MrIcee, ever able to get decent performance on 8k3A?

[xgman]

I remember seeing a post where you were fighting with the 8k3a to get some decent performance. Did you ever figure out how to get this board to work properly? If I run in the 1900mhz range at 201-cas2-2-5-1T,4way with 2719 or 2626 bios, I only get a score of 2750 or so in Sandra. This seems patheticly low for these timings. I know 2312 is the only bios to have "cpu fast command code" enabled, but still doesn't seem right. Any tips?

[MrIcee]

Hi Xgman:)

I played with the registers and found some good bandwidth in the board at 200Mhz+ FSB....but I also found it either did not utilize the bandwidth to its full potential, or the AGP implementation was not as aggressive as Abits. Needless to say...having benched both the 8K3A+ and KX7-333R at my baseline benching MHZ of 1695Mhz...or 8.5x199Mhz...the Abit board smokes the Epox in 3DMark. OPP also found this out awhile ago. Bear in mind that that is with all things being equal, OS, drivers, vid card overclocks....etc. I did also run the fabled 2312 bios using a higher multi, lower FSB to equal or stay close to 1700Mhz, and it was no dice. I spent 2weeks with 8K3A+....it is a great board no doubt...stable etc...but it won't bench with the KX7. I know I will inflame my Epox friends here...but once a KX7 is tweaked properly....nothing will touch it. My final scores at 1695Mhz were 13,356 for the Epox....13,973 for the Abit...My Abit score is posted at MO on the ORB in the 1700Mhz and under category. I know my Epox friends will say no way to my findings...but I was very deliberate and non biased in my testing. If someone else can get that board to equal the Abit at 1695Mhz...I invite you have at it and post up your scores:D

Randi:D

EDIT: I just found my notes from the 8K3A+ evaluation....here are the WPCrset settings I used with the 2619 bios to achieve better bandwidth....3100MB/S+ at 207Mhz, maxed mem timings :

6A=FF Longtime used bandwidth enhancement

97=73 This change enables fast CPU decode at 200Mhz+

B0=07 This change helps to prevent freezes in 3DMark in Dragon and Nature..also on the KX7 if you are suffering that anomlie.

The 93=FF bandwidth enhancement..tho it works on the KX7 and seen as improved numbers...had no effect on the 8K3A+ as far as numbers in Sandra.

[SKATAN]

mricee

and for the comparison benchmarks what where the clocks used on the ti4600?



where you using cpu comand decode ultra or fast on the kx7 for that score ?

[Jeff]

Thanks for the memory tweaks MrIcee, I can't wait to try them. I just recently broke 3k on my 8k3a+ but I needed to run at 212FSB to do it.

[MrIcee]

Skatan:)

Nice to see you my friend:)

For all benches in 3Dmark my Ti4600 is run at 384/778. On the KX7...fast cpu decode is a must after about 175Mhz...it varies from board to board...Ultra only works with AthlonXP's up to 175Mhz if that. So for all intents and purposes..everything was equal...including fast cpu decode being enabled on the Epox. I verified it by the substantial gain in bandwidth thru SiSoft Sandra. That is where the Epox engineers neutered a great board....as by using fast cpu decode it gives it much better performance at high FSB. My only question remains what else is different in Epox implentation of the registers...as I mentioned before..it either wasn't fully using the excellent bandwidth I produced with it...or its AGP implementation is just not as aggressive as Abits. I'm not sure exactly what Abit's "Enhance AGP" feature changes..but its effective:)

Randi:D

[Trevor]

Where can I get the WPCREDIT pcr file for an kx7? edit=11063099.pcr :rolleyes:
MrIcee, what settings would you suggest for the kx7?
cheers :)

[xgman]

Thanks, MrIceee, those are the settings I was looking for to try. I'll give them a try until the Nforce2 board comes out which I hope will give the AMD paltform a decent boost.

[SKATAN]

MrIcee

thankz for clearing me up :)

Just wish I could test a kx7 just to see how good it is :(

[xgman]

I have wpcredit 1.2a and tried setting the registers as described above, but when I selet the register and choose edit, I change the 2 digit Hex # and choose set and then ok to set new register, but all of the suggested settings change right back to what they were. They all say "reserved". I can't even test the settings once. Anyone know why this happens?

[wite]

i also tried the sugested settings but like xgman some of the registers youst jump back for me.

i can change
6A = FF
93 = FF

but the other two 97 = 73 (would love to get that to work but it dont) and B0 = 07 dont work.

this is with bios 2719

[Learn]

hello all

my first post !

i' ve the 8k3a rev 1.0 with 2619 bios

this wpcredit tweak don't work :(

unabled to run " fast CPU decode " at 201.99 fsb

why?

thanks

[xgman]

Some guy on an Epox forum wrote this:

"You cannot manually set bit 1 of register 97 to 1 on the 8k3a+, its function is purely controlled via the bios...on all 8k3a bios not just later ones.

Im using 2719 bios and can happily set 6A to FF and 93 to FF, and get a corresponding bandwith boost.

Ihave no use for the B0 tweak as my 3dmark is stable without it (although your are correct in that this register is also set by the bios)"

[MrIcee]

Guys..

I am aware that the setting for Fast Decode will not set in WPCredit...why do you think I suggested you set it in WPCrset?? The bios keeps kicking it back to its stock setting. I gave up and just put it in WPCrset. I couldn't tell that it was changed until I verified it with Sandra memory benches....and there was a boost
in bandwidth. If you need to see the numbers...go to my thread in this forum where I posted the Sandra numbers at 207Mhz...I don't believe you will achieve that without fast cpu decode...as I couldn't previously with all timings maxed and 6A=FF engaged. Again...I saw no improvement in MY setup with 93=FF...I mean no change in numbers at all...but I also had 97=73 enabled FIRST, then tried 93=FF.

Again..and I posted this in another thread weeks ago....you cannot set bit 1 of register 97 via WPCredit....this is a known problem. Use WPCrset and just set it, then check your bandwidth before and after the reboot.

Randi:D

EDIT: Heres a link to the thread..read the 3rd paragraph: http://www.xtremesystems.org/forums/showthread.php?s=&threadid=1674 I believe still thats the highest 8K3A+ bandwidth I've seen posted.

[Learn]

thank you for the answer

I had not seen your thread

;)

[MrIcee]

You're very welcome my friend:)

Randi:D

[wite]

ok sorry i didn't know that u could enable settings via wpcredit even if it didn't work in wpcredit.

I have tried it now by youst adding it to wpcrset and im sorry to say that it didn't boost my scores at all. but maybe its because i have such a bad stick of ram, have to use 2.5 3.6.3.2 to get it working at 200mhz fsb.

[Jeff]

Only one that did anything noticable for my "numbers" was 6A=FF. 97=73 via WPCRSET didn't change my memory bench.(or SuperPi) at all.

I did however break 3.1k... but again... that was at a higher FSB than your 3.1k. And the 214MHz bus needed for that is a little too much for me to run 24/7 stability wise.

Again, thanks for the tips.

[xgman]

in wpcreset, 6A and 93 (both to FF) 97 to 73 gave no difference whatsoever in Sandra mem bench. I don't know what is the matter with my MB. Also set was the lower temp change 92=E9 and 95=1E. Do you recommend it? Thanks.

[sysfailur]

Omg, thank you MrIcee, I am going to check some of these settings tomorrow. I've been at my 8k3a+ all night (posted results in my thread in the amd forum). I'll fiddle w/ these settings mentioned here tomorrow for sure!

[MrIcee]

Sys:)

Just be aware that the settings I used and workedfor me are not working for everyone. I cannot explain it but as you see in my link to the 8K3A+ thread...those settings definitely helped me achieve that bandwidth...and at only 207Mhz. I still haven't seen higher yet at similar FSB speed.

Ok...I think I might have a theory on the fast cpu decode thing. I noticed when trying to set bit one, when observing the data change when the bit is enabled, that the data number CHANGED depending on what FSB I was running. This may explain why at 207Mhz it worked for me....but with other guys who may be running, say, 200Mhz, it doesn't. I can honestly tell you all that I saw a jump in Sandra when I set the bit in WPCrest.

Here is what all you Epox guys can try before throwing out the fast decode change. Set your system up at the multi and FSB you wish to run. Open WPCredit and go to register 97. Enable bit one and observe the data "number" as it changed from stock, write it down....close out WPCredit and start WPCrest. Enter 97= whatever data number you wrote down. Reboot and check Sandra bandwidth. Remember to take a baseline bandwidth test before trying this procedure so you can verify OR discount the change you made.

Quite a few times I saw no change..but after several sets of WPCrest it appeared to work, as my bandwidth jumped up. Give it a try guys...I really wanted for this to help you guys gain some bandwidth with your Epox boards as I did.

As an aside...in some memory evals I was doing last night...with a stick of Corsair PC3200 C2 I exceeded my Epox posted bandwith at only 204Mhz and 1495Mhz on my KX7 by about 40 points ALU. Going to keep on testing:)

Randi:D

[SKATAN]

Mricee did you check with wpcredit if fast cpu comand decode was enabled when using wpcrset or just checked the bench and assumed it???

[Jeff]

Can't wait to try your further suggestions later tonight or tomorrow. And I just wanted to thank you again for sharing your findings. :toast:

[MrIcee]

Skatan:)

I just set it and checked bandwidth differences....when I saw jump of 50-100 points after a WPCrset change...I assume that change worked..if no change then it didn't work. Several early attempts were checked via WPCredit and verified not to work. After that when I did gain a chunk in bandwidth...it still showed no change in the bit...tho the data value showing was NOT what was set in WPCrset....strange indeed.

I no longer have the board to play with....so I can't experiment as I did before. I can verify the only data changes that increased bandwidth were 6A=FF and 97=73. If it had not worked I would not have posted it on the boards. But when I set the bit 1 via WPCrset...I went from high 3000's to the figure of 3134 or whatever it was in my Epox thread. If it didn't have to do with the bit change....or attempt at changing it...I have no other explanation for the bandwidth increase.

Randi:D

[SPQQKY]

I wish you had run PCMark2002 while you had the board. I am curious to what a real world bench like that would have show in the memory department. I am a little hesitant about Sandras memory bench. I still think Sandra 2001 is better, unless you get the pro version of Sandra and disable the modules for SSE.

[xgman]

Originally posted by MrIcee
Sys:)

Here is what all you Epox guys can try before throwing out the fast decode change. Set your system up at the multi and FSB you wish to run. Open WPCredit and go to register 97. Enable bit one and observe the data "number" as it changed from stock, write it down....close out WPCredit and start WPCrest. Enter 97= whatever data number you wrote down. Reboot and check Sandra bandwidth. Remember to take a baseline bandwidth test before trying this procedure so you can verify OR discount the change you made.

Randi:D

I am trying to do this on the 8k5a2+, and I am not exactly sure what you mean by "enable bit 1" when I type in 73 for register 97, the bit changes to 30. So exactly where do I put this info when I go to edit my 97=73 register setting in wpcset. Thanks.

[sysfailur]

Originally posted by xgman


I am trying to do this on the 8k5a2+, and I am not exactly sure what you mean by "enable bit 1" when I type in 73 for register 97, the bit changes to 30. So exactly where do I put this info when I go to edit my 97=73 register setting in wpcset. Thanks.

Ahng you got one?! Benches!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

[xgman]

It seems to be just cleaned up a bit since they have had a while to get the bugs out of the 8k3a. But, performance doesn't seem to be any different at least as far as benching. It may be more stable though. What I don't like about it is that the ide and raid and floppy connectors are all jammed together at the far end of the board, next to the power on connectors making it almost impossile to connect 4 devices. I use the promise 133 card and it loads faster anyway so I only have 1 ide connected to cdrw/dvd/s anyway. The other thing is there is no 2312 type bios for it for better memory bandwith.

[Learn]

hello

the bios 2812 for the 8k3a /+

the fast cpu decode it activived !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

see aoa for the download

[xgman]

Yeash, and there is one for the 8k5a2 there too! (update: doesn't work)

[Silver]

Originally posted by Learn
hello

the bios 2812 for the 8k3a /+

the fast cpu decode it activived !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

see aoa for the download

and who is aoa?

[AndrOvr]

Check www.Aoaforums.com or www.aoafiles.com :)

[DJ_Datte]

AOA = Aoafiles.com / aoaforums.com , The Alliance of Advacement(SP!) of Overclocking Arts.

its a great site, very good for Epox users especially ! :)

/Damir

[Silver]

Thanks guys, really appreciate it. Take care.:)

[Silver]

Originally posted by Learn
hello

the bios 2812 for the 8k3a /+

the fast cpu decode it activived !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

see aoa for the download

Is the 2812 a typo?:(

[SPQQKY]

Here you go! Great BIOS, fast as the 2312 and no tearing at 199 fsb and over.

8K3A2812 (http://home.attbi.com/~spqqky69/8k3a2812.zip)

[sysfailur]

Originally posted by SPQQKY
Here you go! Great BIOS, fast as the 2312 and no tearing at 199 fsb and over.

8K3A2812 (http://home.attbi.com/~spqqky69/8k3a2812.zip)

Yeah it owns. :)

[xgman]

What is the highest fsb anyone has gotten with this new bios, and at what timings?

[Jeff]

I've done 196 with max'd out timings and 3.1v on my Corsair XMS3000. I thought I read somewhere that the timings automatically relax once you hit 200 and I plan to test that theory when I get home from work tonight.

[Silver]

Originally posted by SPQQKY
Here you go! Great BIOS, fast as the 2312 and no tearing at 199 fsb and over.

8K3A2812 (http://home.attbi.com/~spqqky69/8k3a2812.zip)

Thank you very much. I shall give it a good run. Unfortunately last night I fried the Ti4400. Screen just closed like the end of a movie and system would post no more. Took all apart today thinking it might be the dual pelts but alas no, it is the video card. Till another comes in I am stuck here on an mx pci card. Better than nothing I would geuss though slow as xxxx.:(

[SPQQKY]

For some reason my unlock went bad on my XP and by the time I got it fixed, I ran out of time to test the BIOS to it's fullest. I read over at AOA that the fast cpu decode is disabled again after 199MHz fsb. Have seen no proof and can't test it myself right now. Anyone confirm this?