The Book of Bulldozer - Revelations: Episode 2 (SuperPI / x87)

[zvonexp]

Works perfectly

Without


With

[[XC] Oj101]

I have the new version mirrored here, it's exactly the same file as The Stilt uploaded and the MD5 hash can be verified at OnlineMD5. No pop-ups, no adverts, no waiting

[EnJoY]

This is really great Stilt, amazing work! One thing I wanted to point out. I have an FX 8350 in an ASRock 990FX Extreme9 running BIOS 1.01 (which was pulled months back). When I opened your tool, it gave me the option to apply the ERRATA fix and said I needed to update my BIOS. Is it really possible that ASRock's BIOS doesn't contain this ERRATA fix? I know the BIOS sucks and is buggy, and they have failed to release even a single update for it, but this would really put me over the edge if it's true.

[The Stilt]

This is really great Stilt, amazing work! One thing I wanted to point out. I have an FX 8350 in an ASRock 990FX Extreme9 running BIOS 1.01 (which was pulled months back). When I opened your tool, it gave me the option to apply the ERRATA fix and said I needed to update my BIOS. Is it really possible that ASRock's BIOS doesn't contain this ERRATA fix? I know the BIOS sucks and is buggy, and they have failed to release even a single update for it, but this would really put me over the edge if it's true.

Checked both of the bioses (1.00 & 1.10) for ASRock 990FX Extreme9.
Both of the bioses are based on an old AGESA version (V1.2.9.0), while the official launch AGESA version for Vishera was V1.5.0.0.
Due the old AGESA version the microcodes are outdated too:

Zambezi (OR-B2) - Installed: 06000626 (01/04/2012), The most recent version: 0600063D (03/06/2013).
Vishera (OR-C0) - Installed: 06000817 (05/24/2012), The most recent version: 06000822 (01/03/2013).

Version 0600063D (OR-B2) contains 10 errata fixes, while the 06000822 (OR-C0) fixes 8.

[FlanK3r]

in 32M its awesome boost, I tried with decent summer stable OC profile...




more than4m 30s better result!!!! Woow. I think, 32M with LN2 could be quickly fun now with Vishera, maybe around 10minutes only.

[Oese]

Many many thanks, this is much appreciated and something I didn't even dream about!

This goes into AMD Hall of Fame I wonder what will be the response from both the community and AMD engineers

[zalbard]

I wonder what will be the response from both the community and AMD engineers

They better be ashamed...

[AzureSky]

I wonder if it would be possible to mod these kinda fixes into bios like they do with SLIC tables and SLI certs (if you dont know what those are, head over to mydigitallife.info and read the bios forums)

[EnJoY]

Checked both of the bioses (1.00 & 1.10) for ASRock 990FX Extreme9.
Both of the bioses are based on an old AGESA version (V1.2.9.0), while the official launch AGESA version for Vishera was V1.5.0.0.
Due the old AGESA version the microcodes are outdated too:

Zambezi (OR-B2) - Installed: 06000626 (01/04/2012), The most recent version: 0600063D (03/06/2013).
Vishera (OR-C0) - Installed: 06000817 (05/24/2012), The most recent version: 06000822 (01/03/2013).

Version 0600063D (OR-B2) contains 10 errata fixes, while the 06000822 (OR-C0) fixes 8.

Wow, ASRock is terrible. Is this at all normal for a 990FX board to be using such an outdated AGESA? Thanks for your time.

[AzureSky]

Wow, ASRock is terrible. Is this at all normal for a 990FX board to be using such an outdated AGESA? Thanks for your time.

kinda wondering if the asrock 990fx extreme4 has the same outdated AGESA....used a few for friends builds......they clock damn well.....

if anybody knows please tell me.

[The Stilt]

Wow, ASRock is terrible. Is this at all normal for a 990FX board to be using such an outdated AGESA? Thanks for your time.

kinda wondering if the asrock 990fx extreme4 has the same outdated AGESA....used a few for friends builds......they clock damn well.....

if anybody knows please tell me.

This is for 970-990 chipset series motherboards only (althought some of the ASUS lower end models managed to slip in too).



Zambezi microcodes marked with red > 06000629 still have the Valve CEG fault present.

[Astennu]

Dammm That looks awe full. So much outdated or even worse :'(
I wonder if the Intel platforms have similar problems.

[EniGmA1987]

Strange that higher revision Gigabyte UD3 boards would revert back to older AGESA code than the older revision board.

[Oese]

Mhm i get an error "cannot initialize driver" on Sabertooth 990 FX R1.0 1604 Bios.

Anyone has a clue?

[The Stilt]

Mhm i get an error "cannot initialize driver" on Sabertooth 990 FX R1.0 1604 Bios.

Anyone has a clue?

Run as Admin.

[The Stilt]

Strange that higher revision Gigabyte UD3 boards would revert back to older AGESA code than the older revision board.

Check the compile date of the bioses.

[Oese]

Run as Admin.

Ouch! Works perfectly.... My Bad

-16 % in SPI 1M

[EnJoY]

This is for 970-990 chipset series motherboards only (althought some of the ASUS lower end models managed to slip in too).

Zambezi microcodes marked with red > 06000629 still have the Valve CEG fault present.

Thanks again for this. Looks like I shouldn't have gambled with ASRock.

[ga1ve1an]

Threw mine in here also. I ran each run without patch (the top Super PI) and with the patch (the bottom one). Very cool update. THis was with my 24/7 5ghz vishera. My chip uses only 1.475v to get the 5ghz and it shaved off 4 minutes from the 32m, so I am gonna play with the higher clocks it can do. Will throw some updates in here shortly. Good job STILT.



Here is the updated with Vishera at 5200mhz. Broke 15mins on 32m finally. LOL. If this improves performance on other programs that might have some similar code to super PI that would be nice. I read some said PhysX.

[Ace123]

are we not able to manually update agesa code in bios as we used to in the past?

[The Stilt]

are we not able to manually update agesa code in bios as we used to in the past?

As far as I know, nobody has ever done it successfully.
You would need to have access to the source code of the bios (vendor / model specific) project.

[Daveburt714]

Thanks again for this. Looks like I shouldn't have gambled with ASRock.

I know this is OT, but I've been playing with ASRock boards myself lately...
The specs and hardware look good, but the bios support has been questionable to say the least!
I bought a 990FX EX9 and was pretty disappointed with it. Bios support is seriously lacking, and the fact that they haven't even bothered to update the AGESA code is pititful.
It's pretty sad when you buy a companies top of the line board and there hasn't been a decent (lasting) bios update in six months...
You can get decent clocks from the EX9, but it takes a lot of effort, and winds up being a random luck of the draw setting that makes no real sense to a decent overclocker. ...
That being said, the bios does fine at close to stock. Guess I need to put together a rig for Craigslist and get rid of this board.

For whatever reason (MicoCenter Combo deal?), I went with ASRock once again and grabbed the 6600K/ASRock EX6 (FM2) board.
I have to admit that it's heads above the EX9, and at least there have been a couple bios updates that have stuck and were actually usefull...

Maybe ASRock is short on bios talent and they're directing it all towards FM2. Reguardless, I'd recommend the EX6, but steer folks away from the EX9!

[zir_blazer]

I also agree that on current Motherboards, the most limiting factor are BIOSes that are missing features or have them badly implemented. You have examples like this, and this, of what issues are present when BIOS aren't fully compliant with specifications. Manufacturers usually don't go back to fix BIOS-related issues, even if from the Hardware side the Motherboard is perfect. This sometimes leaves you with non functional features (AMD-Vi for IOMMU virtualization comes to mind) because the BIOS developers seems to be ridiculous lazy. This include even if AMD wants to patch this via microcode, even if earlier AM3+ Motherboards could work with it, I doubt many will receive BIOS upgrades to do so.
How I would love that Coreboot could become more accessible to get rid of that propietary code crap...


Now, for some offtopic...

@ The Stilt
How in-depth is possible to analyze a BIOS BIN/ROM file with automatic tools and none to little knowledge? I suppose that at least someone must be able to easily retrieve AGESA version from a BIOS file to make the table you did there in a timely manner. Doubt someone could physically test THAT many Motherboards just for that purpose.
I'm looking for creative ways to be able to check if a Motherboard has proper AMD-Vi support or not by checking if it has a ACPI IVRS table (Or ACPI DMAR for Intel VT-d, for that matter), however, I have no idea of how exactly it should be inside a BIOS file to know what I should be looking for (Besides that I should also uncompress the BIOS modules first, I suppose). The idea is that anyone interesed in this feature for virtualization, would be able to discard Motherboards that don't have it functional before purchase by checking the BIOS for that table and at least some insight if the values retrieved seems to be valid.
Yep, absolutely non-related to x87, but its worth a shot.

[SEA]

I wonder if there are any other than SuperPI applications that got improvement?

Another question, does the older BIOS AGESA microcode impact performance, or it is x87 only?

[The Stilt]

A small update: Bulldozer Conditioner R1.02B

Original package checksum (MD5): BCC929498EF1B01B7F99B8F2DA805F46

Bulldozer Conditioner R1.02B

Changes:

- Enhanced the NRAC fix
- Added a UAC prompt (admin rights) for Windows Vista / 7 & 8.
- Updated the AGESA version info

The enhanced fix included in BDC R1.02B is around 10 seconds faster in SuperPI 32M than the one in R1.00 / R1.01 versions