New Z - serious problem with additional SSD/HDD

OK, I found out my cause for shortened battery life

And, as I feared, it is connected to the Lenovo Ultrabay adapter

After careful analysis, I found out the reason - it looks like when the additional drive is connected via DVD SATA port, BIOS does not expose P-states correctly.

What happens then is Windows (or Linux for that matter) cannot scale frequencies down - your CPU will constantly run at the highest frequency!!

As soon as I disconnect the additional drive, and run the Windows DDK PwrTest /info pm - I get normal values.

If the additional drive is connected, Windows sees only Throttle states!!!



This looks like some strange BIOS bug - I cannot possibly see connection between additional drive and CPU, except that something gets corrupted in the ACPI tables.

I am still investigating this, but as it seems right now we will have to choose between long battery life and another HDD/SSD drive.

You can also verify this even without PwrTest tool - just go to Task Manager and then to Resource Monitor, and then to "CPU" tab - if it says "100% maximum frequency" all the time, this means the P-state transitions do not work.

@EDIT - as it turns out, this problem affects people with QUAD SSD setups only (5 drives total when extra SSD/HDD is added)

 

Tested with:

* Intel X25-E
* Intel X25-M
* Old Samsung SLC (MCCOExxxx)

All in Lenovo UltraBay slimline SATA to HDD adapter.

I am still investigating this..

@EDIT1 - UltraBay is not responsible for this - same error happens even if I connect the SSD directly to the slimline SATA and external +5V source...

 

Noooooooooooooooooooooooooooo I just ordered a ultrabay adapter this week! D:

 

Switching the controller to AHCI does not help either

Now this must be one of the stupidest and most frustrating bugs I've seen...

This is just... ridiculous, there is no other name for it.

 

Confirmed on mine - 100% Max Frequency.

 

Yep Looks like we are f***ed....

I think I know what is going on - most likely Intel RST firmware is stomping over _PSD and _PSS ACPI tables and basically messing this up.

Unfortunately, I don't know how to fix at this point...

I remember that if you use acpi.sys from debug version of Windows (checked build) it is possible to load custom ACPI table from the registry. So, in theory if we capture the correct one we'd be able to avoid this issue... maybe.

 

OK, it is DEFINITELY related to some corruption

I just removed built-in SSD (disconnected it) and left second SSD on.

There is no problem..

So, it just looks that having 5 SSDs in the table is stomping over ACPI P-state data...

 

Are you certain it's the rst firmware and not the intel rst or ahci drivers? One test would be to force ahci mode (which you've done) but then force windows to use the standard msahci.sys driver as opposed to the iastor.sys driver. thoughts?

 

Nope

I am using Windows install disc to test, which boots into GUI without any 3rd party drivers... same goes with Linux.

I just did another test - I diconnected the second SSD, booted the Z and then connected it again before Windows Setup loaded..

SSD is visible and performance states work... damnit... It is definitely BIOS that does the corruption

 

That is just plain wild.

 

I wonder if the people who ordered hdd/no-dvd units from sony experience the same issue? Does a spinning drive cause the same behavior? Perhaps those with a factory hdd/no-dvd option have a slightly different bios?

 

To add more irony... CPUGenie, which is not even designed for Nehalem/Westmere can see all P-states in "Power Monitor -> ACPI P state info"...

But it cannot be used to control the CPU as the interfaces are changed completely for Nehalem/Westmere...

Linux also sees P-states, but cannot switch - I guess ACPI information is partially missing.

 

I don't think so. The hdd (in the hdd/non-dvd version) is connected to the same connector as the SSD RAID (in the ssd/dvd version).

 

There is no problem if you disconnect the factory SSD - your 2.5" drive will work, and P-states will work, too.

The problem is - if you have 5 drives (quad SSD + another HDD/SSD), it appears that the ACPI processor data is not correct anymore.

I can totally picture Sony engineers not even testing this case, as this kind of configuration (5 SSDs or 4 SSDs + HDD) is not even offered.

Which also means it will be next to impossible to get Sony to fix this issue - if you report it, they will just say "you're using unsupported configuration"

Damnit... I am really starting to value the Open Source now... if the firmware was Open Source I'd be fixing this in notime

So, basically we have two choices now:

- Add another 2.5" HDD or SSD and live with shorter battery life and higher temps

Or

- Be happy with only factory quad-SSD, and forget the upgrade...


I'm going to try some more things today... but I guess it will be very hard to fix this unless BIOS code is altered not to recognize this SATA port and leave it up to Windows to deal with it.

 

I see. Looks like it's either SSD RAID OR HDD/SDD.

 

Hmm

But the really weird thing is that everything seems to work with Sony's DVD drive. I don't have it here (I left it @work) but it could be that it does not have a serial number? Maybe it's string ID data is just long enough not to trigger the bug...

 

Serial number? You mean the drive's (dvd/ssd/hdd) serial no.?

 

Yep? Can you check in your RAID setup (ctrl + l during boot?)

I am trying to understand what the hell is the difference between a SSD and Sony DVD which makes DVD work and SSD screwing up the processor data

 

I don't have the DVD installed. Have to open the notebook and put it back in...

Anyone who has the DVD drive can help us?

 

Not sure if this helps:

I look at my DVD drive and it has a serial number of 11 digits long. My SSD's 20 digits long, same as the Sony SSD RAID

 

Here is some info from Microsoft:

So, in theory we could grab the ACPI tables from properly working Z and use them in Z with corrupted BIOS data...

 

Sounds positive. Where can the tables be found?

 

Well, they could be dumped with some tool I guess but I'd rather try some other things first... debug acpi.sys is probably slower and dumping a lot of messages, not really the most elagant solution.

I noticed that one of the hidden options in BIOS is ACPI mode selection - and it offers:

- ACPI 1.0b
- ACPI 3.0
- ACPI 4.0

I will try changing this and see if it helps. But, if the SSD data is actually trashing buffers not meant for it, it might trash something else

 

How are you testing this?

I have a Z with stock + add on SSD and am interested in seeing if I have the problem as well. I'm using the Z with 256GB via 128x2, so I only have 3 drives in it, total.

Am at work with the Z but no bootable DVD/CDs.

 

You can easily see the problem if you run Resource Monitor (click on "start" and type Resource Monitor) and go to CPU page.

If the CPU is always on 100% maximum frequency, this means the power management is not working.

 

Did you run Intel rapid storage config (unhide RAID config in BIOS and then enter RAID config during boot) to check that 5th drive is a stand-alone drive?

It is possible that Intel RST BIOS that loads after main BIOS overlaps main BIOS with its data, hence the problem. I experienced similar things on desktop where I would plug a RAID card (with its own BIOS), enable in main BIOS loading extra RAID BIOS and get overlap that start screwing main BIOS in various ways.

So, it may as well be Intel problem, not Sony.

 

It's clearly not on max freq.

Mostly hovers around 53%.

My SSD is a Corsair P128 I believe. This reinforces that it's probably a 5 drive issue.

 

Indeed. I presume yours is the HDD turned SSD version?

 

He has a dual setup (2 x 128)

Yep... so we with 4 x 64 are screwed... properly

It could well be Intel's option ROM that is causing this... As soon as I back things up I'm gonna start fiddling with RAID BIOS and see if anything helps.

Btw, turning on ACPI 1.0b or ACPI 4.0 did not fix anything...

 

Bloody hell - setting controller to IDE mode makes P-states work again

You just have to hate this bug...

 

OK, I give up... for now

I tried absolutely everything I could think of... I guess we'll have to wait for Sony to issue a bios update for the new Z, and just hope they'll fix this idiocy...

So, for now, here is what I've learned:

- This bug affects only "quad SSD" models, not dual SSD
- The bug is triggered by putting additional (fifth) SSD (or, I guess, HDD)
- The bug manifests as CPU not being able to switch to lower than highest frequencies

- If you switch the SATA controller to "IDE", bug disappears

So, most likely this bug is related to AHCI/RAID option ROM, which probably trashes vital data related to the CPU performance states.

 

Mine is the 116 from Amazon.com - it's a standard US model. 128x2 with DVD. I did the Ultrabay 3 thing (further proof that the Ultrabay adapter isn't the issue).

 

I just finished installing Windows again on the old SSD (factory one)

Power drain (reported by Everest):

* With bug: 17-19 W
* Without bug: 8-12 W

As it could be seen - there is a huge difference caused by CPU not entering idle.

I guess people with 4xSSD configuration will have to choose between good battery life and additional SSDs.

 

I have only 128GB (2xSSD). I just got my Crucial C300 256GB and Lenovo Adapter. I plan to install the Crucial SSD to the CD Bay. Do you think that I will have the same issue as you?

 

You should experience what I did, which is to say, "no".

From what we can tell, you need 4 drives in the Sony array to see the problem.

 

So ok - 100% freqs are normal when plugged in. I'm just going to assume the outcry is because it STAYS at 100% CPU freq despite being on battery.

Can I ask whether people here are using the 5th SSD as is, or RAIDed with some other drive(s) from the quad setup?

I'm not the most experienced with RAID setups, but if the AHCI / RAID ROM seems buggy, why not just get rid of the RAID setup, and just use as JBOD (though most people here might not like that option due to performance issues that brings about). Like, each SSD has it's own letter allocation in an OS and in BIOS. That seems plausible considering it isn't the adapter causing the extra drainage.

 

It seems the issue is related to the number of drives. If they are above 4 the problem occurs.

Mine has the quad SSD (in RAID) and an OCZ drive (in JBOD), I can see the issue as described by psyq321.

 

Works fine in triple-SSD config here. ~53% max frequency even when plugged in.

 

Well looks like I am going to ditch the OCZ drive and use the quad SSD instead.

 

Well... yes and no. True, you wouldn't mind too much CPU being stuck @2.66 GHz (or whatever) while on AC if you consider performance only...

BUT there is another issue here, and that's thermal management... CPU stuck @highest freq can't go to idle and can't utilize low-power states. This means the temperature will be considerably higher than normal.

Note that measured power draw is ~10W higher with the bug. Those 10W are certainly going to cause some extra unwanted heating and fan noise

I was using 5th SSD drive separate from the RAID setup.

But unfortunately your idea wouldn't work anyway >unless< you are setting the SATA controller to "IDE" mode.... which kills performance (and performance was the whole point of the 5th SSD idea)

Problem lies in the fact that:

- IF you have Quad SSD + extra SSD
and
- You use RAID or AHCI

Problem happens. And, even if you don't wish RAID (you can do software stripping in Windows Server 2008 R2, for example) you still want to keep AHCI for the good performance.

It looks like same Intel "option ROM" is responsible for both RAID and AHCI (which is to be expected, actually) and if any of those two modes are selected, you'll be getting the CPU bug if you have all 5 available SATA devices filled in.

 

So the best solution for people who are interested in replacing the optical with SSD is to go with 128GB, 192GB, or 384GB, and 256GB with absolute certainty that it's 128x2 not 64x4, and avoid 512GB altogether?

EDIT: change HDD to SSD, I am assuming HDD is fine?

 

I didn't test any HDD, but I don't believe things would be different in 4xSSD+HDD combination, it will probably trigger the same problem.

If you plan to expand with extra drive, then yes - it definitely makes more sense to go with dual SSD as quad SSD will probably trigger CPU power-bug when another drive is added.

If anyone knows a way how to help escalate this issue to the Z product design team, please let me know. I am sure the fix would be quite trivial, but it requires changes in the EFI/BIOS - and thus it requires Sony engineers to do it.

 

Guys is the advanced bios configs (advanced menus enabled) there are individual power management settings for each of the sata ports once the mode is changed to ahci

in there I noticed that the last sata port has a different power setting by default than the others. Perhaps that needs changed?

 

OK I just checked and in the the IDE configuration page once AHCI is selected you will see a "hot plug" and "spin up" setting for each port. The default for port 5 is different than the other 4 ports. Perhaps change it to match?

 

Wouldn't triple-SSD make more sense to step down to than dual? I sense some discrimination since it's not mentioned much.

 

Unfortunately no, it does not change anything... Port 5 is actually not even used - only ports 0 (SSD), 1 (DVD or your own SSD/HDD), 2 (SSD), 3 (SSD) and 4 (SSD)

AHCI Option ROM is needed to see ports #4 and #5 - without AHCI Option ROM only ports #0, #1, #2 and #3 would be seen and this would not work for Sony, as they need (up to) 4 ports for SSDs and one more for DVD.

I guess if AHCI option ROM would be somehow disabled (I cannot do it via BIOS settings) maybe the bug would not appear - but then, even if we could do it, we wouldn't be able to "see" the last of quad SSDs.

Really frustrating bug..

 

ok just thinking out loud here.

The sony libraries/utilities control power on the sata port normally occupied by the DVD-ROM. We know this because the sony sw can turn the DVD on/off depending on your power profile.

Is it possible that the Sony power software is causing the issues? Have you tested on a clean install and not installing any of the sony utilities/drivers?

 

I am always using Windows setup DVD to do tests - so no Sony drivers at all.

Also, Ubuntu 10.04 Live DVD also reports that CPU has no power management in this case.

So, no, unfortunately the problem is BIOS-bound.

This can be proved even further if you connect the SSD to external power via longer cable, and power it on >after< BIOS POST. Everything works then flawlessly - it is just the stupid AHCI ROM that is doing this

 

I think I got the idea how to fix this issue maybe...

It is possible to force controller to other mode after boot - this is what was done for early Apple hardware that was forcing controller to IDE mode in legacy BIOS.

Basically there was a GRUB patch that changed controller mode back to AHCI.

Now, this would work if and only if switching controller back to AHCI or RAID mode does not break the capability agian... and this has to be tested.

I don't have time until the next weekend unfortunately but I might write a small piece of code to test this.

 

A possible workaround is to disable one half of the quad ssd by unplugging one of the connectors to the ssd. This should work but you lose half of the ssd capacity.

One extreme measure is to block one of the ssd drive (out of the 4) by cutting part of the data section of the connector. This would give you 3/4 of the capacity.