Who knows something about PCI multipliers on NF3?

[Radical_53]

Hi!

Now, my new platform seems to run. But, there's still a problem. FSB233 is running flawlessly, as I heard rumors there would be the first divider to get PCI back to 33MHz.

Does anyone know, especially on my Shuttle Board, if there are other multipliers and where they are?

I'd like to max out my RAMs which are running at 1:1 with 2.7V right now, but I don't want to ruin my Raid0 setup.

[Penti]

1/7 PCI divider? I dont think so. Only 1/6.

[Radical_53]

Only as far as I know. How else could some people achieve 300+? Can you explain that? Would bei 50MHz+ on PCI, and I'm sure no HDD or card could do that.

[Penti]

My Seagate Barracuda IV can do 50MHz PCI without corruption, thats at ATA100 so i'm sure it can do much higher even.

But they run a 66MHz pci ide-controllercard to run that speeds, then they dont have to worry about PCI bus as much.

And it has asynchronous/fixed AGP.

[Radical_53]

Yes, I know about AGP. That's one for good at least.

But: Do you know if every PCI-controller card is 66MHz-capable? Won't be, right?

I'm just curious about my Raid and Audigy2. Can't go without those two.

[Penti]

Naw not all controllercards are compatible with the 32-bit/66MHz PCI specs. But for example Promise TX2 are.

Check the datasheets on the products you have it will be written there. Soundcard dont support it though i think.

[Radical_53]

Well if the sound can't take it, I won't have to look for that. At least, I do know that my sound card can do around 36MHz PCI at least.
So 250MHz FSB would be at least possible, which won't be too shabby, also my RAM could do that 1:1.
After that, I could still look if there's a multiplier at 266MHz or 300MHz.

[Penti]

250MHz would be 41.66MHz PCI?

Edit:
216MHz would be 36MHz even. So you can test it. Just run something between 216 and 232MHz to test.

[Radical_53]

Can't be. Right now I'm at 233MHz FSB, and there must be a divider, as everything runs just as "normal".
If I had a higher PCI bus at this frequency, my Audigy would be a pain in the ass.
But nothing like that. So from my point, there is a 1/7 divider.


That's what I'm searching for. People that can tell me where those dividers are. I'm sure that there are more.

[buff]

only SB Live had probs at high fsb.
And I wouldn't be saying the is a 1/7 divider just cause ur hardware can take it:) There aint a 1/7, unless Shuttle implemented one separate from the VIA Norm.

[Radical_53]

VIA?

I ain't got no VIA crap, it's the NF3 150! Ok?


I do still believe that there are dividers, as it's running just more than smooth.
And normally, I'm never that lucky... ;) Know what I mean?

[Penti]

You can still do some tests to see if its there or not anyhow. I dont belive it has any 1/7 divider.

[MrIcee]

If you do a search here on the forums, you'd find many topics addressing this very issue.

There is in fact NO PCI lock on the nForce3 chipset. It has been confirmed by several sources(Xbit labs for example I believe) thru the use of a device called the PCI geiger which measures PCI frequency as you scale the FSB.

The AGP is locked. but NOT the PCI bus.

I found this out quickly in my testing of the AN50R, and I was relegated to using my trusty Promise Ultra133 TX2 66Mhz capable IDE controller card in order to scale the FSB up asynch to 340Mhz.

Randi:D

[Radical_53]

Well, yes. I do already know that it's not locked, no one was talking about a locked PCI bus.

I was just talking about dividers, resetting the PCI clock to 33MHz every time it reaches a certain point.

I for myself am sure that there are such dividers. I'm just searching for any evidence to proof this, also others that might know where those dividers take place.

I'm sure that I've read several times that there is a divider at 233MHz, but I'm sure there are more.

Could you tell me something about that?

Just FSB is no use for me, if I can't use my Audigy2 card without sound corruption, or my SATA Raid0 stripe without the data getting corrupted.

[MrIcee]

At 200Mhz FSB the the PCI is at 33Mhz. From there the PCI bus scales upwards as you increase the FSB. Thats a fact substantiated by testing with PCI geiger.

Reference to dividers at 233Mhz have many times been mentioned in reference to the VIA based A64 solutions, but those boards also have no other divider other than 1/6th beginning at 200Mhz.

I haven't seen mention of dividers here at Xtreme in reference to the nForce3 boards tho it could have been and I missed it.:)

Randi:D

[Radical_53]

Well, I guess I'd have to look that one up, but I'm sure I read about it.

Had the same thing with my KR7-Raid back in the times. Everyone said it had no other divider, but how else could I've achieved ~180MHz with a board that has a 1/4 divider?

I'll check that about the 1/7 on the Shuttle...

[MrIcee]

I reached 200Mhz FSB on the KR7A-Raid for benching 3Dmark back in the day, and OPP and I were the first capable of benching at that FSB speed. He and I both ran either RAid or IDE controller cards for the hard drives, and yes..the Nvidia and ATI cards at the time had a rough time benching at 100Mhz AGP frequency:)

Randi:D

[Radical_53]

Well, I used that 24/7, not only for benching ;)

I guess that's the difference. If my only interest would be benching, there wouldn't have been any problem with this.

But, I do have critical data on my raid, and I don't want bad sound from my Audigy.
So, PCI mustn't be too high.

I do always limit myself to like a 90% overclock, but which is stable for the whole time of usage. Lifetime should be around half a year, which is sufficient for my cause.

[Holst]

Your still all assuming that the IDE controller is running on the same clock as the PCI bus.... I suspect it is not.

[Radical_53]

Well I don't care about IDE.

My sound card is bound on PCI, my SATA discs also. So why should I care about IDE?

[Penti]

I did myself run 172-178MHz tight timings on a KT266A board (4:2:1 dividers) Onboard IDE-controller 24/7. Did run at 200MHz for a shorter while. But on VIA chipsets the onboard IDE-controller is in fact on the PCI bus. I did have a WD 60GB 5400rpm drive also back then and on this setup wich i runned the Seagate Barracuda IV at 50MHz the WD drive went corrupted when running 37.5MHz PCI and over (If i remember correct the onboard sound on that board didn't like 50MHz PCI ;)). On nforce chipset it is probably i different story, but i've only owned a couple nF2 boards wich did have asynchronous PCI bus. Maybe the onboard ide controller run at a different clock on nF3 150 but that doesnt matter in your case as you said.

The divider limitations isn't motherboard releated its chipset, many kt266a board for example did have a clock generator with 1/6 divider support but that does nothing good if the chipset doesn't have support for that dividers. "Locked"/Fixed AGP/PCI is only possible if the chipset has support for asynchronous AGP/PCI speeds and so on. The nf3 motherboards does have clock generators that support asynchronous PCI. So for me it seams that the nf3 chipset only has support for async AGP and only synchronous pci for some unknow reason. Then i don't see any reason for them to have implanted a higher divider then 1/6.

I have a KT333 board wich i have been running i did run 42MHz PCI for a while 24/7 even with 3 drives. All depends on the hardware, that you run fine at 233MHz doesnt prove there is a divider for it. With a 1/6 divider the speed will only be 38.83MHz PCI wich isn't that high really. I've runned higher then that with the same hardware for everyday use over a year now.

Now i will eat some chrismas ham :D regards.

[Radical_53]

Hm. Interesting. I'll be investigating this thing further, but the more information I get, the more confused I am.

Trying some FSBs higher than 240 tomorrow. The board itself seems to do just fine. Strange.

[s e t h]

backup your critical data. especially c:\winnt\config folder. often it will be the first to corrupt forcing a repair or reinstall. if you replace the corrupted config folder with a backed up one you should be able to prevent wasted time.

[Radical_53]

Well, I can't backup my 150GB on the second partition on my Raid.... ;)

THAT is the problem. My installation of XP got cracked when I first put in the board, the onboard Raid controller seems to be buggy. So, it's fresh anyhow.

But still, a good hint to save time. ;)

[STEvil]

make a 2gb partition for fsb testing.

[MrIcee]

Originally posted by Holst
Your still all assuming that the IDE controller is running on the same clock as the PCI bus.... I suspect it is not.

I hit a logjam in FSB overclocking with the Shuttle board, until I installed the Promise PCI IDE controller card. At that point I was able to go well over 300Mhz FSB asynch, leading me to believe that at least on this board the native IDE controller is running on the same clock as the PCI bus.:)

Randi:D

[Radical_53]

Well, still I don't have anything on native IDE that would be somewhat important ;)
I guess my SATA-PCI card does already help me out somehow.

[Penti]

What SATA-controller card are you running?

[Radical_53]

Got it from Ebay. Just a small card, PCI 2.2, with the SI chip on it.

I can post a picture, if that helps. But no "brand" card.