PCI-e to PCI extender for Scope ( experiment going to be a solution :) )

[snoopy4ever]

oh well..., yes I'd like ( and probably one day ) to put my hands on a Xite-1 , but on the meantime I've been lucky enough to get ( slowly ) with 3 Dsp cards, 2 Scope and 1 Pulsar 2. At first I had the P2 then one scope and recently another Scope ( 15 dsp). Naturally I had a MoBo with ONLY 2 PCI slots ... (Asus P6T ) great piece of tech ( a bit old for today standars but what the heck it works like a charm ).
So I deciced to use my 2 scopes and then a friend of mine said I culd get a Sintech PCI-E Express X1 to Dual PCI Riser Extender Card.

810OEJfx8RL._SL1500_.jpg (110.2 KiB) Viewed 12651 times

Yes !!!.. it implied to change my case , since I did not even had any space or slot left to put anything.,

Anyway... I decided to give it a try and see if It could work , the extender conects directly to the PCI-e and feeds the 2 pci slots with a usb 3.0 cable .
And guess what? .. seems to be working just fine.

20190626_225406.jpg (6.44 MiB) Viewed 12651 times

As you can see the inside fitting of the card here :

20190627_123812.jpg (4.64 MiB) Viewed 12651 times

20190626_184655.jpg (4.91 MiB) Viewed 12651 times

I know some people say the pci overhead and such, It may happen, this is just a preliminary report.

I Plan to star "bombing" with sinths and recording, and such ... in order to test if this really works...
But if any of you can give me TIPS OR WAYS to test this setup so we can have a clear conclusion of this I would really appreciate it.

Again, thanks a lot

[dante]

Looks awesome ! There might be whitecoat testing methods such as loading up Masterverbs or adding polyphony to synths to find croak points but just try building the normal projects you would want to use - eg test it musically.

[at0m]

I'm amazed this works at all, looks promising.

AFAIK, all PCI traffic is dealth with by the first card as per cset.ini's boardID setting. So as long as one of the cards on your "real PCI" bus get to deal with that, that could be working - proof is you're doing it. Beyond that, the other cards are communicated with over the STDM cables, IIUC.

Now the real test (!!) would be to change to just 2 PCI cards on the convertor board, and none on the onboard PCI slots. If you can get that running ok (see masterverbtest), then this looks like a great solution, or at least an experimental option, for motherboards with /no/ onboard PCI, only PCI-E.

[snoopy4ever]

I'm amazed this works at all, looks promising.

AFAIK, all PCI traffic is dealth with by the first card as per cset.ini's boardID setting. So as long as one of the cards on your "real PCI" bus get to deal with that, that could be working - proof is you're doing it. Beyond that, the other cards are communicated with over the STDM cables, IIUC.

Now the real test (!!) would be to change to just 2 PCI cards on the convertor board, and none on the onboard PCI slots. If you can get that running ok (see masterverbtest), then this looks like a great solution, or at least an experimental option, for motherboards with /no/ onboard PCI, only PCI-E.

Great Idea at0m !! ... the problem is .. oh well as you can see , I had to do some "engineering " on the case and move the Power Supply in certain position, becasue the base card ( the 2 pci one ) has a strange width , so one of the slots is pretty much on a hard position to use ... BUT .. NEVER MIND!...., let me just get used to this..and I'll try to use both pci's

[snoopy4ever]

Looks awesome ! There might be whitecoat testing methods such as loading up Masterverbs or adding polyphony to synths to find croak points but just try building the normal projects you would want to use - eg test it musically.

You are totally rigth Dante!!! Musically is the key !!!

[garyb]

masterverb tests are pretty accurate and functional measures of performance.

[snoopy4ever]

masterverb tests are pretty accurate and functional measures of performance.

thank you garyb !! .. will do it

[snoopy4ever]

masterverb tests are pretty accurate and functional measures of performance.

Hello garyB

The masterverb resuts are 13 masterverbs and pci limit reached on the 14th. ( empty project of course )

Do you think is a good result ?

Regards!.

[at0m]

Great result for MV test!

Small remark though - you may have tested the onboard PCI's performance, not the extender's performance. To measure the latter, I can think of 2 ways to force the MV's to load on the extender: 1) load them in a mixer (dynamixer would do) that's forced to your Pulsar2 and hope it doesn't reach DSP limit first or 2) remove cards in the onboard PCI slots, maybe even temporarily put the 2 15DSP cards there. I understand this last is quite a Big Job that few people would like to take on, but in my opinion it's the proper way to test the performance of the actual extender card.

[snoopy4ever]

Great result for MV test!

Small remark though - you may have tested the onboard PCI's performance, not the extender's performance. To measure the latter, I can think of 2 ways to force the MV's to load on the extender: 1) load them in a mixer (dynamixer would do) that's forced to your Pulsar2 and hope it doesn't reach DSP limit first or 2) remove cards in the onboard PCI slots, maybe even temporarily put the 2 15DSP cards there. I understand this last is quite a Big Job that few people would like to take on, but in my opinion it's the proper way to test the performance of the actual extender card.

thanks at0m !.. I'll go for the 1 and I'll let you know
snoopy4ever

[snoopy4ever]

Great result for MV test!

Small remark though - you may have tested the onboard PCI's performance, not the extender's performance. To measure the latter, I can think of 2 ways to force the MV's to load on the extender: 1) load them in a mixer (dynamixer would do) that's forced to your Pulsar2 and hope it doesn't reach DSP limit first or 2) remove cards in the onboard PCI slots, maybe even temporarily put the 2 15DSP cards there. I understand this last is quite a Big Job that few people would like to take on, but in my opinion it's the proper way to test the performance of the actual extender card.

hello at0m.,

just did what you asked me ... and I think is crazy enough.. Please let me know if this is correct :

I loadded a dynamicMixer and set 10 stereo channels and inserted 20 MV's on the insert slots.... so far no pci problem .. I'm attaching an image, pls letme know if you see something wrong.

20Mv-on-dynamicMixer.jpg (193.55 KiB) Viewed 12496 times

[astroman]

I don't know how Masterverbs perform at insert position - they may not be 'engaged' at all while zhe channel is inactive.
The usual test is to just load them into the project and they are activated even without anything connected to them.
But that test can be misleading (you can load Masterverb after Masterverb) as has been reported ages ago.
That case is a kind of a short-circuit situation, where the system simply hangs at an early stage and seems to ignore all Masterverbs that are loaded after this event. There's no error report of any kind iirc.

But 13 is a great result in any case, as the slot extender is at least engaged and provides DSP power

[snoopy4ever]

I don't know how Masterverbs perform at insert position - they may not be 'engaged' at all while zhe channel is inactive.
The usual test is to just load them into the project and they are activated even without anything connected to them.
But that test can be misleading (you can load Masterverb after Masterverb) as has been reported ages ago.
That case is a kind of a short-circuit situation, where the system simply hangs at an early stage and seems to ignore all Masterverbs that are loaded after this event. There's no error report of any kind iirc.

But 13 is a great result in any case, as the slot extender is at least engaged and provides DSP power

Thank you astroman !!!.. yes seems to be working and I hope this option may work for other scopers that are in need of using 3 cards on a 2 or less PCI slots .

[garyb]

doesn't seem too bad....

[dante]

20 x MV is more than I ever got - depending on sample rate of course. Still, its a good number.

[snoopy4ever]

doesn't seem too bad....

Thx garyb . ! Not sure about the dynaMixer test but other than that I think I've got some music to do

[snoopy4ever]

20 x MV is more than I ever got - depending on sample rate of course. Still, its a good number.

Thanks dante !!!! I do appreciate all your help and feedback

[astroman]

It just entered my mind (never checked it myself):
if you add Masterverbs, the Async load in the DSP meter should increase with every item added
(it's about stuff that travels between PCI bus and computer memory within a certain time frame)
So if that figure doesn't increase anymore, only the control surface is loaded but no resources are allocated.

[at0m]

[...] load them in a mixer (dynamixer would do) that's forced to your Pulsar2 [...]

hello at0m.,

just did what you asked me ... and I think is crazy enough.. Please let me know if this is correct :

I loadded a dynamicMixer and set 10 stereo channels and inserted 20 MV's on the insert slots.... so far no pci problem .. I'm attaching an image, pls letme know if you see something wrong.

In your current setup, Scope OS still gets to choose on which board it loads the mixer, and I suspect that's currently on one of the Scope cards on onboard PCI. At the bottom of Dynamic Mixer, between the name of the mixer and the midi channel, there's a drag-and-select option to force the mixer (and hence, the MV's) onto a DSP card on the PCI extender. In your case, drag the board option to board2.

The important part here:
I'm trying to devise a way to measure the actual PCI throughput for the extender. Once that is confirmed, such extender would be an option for those motherboards that do not even have any classical PCI slots!

And maybe someone else could give exclusion on this: how can we be sure the MV's send traffic over the extender PCI bus and not the STDM cables to a card that is on onboard PCI?

I think the method I suggest above would be enough, but for a conclusive test, the 2 Scope cards on the extender card and no DSP card on the onboard PCI. I understand physically moving around cards is a tough and annoying job and forcing the mixer on the DSP card on the extender is an easier solution. Or maybe I should just order such an extender card and find out myself. I'm that curious

[garyb]

it does not matter. if the extender provided poor performance, it would be obvious from the MV test.