can cause BSOD, as the input buffer is overflown so easily...

(downclocking by nVidia control pannel is the biggest problem)

 

DOS utility that can do it is found there:

http://www.vogons.org/viewtopic.php?f=24&t=38613

Edited August 16, 2015 by trodas

[TerraRaptor]

No, you won't be able to enable them in windows as bios first need to initialise them

[trodas]

But in DOS it works, according to the claims on link that I just added to the post... And on Amiga I do it all the time. Sometimes even halving datacache on 68060 make demo works (previous 68040 have exactly half the datacache, so if demo fail on 060, then that was the next logical thing to do).

 

BTW, there is descripted the asm code to do that in Intel CPU's:

http://stackoverflow.com/questions/1108485/disable-l2-l1-caches

 

Intel Architecture manual A.3 indicates that cr0 register can be set to disable cache by setting bit 30, i wrote the above code then :

 

invd

mov eax,cr0

mov eax,40000000H ;set bit 30

mov cr0,eax

Edited August 16, 2015 by trodas

[Mr.Scott]

Amiga is not Athlon/Duron.

Disabled caches are a hardware limitation, although there are hard mods that will sometimes enable disabled caches.

Do some freakin research before you open your trap.

[trodas]

Amiga is not Athlon/Duron.

 

No-one claims that, I just used that as explaination, why I believe that enabling and disabling caches can be done during the run of the CPU. That's all.

 

Disabled caches are a hardware limitation, although there are hard mods that will sometimes enable disabled caches.

 

Are you sure? Because if the caches can be disabled and re-enabled by software, then it does not looks like hardware limitation at all. Also all older mainboard can on/off the caches, witch (especially L1 is a killer) greatly affect the speed.

Another thing is, that the program need to run in protected mode, so at kernel level...

 

Do some freakin research before you open your trap.

 

That was uncalled for (help if you can, don't reply or bother if you cannot help), and the irony is, that it come from someone, who claim that caches are HW limitation...

 

But it is time to show you more of what I already gather:

 

Included file for L1 cache control in DOS:

http://cyberia.dnsalias.com/Cyb.11.Htm

 

cacheon.exe : enables L1 cache from DOS or Autoexec.bat.

cacheonw.exe : enables L1 cache in writeback mode if available.

cacheoff.exe : disables L1 cache from DOS or Autoexec.bat.

 

Note: These cache control programs wont run in protected mode. You need to either exit to DOS or run them in Autoexec.bat if using win32 (win95/98 etc).

 

...

 

Similar things are done since i486...

http://www.rigacci.org/docs/biblio/online/firmware/cacheoff.htm

...and the last post there contain concept, how it can be done on Win95/98 system:

 

http://batboard.batlabs.com/viewtopic.php?t=89129

 

You can run DEBUG from the command prompt or in "Run" in Windows. The files get written to whatever your default "My Documents" folder location is. After that, you can copy them over to your RSS machine via floppy or whatever.

 

For protected mode kernels systems (anything based on NT) it is not usable, I think.

 

...

 

 

So, did not that just shut your trap? :celebration:

[Mr.Scott]

http://fab51.com/cpu/palomino/athlon-e5.html

This might be a good start.

[basco]

http://www.ocinside.de/workshop/amd_socketa_overclock_d/

[Mr.Scott]

http://www.ocinside.de/workshop/amd_socketa_overclock_d/

Yup. Another one of my favorite sites. Thanks.