SFP and AMD X2 ?

Shayne White · Post by **Shayne White** » Mon Nov 21, 2005 9:14 am

On 2005-11-14 14:20, MD69 wrote:
Hi all,
I just received my Gygabyte MB and I'll be buying my X2 next week. I'll let you know what happens here.

Best regards

Michel

Hey Michel, did you ever get the x2 chip? How does it work?

Thanks!

Shayne

MD69 · Post by **MD69** » Mon Nov 21, 2005 1:57 pm

Hi Shayne,

Not yet... And to the light of your problems, I am not sure I will. Current
ly, AudioDan and You both got problems with AMD dual core or Intel Dual CPU. If I am correct, both of you use XTC, which is the way I use my boards (2 powerpulsar + srb). I am a bit confused and I am searching more infos. I need to upgrade because I only have 48 voices remaining for my Halion on a PIV 3.2 Northwood. Changing to a new Intel (PIV 3.

will improve my performances of at most 20% while it will cost me about 1200€. Changing to a ATHLON FX 57 will give me about the same increase of performance to about the same price!

What make me a little positive is:
- Either XTC and SFP use the same Drivers. As there is some users reporting that SFP is working correctly on a Dual CPU computer. I do not see any reason why the driver behave differently if its XTC or SFP. Only the way the app interact is changing.
- XTC had some stability problem with Hyperthreading. I use XTC with Hyperthreading activated and low latency (128 samples) and I have a workable config (with some limitations but workable...)since I have "lost" Midi connections from the boards (no midi communication involving the boards except Scope's VSTis).

I am currently looking at restoring midi communication in order as a debugging purpose. If I find something in the configuration which restaure this and I have a stable config I'll let you know, but in the meantime I'll postpone my decision.

I think One ultimate solution could remains: to write an ASIO4ASIO driver (kind of ASIO4ALL driver) which will manage app to driver interaction

Best regards

Michel

Shayne White · Post by **Shayne White** » Mon Nov 21, 2005 2:01 pm

Hi Michel,

I don't use XTC, I use the regular Scope mode with ASIO drivers. Good luck with whatever new system you get!!

Shayne

MD69 · Post by **MD69** » Mon Nov 21, 2005 3:08 pm

Hi Shayne,

I PM you direct about this

regards

Michel

MD69 · Post by **MD69** » Mon Nov 21, 2005 3:57 pm

Hi Shayne,

According to your info, built an SFP project for my sonar 3.1.1, launched Sonar, changed ULLI setting in SFP .... and BOUM! system freezed!

Best regards

Michel

alfonso · Post by **alfonso** » Mon Nov 21, 2005 4:25 pm

On 2005-11-21 15:57, MD69 wrote:
Hi Shayne,

According to your info, built an SFP project for my sonar 3.1.1, launched Sonar, changed ULLI setting in SFP .... and BOUM! system freezed!

Best regards

Michel

You cant change ULLI with any DAW already loaded. When you change ULLI you unload and reload the drivers. DAWs "sit" on drivers, unloading the drivers with an active DAW is like removing the chair from under someone sitting on it, a crash is inevitable.

MD69 · Post by **MD69** » Mon Nov 21, 2005 4:46 pm

Hi Shayne,

You should look at this if you have Sonar 3.1.1:

http://forum.cakewalk.com/tm.asp?m=6284 ... E1&#636424

Seems Sonar 3.1.1 is not compatible with hyperthreading (So dual core ...)

Best regards

Michel

Shayne White · Post by **Shayne White** » Mon Nov 21, 2005 5:23 pm

I have Sonar 5.

No, I've never changed the ULLI setting with Sonar open!! It's always when Sonar (and all other programs) is closed.

[ This Message was edited by: Shayne White on 2005-11-21 17:26 ]

MD69 · Post by **MD69** » Tue Nov 22, 2005 1:02 am

Hi Shayne,

So let's recap. a bit!

Installation:
Motherboard: GA K8NS Ultra 939-> rev 1.002, BIOS rev F9, chipset, USB, SATA and RAID Driver updated...
memory: no info from your config, SMC setting should be 2, 2, 2, 5 for AMD X2
disk: Raid?
Grphic card: ATI Radeon 8500 ->Is your driver compatible with AMD/Dual Core? Anyone on this?
power supply: 400W, ATX V2.0?, V2.01?, V2.02?, V2.03?
-> Sound a little bit low to me! Particularly if its a V2.0, you have raid...
XP SP2 installed, ACPI and APIC AMDCPU cool&Quiet patch installed,
->Power management (enabled?, disabled?) compatible with creamware driver option ?

-> Is there any bended pin on your CPU? (it happens sometimes...)
-> DAW Only or is there other App?

SONAR Stuff:
SONAR 5. -> Cannot help I only have Sonar 3.1.1! Anyone on the forum with SONAR 5 and Dual Core ????

I've been spending time doing audio performance tests because I ran into an old problem I've had since Sonar 2.2 (when ASIO was implemented, actually) where if I have too many envelopes running at once, I start getting clicks and pops. Usually I would correct the problem by setting envelopes to "Jump" mode when they're not changing values (in Jump mode envelopes are not processed).

->Sound like a single threaded SONAR! did you had a single threaded CPU before?

Well, during experimenting, I decided to see what would happen in dual-core mode if I did NOT change the latency setting to make it crash. It didn't seem to crash for a long time, but twice now I've run into a curious thing: when I'm playing a project in Sonar, all of a sudden the ASIO engine and the mouse cursor will freeze (a quarter-second of audio will start repeating over and over), and it's like the usual latency setting crash
-> Sonar have an option for multi processing, is it checked? also, is the app. priority given to background task in XP?
-> Number of midi ports from creamware board in your project compatible with creamware driver setting?

SFP Stuff:
->Is the boards ordering correct?, (and the project rebuild on this config or borrowed from your previous config?)
-> Is the master board the first on the PCI bus?

As you might see from the recap there is some points which need clarifications (not to burden you, but to check which assumption I/we can make!). Particularly, if your reused components/App was working correctly in a dual CPU or hyperthreaded system.
On my side, I can only be of a limited support as I do not run the same computer hardware nor the same App. Anyway, I'll try to reproduce your bug on my system and let you know!

->Anyone with better/more infos welcomed! (particularly, if you appreciated his freq monster, Route...!!!)

Best regards

Michel

Shayne White · Post by **Shayne White** » Tue Nov 22, 2005 11:21 am

Hi Michel,

OK, I'll explain everything from scratch!!

-> Motherboard: Was EPoX EP-9NDA3+ when I first tested the AMD dual-core CPU, exchanged it for GigaByte K8NSC, which I generally like better
-> CPU: AMD Athlon x2 3800+
-> Memory: 2x 512MB PC2700 DDR
-> CreamWare cards: Scope Pro, Pulsar 1, Scope Home (PowerSampler I), with Scope 4.0 software
-> Video card: ATi Radeon 8500 128MB AGP
-> Hard disk: Western Digital 200GB ATA100 parallel, no RAID
-> Power Supply: Nexus 400W, not sure what ATX revision it is (I bought it last year), but it has the extra 4-pin power connector for the motherboard, which I'm using
-> BIOS: I was using the absolute latest version on the EPoX board, not using the absolute latest version on the GigaByte, but they're both compatible with the x2 CPU. Left most settings at default, but disabled things I didn't need, such as LPT1 and audio
-> Using ACPI/APIC; the nVidia drivers are not compatible with Standard PC (which is single-core anyway)!
-> Left all the BIOS power management settings at default; Windows is set to be always on
-> All the CPU pins are straight; it wouldn't go in otherwise!
-> Sonar has no bearing on the matter. If Scope is the only program running after a fresh boot, and I change the ULLI setting, it crashes (only with dual-core, of course)
-> Windows set to prioritize background services; I also turned off a few unnecessary services
-> OS: Windows XP SP2

For a recap, I'll relate the entire story of my testing from the beginning. The first thing I did was I took out my original P4 2.5GHz system from 2002 and put in the new AMD system with the EPoX board. Windows (which was SP1 at that time) had to do a repair to recognize the new stuff. After everything was ready, I installed the nVidia drivers and made sure everything was up and running. Once that was done, I ran Scope -- the only active program (other things were running in the background, of course). My first task was to change the latency setting to test my new fast CPU. Boom! Windows locked up. The mouse cursor froze, the hard drive light went on solid after a moment, and if I kept waiting long enough, eventually I would get a blue screen (I think it was MEMORY_NOT_LESS_OR_EQUAL, or something like that; it never specified what driver had actually caused the crash).

Figuring it was because Windows doesn't repair itself very well, I reinstalled Windows from scratch, this time upgrading to SP2. Everything was a fresh install; I installed the nVidia drivers and then Scope before anything else. Still crashed with the same action.

Then I played around with removing one or more CW cards and moving the remaining cards to different PCI slots. It seemed to work better with using only two cards, but if I kept playing with it long enough, it would crash; maybe within 20 minutes. I didn't try one card only for that long, though. That's the one thing I haven't tried.

Then I tried installing Windows from scratch on another hard drive, but it had the same result.

Then, wondering if it were a power problem, I tried disconnecting extra system fans, leaving only the CPU fan running (the case was open so it was OK). Still crashed.

One of the PCI slots shared an IRQ with the FireWire chip; the others were free. I removed the card from the slot that shared the IRQ, but it still crashed upon testing.

I think it was at that point that I tried switching to Standard PC. The hard disk drivers wouldn't load properly (Windows switched to standard drivers), but I was able to run Scope. It didn't seem to crash, which I found out later was because Standard PC isn't multiprocessing.

I eventually switched back to ACPI multiprocessing. At that point I think I tried ACPI singleprocessing. It worked! After a long time of testing, it didn't crash.

After that, I returned the EPoX board (which I didn't like regardless of the crashing; it was a very bad design) and got the GigaByte board, which is much better. It still crashes with the same setup, however.

I tried removing all my memory and trying another stick of 512MB PC3200 memory from a PowerMac G5. That had no effect.

At that point, I decided to see what would happen if I did NOT change the ULLI setting in dual-core mode. I tried playing with Scope only for a long time, and it didn't seem to crash. I was adding devices, removing devices, loading samples, changing voices, loading presets, etc. Scope itself worked fine.

I fired up Sonar 5 and started playing some of my projects. I noticed I was getting terrible audio performance even at high latencies -- I would get occasional clicks and pops, even if the CPU load were not high. This would fix itself when I removed all the envelopes in my Sonar projects. Turning off multiprocessing in Sonar didn't seem to make much of a difference, but flipping Windows back into ACPI single-core mode actually seemed to improve performance!! (It would handle the envelopes better -- no clicks or pops until the CPU really started pushing itself too high.) I also have Sonar 4 installed; there was no difference in performance. So I think that issue is not with Sonar's processing synths and effects, but it's a low-level driver issue.

Anyway, I just tried playing my projects with no envelopes in dual-core mode, and, after a few minutes, suddenly the ASIO audio engine and the mouse cursor would freeze, in the same manner as if I had changed the ULLI setting. But the MIDI engine kept playing the MIDI tracks into my hardware synths and Scope, with no hiccup whatsoever! I have not experienced such a freeze in single-core mode.

So what's causing the problem? Not a wrong BIOS setting; why would everything work fine in single-core mode? Not a defective Scope card; dual-core would still crash with any combination of two boards. Not the motherboard; I've replaced it. Not the memory; I temporarily replaced it. Not the hard drive or a bad Windows installation; I tried a fresh installation of Windows on another drive. Not an IRQ conflict; I've determined that. Not a conflicting program, such as Sonar, open; Scope is the only program running when it crashes. Not the video card; I don't think it would make the ASIO engine and the mouse freeze and let the MIDI keep going. Defective CPU? Perhaps, but I'm not about to return it. Not enough power? Perhaps, but it doesn't seem right to me. I don't have the money to buy another PS right now anyway. I suppose I could try one Scope card only, which would take a load off the PS, but I won't be able to try that for a few days.

Did I tell you: Scope will crash with a blank project? It's not a weird MIDI problem, obviously.

ACPI single-core has always worked correctly through all of this.

I think I covered everything. If there's anything I missed, let me know!

Thanks!

Shayne

P.S. I should probably explain that envelopes problem a bit more. It's a problem I've had since Sonar 2.2. If I have too many envelopes running at once, I start getting clicks and pops. (It should be noted that this happens only under ASIO; WDM works fine.) The usual remedy is to set envelopes to "Jump" when they're not changing values; that way, only 1-2 envelopes are active at any one time, and it works OK. However, when I was playing my projects in dual-core mode, it didn't even seem to be able to handle that! Single-core seemed to be able to handle them much better. That's why I said it was a low-level driver issue, not Sonar's internal processing.

_________________
Melodious Synth Radio
http://www.melodious-synth.com

Indieanna: Integrated Solutions for the Independent Musician
http://www.indieanna.com

[ This Message was edited by: Shayne White on 2005-11-22 11:38 ]

MD69 · Post by **MD69** » Wed Nov 23, 2005 11:17 am

Hi Shayne,

Tried to reproduce it there, without success

Best regards

Michel

ScofieldKid · Post by **ScofieldKid** » Sat Nov 26, 2005 3:28 pm

Found a link to the hotfix on the Anandtech forums.
http://forums.anandtech.com/messageview ... TMP=Linear

So.. it seems like the obvious things to try if you are X2 and having grief,
1) disable Hyper Thread [ good advice in any situation ]
2) disable Cool-and-Quiet [ the X2 runs quite cool regardless ]
3) install the Microsoft Hotfix
4) punt back to single-proc

I did punt. I think it's going to take some time for all the software companies to get their drivers to be solid under dual-core X2. Things are working great in single-core.

[ This Message was edited by: ScofieldKid on 2005-11-26 21:38 ]

synthetic · Post by **synthetic** » Wed Nov 30, 2005 11:13 pm

Shayne:
>when I'm playing a project in Sonar, all of >a sudden the ASIO engine and the mouse >cursor will freeze (a quarter-second of >audio will start repeating over and over), >and it's like the usual latency setting >crash. But THE MIDI KEEPS GOING. The MIDI >tracks that output to Scope synths continue >to function and play.

I had the same problem as this, I did a re-install of Win XP Pro, Adaptec driver, Scope-driver then scope.
If I had both my Pulsar-Cards connected when installing I still have this problem (or more) so I re-installed it without my 2nd card, run it for a while to check how it worked, then added the 2nd Pulsar, windows showed me a popup that it found new HW, auto-search for drivers and now it works great.
I did a manually check of cset.ini, nothing about NoOfBoards=2 bla bla bla...

I have disactivated Hyperthreading but I run Win as ACPI... Abit IS7 Intel P4 3.0 2.5Gb RAM

jea · Post by **jea** » Wed Dec 07, 2005 2:07 pm

Hi Shayne,

have you tried this:

/INTAFFINITY
Directs the standard x86 multiprocessor HAL (Halmps.dll) to set interrupt affinities such that only the highest numbered processor will receive interrupts. Without the switch, the HAL defaults to its normal behavior of letting all processors receive interrupts.

.... in boot.ini

plus ..... setaffinity for sfp?

ScofieldKid · Post by **ScofieldKid** » Wed Dec 07, 2005 3:06 pm

I'd like to see more documentation on the /INTAFFINITY switch for boot.ini. I have to wonder if effectively that turns your machine into single-proc, so not sure what the value of that is. There is a "/ONECPU" flag as well. Lots of people use that for crippling down to one cpu for gaming.

I was reading the the Entech Powerstrip utility is able to set affinity based on process name. So if you could figure which processes needed to be colocated on the same CPU, then that would make it automatically.

In any case, I think the problem is bigger than any of this. For example, the hotfix that Microsoft released pointed up some of the complexity of all of this. So whether we are looking at 3rd party application, 3rd party driver, or even OSS level issues, it's pretty hard to say where the problems are at this point.

jea · Post by **jea** » Wed Dec 07, 2005 9:36 pm

Well, I heard it from a friend, and he told me about where it was explained:

http://www.sysinternals.com/Information/bootini.html

And there is other flags as well that is interesting, like this one:

/PCILOCK
Stops Windows from dynamically assigning IO/IRQ resources to PCI devices and leaves the devices configured by the BIOS. See Microsoft Knowledge Base article Q148501 for more information.

But /INTAFFINITY should get the driver (beloniging to that irq) loaded to the one core as well, I think.

Shayne White · Post by **Shayne White** » Thu Dec 08, 2005 3:53 pm

Is that all I need to put in? /INTAFFINITY?

I'm rather afraid to muck around with boot.ini in case something goes wrong and I can't get back into Windows. Are you sure this will work?

Thanks for the tip,

Shayne

Shayne White · Post by **Shayne White** » Thu Dec 08, 2005 10:07 pm

OK, I found the page where you got the hint from...I tried both /intaffinity (setting interrupts to one cpu) and /onecpu (turning off multiprocessing) -- not at the same time, of course. To my surprise, it sill crashed!! It has nothing to do with the CPU. The only other explanation is the fact that in ACPI multiprocessing mode, it's using more than 16 IRQs...in ACPI singleprocessing, it's not using more than 16. I tried using /pcilock, but that didn't chnage anything. Well, I'm glad my CPU isn't defective, but I don't know how to make this work. Any ideas? My motherboard doesn't have "force an IRQ number" options, and Windows refuses to follow the BIOS IRQ settings in multiprocessing mode.

Shayne

fra77x · Post by **fra77x** » Fri Dec 09, 2005 9:12 am

I can't also change the Ulli without an instant freeze, but i can work with it, as i just change the ULLI after SFP restart. Except that my system works without any problem.

jea · Post by **jea** » Fri Dec 09, 2005 11:16 am

Hi,

Shayne, did you use GigaByte K8NS Ultra-939?

You can specify the irq's in bios if so.

/INTAFFINITY in boot.ini
/SETAFFINITY for sfp to the highest cpu nr.