Explanation of the NB/CPU speed bugs
Aha you see gOtVoltage, things are quite weird and you have to compare to known values. You should always test any setup to know facts. Software reading vs. Physical speed change are two different things. ;)
Have you seen my 14GHz K7.5 benchmarked? :lol2:
I've seen and known about these bugs on both AM2 and AM2+ with Phenom for a long while. They're very constant with AM2 users but easy to catch by the performance. They're typical BIOS bugs. I've spoken and shown them many times.
Keep two basics in mind:
1) Registers can read anything => software which reads registers only, if register is wrong, software which relies on it is wrong.
2) Phenom systems are known to read wrong registers and thus speed values many many times.
3 i) While the actual physical speed/multi of the system can be very different or not changed at all. You can check this by benchmarking only.
Let's take some examples.
You posted this oc 02-24-2008:
http://i168.photobucket.com/albums/u...0htcpu2300.jpg
The EVEREST Ultimate benchmarks of it you posted after (02-10-2008): NB 2800 vs. NB 5600
http://i168.photobucket.com/albums/u.../NB2800mhz.jpg http://i168.photobucket.com/albums/u.../NB5600mhz.jpg
Quick comment: your IMC scores are very low even for 1800MHz IMC, the read is too low. Something seems wrong with the NB clock controller and/or register. Your scores should be higher even if those 1800 was real, even more with X64.
Now me and Achim compared MB RAM/Cache/IMC performance at the same clocks/timings. Have a look at these at 2200/1800/800. I'll choose to compare Achim's 9500 770 scores as the 790 boards were a bit better.
M3A results: http://www.xtremesystems.org/forums/...&postcount=259
K9A2 Platinum results: http://www.xtremesystems.org/forums/...&postcount=950
PC-AM2RD790 results: http://www.xtremesystems.org/forums/...&postcount=566
If your speeds were above 2200/1800 IMC, you should naturally have higher L1/L2 and L3 scores and lower latencies by far. You should have lower memory bandwidth and higher memory latency though due to us having tighter timings =>
http://www.abload.de/img/everest-gangedtp4.jpg
If you compare, your results are much lower than a 1800MHz IMC. The results mainly affected are L3 read bw and latency. Both of yours are low for even 1800 MHz (this bug is giving you poor IMC performance, even worse than stock). If you ran L3 latency which depends on read bw, it would be slow for even 1800MHz. If you look at the other's, they're even higher.
Now I'll show you the very obvious difference of real IMC clocks as Sami (AMD worker/pioneer allround overclocker) showed us a while back and how you can see them done in typical mode, not with services/processes/themes disabled but just the AV/FW paused from an everyday OS.
TEST REAL IMC SPEED
1809MHz (9x) vs. 2211MHz (11x)
EVEREST Ultimate (check the L3)
http://img526.imageshack.us/img526/2...2stock2dy1.jpg
vs.
http://img522.imageshack.us/img522/5...22stockvl2.jpg
WinRAR
http://img526.imageshack.us/img526/7...6winrarte4.png vs.
http://img517.imageshack.us/img517/2...2211winyq8.jpg
See the glaring difference and perf. gain? ;)
I also have near 2.6GHz real speeds benchmarked and plus 7GHz bugged ones benchmarked. Like I said, actual oc IMC requires high volts and is hard.
Compare the 7GHz benchmark, so you know what a bugged reading looks like (actual speed has obviously not changed from 1809).
http://img138.imageshack.us/img138/4341/nbbwevta9.jpg
No real change from stock (minor fluctuation is normal).
Re SPi: your score is actually still not optimized and fair slow for 2300/1800/800. It could get much faster leaving all speeds/timings the same. If even you managed 2200 IMC real, your 1M would be much lower.
HIH. :)
Ok new BETA BIOS out 228.BSTbeta..
Linky
http://www.lejabeach.com/Biostar/TA770A2+/ 228.BST (BETA)not official yet...Stock Biostar
Polygon has a TLB fix already as well..
http://www.rebelshavenforum.com/sis-...;t=000463;p=44
NPTfid and NB working with 9600BE...Thes Bioses still have a long way to go but keep getting better testing the (TLB disabled fix Bios) .. the 228.TBL MOD Works like a charm the NPTfid and NB clocks but memory @ 1066 can be set using 1066divider.. It will default to 800mhz.
NB settings seem to like 2000mhz and lower with HT being set the same...Before this Bios i had to use FSB or AOD to Oc...Now i can raise/lower the Multi along with the FSB:yepp:
http://i168.photobucket.com/albums/u...NPTfidBios.jpg
I am on X64bit..that is the only reason why i am not posting Oc's over 2.5ghz..It just isnt 100% possable ATM...It has to do with the 64bitOS and Bios..Those of you with 32bit OS should be able to crack 2.5ghz with out a sweat on the TA770 with the (Stock Beta228.BST) or (Mod Beta 228.TBL..TLB disabled BIOS..
Like always many thanks to :up: Polygon :up: for his TLB work and other optimizations..Funny he is way fast with these updates.
UPDATE using 228.TLB BIOS MOD from Polygon
The TA770 can do 270+fsb with Phenom! I maxed out my ram 1080mhz @2.3volts with this Bios or i would of gone more...The Biostar Bioses need to be Optimized more fo Higher ram and CPU above 2600mhz!...The 228 Bios is Biostars latest BETA and Polygon has added only the TLB disable option to it...
The screen is CPU@ 2300mhz(8.5x270fsb) im now using 2400mhz (9x270fsb) havnt had chance to Oc more yet...What is nice is before this the TA770 would only do 230/240 max .....Its definatly a learning process:yepp:
http://i168.photobucket.com/albums/u...fsb2300mhz.jpg
Screenshots...look over then maybe try other Bios...
Ok ....This is definatly wierd....
Between what KTE and me and other have said should of fixed that.......
Maybe you Could you try the 125.BST or the latest 222.BST Official Bioses....set the defaults for your Chip...It has the TLB fix enabled but ,,im pretty shure the FIX or TLBdisabled is not your problem in the 228.TLB BIOS....Almost seems CnQ is not Off/disabled in Bios or Windows when using NPTfid even though you have it set to disabled.
It should read normal values of 11x200FSB...with (CnQ on in Bios) . Once you turn CnQ (off with 125.BST or 222.BST or even 228.TLB) it should now show half clocks unless you have NPTfid set to 11x2200mhz with the 228.TLB.
Polygon at RebelsHaven is using a 9500 and same Bios....Im shure if you post in this thread http://www.rebelshavenforum.com/sis-...;t=000463;p=49 he could give ya the much needed help...
I believe Sam2008 and Orion is using a phenom and TA770 and same Bios as well ...In all its probally something simple were all overlooking;)
NOTE:,,
Also have you tried another NPTfid instead of the 8x...
I was only able to get use of 8.5x or Higher NPTfid and using HT/NB as low as 5x = 1200mhz...
Try the 9xNPTFID with 250fsb = CPU@2250mhz use 1400mhzNB and HT 2gig = 1400mhz...this will make HT/NB 1750mhz...Use a low memory divider like 400/533..
You may have come across a MOBO bug using a certain NPTfid multi who knows...
Before trying another BIos Check the settings in the screens posted...Differance you should have will be the use of a NPTfid Multi ,,FSB ,, Memory divider,,Voltages. For the Most part everything else should remain on Auto except for things you dont use like USB's Esata and maybe the BOOTLAN(TLBfixOFF) ..
If using SATA make shure you Use the ONBOARD Sata in IDE mode...
Hope this helps ya..
DISABLE VIrtulization..
http://i168.photobucket.com/albums/u...sierra/129.jpg
DISABLE SmartT FAN
http://i168.photobucket.com/albums/u...sierra/128.jpg
HT link width 16x....
HT 200 = 200mhz ,, HT 400 = 400mhz ,, HT 600 = 600mhz,, HT 800 = 800mhz,,
HT 1gig = 1000mhz,, HT 1.8gig = 1200mhz,, HT 2.0gig = 1400mhz,, HT 2.2gig =1600mhz,,
HT 2.4gig = 1800mhz max 2.6gig is same as 2.4gig...
Note; Raising FSB above 230 you will lower HT and NB the same from 1800mhz to say 1600mhz...and above 240fsb to say 1400mhz/1200mhz..The fsb will bring this value back up....You want to try to keep it around 1800mhz for your Oc..Mobo dosnt like below 1200MHZ with High FSB..
http://i168.photobucket.com/albums/u...sierra/127.jpg
TIMING MODE under MCT must be set to manual to allow changing the Memory Dividers 400/533/667/800works and 1066 defaults to 800mhz..Set your RAM DIVIDER lower than 800mhz before you Oc or you could max your ram if it dosnt clock well.
http://i168.photobucket.com/albums/u...sierra/124.jpg
MCT Timing i keep AUTO for first boot ..ehn go back and set to manual if you want to tighten timing...
http://i168.photobucket.com/albums/u...sierra/125.jpg
If overclocking and manualy setting the Ram Dividers make shure (AUTO TWEAK and OPTIMAL PERFORMANCE MODE EPP is OFF) It can set Timing too low and Crash the system..i prefer manual timinig anyway..
http://i168.photobucket.com/albums/u...sierra/126.jpg
CnQ must be off to access these options Below....
The 228.TLB BIOS Beta
CPU NB-fid ...
CPU DID = dont touch...keep it at 1...
CPU fid...Set value of CPU Higher or lower....
you can use this in conbination of raising FSB/CPUclock...
In the BIOS on boot it will only show what CPUNPT multi freqency your using it will not show FSB..Its Bios glitch...In windows it will show correct clocks or if you use the Memtest in Bios..
last Voltages will vary for some cpu's over others...HT volts will need to go up alittle and maybe NB/SB....its trial and error really.....
http://i168.photobucket.com/albums/u...sierra/123.jpg
If you have any more qestions or need more screens of other Bios settings just let me know:up: ...
May make B2 run decent ,,use caution,,hahahaa
the 612 BIOS unleashed some NB/L3power in my B2(it was never stable @1800 while @2.58!!!!!!!!
612.XXX Coretemp CPU-z Everest.(trial version)you can see the BIG differance very easily.
Looking below..
@2583mhz( MAX on the 514bios)while using 1800NB and a whopping 1.32vNB with a huge bump 1.44vcore just to get it stable...Thats alot of Vcore and only 83mhz more from preveouse setting of (@2.5mhz with 1800nb 1.20vNB //CPUvcore@1.23v. this was my 24/7 setting...
Now just by changing to the (612.XXX Bios) 2.6mhz,, 13x200fsb and 2200NB was achieved with little hassle..NBVID was set to 1.20v(1C) and default Vcore!
The sytem increased its speed a fair amount...So with this Bios allowing me to raise the NB 400mhz more than stock and using lower volts 1.20vNB CPUVcore was default Vcore..
http://i168.photobucket.com/albums/u...nb_612bios.jpg
http://i168.photobucket.com/albums/u...00_612bios.jpg
http://i168.photobucket.com/albums/u...ierra/ram1.jpg
LOOK close because even though there is only a (17mhz speed differance on the COREs),,The NB is Running at 2200MHZ with 1066mhz(612XXX)VS 1800mhz and 1100mhz/ram(same timing) as in th (514.Bios)
Even with Ram and FSB@200 with the Old Bios i could not raise the NB above 1880 let alone achieve 2.6 just by using the 13.5 multi..There are suttle diferances in this new Bios which seem to improve the Stability for B2 bottom line..
Note: More Speed ,,less volts and Higher Oc...OMG :40beer: I know i am not doing Extreme benching , Yes this chip is a COLDbuggered one,No boot below 12c with 1.3volts and stock @2300mhz setting was Problamatic with X64!
:D Achieving 2400mhz+ was the goal with this particular chip..2600mhz was well = to going to the MOON for me.. :D
612XXX BIOS @2600mhz//2200mhzNB stock 1066mhzRam
http://i168.photobucket.com/albums/u...00_8840mbs.jpg
514.XXX BIOS @2583mhz//1800mhzNB 1100mhzRam
http://i168.photobucket.com/albums/u...18_2583mhz.jpg