Thuban-ready bios - Uncovering secrets?

[Massman]

I am not sure who have this board, but here is the latest bios:

 

http://sf3d.pp.fi/CrosshairIV-Formula-ASUS-0502.rar

 

Since SF3D posted the latest Asus CIV bios on XtremeSystems, I decided to have a look inside the bios to check whether something interesting can be found in relation to the upcoming release of the Thuban series. In the past, I've done this already and it yielded me some decent results back then.

 

This time, again, some interesting bios strings could be found.

 

 

3 things to note in this screenshot:

 

1) It appears that ASUS has integrated CPU multipliers upto 35X, which is higher than the current physical limitation of 31,5x

2) The DRAM frequency multipliers go up to 2000MHz, whereas the highest memory multiplier on the Deneb processors was 1600MHz ... higher memory frequencies possible?

3) ECC support (not sure whether this was already in Deneb as well, just mentioning)

 

Especially the second point seems to be 'real' (I have no idea since I have neither the mainboard nor the CPU) since it comes back several times in the list of bios strings. Do note that there are several undocumented DRAM multipliers for the Deneb as well, but according to the mainboard manufacturers I've contacted, these were hardlocked by AMD. Given that they hardlocked it in the past, I can't imagine they wouldn't hardlock it again if the multiplier isn't supposed to be used ... so, DDR3-2000 might be a real divider?

 

 

For some reason, ASUS has added bios strings for a processor package containing 8 processor cores. This might just be some leftovers from the server CPU microcode (although that would be strange), so I've got no idea what it means. Maybe I should just be optimistic and hope there will be 8-core AM3 chips in the future (Bulldozer?).

[El Gappo]

Very interesting

________

Ford ranger ev

Edited May 13, 2011 by El Gappo

[redhat_ownage]

i really like where this is going

[Alriin]

Very interesting

 

Great News for AMD-Fans. Nice, Massman.

 

I have CIIF and CIIIF, so i must have CIVF. I await it eagerly. :-)

[Djmaddox]

This is looking real nice.

[Antinomy]

I don't want to spoil the party, but what you are reading is just the BIOS strings - they are standardized and may be not implemented at all.

 

But what I can say is that the board has the unlocking option for all six cores (and locking, as I think).

I can add that the board has Core Performance Boost control - allows to select different modes (4 states)

 

And yep, I can confirm that the DDR3-2000 string is used in the BIOS (can't say whether on all CPUs or only some of them).

 

There's a great bunch of voltages also but I think they are quite common.

[Massman]

I don't want to spoil the party, but what you are reading is just the BIOS strings - they are standardized and may be not implemented at all.

 

Yes and no.

 

First of all, AMD and Intel first sample the bios to Award and AMI to create a basic bios which contains all the code you need to get the system up and running stable. Then mainboard manufacturers receive the code from Award and AMI, but in most cases without any a lot of added information. So, for instance, it's possible that mainboard manufacturers have to figure out a lot of options on their own (eg: B2B Cas Delay and QPI multipliers).

 

Now, a bios can only hold a certain amount of code, so it's important to not waste space on useless stuff. For instance, you could provide code to support temp readings upto 16 cores, but when there's no CPU release that uses 16 cores, it's pretty useless code. The same goes for the memory multipliers. As said, in the AMD white papers you can find registry keys for different DRAM multipliers on the Deneb, but since these are not supported by the CPUs, or just not interesting (who wants 1:1 with DDR3?), so mainboard manufacturers didn't use space to integrate them.

 

Now, remember: the Asus Crosshair IV mainboard is a high-end mainboard with loads and loads of bios settings. This means that it's very likely Asus will have to make a choice what settings to integrate in the bios; in other words remove all the code which is not going to be used anyway. The fact that some bios strings remain in the bios is significant enough to assume(*) that they will be used at some stage. I obviously don't know what's really going on, so I leave the opening post open for speculation, but I don't think this should all be considered as 'standard' options that will not be used. Also, since you're kinda confirming my assumptions.

 

(*): assumption because it's always possible that an unimportant string has not been removed yet or doesn't need to be removed because there's already space enough.

[K404]

DDR3-2000 has been hit with the Crosshair 4 Formula. The divider is real

 

http://www.ocxtreme.org/forumenus/showthread.php?t=5812

[Massman]

DDR3-2000 has been hit with the Crosshair 4 Formula. The divider is real

 

http://www.ocxtreme.org/forumenus/showthread.php?t=5812

 

That's the DDR3-1600 divider Coolaler is using.

 

(250MHz * 4) / 1 = 1000MHz

[El Gappo]

Would be very hard/impossible to boot into windows with any current chip at 2000mhz ram.

________

Marijuana Vaporizers

Edited May 13, 2011 by El Gappo

[Massman]

Just went through a couple of MSI bioses. No trace of the new DDR3-2000 divider; but several options that indicate 8-core CPUs.

[Antinomy]

Yes, I know that, Massman and what I wanted to say is just that this method can be inaccurate.

who wants 1:1 with DDR3?

Sempron does

 

Checked the Gigabyte ga-890gpa-ud3h - has only DDR3-1600 (the advertised 1866+ is via overclocking). Nothing interesting though - it's a board with integrated video.

Then the 890XAUD3 - 790X+SB850. This one has DDR3-1600 too. Though it has core managing for six cores (deactivating, can't say whether this option can activate cores). The rest is not interesting.

Edited April 5, 2010 by Antinomy

[flanker]

2000MHz dual is possible now, right (but hard work think),Hope IMC Thubans will better than Phenoms

[Massman]

Sempron does

 

Actually ... you are right! The moment I was writing the message, I already thought

 

Checked the Gigabyte ga-890gpa-ud3h - has only DDR3-1600 (the advertised 1866+ is via overclocking). Nothing interesting though - it's a board with integrated video.

 

Then the 890XAUD3 - 790X+SB850. This one has DDR3-1600 too. Though it has core managing for six cores (deactivating, can't say whether this option can activate cores). The rest is not interesting.

 

The management for six cores has been in the bioses for quite some time (check link in opening post) and can also be found in the MSI 890GXM bios. Also found references for 8-core cpus ... someone an idea?

 

As for the advertised DRAM frequencies ... well ... hehe

[Antinomy]

Yes, I've seen the info on Asus - I just wanted to check another headline manufacturer.

The idea on eight cores is unification - for the quad core we had to make a two bit machineID (haven't read the BIOS developers guide for ages so I quote as I remember it). When we go to six cores, we have to change the routine. So if we implement a three-bit, why not make an eight core? Anyway it will come out, so no double job. That's my idea.

 

What can be said - this is obviously nothing to deal with server CPUs - already 12 cores there. Think it's for future desktop, lower tech. process for sure.

[Antinomy]

I've checked in Biostar TA890GXE BIOS - http://www.biostar.com.tw/app/en/mb/bios.php?S_ID=474

It shows the ability to

enable or disable Core Performance Boost

Disable CPU cores

Activate disabled cores (BIO-unlocKING) - which for now, seems, only is able to unlock third and forth cores only - not for Thuban.

The max multiplier in Biostar is 31.5x. More than enough, though

 

 

Massman, as for the ECC - AMD had this one for ages - since the presentation of AMD Hammer family it's just hidden in BIOS until you plug in a ECC memory stick

I wanted to clear this as the are newsmakers beginning to reprint your words

Edited April 5, 2010 by Antinomy

[Massman]

The idea on eight cores is unification - for the quad core we had to make a two bit machineID (haven't read the BIOS developers guide for ages so I quote as I remember it). When we go to six cores, we have to change the routine. So if we implement a three-bit, why not make an eight core? Anyway it will come out, so no double job. That's my idea.

 

What can be said - this is obviously nothing to deal with server CPUs - already 12 cores there. Think it's for future desktop, lower tech. process for sure.

 

Aha, good info there!

 

Unification does make sense, I guess. There are no real features build around 8 cores, just the very base code is present.

 

I've checked in Biostar TA890GXE BIOS - http://www.biostar.com.tw/app/en/mb/bios.php?S_ID=474

It shows the ability to

enable or disable Core Performance Boost

Disable CPU cores

Activate disabled cores (BIO-unlocKING) - which for now, seems, only is able to unlock third and forth cores only - not for Thuban.

The max multiplier in Biostar is 31.5x. More than enough, though

 

I guess there's no Thuban microcode inserted into that bios yet. 31.5x is, by the way, the maximum useable multiplier for Phenom-II. I've seen some manufacturers add higher multipliers in the bios, but that's just wasting space.

 

Maybe they work for Thuban - no idea.

 

Massman, as for the ECC - AMD had this one for ages - since the presentation of AMD Hammer family it's just hidden in BIOS until you plug in a ECC memory stick

I wanted to clear this as the are newsmakers beginning to reprint your words

 

Aha! Good to know as well. Haven't played with ECC memory since ... Skulltrail, so no idea

[flanker]

I think, maybe is some code for Bulldozer. I believe, samples Bulldozers are ready maybe 1-2 months.

[Antinomy]

I guess there's no Thuban microcode inserted into that bios yet. 31.5x is, by the way, the maximum useable multiplier for Phenom-II. I've seen some manufacturers add higher multipliers in the bios, but that's just wasting space.

You were a bit inattentive

enable or disable Core Performance Boost

this means Thuban support, I'm sure you know, just have missed that line.

[flanker]

today i was looking second at this image and...watch at error reports...Its look like something new architecture-think, its Bulldozer

[chew*]

2000 mhz ram fails to impress me at 9-9-9........seeing as 6-7-6 was doable at 970.... I doubt there will be a benefit to running 9-9-9.

[NaplesDragoon]

Massman what software are you using to read out the information. I am not sure to upgade the BIOS of my motherboard since i am not sure if the new bios includes the thuabn support which was given in the beta BIOS. i am using an asrock a770 Crossfire running a 1055T with the L1.31 BIOS. the new bios is 1.40 but the changelog doesnt state a new cpu code update.

[7ate9]

keep channel interleaving turned off.

And the memory must be set to run in Ganged mode

add ECC to non-ECC memory

 

select ECC configuration

ECC to allow custom profiles or user set

DRAM SCRUB REDIRECT should be disabled and 4-bit ECC mode enabled, otherwise known as killer ECC mode.

These settings have ensured the memory to become stable.

 

DRAM BG scrub set to 1.31MS

 

Data cache BG scrub set to 5.12US

 

L2 cache BG scrub set to 5.12US

 

L3 cache BG scrub set to 5.12US

 

You can play around with it to see what get you stable with ram at 2000mhz

[whiteoverclock]

news ????