9 replies to this topic

[forch]

Ello, seems someone has heard my cry for not able to start new topics
Been posting on anyone running protools thread for a while.

I have successfully configured a hackintosh, running Protools HD 8.0.3 with 1 PCI Core and 1 PCI Accel card. I made a stress test session with 72 stereo tracks, some 40 TDM plugins and over 60 RTAS plugins. Although Protools is pretty loaded, the CPU isn't really bothered with any of it. Only drawback was the harddrive not able to deal with the load on the I/O on loop cycle (get a drive with higher than normal burst ratio). Solid as a rock.

Alot of people seem to be skeptic, because 'DigiDesign' states that it does not work. I do not blame anyone for being skeptic, but a little internal computer knowledge before stating anything would be very appreciated .

First of all, localization of hardware (interupt requests) is done by the BIOS, not the OS. This does not differ from any Macintosh. The drivers written to communicate with some protocol, aren't written to specifically communicate with a slot (let it be VESA, ISA, PCI, PCIe), they are written to communicate with a certain hardware layer.
-- edit
Unless of course, the manufacturer has specifically put some eprom, or whatever rom to recognize the slot, therefor the driver could decide to work, or not. But in many cases this is just some luxury overhead many manufactures do not really care about and do not implement unless the card is made for some specific performance issue, and requires therefor some sort of detection mechanism.
--

To make things worse for a computer. If a PCI slot and a PCIe slot share the same bus (which in most cases is a fact) the driver doesn't even know in which slot the card is in. This might sound very stupid, but it's a fact. There is only one communication channel to the bus, not every slot has it's own bus with it's personal communication channel. You can read the problems some people have when they have this specific order of which card goes into which slot in order to get the system stable and work properly. This is the side effect of sharing a bus over several slots. Although bus sharing is not the real problem, the driver written to communicate with the card could get in trouble when more cards are calling the same communication line (bus) and the company/programmers/idiots didn't properly write the driver to distinct the two. You can write a driver (which in most cases on every University is a standard procedure for students who start writing abstraction layers for hardware) where you claim the bus for just one card. Most test cases only have one card (the test card) and therefor, they never run into any trouble. Until you put a ethernet card on the shared bus, and you'll at some point, notice instability.

I could go on and on about this topic, but I won't

The VirtuaVia (or any expansion company, Magma, Cyclone, etc) PCIe => PCI expansion, unless the expansion (or any expansion) was poorly designed, which would make it impossible/improbable for the BIOS to detect any hardware, shares the same principle.

The PCIe bus get's shared over multiple PCI's. In theory, if the PCI cards work on a shared bus locally (mainboard) on the computer, they should work on the shared bus in the PCIe expansion. The principle is the same because if the expansion is a 4x PCI to 1 PCIe slot, you share 4 PCI's on one PCIe bus. If the PCIe bus is again locally (mainboard) shared with another slot, and this one is filled too, you are just asking for trouble.

Always check the specifications of your motherboard how the setup of the busses are configured. Every mainboard has one

A VirtuaVia PCIe => 4x PCI ATX expansion kit (24v ATX connector) is on it's way from france. Will post some detailed tests and performance issues when it arrives.
-- edit
It arrived and worked, check post below
--

Cheers

[forch]

So today the VirtuaVia ATX expansion kit arrived. We build it in an empty chassis laying around,
used a newer 500w power supply (the small board uses a ATX20+4 connector) and amazingly, everything worked out of the box.

At first, we just had one Core card installed (and therefor I guess only 2 DSP's?). While testing the whole environment, the load on the PCIe 1x slot (which Protools System Usage displays) was pretty high though nothing really fancy was loaded. Then after a short 'celebration' we installed the Accel card and there the trouble started. Some DAE errors (-9073 and -6042, randomly) appeared after we tried to load our setup again (72 stereo tracks, some 40 TDM's and some 60 RTAS). It could play for about 2 seconds and stopped. Sometimes the system just freezed.

So I switch the PCIe VirtuaVia card which connects to the expansion board to a faster PCIe 4x slot, and everything was smooth. No performance decrease, whatsoever. We've tested it a few hours (prolly 4 or so) and solid as a rock.

I can only state from a few (very few) posts I've read, stating this problem where PT HD wasn't able to load more then 48 voices, they probably used the smaller 1x slot. I highly recommend not using this slot because the peak bandwidth isn't that high.
-- edit
The 4x still gave some -6042 DAE errors (PCI load), not as many as the 1x. Later on tried the 16x slot, again less
DAE errors, but still some occasionally (specially when the loop was around 20 seconds, everything above is ok). Currently the graphics (9400GT) used the 8x slot, switched it to the second 16x slot. When looping around 20 seconds, the only DAE error remaining is the 9073 (Disk load) which would make sense because one harddrive is a bit old already, though this happens randomly between 5 and 60 minutes. When looping more than 20 seconds, everything works smooth, no errors. Been running continuously for hours now.
--
Though they tend to clock back to the rate of the card that's installed, the peak's in the bandwidth has it's full capacity.

The gear setup is in my sig.

Here are some pictures of the setup.

Protools HD 8.0.3



The interfaces


The VirtuaVia ATX expansion board with the HD cards.


Cables


The hackingtosh (on this picture, still in the 4x, replaced it to the second PCIe 16x (2.0) slot (blue ones). Graphics now in the first 16x slot. Last slot is a 8x where the graphics card was installed (slowed down the graphics a lot) .. ).

[d.t.heck]

forch, on Mar 27 2010, 01:12 PM, said:

Ello, seems someone has heard my cry for not able to start new topics
Been posting on anyone running protools thread for a while.

I have successfully configured a hackintosh, running Protools HD 8.0.3 with 1 PCI Core and 1 PCI Accel card. I made a stress test session with 72 stereo tracks, some 40 TDM plugins and over 60 RTAS plugins. Although Protools is pretty loaded, the CPU isn't really bothered with any of it. Only drawback was the harddrive not able to deal with the load on the I/O on loop cycle (get a drive with higher than normal burst ratio). Solid as a rock.

Alot of people seem to be skeptic, because 'DigiDesign' states that it does not work. I do not blame anyone for being skeptic, but a little internal computer knowledge before stating anything would be very appreciated .

First of all, localization of hardware (interupt requests) is done by the BIOS, not the OS. This does not differ from any Macintosh. The drivers written to communicate with some protocol, aren't written to specifically communicate with a slot (let it be VESA, ISA, PCI, PCIe), they are written to communicate with a certain hardware layer.
-- edit
Unless of course, the manufacturer has specifically put some eprom, or whatever rom to recognize the slot, therefor the driver could decide to work, or not. But in many cases this is just some luxury overhead many manufactures do not really care about and do not implement unless the card is made for some specific performance issue, and requires therefor some sort of detection mechanism.
--

To make things worse for a computer. If a PCI slot and a PCIe slot share the same bus (which in most cases is a fact) the driver doesn't even know in which slot the card is in. This might sound very stupid, but it's a fact. There is only one communication channel to the bus, not every slot has it's own bus with it's personal communication channel. You can read the problems some people have when they have this specific order of which card goes into which slot in order to get the system stable and work properly. This is the side effect of sharing a bus over several slots. Although bus sharing is not the real problem, the driver written to communicate with the card could get in trouble when more cards are calling the same communication line (bus) and the company/programmers/idiots didn't properly write the driver to distinct the two. You can write a driver (which in most cases on every University is a standard procedure for students who start writing abstraction layers for hardware) where you claim the bus for just one card. Most test cases only have one card (the test card) and therefor, they never run into any trouble. Until you put a ethernet card on the shared bus, and you'll at some point, notice instability.

I could go on and on about this topic, but I won't

The VirtuaVia (or any expansion company, Magma, Cyclone, etc) PCIe => PCI expansion, unless the expansion (or any expansion) was poorly designed, which would make it impossible/improbable for the BIOS to detect any hardware, shares the same principle.

The PCIe bus get's shared over multiple PCI's. In theory, if the PCI cards work on a shared bus locally (mainboard) on the computer, they should work on the shared bus in the PCIe expansion. The principle is the same because if the expansion is a 4x PCI to 1 PCIe slot, you share 4 PCI's on one PCIe bus. If the PCIe bus is again locally (mainboard) shared with another slot, and this one is filled too, you are just asking for trouble.

Always check the specifications of your motherboard how the setup of the busses are configured. Every mainboard has one

A VirtuaVia PCIe => 4x PCI ATX expansion kit (24v ATX connector) is on it's way from france. Will post some detailed tests and performance issues when it arrives.
-- edit
It arrived and worked, check post below
--

Cheers

Same here on X58 with HD2. No problems so far... even the onboard audio is working and a second network card I popped in...

[SaddenTrap]

Nice production setup :jealous: ,congratz

[JustinAiken]

Hey forch,

Your post is a few months old now... Have you been running it smoothly this whole time?  

Do you think it's stable enough to depend on making a living with?

I'm seriously considering moving from a BLA 002 to a HD2 system with Aurora converters, and opening a studio...  Being able to buy cheap PCI cards would help make it financially feasible for me.

[forch]

JustinAiken, on May 26 2010, 04:55 PM, said:

Hey forch,

Your post is a few months old now... Have you been running it smoothly this whole time?  

Do you think it's stable enough to depend on making a living with?

I'm seriously considering moving from a BLA 002 to a HD2 system with Aurora converters, and opening a studio...  Being able to buy cheap PCI cards would help make it financially feasible for me.

Yea it's still running. I'm installing it completely new again as I'm typing this reply.

Some templates for pro tools seem to make the hack unstable. Gray screens with "you can turn your computer off now" appear a lot, not always, but randomly.

Though some other PT HD templates run just fine without trouble. I've been in contact with some Mac Pro (PCIe cards that is) users and they state the same problem. It looks like Avid made crappy software and I can't blame the hackingtosh. Lucky for me I know my way around linux, therefor can do stuff that makes OSx a bit happier.

You need to dig into stuff you normally wouldn't. Because this community has grown so big, and a lot of guys here are able to help (also with scripts), it shouldn't be any problem getting your hack up and running.

For PCI cards and i7-9xx, I wouldn't suggest the UD5, but the UD3, basicaly the same board, only with different slot layouts (the two PCI slots are next to each other). If you have bridge (or link) cables (like the HD cards have). Check the length they can expand from one slot to the other (there are two types, on short, on long (the long one is two times the short length)).

There is a Gigabyte board out there with 3 or 4 (out of 5) usable PCI slots. It has a 1156 socket which fits the latest i5 750 (without hyper-threading). This setup is similar to an i7-9xx setup. The i5's are very stable in overclocking and gigabyte is one of the best boards to overclock (in my experience).

Logic 9 runs superb.

So if you stick with HD2 (I guess 1 core, 1 accel?) take the UD3.

Good luck

[Kaz Muzik]

forch, on Jun 1 2010, 04:54 AM, said:

Yea it's still running. I'm installing it completely new again as I'm typing this reply.

Some templates for pro tools seem to make the hack unstable. Gray screens with "you can turn your computer off now" appear a lot, not always, but randomly.

Though some other PT HD templates run just fine without trouble. I've been in contact with some Mac Pro (PCIe cards that is) users and they state the same problem. It looks like Avid made crappy software and I can't blame the hackingtosh. Lucky for me I know my way around linux, therefor can do stuff that makes OSx a bit happier.

...

I used to consider VirtuaVia's PCIe to 4-slot PCI expansion ATX kit
http://www.virtuavia...on-atx-kit.html
but according to DUC thread http://duc.digidesig...ad.php?t=256828
it seems not stable.

According to http://expansiontech...nsion-units-are
the Pericom Bridge Chip is compatible with Intel Mac.

[gregie]

Hi forch,

which of the Gigabyte-Mainboards with an 1156 socket would you recommend, if you - like me  - would like to have onbord Firewire and eSATA and the option, to install PT HD2 (1 Core and 1 Accel PCIe card)?

Thanks, greg

[forch]

Hi, sorry for the late post, I've been so busy the last couple of months, totally forgot about this place

Um, it's pretty simple with gigabyte boards. Make a list of the boards that fit your needs on specifications, like USB inputs, eSATA, SATA controllers, etc etc. Write a list of all the components they use, and hereby I mean chip-sets. Does it use a Realtek chip for Ethernet, broadcom for wireless, or is it all Intel, etc. Then from each item of the list, search for it on this place (sanely), tonymac, google, yahoo, etc. Try to found out successes, failures, problems, etc. Write down each peace of information you can found about it, even if you don't know what it means you have to do to make it work. Read also the replies people have made on how simple it was, or make a little count on how many successes in general.

If your list is complete, select all the items on the list that 'sort of' or 'work' out of the box. Compare against your desired mainboards and tada.. you can choose your board. If it is so, that none of the boards you desire will work, you really really really should consider buying a mac.

For the 1156, I have no idea, but since you have only two PT PCI cards, I def. would recommend any Gigabyte board that has two PCI's slots next to each other. Else you will face the same problems with the PT HD cards flat cable as I had, they are short, and will not fit otherwise.

For anyone having questions, please send me a private message, they do spawn in my mailbox

Cheers

gregie, on Jun 21 2010, 07:03 PM, said:

Hi forch,

which of the Gigabyte-Mainboards with an 1156 socket would you recommend, if you - like me  - would like to have onbord Firewire and eSATA and the option, to install PT HD2 (1 Core and 1 Accel PCIe card)?

Thanks, greg

@Kaz Muzik

If you were the owner of Digidesign, what would you write? Of course it does not 'support' or does not 'do this' as well as our "Magma" does.

It works. It is important to note that when using PCI slots, you really should decrease your buffer sizes. This means that your hard-drive and CPU has a bit more work to do. But the i7 920+ is just one mean beast.

Kaz Muzik, on Jun 21 2010, 04:45 PM, said:

I used to consider VirtuaVia's PCIe to 4-slot PCI expansion ATX kit
http://www.virtuavia...on-atx-kit.html
but according to DUC thread http://duc.digidesig...ad.php?t=256828
it seems not stable.

According to http://expansiontech...nsion-units-are
the Pericom Bridge Chip is compatible with Intel Mac.

[forch]

@update
We exchange both PCI HD cards with the newer Native HD card. Mainly because recording is being done in Logic and only 8 ins and outs are available using the pro/acc HD cards. Just wanted to confirm that the Native card works like a charm. No problems at all, we exchanged every card on the fly, booted the hack, installed drivers, installed pro tools 9, and after a second reboot everything was detected (2x 192I/O's, 1x96I/O, SyncI/O).

Cheers