Audio Decoding KILLS MacPro (2009) green factor

[VirtualRain]

Sorry for my last post - I tried to insert a picture and the linking bombed...it would show how both data paths get to the CPU in different ways... one directly linked to the CPU and the other running through the ICH10 interface.

So Bill, is it your theory, that using a PCIe USB card or FW card might enable the use of USB/FW audio interfaces without creating the problem?

What slot are you running your card in? My understanding is that the top two (x4) slots are routed through the ICH while the bottom two (x16) slots go directly to the CPU. It would be interested to know if the problem can be avoided using one of the top two slots or only one of the bottom two slots... or either.

Is anyone aware of any USB PCIe cards that would work in a Mac? I guess I'm off to search...
EDIT: Looks like this is worth a try... http://www.newegg.com/Product/Produ...Froogle-_-Add-On+Cards-_-SIIG++Inc-_-15150152

 

[BillB50]

So Bill, is it your theory, that using a PCIe USB card or FW card might enable the use of USB/FW audio interfaces without creating the problem?

What slot are you running your card in? My understanding is that the top two (x4) slots are routed through the ICH while the bottom two (x16) slots go directly to the CPU. It would be interested to know if the problem can be avoided using one of the top two slots or only one of the bottom two slots... or either.

Is anyone aware of any USB PCIe cards that would work in a Mac? I guess I'm off to search...
EDIT: Looks like this is worth a try... http://www.newegg.com/Product/Produ...Froogle-_-Add-On+Cards-_-SIIG++Inc-_-15150152

If your information is correct, then my theory is rubbish - I put the sound card in the third slot counting from the bottom. Again, I looked at a schematic for a regular X58 board which showed all PCIe slots going directly to the CPU. This same schematic showed everything else going through the ICH - that includes on-board audio, USB, and Firewire.

Does anyone know how we can get a schematic for a Mac board so we can confirm?

Also, I exchanged emails with a prominent tech blogger last night who has confirmed the issue and has also spoken to Intel, with no results. This person also believes Apple is aware of the problem but doesn't believe they will act, since it is not "too bad" (Apple's opinion, not the bloggers) ...

 

[smacman]

Does anyone know how we can get a schematic for a Mac board so we can confirm?

Ask and you shall receive...

 

[BillB50]

Ask and you shall receive...

So do I interpret this correctly - I see the 4 PCIe slots connecting to the North Bridge with the South Bridge being the ICH10 hub, although it is not labeled as such in the schematic?

If that were the case, installing a USB or Firewire card into these slots would eliminate our issues as well.

Can someone confirm? Thanks Smacman for the quick response.

 

[Inconsequential]

Ask and you shall receive...

Looking at that, anything on the ich10 bus in theory could set this bug off?

I've also concluded yesterday that USB does it too, but seeing as it doesn't saturate the bus as much as FW800 does it doesn't cause such a big issue

 

[DoFoT9]

Looking at that, anything on the ich10 bus in theory could set this bug off?

I've also concluded yesterday that USB does it too, but seeing as it doesn't saturate the bus as much as FW800 does it doesn't cause such a big issue

does it really? gives a temp increase? get 5 external USB HDDs and copy stuff and see how that goes temp wise

 

[nanofrog]

So Bill, is it your theory, that using a PCIe USB card or FW card might enable the use of USB/FW audio interfaces without creating the problem?

What slot are you running your card in? My understanding is that the top two (x4) slots are routed through the ICH while the bottom two (x16) slots go directly to the CPU. It would be interested to know if the problem can be avoided using one of the top two slots or only one of the bottom two slots... or either.

Is anyone aware of any USB PCIe cards that would work in a Mac? I guess I'm off to search...
EDIT: Looks like this is worth a try... http://www.newegg.com/Product/Produ...Froogle-_-Add-On+Cards-_-SIIG++Inc-_-15150152

In simple terms:

X58/5520 chipsets run all of the PCIe slots.

The ICH10R:

• Audio

• FW

• USB

• SATA ports (in a 2 + 4 config)

• NIC's

• BlueTooth

• PWM controller

So do I interpret this correctly - I see the 4 PCIe slots connecting to the North Bridge with the South Bridge being the ICH10 hub, although it is not labeled as such in the schematic?

Yes.

If that were the case, installing a USB or Firewire card into these slots would eliminate our issues as well.

It should, but we still don't know which device drivers are borked.

Seriously, as the ICH10R is an evolution of the ICH9 family, and that works fine (i.e. '08 systems don't have the issues). Not in OS X or Windows (which uses the same Intel drivers for all the ICH families BTW - IMSM ver 8.9.1023).

Where we may be having the problem, is in the DMI to the chipset then gets converted to QPI. I'm just not sure, as the drivers have mulitple sections, and it may only be in one area (DMI connect). It's possible it is in the chipset drivers, but I can't yet tell what's changed enough to be suspect.

But we're in the right area (North and/or Southbridge).

Looking at that, anything on the ich10 bus in theory could set this bug off?

I'm not seeing it, but I'll continue to look (part comparisions).

 

[eyeruh]

Also, I exchanged emails with a prominent tech blogger last night who has confirmed the issue and has also spoken to Intel, with no results. This person also believes Apple is aware of the problem but doesn't believe they will act, since it is not "too bad" (Apple's opinion, not the bloggers) ...

Do you happen to know if they talked to Apple about the temperature increase, the performance hit or both?

It boggles my mind that anyone at Apple could say a 5-20% performance hit is "not too bad"!

 

[VirtualRain]

So do I interpret this correctly - I see the 4 PCIe slots connecting to the North Bridge with the South Bridge being the ICH10 hub, although it is not labeled as such in the schematic?

If that were the case, installing a USB or Firewire card into these slots would eliminate our issues as well.

Can someone confirm? Thanks Smacman for the quick response.

I was wrong... you are correct... All PCIe slots connect to the NB (and do not involve the ICH). Hence I think using a USB or Firewire card for those of us with USB/FW audio interfaces, should solve this problem.

Is anyone really motivated to test this theory and buy a PCIe USB/FW card like the one I linked at the top of this page?

 

[eyeruh]

Is anyone really motivated to test this theory and buy a PCIe USB/FW card like the one I linked at the top of this page?

I have a PCIe FW card that I used to connect my FW audio interface into my PC before I switched to Mac. Silly me thought I wasn't going to need that anymore. :-|

I'm supposed to be working right now but I'll test it later today. I verified a few days ago that I can reproduce the CPU temp increase and performance drop just by plugging in my FW audio box so this should be a decent test-case.

 

[nanofrog]

I have a PCIe FW card that I used to connect my FW audio interface into my PC before I switched to Mac. Silly me thought I wasn't going to need that anymore. :-|

I'm supposed to be working right now but I'll test it later today. I verified a few days ago that I can reproduce the CPU temp increase and performance drop just by plugging in my FW audio box so this should be a decent test-case.

Thanks. The results could be really helpful.

 

[handheldgames]

Related question based on the GREAT information here.... Could the bandwidth issues we are seeing in ICH10R be a source of the SATA performance issues on the 2009 MacPro?

 

[dryjoy]

Hmmm

I'm wrestling with myself about this, but if it works, I could be SERIOUSLY tempted to go for a USB card. I know I shouldn't have to, but it is an inexpensive solution, and I currently have to use SMC Fan Control to keep my machine within limits, which is not cool (pun intended) when you're recording with a microphone. Fan noise on recordings sucks.

 

[handheldgames]

Another find.

Sonnet Tango 800: USB2.0/FW800 (about $110)

 

[VirtualRain]

I'm wrestling with myself about this, but if it works, I could be SERIOUSLY tempted to go for a USB card. I know I shouldn't have to, but it is an inexpensive solution, and I currently have to use SMC Fan Control to keep my machine within limits, which is not cool (pun intended) when you're recording with a microphone. Fan noise on recordings sucks.

Agreed... but let's not get over excited just yet.

 

[eyeruh]

I have a PCIe FW card that I used to connect my FW audio interface into my PC before I switched to Mac. Silly me thought I wasn't going to need that anymore. :-|

I'm supposed to be working right now but I'll test it later today. I verified a few days ago that I can reproduce the CPU temp increase and performance drop just by plugging in my FW audio box so this should be a decent test-case.

Bad news guys--I installed the PCIe FW card and plugged my FW audio interface into it but I got the same results as I did when using the onboard FW. Same rise in CPU temperatures, same drop in performance. :-(

I tried with the card in slots 2 and 3 just in case they're handled differently but the behavior was the same in both. FWIW, I used a Siig FW 400 card with a TI-chipset.

 

[nanofrog]

Related question based on the GREAT information here.... Could the bandwidth issues we are seeing in ICH10R be a source of the SATA performance issues on the 2009 MacPro?

It's in previous versions of the ICH families as well, and why the throttling has kept on. Just no one noticed previously, as SSD's either didn't exist, or hadn't been tried.

But it's in the chip itself (DMI is only good for 1GB/s up & 1GB/s down), and full SATA 3.0 Gb/s *6 can exceed that (theoretical = 2.25GB/s, and real would be ~ 1.6GB/s <~270MB/s per port>).

It seems that to keep the bandwidth from being consumed solely by SATA throughputs, it's been forcibly limited (split) to allow some band to be available to the other functions on the part (i.e. PCI, NIC, USB,...).

 

[Inconsequential]

Bad news guys--I installed the PCIe FW card and plugged my FW audio interface into it but I got the same results as I did when using the onboard FW. Same rise in CPU temperatures, same drop in performance. :-(

I tried with the card in slots 2 and 3 just in case they're handled differently but the behavior was the same in both. FWIW, I used a Siig FW 400 card with a TI-chipset.

Ouchies!

Frankly I wasn't going to accept that as a solution anyway.

Still no word from engineering, I think I have them stumped! Although not going to count my chickens yet.

And if this bug effects the ICH10 that means the PWM controller could be effected too? Hence why fans etc don't ramp up?

 

[VirtualRain]

Bad news guys--I installed the PCIe FW card and plugged my FW audio interface into it but I got the same results as I did when using the onboard FW. Same rise in CPU temperatures, same drop in performance. :-(

I tried with the card in slots 2 and 3 just in case they're handled differently but the behavior was the same in both. FWIW, I used a Siig FW 400 card with a TI-chipset.

Bad news indeed. It's surprising that a PCIe audio card will work fine, but audio over PCIe FW will not. Can anyone try a USB card to rule that out?

 

[BillB50]

Bad news guys--I installed the PCIe FW card and plugged my FW audio interface into it but I got the same results as I did when using the onboard FW. Same rise in CPU temperatures, same drop in performance. :-(

I tried with the card in slots 2 and 3 just in case they're handled differently but the behavior was the same in both. FWIW, I used a Siig FW 400 card with a TI-chipset.

Just asking a question here because I am surprised by these results - when you plugged in your FW audio, did you have to change the playback within the preferences/sounds panel? I am trying to rule out the possibility of additional audio processing done through the on-board codec which would put you back to square one.

 

[nanofrog]

And if this bug effects the ICH10 that means the PWM controller could be effected too? Hence why fans etc don't ramp up?

It's possible.

According to the block diagram smacman posted of the MP, it appears they did connect the PWM controller to the ICH as most other board do (saves costs and is easier to implement IMO).

 

[eyeruh]

Just asking a question here because I am surprised by these results - when you plugged in your FW audio, did you have to change the playback within the preferences/sounds panel? I am trying to rule out the possibility of additional audio processing done through the on-board codec which would put you back to square one.

No, I didn't have to change anything. Plugging in the audio interface results in it automatically being selected as the current sound input/output. I just doublechecked to verify this (and I already see the temps climbing).

And yes--I'm surprised too! After seeing your results, I really thought this would work.

 

[BillB50]

No, I didn't have to change anything. Plugging in the audio interface results in it automatically being selected as the current sound input/output. I just doublechecked to verify this (and I already see the temps climbing).

And yes--I'm surprised too! After seeing your results, I really thought this would work.

I'm not familiar with external sound devices - how does it work, can you provide a brand and model number?

 

[eyeruh]

I'm not familiar with external sound devices - how does it work, can you provide a brand and model number?

It's a Presonus Firebox. It's an outboard box that contains A/D converters and sends the digital audio data to the computer via FW. Essentially the same thing that your Maya card does except the Maya has the A/D converters on the card itself.

The only thing I can think of that's really different between them is there may be some extra power running over the FW connection that wouldn't be there with the Maya. I'm using an external power source for the Firebox but it seems at least possible that there's still current running over the FW cable. I really don't know one way or the other though & I'm not sure how that would impact our problem.

 

[BillB50]

It's a Presonus Firebox. It's an outboard box that contains A/D converters and sends the digital audio data to the computer via FW. Essentially the same thing that your Maya card does except the Maya has the A/D converters on the card itself.

The only thing I can think of that's really different between them is there may be some extra power running over the FW connection that wouldn't be there with the Maya. I'm using an external power source for the Firebox but it seems at least possible that there's still current running over the FW cable. I really don't know one way or the other though & I'm not sure how that would impact our problem.

I read through the manual for the device. There seems to be some interaction with the audio built into the Mac/PC. For instance, on the first troubleshooting page of the manual the it describes how the Firebox synchronizes digital word frame rates with your Mac; with the blue light indicating the Firebox is sync'd with your Mac. I don't know if the software can interact directly with Core Audio - or perhaps, Core Audio is processed on the built in audio adapter. The true test would be to install a Maya, set the sound preference to the Maya and then plug in the Firebox. I wish we lived next door - I could let you borrow my card to test! I assume your speakers are hooked to the Firebox?