exsys pcie to pci adapter 2 pulsars with x3d cpu working great

[panaman]

hi, after roughly a year of daily usahge i can confirm the exsys pcie to pci adapters are working fine with 2 pulsars, 30 dsp in win10 21hxx.

system:
7950x3d at around 5ghz. ram freq 5800mhz. asus 670e hero mainboard. 3060rtx grafix-
exsys ex1095 pcie to pci adapter card on the pcie x1.
ex1035 expansion board with 4 pci slots in a separate midi tower, always on.

i-m not a power user, don-t need dozens of reverbs. wouldn-t even know how to test the scope bus.
instead i use the many inputs of the a16ultra and a 1696 box, 4 stereo asios into reaper and back. reaper reports 1.3ms @ 96khz, makes me happy.
of course the 2 exsys cards will set you back some 400€, well worth it for me considering i carried my scope investment onto the latest amd platform, and well into the 21st century.

in latency monitor interrupt to process latency is zero (2 squares max). isr ex.time is zero, dpc is 3 squares,

in the course of this year there have only been 2 hickups, once a loud hiss while listening to winamp which needed a scope software restart, and once a no connection when booting, cured with a power off reboot.

i-ll be asking about my winamp headache in another thread.

[garyb]

to test PCI performance, put masterverbs into an EMPTY project and see how many will fit before an error message. this is not to see how many reverbs you can use, but to see how much traffic the system allows. typically, the adaptors severely reduce PCI perfromance, but this may not be an issue. if you can get 6-8 masterverbs, i would say that it works pretty well.

if you are happy, that is the ultimate test.

[panaman]

ok where do the masterverbs go, sends from the mixer or standalone?

[Bud Weiser]

to test PCI performance, put masterverbs into an EMPTY project and see how many will fit before an error message.

Gary, up to now I thought the masterverbs must have been connected to something (hardware I/Os ?) in the project and for the test,- am I wrong ?
Just only load and see how many ?



Bud

[panaman]

ok i got 4, they autoconnected to my mixer and the rest of my project

[garyb]

EMPTY project.

4 is quite poor, but if it's enough, then no worries!

no, they don't need to connect to anything. just having one in the routing window means 1000s of delay lines running across the PCI bus from card to system memory and back. this is a good stress test for the PCI bus.

[Bud Weiser]

EMPTY project.

... they don't need to connect to anything. just having one in the routing window means 1000s of delay lines running across the PCI bus from card to system memory and back. this is a good stress test for the PCI bus.

Thank you,- good info !

I got plenty of Masterverbs loaded,- PCI (and XITE anyway).
But I´ve made connections ...
Did it make a difference ?



Bud

[garyb]

probably not.

[panaman]

thanks Gary for clearing that up.
just one more thing, do you think it might improve if i tried using the 2nd graficcard pcie x 16 slot instead? i know it will still be a x1 adapter, but my x1 is connected via southbridge while the x16 is direct to cpu.
or even use 2 of the adapter cards in the 2 slots for my 2 cards.
i,m actually serious.
thanks again

[garyb]

no, that won't improve PCI performance.

[the19thbear]

hi, after roughly a year of daily usahge i can confirm the exsys pcie to pci adapters are working fine with 2 pulsars, 30 dsp in win10 21hxx.

system:
7950x3d at around 5ghz. ram freq 5800mhz. asus 670e hero mainboard. 3060rtx grafix-
exsys ex1095 pcie to pci adapter card on the pcie x1.
ex1035 expansion board with 4 pci slots in a separate midi tower, always on.

i-m not a power user, don-t need dozens of reverbs. wouldn-t even know how to test the scope bus.
instead i use the many inputs of the a16ultra and a 1696 box, 4 stereo asios into reaper and back. reaper reports 1.3ms @ 96khz, makes me happy.
of course the 2 exsys cards will set you back some 400€, well worth it for me considering i carried my scope investment onto the latest amd platform, and well into the 21st century.

in latency monitor interrupt to process latency is zero (2 squares max). isr ex.time is zero, dpc is 3 squares,

in the course of this year there have only been 2 hickups, once a loud hiss while listening to winamp which needed a scope software restart, and once a no connection when booting, cured with a power off reboot.

i-ll be asking about my winamp headache in another thread.

Can you link to the product? And does it still work in 2025 on your machine?

Thanks

[Bud Weiser]

no, that won't improve PCI performance.

With my AsRock Z97 mainboard, I recognized significantly faster Scope project load times and was able running larger projects (more devices) w/ XITE-1 PCIe card inserted into the PCIe x4 slot being connected directly to CPU instead using the PCIex1 slots being connected to chipset.
Our friend Jimmy (R.I.P.) was the 1st one coming up w/ this trick when using XITE and the same mainboard.
As a side effect,- XITE-1 PCIe card now got it´s own/individual IRQ,- which was never the case when being inserted in any of the available PCIe x1 slots,- where 5, 6 or 7 devices shared the same IRQ #16 or #19, USB controllers and s##t.

And a RME Raydat works very well in that slot too !

So,- it MIGHT work w/ better performance,- depending on the mainboard´s architecture and it´s hardware configuration.
I can imagine it doesn´t work that good w/ every mainboard out there.

Unfortunately it´s hard to "try before you buy" a new computer mainboard.



Bud

[fra77x2]

Jimmy was great and was dedicated to music and computing. He was a real artist with a passion for life. He also adored scope ( for live ) and perhaps his addiction to it contributed to his health issues.

Its not strange he had found motherboards with good performance. He was giving plenty of money and effort for his career

[valis]

When an IRQ sharing really hurts the Scope cards is when PCI or PCIe lanes are shared as well. This is typically documented in the manual or by contacting support.

A shared IRQ (virtual--which means not electrical but just address range) can still be a concern but as Bud notes it will mostly impact the bus bandwidth the card can realize due to IRQ handshaking eating some of the PCI bus clock cycles (time to release old IRQ, negotiate with Scope card and transmit etc). Best to resolve both of course, but the former is more critical.

(In the post north/southbridge era we have had so many changes in how the PCI bus is connected to a host system there's no single post response I can encapsulate best practices in for everyone, so not going to attempt that here)