PDA

View Full Version : nForce2: tweaks and configuration


fahlen
05-05-2003, 01:14 AM
This is to all nForce2-DAW users!

There is a document uploaded at http://www.rme-audio.de/english/techinfo/nforce2_tweaks_01.htm on how to configure and tweak the nForce2 chipset for DAW use. I've written the
document with the prime objective to spread information on how to configure and fine-tune nForce2 motherboards, in combination with Athlon CPUs for use in digital audio/recording. The Asus A7N8X is used as main example as well as users' experiences in combination with RME soundcards:-)

I know there are quite a number of nForce2 users at the audioforum so this may be of
interest.

Cheers,
Anders Fahlén
Agronova
Sweden

Conejo
05-05-2003, 08:00 AM
Hi,
I'm choosing every single element, planning to buid a DAW. Thanks for this enlightning article, any help is essential for a guy who just wants to buid his own musical freedom, despite of the fact that he isnt a computer freak...
But i'm changing...
So you think I should go with this configuration? Any performance comparison between these tweaked system and a dual MP with Tyan? What about the best HD choices and connections? SATA or PATA? RAID? SCSI? If it's not necessary, i'd rather keep it simple...
Well, you just helped a lot, thnks http://www.audioforums.com/forums/smile.gif
rodigo coelho

fahlen
05-06-2003, 08:07 AM
Rodrigo,

For the public record these are my summarized recommendations:

- You'll got more FPU power per monetary unit by selecting the nForce2 rather than the Tyan MP/MPX dual cpu mobos. The latter are coming close to the end of the production cycle though if you are a heavy user of multitasking operations the dual cpu route will be snappier.

- Right now I would select the Abit NF7-S/NF7 as these NF2 boards probably will achieve robust O/C better than any other NF2 board and their RMA support seem to be at least acceptable...

- Assure that you get the new revision 2 board as they will support the coming 400 FSB MHz Barton cpu out of the box (though Abit even supports this on previous 1.x revision boards via BIOS upgrades)

- Assure that you get compatible,brand-named DDR-RAM, high-quality PSU and a Audio-dedicated hard disk (the Seagate Baracuda V with 8 MB cache would fit nicely).

Cheers,
Anders Fahlén
Agronova
Sweden

nakedjackson
05-06-2003, 04:03 PM
how do i know what version my a7n8x delxe mobo is? How do i know if my corsiar xms 3200 memory (1gig) is CAS or not.

Also great tutorial, but i found it somewhat hard to find exactly how to o/c my setup (XP2700, asus A7n8x, 1 gig of corsair xms 3200, true 400 watt power) Could you help me out with some specifics? Also, could i get away with overclocking my cpu with the stock AMD fan, or do i need a new fan?

fahlen
05-08-2003, 12:29 AM
Nakedjackson,

You receive information if you have a revision 2 board via the updated BIOS string. It is also printed on the PCB (but it is not printed on the box). The revision 2 boards will support 200 MHz FSB setting out of the box whereas 1.x revision Asus NF2 boards most probably will not (whereas Abit boards will via BIOS upgrades).

Regarding the Corsair XMS 3200 CAS2 - you will immediately be aware of any problems if your sticks come from the early problematic batches as the computer will not post. If so, read the manual and you'll have instructions on how to rectify the problem.

My document deals with air-cooled, robust O/C only. Check out the examples tabulated at page 3, incl. the adjoined text, and you'll have guidelines on tweaked settings - as a starter. Unfortunately, there is a variability in O/C capacity as a function of NB/SB stepping, even down to the individual board level. Therefore, you'll have to do the tweaks on your specific board to find the appropriate robust setting. Revsion 2 boards and updated BIOS versions will though make O/C easier than it initially was.

Cheers,
Anders Fahlén
Agronova
Sweden

TeleCarlos
05-08-2003, 03:54 PM
Well, I have a problem. My setup recently developed a bad case of the crackling problem while playback and while playing vstis or Reason

My setup:
Asus A7N8x Deluxe (ver.????)
Bios..???...as a 1001.G sticker on Mobo...also on the boot up screen
AMD 2100 xp (1.73 ghz)
Corsair DDR 2100 512mb
Asus V9180TD 64mb (GeForce 4 MX 440 8x)
Delta 66 w/Omni I/O (using 5.00.27...driver)
Edirol PCR-50 (driver is the only one at website)
Vantec aeroflow CPU fan
Antec True Power 430 watt psu
Western Digital 40 gig 7200 rpm (2mb buffer, OS & progs)
Western Digital 80 gig 7200 rpm (8mb buffer, for Audio files)
Plextor 12/10/32 cdrw

Software:
Win xp (w/SP1...came in the CD)
Cubase 5.1r1
Reason 2
Fruity Loops 3.... never use it anyways
Easy CD creator 5 Platinum

Disabled:
Did most of the XP tweaks
LAN, firewire, etc. disabled in the bios
No Modem installed

Delta 66 in IRQ 5 and not shared with anything else (I believe...so says the booting up screen)

This should be a kick ass system and it was for the first six weeks. All of the sudden it has developed this crackles and pops. No new software or anything installed, so it really is strange.

I'm no computer whiz but could it be any of these????:

-A bad RAM stick??
-Some PCI bus transfer data rate thing...whatever it is...???
-Do I need newer BIOS....and what version??
-Need for better / newer USB drivers....???
-Should I try older Delta drivers???
-Need updated drivers for the Mobo??

These a few that came to mind, but who knows?
Anyway, I'll appreciate your input since you are the Nforce2 expert..

Thanks in advance.
Carlos

maggotcontrol
05-09-2003, 07:16 AM
Let's look at the obvious things first..
What's the Asio buffer/latency set to? It's never too low for any system! (or too high?)

Is the BIOS properly configured? If you're not sure, list all of your settings.
You should install Windows XP in ACPI mode on these boards. It'll create more IRQs as needed. Don't sweat the on-board stuff on these boards. They haven't been an issue with the Delta cards on that board.
Try the 4th PCI slot. (2nd from the outer edge of the board)
Kill that silly Easy CD creator program and upgrade to Nero burning ROM.

Yoshi
www.studiokobari.com (http://www.studiokobari.com)

fahlen
05-09-2003, 07:43 AM
Apart from what Maggotcontrol suggested:

- Have you optimized the OS for "Background services"?
- What menu latency setting are you using?
- Do crackles disappear if you increase latency setting?

Cheers,
Anders Fahlén
Agronova
Sweden

nakedjackson
05-09-2003, 12:21 PM
maggot, what do you mean its never too low for any system??? Also, i have an aardvark q10 and cubase sx. The q10 allows me to change asio latency in the q10 control panel, and cubase allows me to change asio latency in the asio settings control panel (c drive->program files->steinberg->asio->asio settings) what is the relationship between these two settings. Also is it possible to change the asio settings while a cubase session is running?

EDIT: Maggotcontrol, why do you say to install in ACPI mode? isn't that crappy for DAW's? Everywhere i read says that standard PC is better for DAW's. There's even an article on the Steinberg website entitled "ACPI Kills Audio Performance"

[This message has been edited by nakedjackson (edited 05-09-2003).]

TeleCarlos
05-09-2003, 03:50 PM
Thanks alot for the heads up. In order of appearance, this it:

-Used it for a couple of weeks with latency around 254 buffer/6ms.latency in the Delta Control Panel and it worked. Did change it to a higher(378, then 512) and it still cracks.

-Acpi; Installed XP like all the guides suggested, ACPI disabled

- Background services, yes the box is checked

-Menu Latency....Don't know what that is???

As previously stated, it WORKED LIKE A ChARM for 2 months, then developed the crackles.

Ireally appreciate your suggestions and will be trying anyways those ideas first on Sunday (no time right now...).

My possible solution: get new HD, re-install again using newest drivers available, beat it to death till it cracks!!!
If it dosn't crack, then it works.

Where can I find my version of the board (in what part is it printed on?) Which BIOS is everyone else using?? Which USB drivers??? USB 2.0 drivers???

I'll be back Saturday to see your replies.
Thanks to everyone.

maggotcontrol
05-10-2003, 09:40 AM
Originally posted by nakedjackson:
maggot, what do you mean its never too low for any system??? Also, i have an aardvark q10 and cubase sx. The q10 allows me to change asio latency in the q10 control panel, and cubase allows me to change asio latency in the asio settings control panel (c drive->program files->steinberg->asio->asio settings) what is the relationship between these two settings. Also is it possible to change the asio settings while a cubase session is running?

EDIT: Maggotcontrol, why do you say to install in ACPI mode? isn't that crappy for DAW's? Everywhere i read says that standard PC is better for DAW's. There's even an article on the Steinberg website entitled "ACPI Kills Audio Performance"

[This message has been edited by nakedjackson (edited 05-09-2003).]

For nForce 2 and SiS 735/745 boards, ACPI works better with Win XP unless you have a rare conflict. You must've been reading a 10 year old article.

Find the general sweet spot for asio buffer/latency for the Q10 via the control panel and then let Cubase re-auto detect. It's not a good idea to change buffer sizes while a project is running. If you're not monitoring realtime input through the Q10, you can raise the latency as it won't affect your use.

yoshi
www.studiokobari.com (http://www.studiokobari.com)

fahlen
05-10-2003, 11:11 AM
TeleCarlos,

A few suggestions:

- start with a barebone setup (no unnecessary usb/pci devices installed), assuring that the audio card is installed in either pci slot #2 or 4 (no sharing of physically hardwired irqs), and using latest BIOS version with default settings;

- disable any unneccesary, competing services running in parallel with your audio program;

- again, check that you don't have any active usb device installed via the on-board usb port while doing the test (there are also problems with skipping audio while typing reported with some PS/2 keyboards - solved by using a PS/2 to USB converter);

- confirm that your memory sticks work without errors by running a complete Memtest86 cycle;

- completely uninstall your driver and re-install the latest Delta 66 driver (even if this means re-installing the same version);

- check, if possible, with an alternative soundcard if the problem prevails;

- check that you have the latest Nvidia Standard IDE driver and not the Nvidia TM IDE driver and that your IDE drives have UDMA enabled

The motherboard revision number is printed on the PCB.

Cheers,
Anders Fahlén
Agronova
Sweden

TeleCarlos
05-15-2003, 01:06 PM
Sorry for the delay,

My DAW seems to be ok now. I think it was the fact that while messing around, I put the hd with the OS in the secondary IDE port...I thinkh. Anyway, put it back where it was and now seems OK.

By the way, the version of my mobo is 1.04...it is printed reeeaaally small.

I'm not going to upsate anything since it is working fine for know.

Regardless, it is good to know we have experts at hand for anything in particular in this forum....

Cheers everybody!!!

Michael Quayle
05-21-2003, 06:11 PM
I'm having trouble getting my Barton 2500+ much more than 200Mhz beyond it's core speed, although many many people report at least a 400 mhz gain through clocking. My signature contains my setup, and I have the CPU set up with an FSB of 206mhz and a multiplier of 10x (I can't seem to get ANYTHING from any higher multiplier). RAM timings for the Corsair 3200 (NOT the LL version) are currently 7 3 3 2T. If anyone has any suggestions as to how I can maybe get this higher I'd be very grateful. Ideally, I'd like this thing running at around 2200 mhz. Hopefully that isn't too unreasonable an expectation of these chips.

Regards,

Mike.

Michael Quayle
05-21-2003, 06:13 PM
apparently signatures aren't currently enabled, so here is my setup:

AMD 'Barton' 2500+ clocked to 2.06Ghz with Thermalright SK7 Heatsink
Asus a7n8x Rev 2.0 MBoard
Corsair 512mb XMS3200 CAS2 RAM
Enermax 350w PSU
ATi Radeon 8500 64mb AGP card
WD1000 BB 100GB System Drive
2x WD1200JB 120 GB Data Drives
Teac 24x IDE CD Burner
Coolermaster ATC201 Case
Motu 828
Mackie HR624
Contour ShuttlePro
Logic 5.5
Nuendo 1.6
Wavelab 4.0

maggotcontrol
05-21-2003, 08:18 PM
Michael, 206MHz is pretty much the wall in terms of FSB overclocking on a Barton at a basic level. You're doing quite good really. If you really want to get dangerous, you need to increase the CPU Vcore to 1.7V-1.75V rather than the default 1.5-1.6V which is not enough for serious overclocking.
Note: as you do this, you are significantly increasing the CPU temps since more voltage is being applied. Make sure you have GOOD cooling and watch the CPU and mobo temps.

I'm sure with some careful tweaking and better cooling, you can get 220MHz FSB and a big multiplier.

Are you running the memory in sync with the CPU? If so, what's the memory's rating? If you want to go beyond 200MHz, you may need PC3200 even with Corsair and also increase the VDimm.

I've got an XP 1700+ running at 200x11 - or 2.2GHz right now. That's a healthy 730MHz overclock and I'm sure it could take more.. although the system would get louder.

Grab the Zalman 3100CNPS heatsink with a more powerful 92mm fan and make sure it blows over the northbridge as well as the CPU and see if you can get better results.

As with all nForce 2 boards, be careful not to 'save and exit' after too many BIOS parameter changes. There are many people who have corrupted their BIOS this way and is a known issue. There's just not enough CMOS memory to save all of that data at once. Make 1 to 2 changes and save/exit. Then go back in and do more parameter changes. This is the only safe way of working with these boards at the moment that I'm aware of.

Yoshi
www.studiokobari.com (http://www.studiokobari.com)

Michael Quayle
05-22-2003, 03:57 AM
"As with all nForce 2 boards, be careful not to 'save and exit' after too many BIOS parameter changes. There are many people who have corrupted their BIOS this way and is a known issue. There's just not enough CMOS memory to save all of that data at once. Make 1 to 2 changes and save/exit. Then go back in and do more parameter changes. This is the only safe way of working with these boards at the moment that I'm aware of."


So THAT's what happened! That's exactly what I did yesterday and I thought I had fried my machine. It just would not POST after I F10'd. Thought I'd REALLY screwed things up, but updating the BIOS seemed to do the trick. Nice to know it's a 'feature' of the board and not just me. I would happily lower the FSB if I could get a higher multiplier off the chip, but I simply don't seem able to manage it. As for the memory (which IS 3200 Corsair) I looked and looked for the sync option, but all I could find was 'Auto' and 'by SPD'. I think I currently have it set up as 'by SPD'. Also, will the result of the raise in CPU vcore be reflected in the POST report processor speed, as my minor experiments into vcore adjustments seemed to make no differences there. My CPU is currently running at 42 degs under stress so I'm certain it can perform better, but 2060 mhz is the current highest stable rating I have had off this config.

Thanks,

Michael.

maggotcontrol
05-22-2003, 06:05 AM
I see, you're in the non-expert mode. If the memory is set to auto, it usually is in sync with the CPU in DDR mode. You'd get better performance by increasing the FSB rather than a higher multiplier. It's not the total MHz that matters. It's how wide that data path is between the CPU and memory. The FSB is the bottleneck here.

The proper way to overclock IMHO is to crank the FSB with a lower multiplier. Then when you hit the max FSB you can achieve with good stability, you can begin to increase the multiplier. Often you won't be able to hit the processors native multiplier because of many factors.

About Vcore, don't raise it above 1.8V or you may risk damaging the CPU.

Yoshi
www.studiokobari.com (http://www.studiokobari.com)

Michael Quayle
05-22-2003, 07:58 AM
That seems to confirm what I have been finding. After a lot of testing and note making, I have found the most stable and highest clock for my system to be 10x at 206mhz with the memory in sync with the FSB. This provides a solid 2.06 Ghz (which isn't the fastest 1830mhz core chip in the world), but I can run it at 1.650v core and it runs at about 38 celcius. When I clocked the chip up to 2200mhz with a slower FSB, I ended up with lower benchmarks results, so I have to concur that the FSB clock is the all round stronger option, especially given that this Asus board locks the PCI bus to 33mhz.

nakedjackson
05-22-2003, 04:01 PM
maggotcontrol, if i reinstall windows with acpi with standard pc as you suggested, what kind of performance increase will i expect? Is it worth reinstalling windows?

maggotcontrol
05-22-2003, 05:03 PM
Well it can resolve many issues regarding IRQs since it'll allow more than 15 IRQs in a system, so perhaps somewhere along the line you may see a boost in performance or fix glitches since it can eradicate shared resources. A simple reinstallation is all that is needed (without pressing the F5/F6 during the blue setup screen) so perhaps it's worth it.

I've yet to be in a situation where Standard PC mode was necessary with the nForce 2 boards as well as the SiS 735/745/746 boards. It's been awhile since I touched Via so I can't comment on those boards.

What kinds of problems are you having, if any?

Yoshi
www.studiokobari.com (http://www.studiokobari.com)

fahlen
05-23-2003, 12:18 AM
There have recently been some interesting discoveries with the APIC controller on nForce2. It has been found that the APIC controller performs with performance hits on some I/O intensive applications (probably due to higher latencies for individual APIC-controlled devices with APIC mode enabled and altered irq routing & assignment). With APIC mode enabled there seems to be some rather intense irq sharing between ide, usb and LAN whereas lower, odd numbered irqs are not used by the APIC controller at all. For example, the tested nForce2 (revision 2 M/B) was found in some applications to perform with up to 20% improvement in ACPI PC HAL (ACPI + PIC) rather than ACPI Uniprocessor HAL (ACPI + APIC). It remains to be verified what gains are real in digital audio.
http://www.lostcircuits.com/cpu/amd_xp3200new/ http://www.planet3dnow.de/cgi-bin/newspub/viewnews.cgi?category=1&id=1052834357

Cheers,
Anders Fahlén
Agronova
Sweden

maggotcontrol
05-23-2003, 05:41 AM
Interesting Fahlen, that's something I can test out and verify sometime soon. I'll post results when I get the chance.

Yoshi
www.studiokobari.com (http://www.studiokobari.com)

fahlen
05-23-2003, 06:06 AM
Yoshi,

That would be great, i.e. to verify if differences in digital audio are real and not just artifacts of benchmarks used. I'm right now swapping configurations as to allow for a test of the coming nForce3 + Opteron:-) If approved, this will be the next nForce-related audio test to carry out.

Cheers,
Anders Fahlén
Agronova
Sweden

nakedjackson
05-23-2003, 09:40 AM
whats the best ram for n force3 / opteron systems? i have corsair xms3200 Is that good?

maggotcontrol
05-23-2003, 10:09 AM
Corsair LLPT (low latency platinum series) worked great at high FSB and in Dual channel mode on nF2 boards. There are a few variations of the XMS series. The cheaper ones weren't stable at 166FSB dual channel. I don't know about nF3 yet but it should be similar.

Yoshi
www.studiokobari.com (http://www.studiokobari.com)

fahlen
05-23-2003, 10:51 AM
I don't think there is any definite memory compatible guide yet for the nForce3 150/250. Anyway, the on-board memory controller is gone. It is now an on-dye Opteron solution that will bring all time low latency values at the forefront. The separate NB/SB chipsets are also gone. This represents an integrated single chipset solution.

There is some good summarized technical description at http://www.bjorn3d.com/_preview.php?articleID=268

Cheers,
Anders Fahlén
Agronova
Sweden

wogg
05-23-2003, 12:07 PM
It appears that the nForce3 will be utilizing the Opteron on-die memory controller. The Opteron supports registered ECC DDR RAM only. Standard consumer grade chips will not work.