How is your Geil Memory module, what is the highest fsb can you achieve in 1:1 ratio? mine i can get is work as 2.5-4-4-7 at 275Mhz under 2.8v, boot to windows successfully with 275*9=2475Mhz, i have not done any detailed testing yet, only use prime95 to run the blend testing, the machine can keep running 1-3 hours before errors. i try to change to 280Mhz, it can not even display the agp screen and i have to reset the bios. i check the online Geil pc4400 it is running at 3-4-4-8 at 275Mhz with the same chips, but if i change to 3-4-4-8 timing, it is also not stable & even worse. 1-2 hour it will fail.i think the reason is that the abit board k8vpro can only support 2.8v mem voltage so far. i am not sure if abit will release newer bios to support higher voltage for memory(the Geil's working voltage 2.6v-3.1v). now the sandra test result for 2.5-4-4-7 275Mhz is 3700+MB/s and 3-4-4-8 3670MB/s both with 84% efficiency. if it can work stably with 275*9, i will use it rather than push the cpu to 2500Mhz+ with a multiplier @ 10 so that i can utilize the Geil High quality memory module(as i find for my cg stepping athlon64, if speed is faster than 2500Mhz, the whole system will be very hard to be cool down so the cpu itself is a bottleneck for further OC instead of the memory sub system).

do you have any idea and suggestion on that? and thank you for the first post to show me the right way to get a athlon64 cpu board and memory although now i have not tuned it to the optimal status.

Keep in touch, it seems only 2 of us in the forum are using athlon64 and interested in system peformance tuning.

http://www.geilusa.com/proddetail.asp?linenumber=24 actually I take some shot for the memory, the code is "GL1GB4000DC". It can be easily recognized in my photos, I also bought them from Bells, so i believe yours are the same ones with mine. GeIL Ultra Platinum PC4000 500MHz Dual Channel DDR Kit Available only in 512MB (2X256MB) and 1GB (2X512MB) Dual Channel DDR Kits. 184pin, Non-ECC, Un-buffered, high performance DDR memory modules. GeIL 3.5ns 32x8 Density DDR Chip. 6 Layers Ultra Low Noises Shielded PCB with Gold 30u" Plating. Optimized SPD for Intel 875/865 chipset Dual Channel DDR motherboards. Platinum Copper Heat Spreader. Lifetime Warranty . CAS 2.5 7-4-4 @ DDR500MHz PC4000 2.6V-3.1V Retail package with clear acrylic case. Newest innovation design Temperature Thermometers allows users to monitor the temperature of their memory modules.

since I add a home-made memory cooling system for this guy. i find the abit k8vpro 2 mem slots are so close to each other. the inside heat is hard to be blowed off. I use an old coolermaster cpu fan to attach on the memory, and a small cable to fasten it. now even @275Mhz full load for memory, i can not feel it warm when i use hand to tough it( of course, i need to clear the static first by touching the casing :-).) if you are insterested my work, i can take a photo of it, after i finish that i feel the cpu temp is also affected--cooled down by around 8-10 degree. now for the system since i will not oc the cpu further and the heat is not a problem for memory, i believe any instability is caused the hardware physical quality and bios settings (also maybe the microsoft OS disability but yet i have not tested the pc under linux or solaris). now the pc sometimes hangs when playing games @275*9 v cpu@1.55 mem@2.80. I think maybe it is caused by the nvidia driver or ms directx but not very sure if it is caused by hardware settings. do u have any ideas? how about your system? so far any system crash or hang or bluescreen or application crash?

most of them are big size, and even I resize I can not find any online pic lib yet.

I am interested in this. thank you for your advice.

is disabled, you can try to turn on them to see if any stability gain and performance drop and decide.Also i find the nv refresh rate lock maybe a problem for this case. and i use powerstrip maybe also a reason for hang. any alternative solution with same function of powerstrip? i can not find the right crack for that. For memory 273Mhz @2.8V 3-4-4-8 with my home-made cooling, yesterday i tested the prime95 memory torture test. I stopped it after around 19 hours without any errors. i find in this speed the best CAS bootable into windows is 2.5 the the best timing i find is 2.5-4-4-10. the 4-4 can not be tweaked further and the command is auto in my setting( it is actually in 2T mode since if i mannul set 1T even can not show agp screen). for the 10 cycle time in 2.5-4-4-10, it is the most efficient way and give you the best performance for L1&L2 cache under sciencemark 2.0, but it doestn't boost any memory throughoutput under sandra. I will test the 2.5-4-4-10 timing today to see the stability.

as far as i know, a lot of ppl got the same problem with me with the 3-D game hang. everytime it happened, the pc stops responding and screen freezed, but it is only at the time i play some 3-D games, the system dump is showing that the bugcheck is "The computer has rebooted from a bugcheck. The bugcheck was: 0x000000ea (0x865b1b38, 0x8695fd20, 0xf7a53cb4, 0x00000001). A full dump was not saved." this message ,accoding to MS official site, is due to the display video card driver programmed badly and make the system into infinite loop. i have tested the prime95 for memory for up to 19 hours no errors, so i feel if the cooling is good, it will not be the memory problem but the V-card driver. also i suspect the via 4 in 1 drivers, since i find a lot of reports about this error, regardless it is a nv or ati card, seem all via chipset range from kt333 until kt800**. hope you can find out some solution.

will change only at the time you scrub and pull it, after that the shape will be fixed. i think unless it is more than 100 degree else the sticker will not change the shape. so far it is quite ok for me, i check it quite a few times, very solid :-) you need think over about the fan if you want to buy one because you need plan the size and the speed of the fan. mine is CM one 5200 rpm, i feel 2000-4000 will be ok because the heat over Mem is not that much while comparing to cpu. i will try 280-300 fsb tonight---sounds crazy.

280+ for Mem can not start windows and sometimes needs clear CMOS, 279 can go in but the login windows fails, 278 can login but memory testing fail. for both 278&279, they fail at BSOD and tell me the memory management errors. 277 can do most of the things, but not stable. conservative setting is 275Mhz(2 Mhz less than the 277) at this time, even can use 2.5-4-4-7 aggressive timing. at this speed in my rig, it is stable and can pass the prime95 memory test but the cpu will fail after around 1 hour. when playing games, the nv_disp.dll error will come out.@ 273Mhz(274 Skipped) can pass both cpu and mem prime95 test (more than 5 hours testing for both). even so @270 game is always a headache due to the nv_disp.dll.
conclusion and analysis: (1)so far this board can only support 2.8v for mem, in the future if in any chance 3.1v is provided, it is very likely that the mem can run @290Mhz with 3-4-4-8(maybe a little cooling needed). (2) the possible reason why game always trouble the system is maybe because the HTT(hypertransport) link between Mem and AGP. explanation: from 270-273*9, it can pass all the tests including prime95, sisoft sandra burn-in test, 3dmark03 and aqua mark(cpu:mem=1:1 timing is either 2.5-4-4-7 or 3-4-4-8). prime95 can prove that the connection between cpu and mem is ok and both of them can function perfectly @ this speed. i search all of the google about this problem, it seems no solution so far. no matter it is infinite loop error or non-paged area or some other errors ,the error codes returned all point to the same memory location BFxxxxxx which is nv_disp.dll. can we blame Nvidia? No, Ati cards also have the same problem in a lot of cases. it doesn't make sense that under 247*10 D3D game can work but under 273*9 it can not. so i come to suspect the link between agp to cpu and mem---Hypertransport. accoding to the hypertransport consortium, it can run under 1.2volt @ 200 400 600 800 Mhz which are all basing on the 200Mhz just multi-channel. maybe 250Mhz+ is too high for Htt for the base bandwidth. that is why most of the board can adjust the htt voltage under bios settings. Abit K8vPro can adjust htt voltage from 1.2 to 1.35 at 0.05V step. even so, i am not very sure if it can make gaming smoothly @ 270Mhz+ fsb under high voltage. Due to the fact my PSU is just a 400W one, the power drain is so great that my rig can not work stable @ this power consumption. i did not try it too much, only adjust the htt voltage to 1.25 volt but my rig fails again in gaming@270Mhz. (3)finally i would like to say with a good PSU and Casing and maybe a better cpu fan, with higher Vdimm(mem voltage)@3.1v and higher Htt voltage maybe 1.35v or 1.4v, we can make it work @285-295*9 crazy speed for both cpu and mem(ratio 1:1).

Abit, else even we have money to buy the necessary equipment, we still cannot achieve more. for multiplier 10, i believe the highest maybe 250*10 under normal original fan cooling. to increase the fsb further more, we must wait for the newer bios. yesterday i can manage to boot into 300*6 under 6:5 ratio, but it is no value for us. higher fsb will make system less stable. anyway now i am using 246*10 to play safe, everything seems good and goes smoothly. we have already squeeze a lot from this poor 3000+ cpu. I am satisfied now. BUT i am also eagerly waiting for the new bios release because I believe the Geil UP ram has far more potential to be OCed.
One more strange thing, so far I read a lot of articles about the AMD64 3000+ OC, it seems most of them are under 2.3Ghz even in very good conditions of the rest parts(PSU, Mem....). How come ours are so easy to be OCed (with stable preformance)??? you have any ideas? or the articles I read are quite old? or they are all busy with the new Athlon64 FX and Socket 939 processors??

Today I use multiplier 8 cpu voltage 1.575 get the 285*8 successfully. cpu and mem all passed the prime95. the mem BW is crazy pc4560 with efficiency 85%=3870MB/s. the cpu now is 285*8=2280Mhz. but unfortunately it fails the 3d mark03 all 3d related things(even the dxdiag, but it gives me the hint for the reason and solution). For better understanding please refer to the graph for K8T800 Pro chipset http://www.via.com.tw/en/images/products/chipsets/k8-series/k8t800pro_blkdiagram.jpg or this page http://www.via.com.tw/en/products/chipsets/k8-series/k8t800pro/

You can see the important routes are Hypertransport and AGP 8X link, no matter how I adjust the setting, i can not get 3d workable under 285Mhz. Why prime95 can work perfectly but the 3D can not work??? Kula, do you notice a value when you do the sisoft sandra mem bandwidth test. It is named chipset 1 & under it "the Maximum Bandwidth". when I do testing @ 285 the value is 1140MB/s. I suddenly realize this is the key. it should be the "V-MAP" bandwidth in the figure above. according to the VIA, the V-MAP is the link between NB and SB, the bandwidth is as large as 1066MB/s. now you got it? yes the V-MAP speed is related to the FSB speed. When I touch it @ 285 Mhz, I feel it is very hot. So it further proves my point, when it is over roughly XMhz, it will be hardly cooled down. X/285=1066/1140 So X=266MHz. the Abit is really doing a good job on NB of this board, a nice blue heatsink. but Abit is really poor on SB, nothing onside the SB. So even @ speed>266Mhz, we can boot into windows, cpu mem testing, it is just hotter than @266 speed. when you doing the 3D game, it pushes the temp even higher!!!-----Beng! system hang for no reason, the nv_disp.dll can not receive the data it needed. Now not all the texture are in the memory, so this V-MAP link is so hot that nv_disp.dll will fall in a infinite loop to waiting for its large volume texture data to pass it to G-Card through HT and AGP8X link.

Now I am running 267*9( if it is not stable I will adjust it to 266), I believe my pc will be stable even @3D games.
For me 267*9=2400Mhz, the cpu is not so hot (much less than 2500MHz) mem is much much lower than my tested 285Mhz. what i will do is buy a small heatsink and put it on the SB to give SB a passive heatsink (I hv done that when I was doing my amplifier project at NUS). I believe after attached the heatsink, the board can run FSB285 very stable in all kinds of jobs. If possible I will attach a small fan to it to make it active heatsinker :-)

Rules of thumb so far I find
1)to increase the FSB to > 275Mhz, increase the cpu voltage, in my case I only can make it work when I adjust it to 1.575V. Mem timing is 3-4-4-8 at least for 285,actually I have not tested the other timing maybe it is really workable. Mem voltage here is 2.8v is enough for 285MHz. Cooling down the SB(south bridge) for our abit k8vpro board.

2)adjust the Htt mutiplier (3 for fsb>250Mhz) to make the whole bandwidth less than 2GHz.

what we can do in the future after apply the cooling system to SB, the reason why the fsb can not be pushed larger than 300+ is we need cool down the Sourth Bridge VT8237 chip. At that time, possiblly we can do FSB 342 with 6:5 ratio mem 285,342*8=2736MHz with maybe a good cooling fan or water cooling :-)

mine can, is your damn fast 6800 Ultra. It eats large volume texture data in very short time. it overloaded the V-MAP speed limit. your system hang. Now what we need to do is not modding the Vdimm which is highly dangerous to MB, but just easliy and safely attach a heatsink to SB. but one thing I need your help, you have any idea what kind of sticker of geil we need to use for the heatsink? It must be a good heat-conductor.

also no thermal tape or sticker provided, I actually just bought a CoolerMaster chipset heatsinker with a fan from SLS, I will try tonight, but so far I tested, @266Mhz my machine can run 3D mark03 70 loops without any error(around 5 hours). That is why I sacrify my lunch time to buy a CM chipset fan. it comes with a double side thermal sticker under the heatsink and it can use 12v as input or a converter to use 5V i think, it is oringinally designed for NB so it is a litte bit large for the SB, but I tried at Fuwell, it can be put in. it is not expensive, 10S$, for the other shop 15S$.
i would like to use a active fan cooling to make sure all my mother board chipsets are stable.So let us try and find out.

but you need take note one thing, my previous assumption maybe wrong, now SB is well cooled down, but the system hang is still there, I realize it is no more a heat problem because all my parts are in good temp. So I try to minimize the problem to NB and AGP bus link, I try to adjust the ration to 8:2:1 instead of the fixed(66Mhz 33Mhz), after that, everything is fine, it is working!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

Now I am testing the 285Mhz(3-4-4-8) with 8:2:1 at 72Mhz:36Mhz, it is fine. Previously, dxdiag and 3d mark03 or AquaMarks all hang at the first second @ this bandwidth. now all of them can pass the test. Since my cpu can not run @285*9 well. I decide change to 275(2.5-4-4-8) 275*9=2475Mhz now I can pass all the 3D tests. Cheers! At least I make some progress, I now understand the lock apg.pci of Abit is not functioning well. So when we run 3D demanding program, the MB will maybe switch between other different ratio in quite a short time---the driver confused and hanged. I feel my poor card can still work @ AGP=72Mhz, so i believe your king card can surely run without any errors now. As i recommend, you can try 285*9 in your case since you have good PSU and Casing cooling. And the most important thing is do not trust the Abit Fixed Freq but mannuly set to 8:2:1.@285Mhz AGP=72 and PCI=36. Btw I disable the fast write so I do not know what will happen if I enable it. But one thing for sure, even before I disable of enable Fast write, I can not run the 3D testing @276+Mhz, but now I can.For 285Mhz*8, My machine is so far so good.
The SB cooling is nice in my machine, 3 blue lights when running, very steady and now the SB is very cool.
Wish you a good luck.