PCI tester

spacef · Post by **spacef** » Thu Feb 05, 2004 5:31 am

Hi there,
i did a few experiments for breakfast and did this. It is pretty useless, and made for the fun, but it allows to test pci perfomances when you don't have reverbs.
With normal pci performances (one board 1) you can load 2, and a pci overflow happens on 3rd load. then you can load more and resynch and see how pci behaves.

If pci message doesn't happen on 3rd load, you have superior pci performances (for 15 dsp, see below)

I upload it here because it is really useless (eventhough cute to look at for a few seconds), it has nothing to do in the device section, and doesn't deserve mentionning except for alternative pci testings (it doesn't take much dsp, but there is much better already for SFP like in Optimaster or the sts)...
download Pci Timeline.zip

nb : it shows a max of 700 miliseconds of audio approx. (according to the zoom settings).

To make finer measurement of your mobo you can use this smaller device download PCI-Timeline-1UNIT.zip . The "big one should be reffered to as "one pound".

[ This Message was edited by: spacef on 2004-02-05 16:51 ]

[ This Message was edited by: spacef on 2004-02-05 16:52 ]

spacef · Post by **spacef** » Thu Feb 05, 2004 5:42 am

pIV + p4pe (default pci settings) + 1 scope generation 1 = 3 max (pci message on 3rd load).

[ This Message was edited by: spacef on 2004-02-05 05:43 ]

kensuguro · Post by **kensuguro** » Thu Feb 05, 2004 8:16 am

? What is this?
I'm on P4P800 with PIV 3ghz 1gb ram
I get "DSP capacity reached" on the 3rd load. What does that mean?

spacef · Post by **spacef** » Thu Feb 05, 2004 8:25 am

I have dsp capacity reached on 4th load only (but the dsp meter is less than half, so i guess it's a pci error too), and it still works fine with 3, after a resynch.

It could mean that the pci bus is less good than on other motherboard, or that it could stand an optimization in the mobo bios....o rmay be you did reach the dsp capacity (depends on how many dsp you have).

[ This Message was edited by: spacef on 2004-02-05 08:35 ]

valis · Post by **valis** » Thu Feb 05, 2004 8:38 am

Supermicro P4DC6+
2x1.8Ghz Xeon
512Mb PC800 rdram
etc...

I can load 3 of them fine, but when I load the 4th I recieve the following message:

"A capacity limit of your DSP cards has been reached.
The specific cause is:
Not enough DSP memory: cannot connect
PC Master 32k Delay.Tap3 and Multi VU-Meter 16.in3...."
etc

spacef · Post by **spacef** » Thu Feb 05, 2004 8:44 am

nice

. i guess it is exceptional that you can load 4 without an error message. If people are interested i could do a smaller one do allow finer results (?)

[ This Message was edited by: spacef on 2004-02-05 08:57 ]

at0m · Post by **at0m** » Thu Feb 05, 2004 9:02 am

Loading 3 here w/o problems, 4th returns same error as Valis.

Yes, those 3-4 devices is kind of a rough measurement. So a smaller version (only 30% of the load?) would make a nice comparing tool. Good thing tho: this PCI Timeline loads way faster than a masterverb

See User Setups topic for my configuration.

Post by **hubird** » Thu Feb 05, 2004 9:04 am

what is that, the 3rd and 4th 'load' ?

spacef · Post by **spacef** » Thu Feb 05, 2004 10:51 am

how much you can load at the same time (well, one after the other, of course).

[ This Message was edited by: spacef on 2004-02-05 10:52 ]

valis · Post by **valis** » Thu Feb 05, 2004 10:55 am

It sounds to me that the error loading the 4th is related to running out of memory in the dsp registers rather than an actual pci overflow...?

spacef · Post by **spacef** » Thu Feb 05, 2004 11:04 am

The "dsp capacity limit reached" does show that scope will refuse to work anymore until something is unloaded (can be pci or dsp) but it is anyway linked to the pci overflow, because the dsp can usually cope with much more than that(sync % in the dsp meter). Here the device use an excessive amount of delay which communicate with the computer ram. This put hard work on the pci bus, and on scope. It can show differences between mobos under extreme pci conditions (apparently all setups do not react the same)
This device is useless anyway, cute to look at audio through it for 5 minutes though. it wasn't made to cause pci overflows basically

[ This Message was edited by: spacef on 2004-02-05 12:45 ]

Immanuel · Post by **Immanuel** » Thu Feb 05, 2004 11:23 am

spacef · Post by **spacef** » Thu Feb 05, 2004 12:38 pm

To make finer measurement of your mobo you can use this smaller device download PCI-Timeline-1UNIT.zip .
The "big one should be reffered to as "one pound".
The "1 unit" is approx 1/5th of a pound.

Next time, i won't eat frosties at breakfast

[ This Message was edited by: spacef on 2004-02-05 12:45 ]

spacef · Post by **spacef** » Thu Feb 05, 2004 1:21 pm

Hey Vallis , as the delay use also dsp memory, it is possible that you get the dsp message "before" the pci overload because you have less than 15 dsp ? there are other modules in there that run on the scope and which play a role too.... especially on low Zoom settings where almost all delay switch to "dsp mode" (eventhough the dsp part is probably loaded by default...??? mmmh).

..

Answer To Vallis (post Below) : the dev in the zip should now load with Zoom at middle.

[ This Message was edited by: spacef on 2004-02-05 16:55 ]

valis · Post by **valis** » Thu Feb 05, 2004 4:10 pm

I was zoomed completely out, and I only have 10 dsp.

Post by **hubird** » Thu Feb 05, 2004 5:42 pm

so that was a dull question, the one about the fifferent loads

I was set on the wrong train I gues coz I thought a PCI tester is a small application, otherwise it's testing it's own use also.
Anyway, I don't have any idea about these things, as you can see haha, but I learnt something already

But I like the idea of creating something usefull, at the end it never really is

spacef · Post by **spacef** » Thu Feb 05, 2004 6:20 pm

(there should be a way to identify all channels and make a real gauge though. but that wasn't really the direction)

(edited : well, it seems it has "some kind" of use? (nice perf astroman!).

[ This Message was edited by: spacef on 2004-02-05 18:39 ]

astroman · Post by **astroman** » Thu Feb 05, 2004 6:36 pm

same here as at0mic and valis

...not exactly:
Pulsar One, Celeron 1Gig on MSI all-in-one Intel-815 376 KB SDRAM

cheers, Tom

samplaire · Post by **samplaire** » Thu Feb 05, 2004 7:49 pm

I got: 3 instances - loading a fourth one causes:

Mac G4 800/1GB RAM Pulasr1 + LunaII and the dsp usage meter showed below a half of the max before loading the fourth device

_________________
Sir Sam Plaire Scopernicus

[ This Message was edited by: samplaire on 2004-02-05 19:55 ]

astroman · Post by **astroman** » Thu Feb 05, 2004 8:53 pm

On 2004-02-05 18:20, spacef wrote:
(edited : well, it seems it has "some kind" of use? (nice perf astroman!).

well, for a laugh - here it is

I think I name the box Dusty Springfield

not EMV compliant btw, positioned under the table with the my ELKA keyboard - almost impossible to photograph.
I could even dive deeper into low performance.
There's a Compaq PII-350 BX in a corner...

But seriously, if such a device is trustworthy it might add well to Pulsar's toolset.
I really like the idea, worth elaborating a bit.

cheers, Tom

[ This Message was edited by: astroman on 2004-02-05 20:58 ]

[ This Message was edited by: astroman on 2006-08-18 18:23 ]