Cant get into windows. With 1.37mod is 101% stable, prime 270x10 1:1 all night long...
Memtest works OK....that's why I think that is strange.
Thanks again.
Printable View
Cant get into windows. With 1.37mod is 101% stable, prime 270x10 1:1 all night long...
Memtest works OK....that's why I think that is strange.
Thanks again.
If you look at AMDs BIOS and Kernel developers guide you will see that for higher clockspeed a lower Read Preamble is chosen. For a Board with 4 DIMM Slots it goes like: 9ns 100Mhz, 7ns 133Mhz, 6ns 166Mhz, 5.5ns 200Mhz. If you continue this to 300Mhz a Read Preamble of 5.06ns would be optimal (5.25ns 233Mhz, 5.125ns 266...). Have you tried setting Read Preamble to 5ns?Quote:
Originally Posted by CD 1986
Give me a screenshot of A64Tweaker, I will try and mod a newer BIOS to have those timings, maybe then it'll work. :stick:Quote:
Originally Posted by .: Eru :.
Here it is.
http://img295.echo.cx/img295/4134/cpuztweaker7tg.jpg
Thanks to all, we'll win this battle :P
Max Async Latency = 9ns <----------- Might help when you exceed DRR650!
Read Preamble = 6.0ns <---------- 5.5ns is the sweet spot on Neo2 :)
Thanks for helping
Now, lookie here http://www.swjka.com/k8nneo2bios/?dir=./need.for.mhz
All BIOSes has been revised to Rev.3 :woot:
What's new in Rev.3:
The HH edition is no longer revised because "Rev.3" BIOSes feature same tweaks that HH edition hadQuote:
Changelog Rev.3:
- trrd = 3 clks
- twr = 2 clks
- trwt = 2 clks (trtw in A64 Tweaker)
- tref = 133Mhz 7.8us (removed)
NEED SOMEONE TO TEST THOSE AND VERIFY ^^^ THOSE SETTINGS IN A64 TWEAKER. THANKS !!!
EDIT: Oh, forgot to thank sideeffect for ftp access, thanks mate :toast:
No problem mate.
Ill test the san diego rev 3 now and see what happens :banana:
edit
Rev 3 for san diego works fine all timings show in a64tweaker as they are listed in changelog :D
Testing out the 166 divider now to see what happens.
The timings that you changed all set fine.
Write to read delay time twtr sets to 1 with 166 divider. 200 divider its 2 also read preamble sets to 6.5 with 166 divider. With 200 divider its 5.5.
Maybe they should be set to 200 values
Thanks for testing, 166 divider in Neo2 is totally messed up :banana:Quote:
Originally Posted by sideeffect
So in next release I will force twtr to always set at 2 clks and read preamble at 5.5ns. 1 clk for twtr is way to tight. Probably those caused boot failures when using 166 divider. Thanks for noticing those :)
K fully tested 166 divider and its still not even any where near stable set read preamble to 5.5 and twtr to 2 with a64 tweaker. Htt was set to 290 multi at 10 Resulting memory frequency was 243 mhz failed prime in about 10 seconds and windows memtest in about the same amount of time.
260 mhz ram speed is stable in 1:1 with exactly the same timings. Pc is also stable at 290 htt when i use 133 divider.
am i missing something here. Could this be a high Htt issue rather than a ram issue or are their other timings that a64 tweaker just isnt showing. Also could this be a bad memory controller on my san diego does it have to deal with 290htt or 243 when is the divider put in place.
Why would 290htt at 133 be stable if it was high htt issue with the cpu. Dont really think anyone can answer this was just saying what i have found with my setup.
Im so hoping this is just a san diego issue that will magically dissapear with the 1.9 bios release which is taking forever to come out.
Can you post an A64 Tweaker screenshot of 166 divider v 133 divider @ 290HTT (without tweaking anything). Unless they are exactly the same timings of course. :)
Only visible difference is the tref refresh rate so im focussing on that now to see what works with 166. Maybe 133mhz 3.9us will work with 166 divider.
What tref rates do you get for 133 and 166 divider? I think tref values might be read wrongly in A64 Tweaker however...
http://www.xtremesystems.org/forums/...882#post907882
when I load the
200 divider a64 tweaker shows 200 3.9
166 divider a64 tweaker shows a tref of 166 3.9.
133 divider a64 tweaker shows 133 3.9
I tried setting 200 3.9 in 166 divider before but it doesnt help but i dont think i tried lower settings.
But if the reading is read wrongly then ill have to try all the settings to see if any work.
lol just tried 100 3.9 in 166 divider and memtest crashed bad in like 0.1 seconds :stick:
I'm downloading RMMA (from that thread i posted) to see if the rates reported in the tweaker are different to what the actually are. Best to test all values in any event, since some people get better results with totally different values. :)
I have right mark ill see what it says to.
It shows 7.8 us instead of 3.9 hmm interesting.
Set 133 7.8 us in a64 tweaker and it now shows as 15.6 us in rmma :slapass:
Went through all the refresh rates with athlon 64 tweaker and the lowest that shows in rmma is 7.8 us.
Im going to restart in 1:1 and see what it says then
Ok, i also found that 200mhz 3.9us is actually 7.8us...so the tweaker is wrong. Seems like what ever you select in the tweaker, the actual value is 1 step higher...?
Im in 1:1 now and it still shows 7.8 us refresh interval with rmaa even though a64 tweaker shows 3.9 us. I think rmma is wrong and doesnt read values lower than 7.8.
Or it could be that the actual values are 1 step higher. Looking for another tool that reads the value now :p:
The reason i think rmaa is wrong is because setting 1.95 in a64 tweaker still shows 7.8 in rmaa.
Ok tried with systool which you can get from here http://www.techpowerup.com/systool/ and it shows the same as a64 tweaker but then its very similar to a64 tweaker it could be the same code :p:
Btw systool is ace if you havnt used it because it shows the real voltages of the mobo unlike speedfan and mbm which show wrong values.
I think RMMA is correct, only because...512mb sticks nearly always have a default refresh rate of 7.8us and that is what RMMA is showing on my setup. I could be wrong, but the guys in that other thread seemed to reach the same conclusion.
Tweaker vs RMMA
1.95us ---> 7.8us
3.9us ----> 7.8us
7.8us ----> 15.6us
15.6us ---> 15.6us
Yeh, systool is great...i helped the author set it up for the nforce3 chipset, since he didnt have an nf3 testbed; reading and changing clock speeds etc... Best all in one tool around! :)
EDIT: have you managed to find any tref value which gives greater stability for your setup?
No ive been to busy messing with rmma and systool lol. Going to try some more now. One thought though if 7.8us is the minimum setting what are the other settings actually doing in a64 tweaker ...... :stick:Quote:
Originally Posted by CD 1986
Ummm....
I suppose smaller sticks (eg. 256Mb) might use a lower refresh rate value. And maybe a lower value might help improve performance in some cases. I'm guessing here though. :p:
I think if the value could be set with a 256 mb stick then it could be set with a 512mb stick even if it caused the 512 stick to crash while setting it.
Someone with a 256mb stick test this and see what reading they get with rmma please :slap:
Also, look in Everest to find the default refresh rate of your memory modules...its in the SPD info section. These were taken from my laptop memory modules...they have default of 7.8us, but they are only DDR333:
LOL this is just plain weird, Ok so im running memtest and it keeps finding errors after a few seconds. I change all the tref values each failure and start again but it still errors. (166 divider)
So I leave tref on 133. 3.9us and change the memclk frequency to 183 which takes my ram up to 263mhz and memtest is running fine now ....
edit- everest shows 7.8us for me as well in spd
Good find there! What happens if you boot using the 133 divider and then having A64 startup settings set to 166 divider? Is it still as unstable as booting with the 166 divider from the bios?
Didn't 1.9b5 have a 180 divider in? If so, you might be able to get around your problem that way when it's officially released. :)