How high can I safely go with the PCIE clock? (to break G41 ~345 MHz FSB ceiling)

[Assimilator1]

1st yea I know it's an old system, we're both scint , it's my sons rig (mbrd Asus P5G41T-M LX) & I'm upgrading it with a Q9550 & just looking to mildly overclock it. I quickly discovered a 343 MHz FSB wall (343 boots & runs, 344 doesn't boot, CPU will do 3.6 GHz btw), googled it to find out that this is a common problem with the G41 chipset, & a few threads/articles mention raising the PCIE frequency raises the FSB ceiling, which indeed it does! (why??).

Anyway, so setting PCIE to 104 MHz raised FSB to 357 MHz, 106 MHz raised FSB to 365 MHz.

I could go higher but I vaguely recall reading the PCIE bus is very sensitive to overclocking, so, typically, how high can the PCIE bus go?
And am I right in thinking the SATA bus is tied to the PCIE bus? The thing I'm most worried about is corrupting data on the HDD! Naturally I don't want to do that, been there in the old socket 7 days .

[edit] lol, looking at my sig I wonder if that's the same issue with my 2nd rig! (P43 IIRC)

 

[JoeRambo]

You are right to be worried about SATA problems, in one rig after rising PCIE clock, SSD was just fine, but when "storage" WD Green was connected, it was starting to have crazy problems (like rar archives failing checksum test immediately after creation). I would not touch those things, extra 10-20FSB is not worth the risk of silent! corruption on HDDs.

 

[DigDog]

if the fsb is a problem, why not just buy a cheap 775 board; my 965p-ds3 was happy to do 400 and would have done more if needed.

 

[BUnit1701]

IIRC, the ceiling for PCIE frequency before running into SATA corruption is 110Mhz. I dont know that I would want to push that far though.

 

[Assimilator1]

Yea I wouldn't want to cut it too close!

It's at 106 MHz atm & so far it's error free on OCCT at 20+hrs, but naturally that's not testing HDD data. That'll be my next test, I think copying a few 100GBs of videos should be a good test, what do you guys think?
Afterwards I'll test the grx.

What does anyone else reckon a (typically) safe PCIE bus speed is?

You are right to be worried about SATA problems, in one rig after rising PCIE clock, SSD was just fine, but when "storage" WD Green was connected, it was starting to have crazy problems (like rar archives failing checksum test immediately after creation). I would not touch those things, extra 10-20FSB is not worth the risk of silent! corruption on HDDs.

What freq was the PCIE at? [update] He PMed me, ~110 MHz.

Well the extra 22 MHz FSB gets me another ~ 200 MHz on the CPU taking it to 3.1 GHz, his old CPU was a dual core Pentium @3.4 GHz. Although he primarily plays Smite (atm), which uses 3-4 cores so would probably still get a boost with the quad at it's default speed, I'd rather try to get it stable at 3.1 GHz so it's at least a similar performance for games using 2 cores.
But obviously I don't want to corrupt data so I'll want to leave a wide margin .

Digdog
He only bought the mbrd just over a year ago when I upgraded his rig, TBH it wasn't a mbrd I wanted to get him but he had a very limited budget. And the combination of the relatively high prices of P35 & P45 mbrds at the time along with DDR2 prices put it out of his budget, & we had a 45nm dual core to hand which I knew could o/c to 3.7 GHz. DDR3 RAM was far cheaper. That being said I didn't know this mbrd was that bad an overclocker!
Re 965 mbrds, must admit I don't think I thought of that at the time, but they don't support 45nm Core 2s anyway do they? Anyhow it would of been DDR2.

Thanks guys so far anyway

 

[Assimilator1]

No more opinions or experiences on how high the PCIE bus could typically safely go?

 

[BonzaiDuck]

No more opinions or experiences on how high the PCIE bus could typically safely go?

For the technology you're working with, I think the advice was to stay below 110%. I'd follow more conservative advice. If the Q9550 system were just a play-toy, I might do it myself. Depends on what you're using the system for.

But none of my systems are just play-toys. SO in that scenario with your LGA_775-Q9550 rig, I'd live within the limitations of the chipset and motherboard -- leaving the PCIE clock alone. And it's really not worth it to replace the board for that CPU. You could do better by putting the money toward current or last-gen CPUs, board and memory.

 

[Assimilator1]

Ok thx, I have no intention of replacing the mbrd, it was new just over a year ago, oh & he's old enough now to spend his own money (when he's sorted himself out), I'll just advise him & build it. That's not going to happen for some while.

So far I've stopped at 106 MHz PCIE, I've just got to do some data 'stability' testing to see if that's ok. If it is, that'll do.
Oh & it is just a gaming rig but nonetheless I don't want it corrupting it's data.

Btw anyone any idea why it won't boot if you set the FSB more than ~3.4x the PCIE bus speed?

 

[SPBHM]

don't know the details but G41 and G31 don't have the appropriate "lock" or "divider" for the FSB over 333 (or is the ICH7 at fault?) anyway, unlike p35 and others you are forced to OC the PCIE clock over fsb 333 (or 340 something),

I think I didn't notice any problems running at 115 (or was it 110!?) for a few months on my G31+ICH7 board with 2 sata HDs, only stopped using that because I upgraded to 1155 at the time.

 

[john3850]

I ran a average of 120 on a 680i with 8800gt sli no problems but I hurt a volt modded sli 7900gt at 140 PCi_E.

 

[2is]

if the fsb is a problem, why not just buy a cheap 775 board; my 965p-ds3 was happy to do 400 and would have done more if needed.

Have you checked availability of 775 boards lately? You can't really go too cheap if you want to get a decent OC on a quad core CPU and every board I've seen (lately) for 775 has VERY anemic power delivery. Buying a good used board is a better option than a new cheap one.

 

[Assimilator1]

Just doing a hash check (I think that's the right term, lol) of ~200 GB of videos I copied from itself, it's checked 454 out 454 & no errors, so it looks like the HDDs happy at that very mild PCIE o/c .
Using Exactfile incase anyone's interested.

don't know the details but G41 and G31 don't have the appropriate "lock" or "divider" for the FSB over 333 (or is the ICH7 at fault?) anyway, unlike p35 and others you are forced to OC the PCIE clock over fsb 333 (or 340 something),

I think I didn't notice any problems running at 115 (or was it 110!?) for a few months on my G31+ICH7 board with 2 sata HDs, only stopped using that because I upgraded to 1155 at the time.

Ok, but why is it choosing a higher PCIE clock allows a higher FSB to work? you'd think it would reduce the FSB ceiling

Oh & any chance you can pin down what clock your PCIE was at? (old threads, screenshots etc).

John3850
What about SATA HDDs?

Btw re the 7950, sure it wasn't the volt mod that did it in?

 

[john3850]

I had no SATA HDDs problems due to slow speeds of 680i.

 

[Assimilator1]

A? Sorry I don't follow, you said the PCIE bus was at 120 MHz, so it & the SATA bus is overclocked......

 

[john3850]

That 680i board was very slow at moving data memory and such compered to any c2d intel mb.
I always ran the PCI-E at least at I20 or more for years.
When I had a cheap or single video card in it I ran PCI E at 140 with no SATA HDDs problems.
I lost one card from two different sli setups in that board but never had any problems with a single card or SATA HDDs.

 

[Assimilator1]

Ah ok, got ya now, so data throughput was slow on it but you were running an o/ced PCIE, I guess you could be right that it's slow throughput kept the SATA drives safe, seems a bit unlikely though, who knows!

Anyone else run an o/ced PCIE with SATA drives?