PCI-E overclock?

W

Willsonman

[H]ard|Gawd
Joined
Jun 11, 2005
Messages
1,195
anybody try to increase this with clockgen or the like? Is it linked to anything? Is it safe to just fiddle with?
 
I would imagine it's like any other bus. You probably should leave it alone....AGP/PCI being out of sync gives you issues. I don't see why OCing the PCI-E bus would be any different.
 
You can increase it slightly which will improve your video card performance (it's a feature of some ASUS boards iirc.) I think anything past 3-5% is asking for trouble though, and it's normally much easier just to overclock the video card directly.

It might be linked to the SATA clockspeed, so increasing it could give disk corruption (I might be wrong on that.)

If you're running any other pci devices (raid card or TV card) it's likely going to screw up their performance though.
 
If you do it please post the results here in first hand.

Thank you.
 
Does thi adjust your card speeds are the PCI e freqeuncy??
 
I'll try to be as thurough as I can.

I'll give some rig settings just for info sake. I'm using my sig rig. The CPU is running at a 8.5 multi and 306 HTT with teh HT link at 4X. The ram timings are at 2-3-2-6. Since this is just for experimentation purposes I UNDERCLOCKED my GPU to a core frequency of 325 and mem frequency of 425. Its an R480 core.

Using ATiTool I started a scan for artifacts and bumped the PCI-E frequency using clockgen 1 mhz every 30 seconds and stopped at 105mhz. I let ATiTool run for another 5 minutes just be on the safe side of things. I ran 3dm06 with all the standard settings and got 1477 at the stock 100mhz on the PCI-E bus and 1479 at 105mhz. Not too much increase and I did not notice any more heat than usual at idle or under load after the 3dm06 ran.

I may play a little more with this at default GPU settings and maybe a little more mhz on the bus... just wanted to update.
 
There have been a number of cases where I've seen the PCI-E clock help out a GPU oc when it (the PCI-E bus) is incrased...but do a lot of before and after testing and watch for general system stability...as it's not always a gaurantee that it'll get you anywhere.
 
just to confirm cornelious....

I just ran the gpu at 400 core and 490 mem. 100mhz PCI-E got 1736 3dmarks and at 110mhz got 1739 3dmarks.
 
Willsonman said:
100mhz PCI-E got 1736 3dmarks and at 110mhz got 1739 3dmarks.
Click to expand...
thats easily within the margin of error

im sure if u reran the test a few times at 100mhz you could get that 1739 or vice versa
 
Willsonman said:
just to confirm cornelious....

I just ran the gpu at 400 core and 490 mem. 100mhz PCI-E got 1736 3dmarks and at 110mhz got 1739 3dmarks.
Click to expand...

I'd say that was within the margin of error of 3dmark.
 
Yeah, thats what I thought and thats why I went to 110 on the second try. Do you think the second set of results iswithin as well?
 
Willsonman said:
Yeah, thats what I thought and thats why I went to 110 on the second try. Do you think the second set of results iswithin as well?
Click to expand...

Yes, more than likely it is...even 3DMark03/05/06 are sensitive enough that it's very easy to get the scores moving around more then 10 points or so depending on background services and whatnot you might have running...so I would almost (pretty much would) put money on the fact that the PCI-E "overclock" didn't do a thing for you.
 
Willsonman said:
Eclipse? any thoughts?
Click to expand...
i'll mess with it when i get my VERY FIRST pci-e card.. tomorrow hopefully :D

what tests would you like to see done, change in performance and change in overclock, if any?
seems like the performance difference is pretty small for now, but we'll see.
 
(cf)Eclipse said:
i'll mess with it when i get my VERY FIRST pci-e card.. tomorrow hopefully :D

what tests would you like to see done, change in performance and change in overclock, if any?
seems like the performance difference is pretty small for now, but we'll see.
Click to expand...

I'm pretty sure I've seen it effect OC'ability a little moreso then straight out performance in the past.

Haha...so I'm not the absolute last person on [H] not to have a PCI-E card. :p
 
Pretty funny that two big participators of this forum are still in AGP land. Just curious what you are getting eclipse?

I would like to see what performance increases are out there as far as changing it. I'm not exactly sure what is a safe limit to go. I also do nto know what all runs on the PCI-E bus... how do those components react to it being a little faster? Is there a better "benchmark" (for lack of a better term) for determining stability and a real increase in performance?

I'm all up for helping to look more into this but I think I'm a little in the dark. Mostly I dunno what the bus controls. I guess I can look into that if you fellas dont want to just give me a short version. I have summer break from school so I have lots more time to do stuff. I hate the heat in NC so I'm always inside. Any ideas let me know.
 
Just looked around a bit and found that is just controls basically video cards, SATA I & II, and ethernet. anything else?
 
Willsonman said:
Just looked around a bit and found that is just controls basically video cards, SATA I & II, and ethernet. anything else?
Click to expand...

Isnt SATA notoriously picky about OC's..Dont feel like reinstalling windows.. :p
 
g1xx3r said:
Isnt SATA notoriously picky about OC's..Dont feel like reinstalling windows.. :p
Click to expand...

Yes...it is...so be careful what you do (anyone)...'cus it's really not worth re-installing XP for the marginal gains that this "oc" might net you.
 
yeah... thats why I am doing this on my 32-bit windows. My regular 64-bit runs off of my SATA drive so I'm not really risking too much by doing it.
 
Willsonman said:
yeah... thats why I am doing this on my 32-bit windows. My regular 64-bit runs off of my SATA drive so I'm not really risking too much by doing it.
Click to expand...

Still...be carefull bro...wouldn't want to see anything happen.
 
Considering no card fully saturates the PCIe 16x lane so far. Woudnt OCing the bus make no difference at all?
 
Postal Dude said:
Considering no card fully saturates the PCIe 16x lane so far. Woudnt OCing the bus make no difference at all?
Click to expand...

You'd think so...but it works differently enough from that that I'm honestly not sure exactly how it does/can effect things positively...all I know is that I've seen people achieve higher GPU oc's after raising their PCI-E clock...don't ask me how, but sometimes it does work. :p
 
Maybe ths routes more power to the card, so it would only work on cards without a 6-pin connecter?
 
Willsonman said:
Pretty funny that two big participators of this forum are still in AGP land. Just curious what you are getting eclipse?

I would like to see what performance increases are out there as far as changing it. I'm not exactly sure what is a safe limit to go. I also do nto know what all runs on the PCI-E bus... how do those components react to it being a little faster? Is there a better "benchmark" (for lack of a better term) for determining stability and a real increase in performance?

I'm all up for helping to look more into this but I think I'm a little in the dark. Mostly I dunno what the bus controls. I guess I can look into that if you fellas dont want to just give me a short version. I have summer break from school so I have lots more time to do stuff. I hate the heat in NC so I'm always inside. Any ideas let me know.
Click to expand...
i was actually in pci land. pf4mx ftw ;)
got a 7600gt, though it's got a few quirks that need to be worked out, namely ram contact issues with the heatsink. gotta bug MSI a bit about it.

a good benchmark for stability would be anything that uses the video card really.. though you should make sure it isn't affecting the pci bus by copying large files and getting an md5 checksum on them to check for corruption
 
firewolf said:
Maybe ths routes more power to the card, so it would only work on cards without a 6-pin connecter?
Click to expand...

Not true in any means...but following what (cf)Eclipse stated earlier would be a good idea if you are going to be "toying" with the PCI-E clock.
 
(cf)Eclipse said:
i was actually in pci land. pf4mx ftw ;)

a good benchmark for stability would be anything that uses the video card really.. though you should make sure it isn't affecting the pci bus by copying large files and getting an md5 checksum on them to check for corruption
Click to expand...

WOW :eek: PCI land. just.... WOW

How large is large enough... 1GB or so?
 
Willsonman said:
WOW :eek: PCI land. just.... WOW

How large is large enough... 1GB or so?
Click to expand...

That should be fine...if you have any full length movies on your comp (not that anyone at [H] believes in warez/pirated software... :rolleyes: ) then something of that nature should also be sufficient.
 
Yeah I was thinking of just making an iso backup of one of my own movies and using that using dvdshrink
 
Willsonman said:
Yeah I was thinking of just making an iso backup of one of my own movies and using that using dvdshrink
Click to expand...

A backup image would definitely be a good idea...reminds me...I really need to start looking at picking up an external hard drive bigger then my internal hard drive...or an external hard drive period...I hate having to partition my main drive to backup information but never having enough room to save everything I want...arg...oh well. It'd be easier to install a second internal drive, but wiring comes first, and I don't NEED the extra internal drive...just want the extra space for temporary backup...money has to be spent I guess.
 
ok, I've got some spare time now so I'm gonna do some tests for file corruption when I OC the bus. I'll try to keep some pretty detailed results.
 
Willsonman said:
ok, I've got some spare time now so I'm gonna do some tests for file corruption when I OC the bus. I'll try to keep some pretty detailed results.
Click to expand...

Be sure to report back, I'll be interested in taking a look at the results.
 
Alrighty, I'm gonna take a break for now... got some other things to do but so far.....

Started off defragmenting with O&O. Tranfered an iso of the fugitive with harrison ford from my storage drive(PATA133) to my Speedy drive(SATAII) then used Advanced Checksum verifier to do an md5 check. I did this in 1 mhz increments up to 117 mhz for now. no errors in checksum and all appears stable in the atitool atifact scanner.

More to follow later...
 
Willsonman said:
Alrighty, I'm gonna take a break for now... got some other things to do but so far.....

Started off defragmenting with O&O. Tranfered an iso of the fugitive with harrison ford from my storage drive(PATA133) to my Speedy drive(SATAII) then used Advanced Checksum verifier to do an md5 check. I did this in 1 mhz increments up to 117 mhz for now. no errors in checksum and all appears stable in the atitool atifact scanner.

More to follow later...
Click to expand...

Good stuff so far, can't wait to see this finish up.
 
yes, im watching this thread.
seeing results should be interesting.
 
Well, another update. I've not hit a wall yet, which is the good news. I'm only at 130mhz :eek: Not that I know what is really expected here but I'm pretty excited about it! No corruption of data and I scanned for artifacts at stock frequencies for the video card for 5 minutes and turned up nothin.

The bad news is that I cannot test for better overclocks of the video card for a couple days. My zalman ramsinks have been transferred too many times so I ordered new ones from the egg. Anyways, things still look good so I'll let this sit for awhile and come back to it later.
 
i'll try the overclocking bit in a while, got something i need to write up first :p
 
(cf)Eclipse said:
i'll try the overclocking bit in a while, got something i need to write up first :p
Click to expand...

No sweat man... I know you have a lot on your table.
 
You must log in or register to reply here.
Back
Top