Basic Overclocking question on an Athlon 64 3200+

[fallguy]

You dont have an AGP/PCI lock, and that can hinder your overclocking. It might be the cause of your artifacts, because the AGP is running in the mid 70's instead of 66Mhz.

 

[rainz]

You dont have an AGP/PCI lock, and that can hinder your overclocking. It might be the cause of your artifacts, because the AGP is running in the mid 70's instead of 66Mhz.

Well i'm not sure to understand what you are trying to say since my problem seems to be fixed with my RAM voltage increasement at 2.8V.

Yeh i agree i dont have any " lock " . And it's a good thing. I tested a lot with the 2.8V in FC ... and never saw this bug again .. and any kinds of bugs somewhere else.

But i still dont get it .. Why i was able to run like 3dmark 01/03 + aquamark like 40 times w/o any bugs and in FC .. BANG ... bugged each times after like 5-10 mins ! I could test all my other games too .. but tested only my benchmarks + FC since FC is prolly the most demanding.

If you have any explanation for this.

RainZ

 

[BRiT]

One should always use a memory tester application when tweaking their system. An invaluable one is MemTest, available at http://www.memtest.org/ . It is based on this other program, memtest86, available at http://www.memtest86.com/ . The system is not stable until you can run all the tests multiple times without ANY errors occuring.

The likely reason why FC exhibited the bug and not the other programs are related to memory usage patterns. The Memtest86 page has an excellent write up on memory access patterns and issues.

There are many good approaches for testing memory. However, many tests simply throw some patterns at memory without much thought or knowledge of the memory architecture or how errors can best be detected. This works fine for hard memory failures but does little to find intermittent errors. BIOS based memory tests are useless for finding intermittent memory errors.

Memory chips consist of a large array of tightly packed memory cells, one for each bit of data. The vast majority of the intermittent failures are a result of interaction between these memory cells. Often writing a memory cell can cause one of the adjacent cells to be written with the same data. An effective memory test attempts to test for this condition. Therefore, an ideal strategy for testing memory would be the following:
1) write a cell with a zero
2) write all of the adjacent cells with a one, one or more times
3) check that the first cell still has a zero


It should be obvious that this strategy requires an exact knowledge of how the memory cells are laid out on the chip. In addition there is a never ending number of possible chip layouts for different chip types and manufacturers making this strategy impractical. However, there are testing algorithms that can approximate this ideal.

Memtest86 uses two algorithms that provide a reasonable approximation of the ideal test strategy above. The first of these strategies is called moving inversions. The moving inversion test works as follows:
1) Fill memory with a pattern
2) Starting at the lowest address
2a check that the pattern has not changed
2b write the patterns complement
2c increment the address
repeat 2a - 2c
3) Starting at the highest address
3a check that the pattern has not changed
3b write the patterns complement
3c decrement the address
repeat 3a - 3c


This algorithm is a good approximation of an ideal memory test but there are some limitations. Most high density chips today store data 4 to 16 bits wide. With chips that are more than one bit wide it is impossible to selectively read or write just one bit. This means that we cannot guarantee that all adjacent cells have been tested for interaction. In this case the best we can do is to use some patterns to insure that all adjacent cells have at least been written with all possible one and zero combinations.

It can also be seen that caching, buffering and out of order execution will interfere with the moving inversions algorithm and make less effective. It is possible to turn off cache but the memory buffering in new high performance chips can not be disabled. To address this limitation a new algorithm I call Modulo-X was created. This algorithm is not affected by cache or buffering. The algorithm works as follows:
1) For starting offsets of 0 - 20 do
1a write every 20th location with a pattern
1b write all other locations with the patterns complement
repeat 1b one or more times
1c check every 20th location for the pattern


This algorithm accomplishes nearly the same level of adjacency testing as moving inversions but is not affected by caching or buffering. Since separate write passes (1a, 1b) and the read pass (1c) are done for all of memory we can be assured that all of the buffers and cache have been flushed between passes. The selection of 20 as the stride size was somewhat arbitrary. Larger strides may be more effective but would take longer to execute. The choice of 20 seemed to be a reasonable compromise between speed and thoroughness.

 

[fallguy]

Yeh i agree i dont have any " lock " . And it's a good thing.

I dont know why you would call that a good thing. Running the AGP/PCI bus' out of spec isnt a good thing. It can cause problems. It might not be your problem, thats why I said it might be the problem.

If more volts fixed it for you thats great.

 

[rainz]

Yeh i agree i dont have any " lock " . And it's a good thing.

I dont know why you would call that a good thing. Running the AGP/PCI bus' out of spec isnt a good thing. It can cause problems. It might not be your problem, thats why I said it might be the problem.

If more volts fixed it for you thats great.

ok ok Fallguy it " might " sorry

RainZ

 

[rainz]

Here the result, i tested memtest86+ like you suggested BRiT and i have to admit .. I'm impressed since i wasnt expecting the best from my Kingston ValueRam. Well i have to now ?

Memtest86+ v1.15
----------------------
AMD Athlon 64 2200 MHz
L1 Cache: 128K 18029MB/s
L2 Cache: 1024K 3984MB/s
Memory: 1023M 1507MB/s
Chipset: VIA K8T800 ( ECC: Disabled )

Settings: RAM: 220 Mhz ( DDR440 ) / CAS: 3-3-3-8 / Single Channel ( 64 bits )

Walltime: 3:09:25 <-----
Cached: 1023M
RsvdMem: 880K
MemMap: e820-Std
Cache: off
ECC: off
Test: Std
Pass: 10 <-----
Errors: 0 <-----
ECC Errs: 0

No errors in more than 3 hours .. that's enough ? I have to admit ... i'm really impressed by my 400mhz Kingston Valueram ..

And now since it's stable i should set it in the Bios ... But i need some advices .. dunno which ones are the good one to set.

I wanna set it like it was tested in memtest86+ ( 3-3-3-8 )

My Bios Setting for the memory:
--------------------------------------

[.....] = Current setting
(.....) = The options i have

Memclock Mode [Limit] --- ( auto / limit )
Memclock to CPU Ratio [2:1 DDR400] --- ( 1:1 DDR200 / 4:3 DDR266 / 3:2 / 5:3 DDR333 / 2:1 DDR400 )
Bank Interleaving [auto] --- ( auto / disabled )
CAS Latency (CL) [auto] --- ( auto / 2.0 / 3.0 / 2.5 )
TRC [auto] --- ( auto / 7,8,9,10,11,12,13 CLK )
TRFC [auto] --- ( auto / 9,10,11,12,13,14,15 CLK )
TRCD [auto] --- ( auto / 2,3,4,5,6 CLK )
TWR [auto] --- ( auto / 2,3 CLK )
TRWT [auto] --- ( auto / 1,2,3,4,5,6 CLK )
TRAS [auto] --- ( auto / 5,6,7,8,9,10,11,12,13,14,15 CLK )
TRP [auto] --- ( auto / 2,3,4,5,6 )
DDR Clock Delay [auto] --- ( auto / 1,2,3,4,5,6,7,8 )

If someone is an expert in the memory setting in the Bios. I wanna know which ones are for the 3-3-3-8 !!!!!

And if it's better to let the "Memclock Mode" at [ limit or auto ] ?
And if it's better to let the "Memclock to CPU Ratio" at [ 2:1 DDR 400 or 3:2 or AUTO ] ?

Thx & Thx !

RainZ

 

[AAlcHemY]

Settings: RAM: 220 Mhz ( DDR440 ) / CAS: 3-3-3-8 / Single Channel ( 64 bits )

I wanna set it like it was tested in memtest86+ ( 3-3-3-8 )

My Bios Setting for the memory:
--------------------------------------

[.....] = Current setting
(.....) = The options i have

Memclock Mode [Limit] --- ( auto / limit )
Memclock to CPU Ratio [2:1 DDR400] --- ( 1:1 DDR200 / 4:3 DDR266 / 3:2 / 5:3 DDR333 / 2:1 DDR400 )
Bank Interleaving [auto] --- ( auto / disabled )
CAS Latency (CL) [auto] --- ( auto / 2.0 / 3.0 / 2.5 )
TRC ( Active to precharge time )[auto] --- ( auto / 7,8,9,10,11,12,13 CLK )
TRFC ( Refresh row cycle time )[auto] --- ( auto / 9,10,11,12,13,14,15 CLK )
TRCD ( RAS to CAS delay )[auto] --- ( auto / 2,3,4,5,6 CLK )
TWR ( Write recovery time )[auto] --- ( auto / 2,3 CLK )
TRWT ( Read to write delay )[auto] --- ( auto / 1,2,3,4,5,6 CLK )
TRAS ( Precharge to active delay ) [auto] --- ( auto / 5,6,7,8,9,10,11,12,13,14,15 CLK )
TRP ( RAS precharge delay )[auto] --- ( auto / 2,3,4,5,6 )
DDR Clock Delay [auto] --- ( auto / 1,2,3,4,5,6,7,8 )

If someone is an expert in the memory setting in the Bios. I wanna know which ones are for the 3-3-3-8 !!!!!

And if it's better to let the "Memclock Mode" at [ limit or auto ] ?
And if it's better to let the "Memclock to CPU Ratio" at [ 2:1 DDR 400 or 3:2 or AUTO ] ?

I'm no expert but the added text should give you 3-3-3-8.
The 'Memclock mode' has to be on auto. Its for limiting the ram module for running over its rated specifications. In your case, in stead of 210Mhz, it wil run at +/- 152Mhz, when 'Memclock mode' is set at 'limit'.

The 'Memclock to CPU Ratio' should be on 2:1.

Use CPU-Z for checking if the correct timings/frequencies are applied.

And to see if there is a actual improvement, Superpi requires a combination of good timings and high bandwith. Run the program on the '1M' setting, a good time is around 39-40sec. ( I got 41 sec with the XP-M on 210x 12/ 2-2-2-3 )

Also, this memory tutorial might be interesting.

Hope this helps.

 

[rainz]

Here.

http://www.xioserv.com/view.php?f=rainz.alchemy3.JPG

39s !!! I'ts good ? ( tested it 5 times ! 39s each times )

But there's a lil prob .. You can look at the screenshot for the info.

1= Setting when everything was set to AUTO by the Bios

2= Setting you said

3= Setting you said + modified 2 setting

CAS# Latency 2.5 to 3.0 ( to be like the first setting )
RAS# to CAS# Latency 5 CLK to 3 CLK ( Really not sure for this one .. but it was the only set at 5 and not like the auto setting )

Everything sounds good now ? ( #3 is the actual setting = 39s )

I'll test with the actual setting with memtest86+ again.

Thx !

RainZ

 

[AAlcHemY]

Here.
http://www.xioserv.com/view.php?f=rainz.alchemy3.JPG
39s !!! I'ts good ? ( tested it 5 times ! 39s each times )

CAS# Latency 2.5 to 3.0 ( to be like the first setting )
RAS# to CAS# Latency 5 CLK to 3 CLK ( Really not sure for this one .. but it was the only set at 5 and not like the auto setting )

Everything sounds good now ? ( #3 is the actual setting = 39s )

Thx !

RainZ

39sec, looks good.

But indeed, I made a little mistake
TRCD ( RAS to CAS delay )[auto] --- ( auto / 2,3,4,5,6 CLK ) and not 5
Only one thing left on the actual settings, CAS# Latency on 2.5. ( You could try 2 though. )

 

[rainz]

Here.
http://www.xioserv.com/view.php?f=rainz.alchemy3.JPG
39s !!! I'ts good ? ( tested it 5 times ! 39s each times )

CAS# Latency 2.5 to 3.0 ( to be like the first setting )
RAS# to CAS# Latency 5 CLK to 3 CLK ( Really not sure for this one .. but it was the only set at 5 and not like the auto setting )

Everything sounds good now ? ( #3 is the actual setting = 39s )

Thx !

RainZ

39sec, looks good.

But indeed, I made a little mistake
TRCD ( RAS to CAS delay )[auto] --- ( auto / 2,3,4,5,6 CLK ) and not 5
Only one thing left on the actual settings, CAS# Latency on 2.5. ( You could try 2 though. )

Hmmm i tested again this morning with ( CAS 3.0 ) 3-3-3-8 .. 4 hrs of memtest86+ ( 13 pass 0 errors ) so i can say it's a winning setting since i can do 39s in the SuperPI program.

And i tried like you said CAS 2.0 ... the system wont even boot ... totally frozen! and at 2.5 same thing .. wont boot .. unless i set the thing you said earlier at 5 .. so i tested with 2.5 + the RAS to CAS delay back to 5 ( when you made the mistake ) and i did the test in 40s ..

So i think it's better to keep my actual setting since it's the faster and after 7hrs20mins of testing no errors. right ?

And i could Overclock a little my ATI 9800 Pro with ATItools too .. but each times i run the test and when i see yellow artifacts ... yeh i guess it's normal since it try to reach the maximum for the mem / core but .. i hate that .. each times i see all the yellow artifacts .. i close the prog

What you think about that ?

Thx again for all your help !

RainZ

 

[rainz]

aah almost forgot..

1: Memclock Mode [Limit] --- ( auto / limit )
2: Memclock to CPU Ratio [2:1 DDR400] --- ( 1:1 DDR200 / 4:3 DDR266 / 3:2 / 5:3 DDR333 / 2:1 DDR400 )

1= i have to let this one to limit cause if i let the option to AUTO the other option ( memclock to CPU Ratio isnt there

2= and i set this one to 2:1 DDR400 like you said

Sounds good for that too ?

THANKS

RainZ

 

[AAlcHemY]

So i think it's better to keep my actual setting since it's the faster and after 7hrs20mins of testing no errors. right ?

Indeed, but in my experience the settings I gave would be faster, looks like the memory is holding you back. So just take the 39' timings.

And i could Overclock a little my ATI 9800 Pro with ATItools too ..
What you think about that ?

Try increasing the core frequency with 10Mhz at a time, then run 3D Mark 2003 for testing the stability. When its stable, 10Mhz up. If you get a hardlock, the core is clocked to high, decrease the frequency with 15-20Mhz then.

Then set the core back to normal and increase the memory clock, again in 10Mhz at a time. When you get artifacts or 'dancing' polygone's, the memory is clocked to high, again set the memory 15-20Mhz lower then. Now you have found a safe frequency for the core and memory.

Somehow I don't really trust those 'automatic' oc tools, its better when you do it yourself imo.

1= i have to let this one to limit cause if i let the option to AUTO the other option ( memclock to CPU Ratio isnt there

2= and i set this one to 2:1 DDR400 like you said

That sounds good, because the ram frequency ( In CPU-Z ) on the screenshot you gave is correct ( 220Mhz ). To be really sure, you can use SiSoft Sandra to measure the bandwith and compare to other systems.

PS, my nick is spelled with 2 A's.

 

[rainz]

http://www.xioserv.com/view.php?f=rainz.RainzCompOC.JPG

Here.. Is it decent AAlchemy ?

What's impress me the most .. i score faster than the FX-51 and almost as fast as the FX-53 Normal ?!

RainZ

 

[AAlcHemY]

What's impress me the most .. i score faster than the FX-51 and almost as fast as the FX-53 Normal ?!

RainZ

Numbers can't lie.

Welcome to the wonderfull world of overclocking my friend.

 

[rainz]

What's impress me the most .. i score faster than the FX-51 and almost as fast as the FX-53 Normal ?!

RainZ

Numbers can't lie.

Welcome to the wonderfull world of overclocking my friend.

YES ! YES ! and YES ! and a part of this good conclusion is thanks to you .

thank you so much you are really nice AAlchemy !

And for my video card .. dunno i think i'll wait :? i'm still affraid by the yellow artifacts And one thing annoying me hard .. i dont have the Temp. monitoring since it's not a XT ..

RainZ

 

[rainz]

AAlchemy any idea what this option in the bios do ?

HT Tristate Enable [ Enable / disable ] ( currently at disable on mine but i just saw this a review and it was at enable .. ?

http://www.amdmb.com/article-display.php?ArticleID=265&PageID=4

Thx

RainZ

 

[rainz]

Hmmm after reading lot of reviews about the benchmarking limit of my Asus K8V Deluxe .. i seems to be really LUCKY to be able to reach 220mhz 100% stable ..

Or .. maybe things have changed since the reviews ..most of reviews use 1002 Bios Version and now it's up to 1007 Bios Version.. Well no idea :?

W/o cpu clock multiplier and w/o lock for the speed on the AGP bus it's the problem. at 220mhz my AGP bus is around ? ~75mhz ? I'm prolly lucky at this speed.

EDIT:
A review with the 1005 Bios Version and the guy was able to reach 220mhz fully stable like me ! Prolly the Bios version help a lot

http://www.motherboards.org/articlesd/motherboard-reviews/1386_4.html

And it seems i might be able to crank it up to 225-230mhz if i set the Vlink voltage to 2.7 .. is it safe to run it 24/7 at this setting ? Cause i'll let my O/C 24/7 since the temp is fine. ( well IF my memory can handle it )

http://www.pcstats.com/articleview.cfm?articleid=1532&page=5

RainZ

 

[AAlcHemY]

HT Tristate Enable [ Enable / disable ] ( currently at disable on mine but i just saw this a review and it was at enable .. ?

To be honest, I don't know what it actually does, but it should increase performance when enabled.

W/o cpu clock multiplier and w/o lock for the speed on the AGP bus it's the problem. at 220mhz my AGP bus is around ? ~75mhz ? I'm prolly lucky at this speed.

You're lucky indeed, ATi cards don't like high agp speeds. Big difference to my Ti4400 who loves 100mhz agp

And it seems i might be able to crank it up to 225-230mhz if i set the Vlink voltage to 2.7 .. is it safe to run it 24/7 at this setting ? Cause i'll let my O/C 24/7 since the temp is fine. ( well IF my memory can handle it )

I would be safe 24/7 with 2.7 Vlink, but the memory is holding you back. :?

( Msn info added in the signature btw.)

 

[rainz]

HT Tristate Enable [ Enable / disable ] ( currently at disable on mine but i just saw this a review and it was at enable .. ?

To be honest, I don't know what it actually does, but it should increase performance when enabled.

W/o cpu clock multiplier and w/o lock for the speed on the AGP bus it's the problem. at 220mhz my AGP bus is around ? ~75mhz ? I'm prolly lucky at this speed.

You're lucky indeed, ATi cards don't like high agp speeds. Big difference to my Ti4400 who loves 100mhz agp

And it seems i might be able to crank it up to 225-230mhz if i set the Vlink voltage to 2.7 .. is it safe to run it 24/7 at this setting ? Cause i'll let my O/C 24/7 since the temp is fine. ( well IF my memory can handle it )

I would be safe 24/7 with 2.7 Vlink, but the memory is holding you back. :?

( Msn info added in the signature btw.)

http://www.hardocp.com/image.html?image=MTA3NjA0NDc3MjVGY1l6bG9oQ1VfMl80X2wuZ2lm

:? just enabled it .. not sure yet if it's better or not .. Trying to search exactly what this option do .. ( Enable link tristate during the disconnecte state of an LDTSTOP ) ... yeh yeh sure !

Tested it at 230mhz .. wont even boot .. at 225 i can boot but i'll test it later if it's stable .. for now i dont want to wait couple of hrs in tests.

RainZ

 

[fallguy]

Here is 230FSB for me,

I can do more, but havent played with it much yet. I just got this setup last Friday, and the stock bios doesnt allow you to change multipliers. There is another out that does... but I havent tried the beta one yet. I can boot at 235FSB, but it wasnt 100% stable. I believe I can get it to do more, and stable once I play with it more, and can change the multiplier.