8RDA+ -- high FSB+ATI=BAD, high FSB+NVidia=Good, WTf?

[sotti]

I've had a GF3ti200 in my system since I got my 8RDA+

I've been running 188x12.5 and it's stable in prime95 and never had a crash in a 3d app

Loaded up a 9700pro this week.

got crashes out of 3dmark constantly, rarely would complete a full benchmark.

The crashes were only to desktop.

I fixed the issue by lowering my FSB to 175.

AGP frenquency was locked at 66mhz
8x and fastwrites where enabled, anybody have any clue about this?

[flytek]

tested a 9500 this week. same thing here. boards no longer do 200 and 3dmark. not enough time to read but i recall no solution at this time. others have much more expirience. no issues with the thinner 9000pro though. it seems the external power supplies to the high end ati's are not working for this purpose anyway.

[sotti]

hopefully someone will come by with words of wisdom

[Flymo]

DISABLE fastwrites in bios with the radeon.

Also, Radeons do not like ANY trace of nvidia drivers in the system.

Look for instructions to remove EVERY trace of drivers or format/reinstall.

My Radeon 9500 modded to 9700 running at 375 core, 333 mem runs flawlessly @ 197 fsb - the board craps out here - I guess I need cooling on the mofsets.

Flymo

[sotti]

but fastwrites can hold you back, I'll check into that.

[Flymo]

Also, increase the AGP voltage in the BIOS one step at a time until you get stable.

diable Fastwrites and set Vagp to 1,6V, that should be enough.
I'm perfectly stable at 225Mhz FSB sync with my modded radeon 9500.

iam to stable at 200fsb with my 9700pro and my 8rda board.. now i did have to bump the agp voltage up a notch but that was all i did.. no problems was able to run toast for 24 hrs straight before giving up cause i was satisfied it was stable at the 200 fsb...

[kiljaden5]

I read somewhere a while ago.. an ATI tech said that increasing Vagp to 1.7 would be a good idea, also make sure that the card is plugged into a cable straight from the power supply and doesn't go through other devices

think that should help

[nfrenay]

I would suggest you to be careful with Vagp. If you need to up it, try to do it as little as possible (1.6V for example). As far as I know, this voltage goes to the video card's own power regulation system, so there is no big advantage in overvolting it, unless you're volt-modding the video card or the motherboard can't provide a steady output.

good luck!

[Rudzer]

per - how the hell you managed to get 15.5k with your Radeon9700PRO at only 2Ghz?

[gryphyn]

Rudzer,
I would hope that his 3Dmark would be 15k or more since I'm about the same score with older generation boards at a smaller mhz

[jamie hayman]

Originally posted by flytek
tested a 9500 this week. same thing here. boards no longer do 200 and 3dmark. not enough time to read but i recall no solution at this time. others have much more expirience. no issues with the thinner 9000pro though. it seems the external power supplies to the high end ati's are not working for this purpose anyway.

I have no problem running at 12X203=2436MHz with my 9500pro...
check sig.....
Fastwrites are disabled
maybe you need a bit more Vcore
My board is not VDD modded yet..
Full load temp playing Nascar2002 is 49 degrees..

[muzz]

As has been stated ( fly) MAKE SURE ALL vestiges of NV vcard drivers are removed from your system, I have seen that cause more problems than anything else.

I've run an 8500 and a 9700p at 238 fsb without issues on 2 different 8RDA+ boards......
My guess is the problem lies elsewhere, unless you have a bad card.

[flytek]

finally got the 9500 up to 200 fsb on the desk. had to swap many componants and run many test over two days. my final recommendations.
do a full system reinstall even if its at 200fsb
make absolutely sure that the cpu and memory are up to speed and stable
stock voltage to memory and agp should suffice if the componants are up to the task.
run 98se for best visual quality and 3dmark stability & performance.

testbed specs 10630 3dmarks, 98se dx8.1, 0 mods yet

stock 8RDA+ 1.1, athlon xp1800+ 307 dlt3c @ 2.2ghz & 1.625V.
2x256mb ddr333 @ 100%7-3-3-2.5, ati9500 dvi, 300watt psu.

[Wildwood]

Had same problem....but found it to be a driver issue...i went back to the Omega's 2.3.92's and it stopped doing it!!!!

[flytek]

ran prime95 all night @ 2.2ghz 1.65V. it will run 3dmark as well time after time except as soon as i set 3dmark to loop it will crash out to the desktop. restarting 3dmark will run smoothly unless set to loop. must have run 3dmark 20 times already with prime 95 still running. when looping it crashes on the second or third loop. could this ddr333 be failing only long term 3d?
any suggestions before i switch off tests and change ati driver?
any hot tips to turn this 9500 way up would be very welcomed???

[Flymo]

run memtest overnight to make sure the mem is ok at that fsb and mem timings.
My twinmos/winbond runes fine @ 195fsb and 3.2.2.2

I need to get some heatsinks on the mosfets tho and active cooling on the northbridge to increase fsb


Flymo

Originally posted by Rudzer
per - how the hell you managed to get 15.5k with your Radeon9700PRO at only 2Ghz?

i got 16137 with this xp1800 oced to 2100 mhz 10.5x200 1.85v.. heck im trying for 17k but i seem to be stuck... all ive done to this set up is the active coolers on the nb and sb and syncs on the mosfits.. i oced the card to 344/324 and thats about it really..
here is the screen shot of last nights benchie this is my highest so far with this xp1800 oced it's the best i can do right now...

[flytek]

funny thing that prime95 will run for 24 hours when a few loops of 3dmark will bring a cpu to its knees at the same settings. the ddr333 on the testbed only does ddr400 @ 9-4-4-3.0 reliably. so with dodgy memory and cpu my stability was wishfull thinking.
with the new memory timings and a 2ghz cpu @1.5 Vcore everything loops fine at 200fsb. i think i'll stick with these cpu setting till i replace this unlapped aluminium heatsink.

edit. learnt my lesson. every day stability is not 3d stability. now totally stable with 2x256 hyperx3000 one bar code apart.